You cannot select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
llm-course/Quantize_models_with_ExLlam...

8505 lines
513 KiB
Plaintext

This file contains invisible Unicode characters!

This file contains invisible Unicode characters that may be processed differently from what appears below. If your use case is intentional and legitimate, you can safely ignore this warning. Use the Escape button to reveal hidden characters.

This file contains ambiguous Unicode characters that may be confused with others in your current locale. If your use case is intentional and legitimate, you can safely ignore this warning. Use the Escape button to highlight these characters.

{
"nbformat": 4,
"nbformat_minor": 0,
"metadata": {
"colab": {
"provenance": [],
"gpuType": "T4",
"authorship_tag": "ABX9TyMkzVB0crkJL3VvXFtgSkwZ",
"include_colab_link": true
},
"kernelspec": {
"name": "python3",
"display_name": "Python 3"
},
"language_info": {
"name": "python"
},
"accelerator": "GPU",
"widgets": {
"application/vnd.jupyter.widget-state+json": {
"da62083d1a3d4d62b76a92abec717a24": {
"model_module": "@jupyter-widgets/controls",
"model_name": "VBoxModel",
"model_module_version": "1.5.0",
"state": {
"_dom_classes": [],
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "VBoxModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/controls",
"_view_module_version": "1.5.0",
"_view_name": "VBoxView",
"box_style": "",
"children": [
"IPY_MODEL_12051bdac0aa466c91122f3cb0e1ab2b",
"IPY_MODEL_e9569646ef72451496087c49a2487ddc",
"IPY_MODEL_23fc55e87153404d90f55930f7f73988",
"IPY_MODEL_903957b2c1404703b2e59a2875da77ec"
],
"layout": "IPY_MODEL_05c1e6ba7977465092a69324888cf59e"
}
},
"9109e78e934040a7a4fedc9e9a2b5751": {
"model_module": "@jupyter-widgets/controls",
"model_name": "HTMLModel",
"model_module_version": "1.5.0",
"state": {
"_dom_classes": [],
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "HTMLModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/controls",
"_view_module_version": "1.5.0",
"_view_name": "HTMLView",
"description": "",
"description_tooltip": null,
"layout": "IPY_MODEL_26073e1c5d284656be9c5fdb55276a25",
"placeholder": "",
"style": "IPY_MODEL_51b1646d7e194c75b1109d16b8291c76",
"value": "<center> <img\nsrc=https://huggingface.co/front/assets/huggingface_logo-noborder.svg\nalt='Hugging Face'> <br> Copy a token from <a\nhref=\"https://huggingface.co/settings/tokens\" target=\"_blank\">your Hugging Face\ntokens page</a> and paste it below. <br> Immediately click login after copying\nyour token or it might be stored in plain text in this notebook file. </center>"
}
},
"e2bf96c3e5e946f5a832f3d5bcf86202": {
"model_module": "@jupyter-widgets/controls",
"model_name": "PasswordModel",
"model_module_version": "1.5.0",
"state": {
"_dom_classes": [],
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "PasswordModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/controls",
"_view_module_version": "1.5.0",
"_view_name": "PasswordView",
"continuous_update": true,
"description": "Token:",
"description_tooltip": null,
"disabled": false,
"layout": "IPY_MODEL_3efa5c88dc2d491c99678d73e4144eeb",
"placeholder": "",
"style": "IPY_MODEL_b1dfa66aec6a4e86bff78e3a62905a3c",
"value": ""
}
},
"d83804efc2c34b628cda200bb40fa8db": {
"model_module": "@jupyter-widgets/controls",
"model_name": "CheckboxModel",
"model_module_version": "1.5.0",
"state": {
"_dom_classes": [],
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "CheckboxModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/controls",
"_view_module_version": "1.5.0",
"_view_name": "CheckboxView",
"description": "Add token as git credential?",
"description_tooltip": null,
"disabled": false,
"indent": true,
"layout": "IPY_MODEL_5e4cc08b7aa44cacae18fcc4131c174d",
"style": "IPY_MODEL_813d16aa8a144cf7b3c1b08d096a2b20",
"value": true
}
},
"401d9fb01f054e6aaaf62b1546ab7686": {
"model_module": "@jupyter-widgets/controls",
"model_name": "ButtonModel",
"model_module_version": "1.5.0",
"state": {
"_dom_classes": [],
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "ButtonModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/controls",
"_view_module_version": "1.5.0",
"_view_name": "ButtonView",
"button_style": "",
"description": "Login",
"disabled": false,
"icon": "",
"layout": "IPY_MODEL_63c15146f75f473394ebcc165ca0d7f6",
"style": "IPY_MODEL_cc10d97351604ee18918afa1d955a089",
"tooltip": ""
}
},
"9f6d67cf6b9748288044a8c7a98cc2e3": {
"model_module": "@jupyter-widgets/controls",
"model_name": "HTMLModel",
"model_module_version": "1.5.0",
"state": {
"_dom_classes": [],
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "HTMLModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/controls",
"_view_module_version": "1.5.0",
"_view_name": "HTMLView",
"description": "",
"description_tooltip": null,
"layout": "IPY_MODEL_6ff94a654ed54915b2082197920e89ab",
"placeholder": "",
"style": "IPY_MODEL_dc9118ad41d247cab135811a183805e8",
"value": "\n<b>Pro Tip:</b> If you don't already have one, you can create a dedicated\n'notebooks' token with 'write' access, that you can then easily reuse for all\nnotebooks. </center>"
}
},
"05c1e6ba7977465092a69324888cf59e": {
"model_module": "@jupyter-widgets/base",
"model_name": "LayoutModel",
"model_module_version": "1.2.0",
"state": {
"_model_module": "@jupyter-widgets/base",
"_model_module_version": "1.2.0",
"_model_name": "LayoutModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "LayoutView",
"align_content": null,
"align_items": "center",
"align_self": null,
"border": null,
"bottom": null,
"display": "flex",
"flex": null,
"flex_flow": "column",
"grid_area": null,
"grid_auto_columns": null,
"grid_auto_flow": null,
"grid_auto_rows": null,
"grid_column": null,
"grid_gap": null,
"grid_row": null,
"grid_template_areas": null,
"grid_template_columns": null,
"grid_template_rows": null,
"height": null,
"justify_content": null,
"justify_items": null,
"left": null,
"margin": null,
"max_height": null,
"max_width": null,
"min_height": null,
"min_width": null,
"object_fit": null,
"object_position": null,
"order": null,
"overflow": null,
"overflow_x": null,
"overflow_y": null,
"padding": null,
"right": null,
"top": null,
"visibility": null,
"width": "50%"
}
},
"26073e1c5d284656be9c5fdb55276a25": {
"model_module": "@jupyter-widgets/base",
"model_name": "LayoutModel",
"model_module_version": "1.2.0",
"state": {
"_model_module": "@jupyter-widgets/base",
"_model_module_version": "1.2.0",
"_model_name": "LayoutModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "LayoutView",
"align_content": null,
"align_items": null,
"align_self": null,
"border": null,
"bottom": null,
"display": null,
"flex": null,
"flex_flow": null,
"grid_area": null,
"grid_auto_columns": null,
"grid_auto_flow": null,
"grid_auto_rows": null,
"grid_column": null,
"grid_gap": null,
"grid_row": null,
"grid_template_areas": null,
"grid_template_columns": null,
"grid_template_rows": null,
"height": null,
"justify_content": null,
"justify_items": null,
"left": null,
"margin": null,
"max_height": null,
"max_width": null,
"min_height": null,
"min_width": null,
"object_fit": null,
"object_position": null,
"order": null,
"overflow": null,
"overflow_x": null,
"overflow_y": null,
"padding": null,
"right": null,
"top": null,
"visibility": null,
"width": null
}
},
"51b1646d7e194c75b1109d16b8291c76": {
"model_module": "@jupyter-widgets/controls",
"model_name": "DescriptionStyleModel",
"model_module_version": "1.5.0",
"state": {
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "DescriptionStyleModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "StyleView",
"description_width": ""
}
},
"3efa5c88dc2d491c99678d73e4144eeb": {
"model_module": "@jupyter-widgets/base",
"model_name": "LayoutModel",
"model_module_version": "1.2.0",
"state": {
"_model_module": "@jupyter-widgets/base",
"_model_module_version": "1.2.0",
"_model_name": "LayoutModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "LayoutView",
"align_content": null,
"align_items": null,
"align_self": null,
"border": null,
"bottom": null,
"display": null,
"flex": null,
"flex_flow": null,
"grid_area": null,
"grid_auto_columns": null,
"grid_auto_flow": null,
"grid_auto_rows": null,
"grid_column": null,
"grid_gap": null,
"grid_row": null,
"grid_template_areas": null,
"grid_template_columns": null,
"grid_template_rows": null,
"height": null,
"justify_content": null,
"justify_items": null,
"left": null,
"margin": null,
"max_height": null,
"max_width": null,
"min_height": null,
"min_width": null,
"object_fit": null,
"object_position": null,
"order": null,
"overflow": null,
"overflow_x": null,
"overflow_y": null,
"padding": null,
"right": null,
"top": null,
"visibility": null,
"width": null
}
},
"b1dfa66aec6a4e86bff78e3a62905a3c": {
"model_module": "@jupyter-widgets/controls",
"model_name": "DescriptionStyleModel",
"model_module_version": "1.5.0",
"state": {
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "DescriptionStyleModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "StyleView",
"description_width": ""
}
},
"5e4cc08b7aa44cacae18fcc4131c174d": {
"model_module": "@jupyter-widgets/base",
"model_name": "LayoutModel",
"model_module_version": "1.2.0",
"state": {
"_model_module": "@jupyter-widgets/base",
"_model_module_version": "1.2.0",
"_model_name": "LayoutModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "LayoutView",
"align_content": null,
"align_items": null,
"align_self": null,
"border": null,
"bottom": null,
"display": null,
"flex": null,
"flex_flow": null,
"grid_area": null,
"grid_auto_columns": null,
"grid_auto_flow": null,
"grid_auto_rows": null,
"grid_column": null,
"grid_gap": null,
"grid_row": null,
"grid_template_areas": null,
"grid_template_columns": null,
"grid_template_rows": null,
"height": null,
"justify_content": null,
"justify_items": null,
"left": null,
"margin": null,
"max_height": null,
"max_width": null,
"min_height": null,
"min_width": null,
"object_fit": null,
"object_position": null,
"order": null,
"overflow": null,
"overflow_x": null,
"overflow_y": null,
"padding": null,
"right": null,
"top": null,
"visibility": null,
"width": null
}
},
"813d16aa8a144cf7b3c1b08d096a2b20": {
"model_module": "@jupyter-widgets/controls",
"model_name": "DescriptionStyleModel",
"model_module_version": "1.5.0",
"state": {
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "DescriptionStyleModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "StyleView",
"description_width": ""
}
},
"63c15146f75f473394ebcc165ca0d7f6": {
"model_module": "@jupyter-widgets/base",
"model_name": "LayoutModel",
"model_module_version": "1.2.0",
"state": {
"_model_module": "@jupyter-widgets/base",
"_model_module_version": "1.2.0",
"_model_name": "LayoutModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "LayoutView",
"align_content": null,
"align_items": null,
"align_self": null,
"border": null,
"bottom": null,
"display": null,
"flex": null,
"flex_flow": null,
"grid_area": null,
"grid_auto_columns": null,
"grid_auto_flow": null,
"grid_auto_rows": null,
"grid_column": null,
"grid_gap": null,
"grid_row": null,
"grid_template_areas": null,
"grid_template_columns": null,
"grid_template_rows": null,
"height": null,
"justify_content": null,
"justify_items": null,
"left": null,
"margin": null,
"max_height": null,
"max_width": null,
"min_height": null,
"min_width": null,
"object_fit": null,
"object_position": null,
"order": null,
"overflow": null,
"overflow_x": null,
"overflow_y": null,
"padding": null,
"right": null,
"top": null,
"visibility": null,
"width": null
}
},
"cc10d97351604ee18918afa1d955a089": {
"model_module": "@jupyter-widgets/controls",
"model_name": "ButtonStyleModel",
"model_module_version": "1.5.0",
"state": {
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "ButtonStyleModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "StyleView",
"button_color": null,
"font_weight": ""
}
},
"6ff94a654ed54915b2082197920e89ab": {
"model_module": "@jupyter-widgets/base",
"model_name": "LayoutModel",
"model_module_version": "1.2.0",
"state": {
"_model_module": "@jupyter-widgets/base",
"_model_module_version": "1.2.0",
"_model_name": "LayoutModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "LayoutView",
"align_content": null,
"align_items": null,
"align_self": null,
"border": null,
"bottom": null,
"display": null,
"flex": null,
"flex_flow": null,
"grid_area": null,
"grid_auto_columns": null,
"grid_auto_flow": null,
"grid_auto_rows": null,
"grid_column": null,
"grid_gap": null,
"grid_row": null,
"grid_template_areas": null,
"grid_template_columns": null,
"grid_template_rows": null,
"height": null,
"justify_content": null,
"justify_items": null,
"left": null,
"margin": null,
"max_height": null,
"max_width": null,
"min_height": null,
"min_width": null,
"object_fit": null,
"object_position": null,
"order": null,
"overflow": null,
"overflow_x": null,
"overflow_y": null,
"padding": null,
"right": null,
"top": null,
"visibility": null,
"width": null
}
},
"dc9118ad41d247cab135811a183805e8": {
"model_module": "@jupyter-widgets/controls",
"model_name": "DescriptionStyleModel",
"model_module_version": "1.5.0",
"state": {
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "DescriptionStyleModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "StyleView",
"description_width": ""
}
},
"3e285ebb0b7d414fb5ab6ee02ccb4f50": {
"model_module": "@jupyter-widgets/controls",
"model_name": "LabelModel",
"model_module_version": "1.5.0",
"state": {
"_dom_classes": [],
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "LabelModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/controls",
"_view_module_version": "1.5.0",
"_view_name": "LabelView",
"description": "",
"description_tooltip": null,
"layout": "IPY_MODEL_ae75d86eb24241d082669a422df4180c",
"placeholder": "",
"style": "IPY_MODEL_f8e648aa342c43aa9e960622f897e5c5",
"value": "Connecting..."
}
},
"ae75d86eb24241d082669a422df4180c": {
"model_module": "@jupyter-widgets/base",
"model_name": "LayoutModel",
"model_module_version": "1.2.0",
"state": {
"_model_module": "@jupyter-widgets/base",
"_model_module_version": "1.2.0",
"_model_name": "LayoutModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "LayoutView",
"align_content": null,
"align_items": null,
"align_self": null,
"border": null,
"bottom": null,
"display": null,
"flex": null,
"flex_flow": null,
"grid_area": null,
"grid_auto_columns": null,
"grid_auto_flow": null,
"grid_auto_rows": null,
"grid_column": null,
"grid_gap": null,
"grid_row": null,
"grid_template_areas": null,
"grid_template_columns": null,
"grid_template_rows": null,
"height": null,
"justify_content": null,
"justify_items": null,
"left": null,
"margin": null,
"max_height": null,
"max_width": null,
"min_height": null,
"min_width": null,
"object_fit": null,
"object_position": null,
"order": null,
"overflow": null,
"overflow_x": null,
"overflow_y": null,
"padding": null,
"right": null,
"top": null,
"visibility": null,
"width": null
}
},
"f8e648aa342c43aa9e960622f897e5c5": {
"model_module": "@jupyter-widgets/controls",
"model_name": "DescriptionStyleModel",
"model_module_version": "1.5.0",
"state": {
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "DescriptionStyleModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "StyleView",
"description_width": ""
}
},
"12051bdac0aa466c91122f3cb0e1ab2b": {
"model_module": "@jupyter-widgets/controls",
"model_name": "LabelModel",
"model_module_version": "1.5.0",
"state": {
"_dom_classes": [],
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "LabelModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/controls",
"_view_module_version": "1.5.0",
"_view_name": "LabelView",
"description": "",
"description_tooltip": null,
"layout": "IPY_MODEL_cfbb97be5d284781b35abeca297bbad9",
"placeholder": "",
"style": "IPY_MODEL_ab04ee5ca65f45fc9368a80847926198",
"value": "Token is valid (permission: write)."
}
},
"e9569646ef72451496087c49a2487ddc": {
"model_module": "@jupyter-widgets/controls",
"model_name": "LabelModel",
"model_module_version": "1.5.0",
"state": {
"_dom_classes": [],
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "LabelModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/controls",
"_view_module_version": "1.5.0",
"_view_name": "LabelView",
"description": "",
"description_tooltip": null,
"layout": "IPY_MODEL_0fd23d208e5942439a7e83797ac9fe64",
"placeholder": "",
"style": "IPY_MODEL_e1da3e30982746959ec7ebd180a3bbb9",
"value": "Your token has been saved in your configured git credential helpers (store)."
}
},
"23fc55e87153404d90f55930f7f73988": {
"model_module": "@jupyter-widgets/controls",
"model_name": "LabelModel",
"model_module_version": "1.5.0",
"state": {
"_dom_classes": [],
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "LabelModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/controls",
"_view_module_version": "1.5.0",
"_view_name": "LabelView",
"description": "",
"description_tooltip": null,
"layout": "IPY_MODEL_0836d18df07244afb40d2cc2f6a3879e",
"placeholder": "",
"style": "IPY_MODEL_79189477733f4aff8e62c4cbf318e91b",
"value": "Your token has been saved to /root/.cache/huggingface/token"
}
},
"903957b2c1404703b2e59a2875da77ec": {
"model_module": "@jupyter-widgets/controls",
"model_name": "LabelModel",
"model_module_version": "1.5.0",
"state": {
"_dom_classes": [],
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "LabelModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/controls",
"_view_module_version": "1.5.0",
"_view_name": "LabelView",
"description": "",
"description_tooltip": null,
"layout": "IPY_MODEL_4a4d70a7a8dd4af4b3147161193b1885",
"placeholder": "",
"style": "IPY_MODEL_b72edebfa2a548b49c47bad5bc7aeecd",
"value": "Login successful"
}
},
"cfbb97be5d284781b35abeca297bbad9": {
"model_module": "@jupyter-widgets/base",
"model_name": "LayoutModel",
"model_module_version": "1.2.0",
"state": {
"_model_module": "@jupyter-widgets/base",
"_model_module_version": "1.2.0",
"_model_name": "LayoutModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "LayoutView",
"align_content": null,
"align_items": null,
"align_self": null,
"border": null,
"bottom": null,
"display": null,
"flex": null,
"flex_flow": null,
"grid_area": null,
"grid_auto_columns": null,
"grid_auto_flow": null,
"grid_auto_rows": null,
"grid_column": null,
"grid_gap": null,
"grid_row": null,
"grid_template_areas": null,
"grid_template_columns": null,
"grid_template_rows": null,
"height": null,
"justify_content": null,
"justify_items": null,
"left": null,
"margin": null,
"max_height": null,
"max_width": null,
"min_height": null,
"min_width": null,
"object_fit": null,
"object_position": null,
"order": null,
"overflow": null,
"overflow_x": null,
"overflow_y": null,
"padding": null,
"right": null,
"top": null,
"visibility": null,
"width": null
}
},
"ab04ee5ca65f45fc9368a80847926198": {
"model_module": "@jupyter-widgets/controls",
"model_name": "DescriptionStyleModel",
"model_module_version": "1.5.0",
"state": {
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "DescriptionStyleModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "StyleView",
"description_width": ""
}
},
"0fd23d208e5942439a7e83797ac9fe64": {
"model_module": "@jupyter-widgets/base",
"model_name": "LayoutModel",
"model_module_version": "1.2.0",
"state": {
"_model_module": "@jupyter-widgets/base",
"_model_module_version": "1.2.0",
"_model_name": "LayoutModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "LayoutView",
"align_content": null,
"align_items": null,
"align_self": null,
"border": null,
"bottom": null,
"display": null,
"flex": null,
"flex_flow": null,
"grid_area": null,
"grid_auto_columns": null,
"grid_auto_flow": null,
"grid_auto_rows": null,
"grid_column": null,
"grid_gap": null,
"grid_row": null,
"grid_template_areas": null,
"grid_template_columns": null,
"grid_template_rows": null,
"height": null,
"justify_content": null,
"justify_items": null,
"left": null,
"margin": null,
"max_height": null,
"max_width": null,
"min_height": null,
"min_width": null,
"object_fit": null,
"object_position": null,
"order": null,
"overflow": null,
"overflow_x": null,
"overflow_y": null,
"padding": null,
"right": null,
"top": null,
"visibility": null,
"width": null
}
},
"e1da3e30982746959ec7ebd180a3bbb9": {
"model_module": "@jupyter-widgets/controls",
"model_name": "DescriptionStyleModel",
"model_module_version": "1.5.0",
"state": {
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "DescriptionStyleModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "StyleView",
"description_width": ""
}
},
"0836d18df07244afb40d2cc2f6a3879e": {
"model_module": "@jupyter-widgets/base",
"model_name": "LayoutModel",
"model_module_version": "1.2.0",
"state": {
"_model_module": "@jupyter-widgets/base",
"_model_module_version": "1.2.0",
"_model_name": "LayoutModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "LayoutView",
"align_content": null,
"align_items": null,
"align_self": null,
"border": null,
"bottom": null,
"display": null,
"flex": null,
"flex_flow": null,
"grid_area": null,
"grid_auto_columns": null,
"grid_auto_flow": null,
"grid_auto_rows": null,
"grid_column": null,
"grid_gap": null,
"grid_row": null,
"grid_template_areas": null,
"grid_template_columns": null,
"grid_template_rows": null,
"height": null,
"justify_content": null,
"justify_items": null,
"left": null,
"margin": null,
"max_height": null,
"max_width": null,
"min_height": null,
"min_width": null,
"object_fit": null,
"object_position": null,
"order": null,
"overflow": null,
"overflow_x": null,
"overflow_y": null,
"padding": null,
"right": null,
"top": null,
"visibility": null,
"width": null
}
},
"79189477733f4aff8e62c4cbf318e91b": {
"model_module": "@jupyter-widgets/controls",
"model_name": "DescriptionStyleModel",
"model_module_version": "1.5.0",
"state": {
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "DescriptionStyleModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "StyleView",
"description_width": ""
}
},
"4a4d70a7a8dd4af4b3147161193b1885": {
"model_module": "@jupyter-widgets/base",
"model_name": "LayoutModel",
"model_module_version": "1.2.0",
"state": {
"_model_module": "@jupyter-widgets/base",
"_model_module_version": "1.2.0",
"_model_name": "LayoutModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "LayoutView",
"align_content": null,
"align_items": null,
"align_self": null,
"border": null,
"bottom": null,
"display": null,
"flex": null,
"flex_flow": null,
"grid_area": null,
"grid_auto_columns": null,
"grid_auto_flow": null,
"grid_auto_rows": null,
"grid_column": null,
"grid_gap": null,
"grid_row": null,
"grid_template_areas": null,
"grid_template_columns": null,
"grid_template_rows": null,
"height": null,
"justify_content": null,
"justify_items": null,
"left": null,
"margin": null,
"max_height": null,
"max_width": null,
"min_height": null,
"min_width": null,
"object_fit": null,
"object_position": null,
"order": null,
"overflow": null,
"overflow_x": null,
"overflow_y": null,
"padding": null,
"right": null,
"top": null,
"visibility": null,
"width": null
}
},
"b72edebfa2a548b49c47bad5bc7aeecd": {
"model_module": "@jupyter-widgets/controls",
"model_name": "DescriptionStyleModel",
"model_module_version": "1.5.0",
"state": {
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "DescriptionStyleModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "StyleView",
"description_width": ""
}
},
"98916f7a1809458e8ebc568f770ed544": {
"model_module": "@jupyter-widgets/controls",
"model_name": "HBoxModel",
"model_module_version": "1.5.0",
"state": {
"_dom_classes": [],
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "HBoxModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/controls",
"_view_module_version": "1.5.0",
"_view_name": "HBoxView",
"box_style": "",
"children": [
"IPY_MODEL_2c6ba20e4b384d6fa46fbd832daf36e2",
"IPY_MODEL_a6920b63b1cb4c1297bc5a18c7e30f7e",
"IPY_MODEL_c18e0920993f4189ab927b167fb9a046"
],
"layout": "IPY_MODEL_13565f1b6e6e4fa7a365160dad1cdb42"
}
},
"2c6ba20e4b384d6fa46fbd832daf36e2": {
"model_module": "@jupyter-widgets/controls",
"model_name": "HTMLModel",
"model_module_version": "1.5.0",
"state": {
"_dom_classes": [],
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "HTMLModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/controls",
"_view_module_version": "1.5.0",
"_view_name": "HTMLView",
"description": "",
"description_tooltip": null,
"layout": "IPY_MODEL_fa3308e0b8fb469d8549199e8bfe2278",
"placeholder": "",
"style": "IPY_MODEL_e7672f0406bd4f2c8a0ef83fd390e49d",
"value": "cal_data.safetensors: 100%"
}
},
"a6920b63b1cb4c1297bc5a18c7e30f7e": {
"model_module": "@jupyter-widgets/controls",
"model_name": "FloatProgressModel",
"model_module_version": "1.5.0",
"state": {
"_dom_classes": [],
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "FloatProgressModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/controls",
"_view_module_version": "1.5.0",
"_view_name": "ProgressView",
"bar_style": "success",
"description": "",
"description_tooltip": null,
"layout": "IPY_MODEL_22b9c33a7c084ccd9e8878bda3748486",
"max": 1638488,
"min": 0,
"orientation": "horizontal",
"style": "IPY_MODEL_2d6aa1ecdf7444b4b4cb81515d3654be",
"value": 1638488
}
},
"c18e0920993f4189ab927b167fb9a046": {
"model_module": "@jupyter-widgets/controls",
"model_name": "HTMLModel",
"model_module_version": "1.5.0",
"state": {
"_dom_classes": [],
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "HTMLModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/controls",
"_view_module_version": "1.5.0",
"_view_name": "HTMLView",
"description": "",
"description_tooltip": null,
"layout": "IPY_MODEL_6cedbad90ca9423f85a48d7e4073a378",
"placeholder": "",
"style": "IPY_MODEL_c893eb5da4e34e4999981b3810a4205b",
"value": " 1.64M/1.64M [00:01&lt;00:00, 22.1kB/s]"
}
},
"13565f1b6e6e4fa7a365160dad1cdb42": {
"model_module": "@jupyter-widgets/base",
"model_name": "LayoutModel",
"model_module_version": "1.2.0",
"state": {
"_model_module": "@jupyter-widgets/base",
"_model_module_version": "1.2.0",
"_model_name": "LayoutModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "LayoutView",
"align_content": null,
"align_items": null,
"align_self": null,
"border": null,
"bottom": null,
"display": null,
"flex": null,
"flex_flow": null,
"grid_area": null,
"grid_auto_columns": null,
"grid_auto_flow": null,
"grid_auto_rows": null,
"grid_column": null,
"grid_gap": null,
"grid_row": null,
"grid_template_areas": null,
"grid_template_columns": null,
"grid_template_rows": null,
"height": null,
"justify_content": null,
"justify_items": null,
"left": null,
"margin": null,
"max_height": null,
"max_width": null,
"min_height": null,
"min_width": null,
"object_fit": null,
"object_position": null,
"order": null,
"overflow": null,
"overflow_x": null,
"overflow_y": null,
"padding": null,
"right": null,
"top": null,
"visibility": null,
"width": null
}
},
"fa3308e0b8fb469d8549199e8bfe2278": {
"model_module": "@jupyter-widgets/base",
"model_name": "LayoutModel",
"model_module_version": "1.2.0",
"state": {
"_model_module": "@jupyter-widgets/base",
"_model_module_version": "1.2.0",
"_model_name": "LayoutModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "LayoutView",
"align_content": null,
"align_items": null,
"align_self": null,
"border": null,
"bottom": null,
"display": null,
"flex": null,
"flex_flow": null,
"grid_area": null,
"grid_auto_columns": null,
"grid_auto_flow": null,
"grid_auto_rows": null,
"grid_column": null,
"grid_gap": null,
"grid_row": null,
"grid_template_areas": null,
"grid_template_columns": null,
"grid_template_rows": null,
"height": null,
"justify_content": null,
"justify_items": null,
"left": null,
"margin": null,
"max_height": null,
"max_width": null,
"min_height": null,
"min_width": null,
"object_fit": null,
"object_position": null,
"order": null,
"overflow": null,
"overflow_x": null,
"overflow_y": null,
"padding": null,
"right": null,
"top": null,
"visibility": null,
"width": null
}
},
"e7672f0406bd4f2c8a0ef83fd390e49d": {
"model_module": "@jupyter-widgets/controls",
"model_name": "DescriptionStyleModel",
"model_module_version": "1.5.0",
"state": {
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "DescriptionStyleModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "StyleView",
"description_width": ""
}
},
"22b9c33a7c084ccd9e8878bda3748486": {
"model_module": "@jupyter-widgets/base",
"model_name": "LayoutModel",
"model_module_version": "1.2.0",
"state": {
"_model_module": "@jupyter-widgets/base",
"_model_module_version": "1.2.0",
"_model_name": "LayoutModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "LayoutView",
"align_content": null,
"align_items": null,
"align_self": null,
"border": null,
"bottom": null,
"display": null,
"flex": null,
"flex_flow": null,
"grid_area": null,
"grid_auto_columns": null,
"grid_auto_flow": null,
"grid_auto_rows": null,
"grid_column": null,
"grid_gap": null,
"grid_row": null,
"grid_template_areas": null,
"grid_template_columns": null,
"grid_template_rows": null,
"height": null,
"justify_content": null,
"justify_items": null,
"left": null,
"margin": null,
"max_height": null,
"max_width": null,
"min_height": null,
"min_width": null,
"object_fit": null,
"object_position": null,
"order": null,
"overflow": null,
"overflow_x": null,
"overflow_y": null,
"padding": null,
"right": null,
"top": null,
"visibility": null,
"width": null
}
},
"2d6aa1ecdf7444b4b4cb81515d3654be": {
"model_module": "@jupyter-widgets/controls",
"model_name": "ProgressStyleModel",
"model_module_version": "1.5.0",
"state": {
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "ProgressStyleModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "StyleView",
"bar_color": null,
"description_width": ""
}
},
"6cedbad90ca9423f85a48d7e4073a378": {
"model_module": "@jupyter-widgets/base",
"model_name": "LayoutModel",
"model_module_version": "1.2.0",
"state": {
"_model_module": "@jupyter-widgets/base",
"_model_module_version": "1.2.0",
"_model_name": "LayoutModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "LayoutView",
"align_content": null,
"align_items": null,
"align_self": null,
"border": null,
"bottom": null,
"display": null,
"flex": null,
"flex_flow": null,
"grid_area": null,
"grid_auto_columns": null,
"grid_auto_flow": null,
"grid_auto_rows": null,
"grid_column": null,
"grid_gap": null,
"grid_row": null,
"grid_template_areas": null,
"grid_template_columns": null,
"grid_template_rows": null,
"height": null,
"justify_content": null,
"justify_items": null,
"left": null,
"margin": null,
"max_height": null,
"max_width": null,
"min_height": null,
"min_width": null,
"object_fit": null,
"object_position": null,
"order": null,
"overflow": null,
"overflow_x": null,
"overflow_y": null,
"padding": null,
"right": null,
"top": null,
"visibility": null,
"width": null
}
},
"c893eb5da4e34e4999981b3810a4205b": {
"model_module": "@jupyter-widgets/controls",
"model_name": "DescriptionStyleModel",
"model_module_version": "1.5.0",
"state": {
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "DescriptionStyleModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "StyleView",
"description_width": ""
}
},
"2a5efce36f38406189e82e2e27b06bab": {
"model_module": "@jupyter-widgets/controls",
"model_name": "HBoxModel",
"model_module_version": "1.5.0",
"state": {
"_dom_classes": [],
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "HBoxModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/controls",
"_view_module_version": "1.5.0",
"_view_name": "HBoxView",
"box_style": "",
"children": [
"IPY_MODEL_3c828089249743e6aa4f3ee60f77caee",
"IPY_MODEL_543b02a2098043ca849ecdda4f4c7a81",
"IPY_MODEL_9b076b5c6cf74e56abb01cbcb59d1deb"
],
"layout": "IPY_MODEL_ac08f05888c64726aabcb9a1cfd65aba"
}
},
"3c828089249743e6aa4f3ee60f77caee": {
"model_module": "@jupyter-widgets/controls",
"model_name": "HTMLModel",
"model_module_version": "1.5.0",
"state": {
"_dom_classes": [],
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "HTMLModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/controls",
"_view_module_version": "1.5.0",
"_view_name": "HTMLView",
"description": "",
"description_tooltip": null,
"layout": "IPY_MODEL_4f48b906c779402187567aad8b4ba488",
"placeholder": "",
"style": "IPY_MODEL_e004498b56524804b28661cde5fdce81",
"value": "output.safetensors: 100%"
}
},
"543b02a2098043ca849ecdda4f4c7a81": {
"model_module": "@jupyter-widgets/controls",
"model_name": "FloatProgressModel",
"model_module_version": "1.5.0",
"state": {
"_dom_classes": [],
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "FloatProgressModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/controls",
"_view_module_version": "1.5.0",
"_view_name": "ProgressView",
"bar_style": "success",
"description": "",
"description_tooltip": null,
"layout": "IPY_MODEL_9931d545676d49039aba007f13904155",
"max": 4735018016,
"min": 0,
"orientation": "horizontal",
"style": "IPY_MODEL_130049623d6744dcac8a7ffa25e1e1d1",
"value": 4735018016
}
},
"9b076b5c6cf74e56abb01cbcb59d1deb": {
"model_module": "@jupyter-widgets/controls",
"model_name": "HTMLModel",
"model_module_version": "1.5.0",
"state": {
"_dom_classes": [],
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "HTMLModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/controls",
"_view_module_version": "1.5.0",
"_view_name": "HTMLView",
"description": "",
"description_tooltip": null,
"layout": "IPY_MODEL_f48ddd7d4d304e10b3fabc3cdc541dbf",
"placeholder": "",
"style": "IPY_MODEL_bcdd93e578fd49e8b1bb76f43bde3483",
"value": " 4.74G/4.74G [02:21&lt;00:00, 32.3MB/s]"
}
},
"ac08f05888c64726aabcb9a1cfd65aba": {
"model_module": "@jupyter-widgets/base",
"model_name": "LayoutModel",
"model_module_version": "1.2.0",
"state": {
"_model_module": "@jupyter-widgets/base",
"_model_module_version": "1.2.0",
"_model_name": "LayoutModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "LayoutView",
"align_content": null,
"align_items": null,
"align_self": null,
"border": null,
"bottom": null,
"display": null,
"flex": null,
"flex_flow": null,
"grid_area": null,
"grid_auto_columns": null,
"grid_auto_flow": null,
"grid_auto_rows": null,
"grid_column": null,
"grid_gap": null,
"grid_row": null,
"grid_template_areas": null,
"grid_template_columns": null,
"grid_template_rows": null,
"height": null,
"justify_content": null,
"justify_items": null,
"left": null,
"margin": null,
"max_height": null,
"max_width": null,
"min_height": null,
"min_width": null,
"object_fit": null,
"object_position": null,
"order": null,
"overflow": null,
"overflow_x": null,
"overflow_y": null,
"padding": null,
"right": null,
"top": null,
"visibility": null,
"width": null
}
},
"4f48b906c779402187567aad8b4ba488": {
"model_module": "@jupyter-widgets/base",
"model_name": "LayoutModel",
"model_module_version": "1.2.0",
"state": {
"_model_module": "@jupyter-widgets/base",
"_model_module_version": "1.2.0",
"_model_name": "LayoutModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "LayoutView",
"align_content": null,
"align_items": null,
"align_self": null,
"border": null,
"bottom": null,
"display": null,
"flex": null,
"flex_flow": null,
"grid_area": null,
"grid_auto_columns": null,
"grid_auto_flow": null,
"grid_auto_rows": null,
"grid_column": null,
"grid_gap": null,
"grid_row": null,
"grid_template_areas": null,
"grid_template_columns": null,
"grid_template_rows": null,
"height": null,
"justify_content": null,
"justify_items": null,
"left": null,
"margin": null,
"max_height": null,
"max_width": null,
"min_height": null,
"min_width": null,
"object_fit": null,
"object_position": null,
"order": null,
"overflow": null,
"overflow_x": null,
"overflow_y": null,
"padding": null,
"right": null,
"top": null,
"visibility": null,
"width": null
}
},
"e004498b56524804b28661cde5fdce81": {
"model_module": "@jupyter-widgets/controls",
"model_name": "DescriptionStyleModel",
"model_module_version": "1.5.0",
"state": {
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "DescriptionStyleModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "StyleView",
"description_width": ""
}
},
"9931d545676d49039aba007f13904155": {
"model_module": "@jupyter-widgets/base",
"model_name": "LayoutModel",
"model_module_version": "1.2.0",
"state": {
"_model_module": "@jupyter-widgets/base",
"_model_module_version": "1.2.0",
"_model_name": "LayoutModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "LayoutView",
"align_content": null,
"align_items": null,
"align_self": null,
"border": null,
"bottom": null,
"display": null,
"flex": null,
"flex_flow": null,
"grid_area": null,
"grid_auto_columns": null,
"grid_auto_flow": null,
"grid_auto_rows": null,
"grid_column": null,
"grid_gap": null,
"grid_row": null,
"grid_template_areas": null,
"grid_template_columns": null,
"grid_template_rows": null,
"height": null,
"justify_content": null,
"justify_items": null,
"left": null,
"margin": null,
"max_height": null,
"max_width": null,
"min_height": null,
"min_width": null,
"object_fit": null,
"object_position": null,
"order": null,
"overflow": null,
"overflow_x": null,
"overflow_y": null,
"padding": null,
"right": null,
"top": null,
"visibility": null,
"width": null
}
},
"130049623d6744dcac8a7ffa25e1e1d1": {
"model_module": "@jupyter-widgets/controls",
"model_name": "ProgressStyleModel",
"model_module_version": "1.5.0",
"state": {
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "ProgressStyleModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "StyleView",
"bar_color": null,
"description_width": ""
}
},
"f48ddd7d4d304e10b3fabc3cdc541dbf": {
"model_module": "@jupyter-widgets/base",
"model_name": "LayoutModel",
"model_module_version": "1.2.0",
"state": {
"_model_module": "@jupyter-widgets/base",
"_model_module_version": "1.2.0",
"_model_name": "LayoutModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "LayoutView",
"align_content": null,
"align_items": null,
"align_self": null,
"border": null,
"bottom": null,
"display": null,
"flex": null,
"flex_flow": null,
"grid_area": null,
"grid_auto_columns": null,
"grid_auto_flow": null,
"grid_auto_rows": null,
"grid_column": null,
"grid_gap": null,
"grid_row": null,
"grid_template_areas": null,
"grid_template_columns": null,
"grid_template_rows": null,
"height": null,
"justify_content": null,
"justify_items": null,
"left": null,
"margin": null,
"max_height": null,
"max_width": null,
"min_height": null,
"min_width": null,
"object_fit": null,
"object_position": null,
"order": null,
"overflow": null,
"overflow_x": null,
"overflow_y": null,
"padding": null,
"right": null,
"top": null,
"visibility": null,
"width": null
}
},
"bcdd93e578fd49e8b1bb76f43bde3483": {
"model_module": "@jupyter-widgets/controls",
"model_name": "DescriptionStyleModel",
"model_module_version": "1.5.0",
"state": {
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "DescriptionStyleModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "StyleView",
"description_width": ""
}
},
"a051f03b2e8340dea8b8d8f0389f1d76": {
"model_module": "@jupyter-widgets/controls",
"model_name": "HBoxModel",
"model_module_version": "1.5.0",
"state": {
"_dom_classes": [],
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "HBoxModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/controls",
"_view_module_version": "1.5.0",
"_view_name": "HBoxView",
"box_style": "",
"children": [
"IPY_MODEL_17c15d87496d4e5ea25814806b8606ea",
"IPY_MODEL_71093ecee7ac45798d8bcb924af2906c",
"IPY_MODEL_5a0d9471e6cd4f2988f042c4a4e48a3d"
],
"layout": "IPY_MODEL_841d7197d6324552a0fbb6d6f5ac7601"
}
},
"17c15d87496d4e5ea25814806b8606ea": {
"model_module": "@jupyter-widgets/controls",
"model_name": "HTMLModel",
"model_module_version": "1.5.0",
"state": {
"_dom_classes": [],
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "HTMLModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/controls",
"_view_module_version": "1.5.0",
"_view_name": "HTMLView",
"description": "",
"description_tooltip": null,
"layout": "IPY_MODEL_3c3d9d10ee9e466ba170b3dd3aa3ecdd",
"placeholder": "",
"style": "IPY_MODEL_fa9d061965d74d0eb6d728c1f4058845",
"value": "Upload 5 LFS files: 100%"
}
},
"71093ecee7ac45798d8bcb924af2906c": {
"model_module": "@jupyter-widgets/controls",
"model_name": "FloatProgressModel",
"model_module_version": "1.5.0",
"state": {
"_dom_classes": [],
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "FloatProgressModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/controls",
"_view_module_version": "1.5.0",
"_view_name": "ProgressView",
"bar_style": "success",
"description": "",
"description_tooltip": null,
"layout": "IPY_MODEL_62de7ed888614d26875a0ecd99cb996f",
"max": 5,
"min": 0,
"orientation": "horizontal",
"style": "IPY_MODEL_89dc60db3b9a4a1587b00130ffde2311",
"value": 5
}
},
"5a0d9471e6cd4f2988f042c4a4e48a3d": {
"model_module": "@jupyter-widgets/controls",
"model_name": "HTMLModel",
"model_module_version": "1.5.0",
"state": {
"_dom_classes": [],
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "HTMLModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/controls",
"_view_module_version": "1.5.0",
"_view_name": "HTMLView",
"description": "",
"description_tooltip": null,
"layout": "IPY_MODEL_9d0851b2c54c4e5cb635cd2652890824",
"placeholder": "",
"style": "IPY_MODEL_7a89919a4de9454aa8a7c134f1b1fa3d",
"value": " 5/5 [02:22&lt;00:00, 57.57s/it]"
}
},
"841d7197d6324552a0fbb6d6f5ac7601": {
"model_module": "@jupyter-widgets/base",
"model_name": "LayoutModel",
"model_module_version": "1.2.0",
"state": {
"_model_module": "@jupyter-widgets/base",
"_model_module_version": "1.2.0",
"_model_name": "LayoutModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "LayoutView",
"align_content": null,
"align_items": null,
"align_self": null,
"border": null,
"bottom": null,
"display": null,
"flex": null,
"flex_flow": null,
"grid_area": null,
"grid_auto_columns": null,
"grid_auto_flow": null,
"grid_auto_rows": null,
"grid_column": null,
"grid_gap": null,
"grid_row": null,
"grid_template_areas": null,
"grid_template_columns": null,
"grid_template_rows": null,
"height": null,
"justify_content": null,
"justify_items": null,
"left": null,
"margin": null,
"max_height": null,
"max_width": null,
"min_height": null,
"min_width": null,
"object_fit": null,
"object_position": null,
"order": null,
"overflow": null,
"overflow_x": null,
"overflow_y": null,
"padding": null,
"right": null,
"top": null,
"visibility": null,
"width": null
}
},
"3c3d9d10ee9e466ba170b3dd3aa3ecdd": {
"model_module": "@jupyter-widgets/base",
"model_name": "LayoutModel",
"model_module_version": "1.2.0",
"state": {
"_model_module": "@jupyter-widgets/base",
"_model_module_version": "1.2.0",
"_model_name": "LayoutModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "LayoutView",
"align_content": null,
"align_items": null,
"align_self": null,
"border": null,
"bottom": null,
"display": null,
"flex": null,
"flex_flow": null,
"grid_area": null,
"grid_auto_columns": null,
"grid_auto_flow": null,
"grid_auto_rows": null,
"grid_column": null,
"grid_gap": null,
"grid_row": null,
"grid_template_areas": null,
"grid_template_columns": null,
"grid_template_rows": null,
"height": null,
"justify_content": null,
"justify_items": null,
"left": null,
"margin": null,
"max_height": null,
"max_width": null,
"min_height": null,
"min_width": null,
"object_fit": null,
"object_position": null,
"order": null,
"overflow": null,
"overflow_x": null,
"overflow_y": null,
"padding": null,
"right": null,
"top": null,
"visibility": null,
"width": null
}
},
"fa9d061965d74d0eb6d728c1f4058845": {
"model_module": "@jupyter-widgets/controls",
"model_name": "DescriptionStyleModel",
"model_module_version": "1.5.0",
"state": {
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "DescriptionStyleModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "StyleView",
"description_width": ""
}
},
"62de7ed888614d26875a0ecd99cb996f": {
"model_module": "@jupyter-widgets/base",
"model_name": "LayoutModel",
"model_module_version": "1.2.0",
"state": {
"_model_module": "@jupyter-widgets/base",
"_model_module_version": "1.2.0",
"_model_name": "LayoutModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "LayoutView",
"align_content": null,
"align_items": null,
"align_self": null,
"border": null,
"bottom": null,
"display": null,
"flex": null,
"flex_flow": null,
"grid_area": null,
"grid_auto_columns": null,
"grid_auto_flow": null,
"grid_auto_rows": null,
"grid_column": null,
"grid_gap": null,
"grid_row": null,
"grid_template_areas": null,
"grid_template_columns": null,
"grid_template_rows": null,
"height": null,
"justify_content": null,
"justify_items": null,
"left": null,
"margin": null,
"max_height": null,
"max_width": null,
"min_height": null,
"min_width": null,
"object_fit": null,
"object_position": null,
"order": null,
"overflow": null,
"overflow_x": null,
"overflow_y": null,
"padding": null,
"right": null,
"top": null,
"visibility": null,
"width": null
}
},
"89dc60db3b9a4a1587b00130ffde2311": {
"model_module": "@jupyter-widgets/controls",
"model_name": "ProgressStyleModel",
"model_module_version": "1.5.0",
"state": {
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "ProgressStyleModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "StyleView",
"bar_color": null,
"description_width": ""
}
},
"9d0851b2c54c4e5cb635cd2652890824": {
"model_module": "@jupyter-widgets/base",
"model_name": "LayoutModel",
"model_module_version": "1.2.0",
"state": {
"_model_module": "@jupyter-widgets/base",
"_model_module_version": "1.2.0",
"_model_name": "LayoutModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "LayoutView",
"align_content": null,
"align_items": null,
"align_self": null,
"border": null,
"bottom": null,
"display": null,
"flex": null,
"flex_flow": null,
"grid_area": null,
"grid_auto_columns": null,
"grid_auto_flow": null,
"grid_auto_rows": null,
"grid_column": null,
"grid_gap": null,
"grid_row": null,
"grid_template_areas": null,
"grid_template_columns": null,
"grid_template_rows": null,
"height": null,
"justify_content": null,
"justify_items": null,
"left": null,
"margin": null,
"max_height": null,
"max_width": null,
"min_height": null,
"min_width": null,
"object_fit": null,
"object_position": null,
"order": null,
"overflow": null,
"overflow_x": null,
"overflow_y": null,
"padding": null,
"right": null,
"top": null,
"visibility": null,
"width": null
}
},
"7a89919a4de9454aa8a7c134f1b1fa3d": {
"model_module": "@jupyter-widgets/controls",
"model_name": "DescriptionStyleModel",
"model_module_version": "1.5.0",
"state": {
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "DescriptionStyleModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "StyleView",
"description_width": ""
}
},
"30fb290999b84d9eb69b8dd720c534b5": {
"model_module": "@jupyter-widgets/controls",
"model_name": "HBoxModel",
"model_module_version": "1.5.0",
"state": {
"_dom_classes": [],
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "HBoxModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/controls",
"_view_module_version": "1.5.0",
"_view_name": "HBoxView",
"box_style": "",
"children": [
"IPY_MODEL_37a67b2e108248858858b72cf2d00aba",
"IPY_MODEL_a038d308888c40bca1cc45c962344ec6",
"IPY_MODEL_70bd55b8cd2f42d6b9ecd76c17972771"
],
"layout": "IPY_MODEL_e18504b6c48a45ef8abe75f2d59147ea"
}
},
"37a67b2e108248858858b72cf2d00aba": {
"model_module": "@jupyter-widgets/controls",
"model_name": "HTMLModel",
"model_module_version": "1.5.0",
"state": {
"_dom_classes": [],
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "HTMLModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/controls",
"_view_module_version": "1.5.0",
"_view_name": "HTMLView",
"description": "",
"description_tooltip": null,
"layout": "IPY_MODEL_b8e8032b75cf4b1eadfe76395f8811ac",
"placeholder": "",
"style": "IPY_MODEL_dd6c2ab6494c48878a7521c0fc0e2109",
"value": "input_states.safetensors: 100%"
}
},
"a038d308888c40bca1cc45c962344ec6": {
"model_module": "@jupyter-widgets/controls",
"model_name": "FloatProgressModel",
"model_module_version": "1.5.0",
"state": {
"_dom_classes": [],
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "FloatProgressModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/controls",
"_view_module_version": "1.5.0",
"_view_name": "ProgressView",
"bar_style": "success",
"description": "",
"description_tooltip": null,
"layout": "IPY_MODEL_6e98141655cc48d48f47f3981fb385e6",
"max": 1677721696,
"min": 0,
"orientation": "horizontal",
"style": "IPY_MODEL_fbf6c4c6ed85465095eb3ae986994c69",
"value": 1677721696
}
},
"70bd55b8cd2f42d6b9ecd76c17972771": {
"model_module": "@jupyter-widgets/controls",
"model_name": "HTMLModel",
"model_module_version": "1.5.0",
"state": {
"_dom_classes": [],
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "HTMLModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/controls",
"_view_module_version": "1.5.0",
"_view_name": "HTMLView",
"description": "",
"description_tooltip": null,
"layout": "IPY_MODEL_95addcc750ad45ecbc37c5c7f463d812",
"placeholder": "",
"style": "IPY_MODEL_0e8403a6428b4387a1ebbdecf7c7aa03",
"value": " 1.68G/1.68G [00:52&lt;00:00, 40.4MB/s]"
}
},
"e18504b6c48a45ef8abe75f2d59147ea": {
"model_module": "@jupyter-widgets/base",
"model_name": "LayoutModel",
"model_module_version": "1.2.0",
"state": {
"_model_module": "@jupyter-widgets/base",
"_model_module_version": "1.2.0",
"_model_name": "LayoutModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "LayoutView",
"align_content": null,
"align_items": null,
"align_self": null,
"border": null,
"bottom": null,
"display": null,
"flex": null,
"flex_flow": null,
"grid_area": null,
"grid_auto_columns": null,
"grid_auto_flow": null,
"grid_auto_rows": null,
"grid_column": null,
"grid_gap": null,
"grid_row": null,
"grid_template_areas": null,
"grid_template_columns": null,
"grid_template_rows": null,
"height": null,
"justify_content": null,
"justify_items": null,
"left": null,
"margin": null,
"max_height": null,
"max_width": null,
"min_height": null,
"min_width": null,
"object_fit": null,
"object_position": null,
"order": null,
"overflow": null,
"overflow_x": null,
"overflow_y": null,
"padding": null,
"right": null,
"top": null,
"visibility": null,
"width": null
}
},
"b8e8032b75cf4b1eadfe76395f8811ac": {
"model_module": "@jupyter-widgets/base",
"model_name": "LayoutModel",
"model_module_version": "1.2.0",
"state": {
"_model_module": "@jupyter-widgets/base",
"_model_module_version": "1.2.0",
"_model_name": "LayoutModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "LayoutView",
"align_content": null,
"align_items": null,
"align_self": null,
"border": null,
"bottom": null,
"display": null,
"flex": null,
"flex_flow": null,
"grid_area": null,
"grid_auto_columns": null,
"grid_auto_flow": null,
"grid_auto_rows": null,
"grid_column": null,
"grid_gap": null,
"grid_row": null,
"grid_template_areas": null,
"grid_template_columns": null,
"grid_template_rows": null,
"height": null,
"justify_content": null,
"justify_items": null,
"left": null,
"margin": null,
"max_height": null,
"max_width": null,
"min_height": null,
"min_width": null,
"object_fit": null,
"object_position": null,
"order": null,
"overflow": null,
"overflow_x": null,
"overflow_y": null,
"padding": null,
"right": null,
"top": null,
"visibility": null,
"width": null
}
},
"dd6c2ab6494c48878a7521c0fc0e2109": {
"model_module": "@jupyter-widgets/controls",
"model_name": "DescriptionStyleModel",
"model_module_version": "1.5.0",
"state": {
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "DescriptionStyleModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "StyleView",
"description_width": ""
}
},
"6e98141655cc48d48f47f3981fb385e6": {
"model_module": "@jupyter-widgets/base",
"model_name": "LayoutModel",
"model_module_version": "1.2.0",
"state": {
"_model_module": "@jupyter-widgets/base",
"_model_module_version": "1.2.0",
"_model_name": "LayoutModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "LayoutView",
"align_content": null,
"align_items": null,
"align_self": null,
"border": null,
"bottom": null,
"display": null,
"flex": null,
"flex_flow": null,
"grid_area": null,
"grid_auto_columns": null,
"grid_auto_flow": null,
"grid_auto_rows": null,
"grid_column": null,
"grid_gap": null,
"grid_row": null,
"grid_template_areas": null,
"grid_template_columns": null,
"grid_template_rows": null,
"height": null,
"justify_content": null,
"justify_items": null,
"left": null,
"margin": null,
"max_height": null,
"max_width": null,
"min_height": null,
"min_width": null,
"object_fit": null,
"object_position": null,
"order": null,
"overflow": null,
"overflow_x": null,
"overflow_y": null,
"padding": null,
"right": null,
"top": null,
"visibility": null,
"width": null
}
},
"fbf6c4c6ed85465095eb3ae986994c69": {
"model_module": "@jupyter-widgets/controls",
"model_name": "ProgressStyleModel",
"model_module_version": "1.5.0",
"state": {
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "ProgressStyleModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "StyleView",
"bar_color": null,
"description_width": ""
}
},
"95addcc750ad45ecbc37c5c7f463d812": {
"model_module": "@jupyter-widgets/base",
"model_name": "LayoutModel",
"model_module_version": "1.2.0",
"state": {
"_model_module": "@jupyter-widgets/base",
"_model_module_version": "1.2.0",
"_model_name": "LayoutModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "LayoutView",
"align_content": null,
"align_items": null,
"align_self": null,
"border": null,
"bottom": null,
"display": null,
"flex": null,
"flex_flow": null,
"grid_area": null,
"grid_auto_columns": null,
"grid_auto_flow": null,
"grid_auto_rows": null,
"grid_column": null,
"grid_gap": null,
"grid_row": null,
"grid_template_areas": null,
"grid_template_columns": null,
"grid_template_rows": null,
"height": null,
"justify_content": null,
"justify_items": null,
"left": null,
"margin": null,
"max_height": null,
"max_width": null,
"min_height": null,
"min_width": null,
"object_fit": null,
"object_position": null,
"order": null,
"overflow": null,
"overflow_x": null,
"overflow_y": null,
"padding": null,
"right": null,
"top": null,
"visibility": null,
"width": null
}
},
"0e8403a6428b4387a1ebbdecf7c7aa03": {
"model_module": "@jupyter-widgets/controls",
"model_name": "DescriptionStyleModel",
"model_module_version": "1.5.0",
"state": {
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "DescriptionStyleModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "StyleView",
"description_width": ""
}
},
"945dffb008b44010807bf11ae1dbfec2": {
"model_module": "@jupyter-widgets/controls",
"model_name": "HBoxModel",
"model_module_version": "1.5.0",
"state": {
"_dom_classes": [],
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "HBoxModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/controls",
"_view_module_version": "1.5.0",
"_view_name": "HBoxView",
"box_style": "",
"children": [
"IPY_MODEL_9a28c85b06bc49ce9c36534766b7650f",
"IPY_MODEL_46964345de59401ab8bc5f51dcfa94fd",
"IPY_MODEL_60b0f63948dd47fcb6d8aaab5bfedf5c"
],
"layout": "IPY_MODEL_edc938cff4ec477b92ecb768f3ab2fab"
}
},
"9a28c85b06bc49ce9c36534766b7650f": {
"model_module": "@jupyter-widgets/controls",
"model_name": "HTMLModel",
"model_module_version": "1.5.0",
"state": {
"_dom_classes": [],
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "HTMLModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/controls",
"_view_module_version": "1.5.0",
"_view_name": "HTMLView",
"description": "",
"description_tooltip": null,
"layout": "IPY_MODEL_7d3c1a673c614c819f5b83b8d9fb61ef",
"placeholder": "",
"style": "IPY_MODEL_7a8b92bd2ea34c2181f6dcaaea9c5dee",
"value": "training_args.bin: 100%"
}
},
"46964345de59401ab8bc5f51dcfa94fd": {
"model_module": "@jupyter-widgets/controls",
"model_name": "FloatProgressModel",
"model_module_version": "1.5.0",
"state": {
"_dom_classes": [],
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "FloatProgressModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/controls",
"_view_module_version": "1.5.0",
"_view_name": "ProgressView",
"bar_style": "success",
"description": "",
"description_tooltip": null,
"layout": "IPY_MODEL_6f83a7cd4e664e8aaf847196344d1256",
"max": 5435,
"min": 0,
"orientation": "horizontal",
"style": "IPY_MODEL_145c52ca456d4052af8b458578858176",
"value": 5435
}
},
"60b0f63948dd47fcb6d8aaab5bfedf5c": {
"model_module": "@jupyter-widgets/controls",
"model_name": "HTMLModel",
"model_module_version": "1.5.0",
"state": {
"_dom_classes": [],
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "HTMLModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/controls",
"_view_module_version": "1.5.0",
"_view_name": "HTMLView",
"description": "",
"description_tooltip": null,
"layout": "IPY_MODEL_b3d7552080874350a3688dab8838452c",
"placeholder": "",
"style": "IPY_MODEL_63e84ef3cff94a929602b8b419cd8071",
"value": " 5.43k/5.43k [00:00&lt;00:00, 8.05kB/s]"
}
},
"edc938cff4ec477b92ecb768f3ab2fab": {
"model_module": "@jupyter-widgets/base",
"model_name": "LayoutModel",
"model_module_version": "1.2.0",
"state": {
"_model_module": "@jupyter-widgets/base",
"_model_module_version": "1.2.0",
"_model_name": "LayoutModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "LayoutView",
"align_content": null,
"align_items": null,
"align_self": null,
"border": null,
"bottom": null,
"display": null,
"flex": null,
"flex_flow": null,
"grid_area": null,
"grid_auto_columns": null,
"grid_auto_flow": null,
"grid_auto_rows": null,
"grid_column": null,
"grid_gap": null,
"grid_row": null,
"grid_template_areas": null,
"grid_template_columns": null,
"grid_template_rows": null,
"height": null,
"justify_content": null,
"justify_items": null,
"left": null,
"margin": null,
"max_height": null,
"max_width": null,
"min_height": null,
"min_width": null,
"object_fit": null,
"object_position": null,
"order": null,
"overflow": null,
"overflow_x": null,
"overflow_y": null,
"padding": null,
"right": null,
"top": null,
"visibility": null,
"width": null
}
},
"7d3c1a673c614c819f5b83b8d9fb61ef": {
"model_module": "@jupyter-widgets/base",
"model_name": "LayoutModel",
"model_module_version": "1.2.0",
"state": {
"_model_module": "@jupyter-widgets/base",
"_model_module_version": "1.2.0",
"_model_name": "LayoutModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "LayoutView",
"align_content": null,
"align_items": null,
"align_self": null,
"border": null,
"bottom": null,
"display": null,
"flex": null,
"flex_flow": null,
"grid_area": null,
"grid_auto_columns": null,
"grid_auto_flow": null,
"grid_auto_rows": null,
"grid_column": null,
"grid_gap": null,
"grid_row": null,
"grid_template_areas": null,
"grid_template_columns": null,
"grid_template_rows": null,
"height": null,
"justify_content": null,
"justify_items": null,
"left": null,
"margin": null,
"max_height": null,
"max_width": null,
"min_height": null,
"min_width": null,
"object_fit": null,
"object_position": null,
"order": null,
"overflow": null,
"overflow_x": null,
"overflow_y": null,
"padding": null,
"right": null,
"top": null,
"visibility": null,
"width": null
}
},
"7a8b92bd2ea34c2181f6dcaaea9c5dee": {
"model_module": "@jupyter-widgets/controls",
"model_name": "DescriptionStyleModel",
"model_module_version": "1.5.0",
"state": {
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "DescriptionStyleModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "StyleView",
"description_width": ""
}
},
"6f83a7cd4e664e8aaf847196344d1256": {
"model_module": "@jupyter-widgets/base",
"model_name": "LayoutModel",
"model_module_version": "1.2.0",
"state": {
"_model_module": "@jupyter-widgets/base",
"_model_module_version": "1.2.0",
"_model_name": "LayoutModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "LayoutView",
"align_content": null,
"align_items": null,
"align_self": null,
"border": null,
"bottom": null,
"display": null,
"flex": null,
"flex_flow": null,
"grid_area": null,
"grid_auto_columns": null,
"grid_auto_flow": null,
"grid_auto_rows": null,
"grid_column": null,
"grid_gap": null,
"grid_row": null,
"grid_template_areas": null,
"grid_template_columns": null,
"grid_template_rows": null,
"height": null,
"justify_content": null,
"justify_items": null,
"left": null,
"margin": null,
"max_height": null,
"max_width": null,
"min_height": null,
"min_width": null,
"object_fit": null,
"object_position": null,
"order": null,
"overflow": null,
"overflow_x": null,
"overflow_y": null,
"padding": null,
"right": null,
"top": null,
"visibility": null,
"width": null
}
},
"145c52ca456d4052af8b458578858176": {
"model_module": "@jupyter-widgets/controls",
"model_name": "ProgressStyleModel",
"model_module_version": "1.5.0",
"state": {
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "ProgressStyleModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "StyleView",
"bar_color": null,
"description_width": ""
}
},
"b3d7552080874350a3688dab8838452c": {
"model_module": "@jupyter-widgets/base",
"model_name": "LayoutModel",
"model_module_version": "1.2.0",
"state": {
"_model_module": "@jupyter-widgets/base",
"_model_module_version": "1.2.0",
"_model_name": "LayoutModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "LayoutView",
"align_content": null,
"align_items": null,
"align_self": null,
"border": null,
"bottom": null,
"display": null,
"flex": null,
"flex_flow": null,
"grid_area": null,
"grid_auto_columns": null,
"grid_auto_flow": null,
"grid_auto_rows": null,
"grid_column": null,
"grid_gap": null,
"grid_row": null,
"grid_template_areas": null,
"grid_template_columns": null,
"grid_template_rows": null,
"height": null,
"justify_content": null,
"justify_items": null,
"left": null,
"margin": null,
"max_height": null,
"max_width": null,
"min_height": null,
"min_width": null,
"object_fit": null,
"object_position": null,
"order": null,
"overflow": null,
"overflow_x": null,
"overflow_y": null,
"padding": null,
"right": null,
"top": null,
"visibility": null,
"width": null
}
},
"63e84ef3cff94a929602b8b419cd8071": {
"model_module": "@jupyter-widgets/controls",
"model_name": "DescriptionStyleModel",
"model_module_version": "1.5.0",
"state": {
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "DescriptionStyleModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "StyleView",
"description_width": ""
}
},
"55b59177fdb6441ba564d5f3f5dfa422": {
"model_module": "@jupyter-widgets/controls",
"model_name": "HBoxModel",
"model_module_version": "1.5.0",
"state": {
"_dom_classes": [],
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "HBoxModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/controls",
"_view_module_version": "1.5.0",
"_view_name": "HBoxView",
"box_style": "",
"children": [
"IPY_MODEL_52d6cbef58b947119609460d41917399",
"IPY_MODEL_264f7e06376748dda35453413b382949",
"IPY_MODEL_1b84d95ea138413d8d8d1ea997741f86"
],
"layout": "IPY_MODEL_8f581db41cd54f89a81f8424042190cf"
}
},
"52d6cbef58b947119609460d41917399": {
"model_module": "@jupyter-widgets/controls",
"model_name": "HTMLModel",
"model_module_version": "1.5.0",
"state": {
"_dom_classes": [],
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "HTMLModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/controls",
"_view_module_version": "1.5.0",
"_view_name": "HTMLView",
"description": "",
"description_tooltip": null,
"layout": "IPY_MODEL_15ca60c283cb4fa389ed92e9433af851",
"placeholder": "",
"style": "IPY_MODEL_ee220712a2e0467a8d07c473749a958c",
"value": "tokenizer.model: 100%"
}
},
"264f7e06376748dda35453413b382949": {
"model_module": "@jupyter-widgets/controls",
"model_name": "FloatProgressModel",
"model_module_version": "1.5.0",
"state": {
"_dom_classes": [],
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "FloatProgressModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/controls",
"_view_module_version": "1.5.0",
"_view_name": "ProgressView",
"bar_style": "success",
"description": "",
"description_tooltip": null,
"layout": "IPY_MODEL_689c0aca3273499e820bad016d3ccaa9",
"max": 493443,
"min": 0,
"orientation": "horizontal",
"style": "IPY_MODEL_a2096f9286ab486e9baad9f349a23a95",
"value": 493443
}
},
"1b84d95ea138413d8d8d1ea997741f86": {
"model_module": "@jupyter-widgets/controls",
"model_name": "HTMLModel",
"model_module_version": "1.5.0",
"state": {
"_dom_classes": [],
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "HTMLModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/controls",
"_view_module_version": "1.5.0",
"_view_name": "HTMLView",
"description": "",
"description_tooltip": null,
"layout": "IPY_MODEL_1bcc515d9d764c908af2988e29eaeeed",
"placeholder": "",
"style": "IPY_MODEL_ca437b02cab741489217ad4d8780f728",
"value": " 493k/493k [00:00&lt;00:00, 22.9kB/s]"
}
},
"8f581db41cd54f89a81f8424042190cf": {
"model_module": "@jupyter-widgets/base",
"model_name": "LayoutModel",
"model_module_version": "1.2.0",
"state": {
"_model_module": "@jupyter-widgets/base",
"_model_module_version": "1.2.0",
"_model_name": "LayoutModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "LayoutView",
"align_content": null,
"align_items": null,
"align_self": null,
"border": null,
"bottom": null,
"display": null,
"flex": null,
"flex_flow": null,
"grid_area": null,
"grid_auto_columns": null,
"grid_auto_flow": null,
"grid_auto_rows": null,
"grid_column": null,
"grid_gap": null,
"grid_row": null,
"grid_template_areas": null,
"grid_template_columns": null,
"grid_template_rows": null,
"height": null,
"justify_content": null,
"justify_items": null,
"left": null,
"margin": null,
"max_height": null,
"max_width": null,
"min_height": null,
"min_width": null,
"object_fit": null,
"object_position": null,
"order": null,
"overflow": null,
"overflow_x": null,
"overflow_y": null,
"padding": null,
"right": null,
"top": null,
"visibility": null,
"width": null
}
},
"15ca60c283cb4fa389ed92e9433af851": {
"model_module": "@jupyter-widgets/base",
"model_name": "LayoutModel",
"model_module_version": "1.2.0",
"state": {
"_model_module": "@jupyter-widgets/base",
"_model_module_version": "1.2.0",
"_model_name": "LayoutModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "LayoutView",
"align_content": null,
"align_items": null,
"align_self": null,
"border": null,
"bottom": null,
"display": null,
"flex": null,
"flex_flow": null,
"grid_area": null,
"grid_auto_columns": null,
"grid_auto_flow": null,
"grid_auto_rows": null,
"grid_column": null,
"grid_gap": null,
"grid_row": null,
"grid_template_areas": null,
"grid_template_columns": null,
"grid_template_rows": null,
"height": null,
"justify_content": null,
"justify_items": null,
"left": null,
"margin": null,
"max_height": null,
"max_width": null,
"min_height": null,
"min_width": null,
"object_fit": null,
"object_position": null,
"order": null,
"overflow": null,
"overflow_x": null,
"overflow_y": null,
"padding": null,
"right": null,
"top": null,
"visibility": null,
"width": null
}
},
"ee220712a2e0467a8d07c473749a958c": {
"model_module": "@jupyter-widgets/controls",
"model_name": "DescriptionStyleModel",
"model_module_version": "1.5.0",
"state": {
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "DescriptionStyleModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "StyleView",
"description_width": ""
}
},
"689c0aca3273499e820bad016d3ccaa9": {
"model_module": "@jupyter-widgets/base",
"model_name": "LayoutModel",
"model_module_version": "1.2.0",
"state": {
"_model_module": "@jupyter-widgets/base",
"_model_module_version": "1.2.0",
"_model_name": "LayoutModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "LayoutView",
"align_content": null,
"align_items": null,
"align_self": null,
"border": null,
"bottom": null,
"display": null,
"flex": null,
"flex_flow": null,
"grid_area": null,
"grid_auto_columns": null,
"grid_auto_flow": null,
"grid_auto_rows": null,
"grid_column": null,
"grid_gap": null,
"grid_row": null,
"grid_template_areas": null,
"grid_template_columns": null,
"grid_template_rows": null,
"height": null,
"justify_content": null,
"justify_items": null,
"left": null,
"margin": null,
"max_height": null,
"max_width": null,
"min_height": null,
"min_width": null,
"object_fit": null,
"object_position": null,
"order": null,
"overflow": null,
"overflow_x": null,
"overflow_y": null,
"padding": null,
"right": null,
"top": null,
"visibility": null,
"width": null
}
},
"a2096f9286ab486e9baad9f349a23a95": {
"model_module": "@jupyter-widgets/controls",
"model_name": "ProgressStyleModel",
"model_module_version": "1.5.0",
"state": {
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "ProgressStyleModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "StyleView",
"bar_color": null,
"description_width": ""
}
},
"1bcc515d9d764c908af2988e29eaeeed": {
"model_module": "@jupyter-widgets/base",
"model_name": "LayoutModel",
"model_module_version": "1.2.0",
"state": {
"_model_module": "@jupyter-widgets/base",
"_model_module_version": "1.2.0",
"_model_name": "LayoutModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "LayoutView",
"align_content": null,
"align_items": null,
"align_self": null,
"border": null,
"bottom": null,
"display": null,
"flex": null,
"flex_flow": null,
"grid_area": null,
"grid_auto_columns": null,
"grid_auto_flow": null,
"grid_auto_rows": null,
"grid_column": null,
"grid_gap": null,
"grid_row": null,
"grid_template_areas": null,
"grid_template_columns": null,
"grid_template_rows": null,
"height": null,
"justify_content": null,
"justify_items": null,
"left": null,
"margin": null,
"max_height": null,
"max_width": null,
"min_height": null,
"min_width": null,
"object_fit": null,
"object_position": null,
"order": null,
"overflow": null,
"overflow_x": null,
"overflow_y": null,
"padding": null,
"right": null,
"top": null,
"visibility": null,
"width": null
}
},
"ca437b02cab741489217ad4d8780f728": {
"model_module": "@jupyter-widgets/controls",
"model_name": "DescriptionStyleModel",
"model_module_version": "1.5.0",
"state": {
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "DescriptionStyleModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "StyleView",
"description_width": ""
}
}
}
}
},
"cells": [
{
"cell_type": "markdown",
"metadata": {
"id": "view-in-github",
"colab_type": "text"
},
"source": [
"<a href=\"https://colab.research.google.com/github/mlabonne/llm-course/blob/main/Quantize_models_with_ExLlamaV2.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
]
},
{
"cell_type": "markdown",
"source": [
"# ExLlamaV2: The Fastest Library to Run LLMs\n",
"\n",
"❤️ Created by [@maximelabonne](https://twitter.com/maximelabonne) as part of the 🗣️ [Large Language Model Course](https://github.com/mlabonne/llm-course)."
],
"metadata": {
"id": "QzUdY6GPiZXG"
}
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {
"id": "4niZ8igkiXZb"
},
"outputs": [],
"source": [
"# Install ExLLamaV2\n",
"!git clone https://github.com/turboderp/exllamav2\n",
"!pip install -e exllamav2"
]
},
{
"cell_type": "code",
"source": [
"MODEL_NAME = \"zephyr-7b-beta\"\n",
"BPW = 5.0\n",
"\n",
"# Download model\n",
"!git lfs install\n",
"!git clone https://huggingface.co/HuggingFaceH4/{MODEL_NAME}\n",
"!mv {MODEL_NAME} base_model\n",
"!rm base_mode/*.bin\n",
"\n",
"# Download dataset\n",
"!wget https://huggingface.co/datasets/wikitext/resolve/9a9e482b5987f9d25b3a9b2883fc6cc9fd8071b3/wikitext-103-v1/wikitext-test.parquet"
],
"metadata": {
"colab": {
"base_uri": "https://localhost:8080/"
},
"id": "elHDPmXAil0c",
"outputId": "5b732dac-fead-4ab2-f3e8-7b3622d3c690"
},
"execution_count": null,
"outputs": [
{
"output_type": "stream",
"name": "stdout",
"text": [
"Git LFS initialized.\n",
"Cloning into 'zephyr-7b-beta'...\n",
"remote: Enumerating objects: 55, done.\u001b[K\n",
"remote: Total 55 (delta 0), reused 0 (delta 0), pack-reused 55\u001b[K\n",
"Unpacking objects: 100% (55/55), 534.67 KiB | 4.73 MiB/s, done.\n",
"Filtering content: 100% (10/10), 13.48 GiB | 129.35 MiB/s, done.\n",
"rm: cannot remove 'base_mode/*.bin': No such file or directory\n",
"--2023-11-03 18:05:02-- https://huggingface.co/datasets/wikitext/resolve/9a9e482b5987f9d25b3a9b2883fc6cc9fd8071b3/wikitext-103-v1/wikitext-test.parquet\n",
"Resolving huggingface.co (huggingface.co)... 65.8.178.27, 65.8.178.93, 65.8.178.118, ...\n",
"Connecting to huggingface.co (huggingface.co)|65.8.178.27|:443... connected.\n",
"HTTP request sent, awaiting response... 200 OK\n",
"Length: 721735 (705K)\n",
"Saving to: wikitext-test.parquet\n",
"\n",
"wikitext-test.parqu 100%[===================>] 704.82K --.-KB/s in 0.1s \n",
"\n",
"2023-11-03 18:05:02 (5.57 MB/s) - wikitext-test.parquet saved [721735/721735]\n",
"\n"
]
}
]
},
{
"cell_type": "code",
"source": [
"# Quantize model\n",
"!mkdir quant\n",
"!python exllamav2/convert.py \\\n",
" -i base_model \\\n",
" -o quant \\\n",
" -c wikitext-test.parquet \\\n",
" -b {BPW}"
],
"metadata": {
"colab": {
"base_uri": "https://localhost:8080/"
},
"id": "jigQLg8Fis1f",
"outputId": "5f2e97ee-f823-45f7-f39a-f4c6a774b587"
},
"execution_count": null,
"outputs": [
{
"output_type": "stream",
"name": "stdout",
"text": [
"\u001b[1;30;43mStreaming output truncated to the last 5000 lines.\u001b[0m\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01444\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01401\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.01332\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.00876\n",
" -- Time: 3.78 seconds\n",
" -- Linear: model.layers.7.self_attn.o_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.19280\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.17356\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.16492\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.14247\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.09048\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.08221\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.10930\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.09792\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.09219\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.07561\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.07156\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.05653\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.04767\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.04434\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.04352\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.02850\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.02416\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.02348\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.02138\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.02088\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01587\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01675\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.01480\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.01294\n",
" -- Time: 6.21 seconds\n",
" -- Layer: model.layers.7 (MLP)\n",
" -- Linear: model.layers.7.mlp.gate_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.17192\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.16163\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.15815\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.14380\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.08135\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.07770\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.09087\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.08363\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.08210\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.07280\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.06958\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.04646\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.04014\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.03912\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.03888\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.02327\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.02036\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.02012\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.01866\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.01851\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01246\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01264\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.01212\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.00882\n",
" -- Time: 16.78 seconds\n",
" -- Linear: model.layers.7.mlp.up_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.24041\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.22653\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.22191\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.20179\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.11401\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.10911\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.12697\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.11680\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.11501\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.10219\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.09774\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.06494\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.05600\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.05474\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.05444\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.03253\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.02831\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.02799\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.02593\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.02574\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01738\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01727\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.01695\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.01176\n",
" -- Time: 16.71 seconds\n",
" -- Linear: model.layers.7.mlp.down_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.18 bpw rfn_error: 0.20930\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.18827\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.17987\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.16126\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.09629\n",
" -- 0.2:6b/0.8:3b 32g s4 3.71 bpw rfn_error: 0.08799\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.11395\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.10350\n",
" -- 0.05:4b/0.95:3b 32g s4 3.18 bpw rfn_error: 0.09877\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.08309\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.07956\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.05813\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.04962\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.04637\n",
" -- 0.1:6b/0.9:4b 32g s4 4.32 bpw rfn_error: 0.04557\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.02919\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.02440\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.34 bpw rfn_error: 0.02402\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.02185\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.02135\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01579\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01587\n",
" -- 0.1:8b/0.9:6b 128g s4 6.23 bpw rfn_error: 0.01471\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.01128\n",
" -- Time: 26.80 seconds\n",
" -- Layer: model.layers.8 (Attention)\n",
" -- Linear: model.layers.8.self_attn.q_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.07321\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.06490\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.05988\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.05294\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.03374\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.02962\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.04266\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.03904\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.03463\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.02839\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.02717\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.02166\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.01866\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.01627\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.01567\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.01085\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.00853\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.00809\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.00748\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.00708\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.00570\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.00571\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.00489\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.00384\n",
" -- Time: 6.22 seconds\n",
" -- Linear: model.layers.8.self_attn.k_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.06478\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.05636\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.05113\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.04496\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.02945\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.02518\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.03836\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.03491\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.03047\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.02440\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.02347\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.01943\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.01666\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.01420\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.01356\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.00974\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.00741\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.00695\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.00641\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.00598\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.00506\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.00499\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.00419\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.00326\n",
" -- Time: 3.87 seconds\n",
" -- Linear: model.layers.8.self_attn.v_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.18482\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.16487\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.15686\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.13830\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.08576\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.07784\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.10130\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.09213\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.08768\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.07196\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.06795\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.05156\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.04405\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.04111\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.04040\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.02577\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.02116\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.02051\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.01835\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.01788\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01333\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01308\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.01217\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.00845\n",
" -- Time: 3.83 seconds\n",
" -- Linear: model.layers.8.self_attn.o_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.20113\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.18273\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.17538\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.15457\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.09416\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.08692\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.11035\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.10015\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.09581\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.08035\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.07590\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.05670\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.04819\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.04548\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.04482\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.02840\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.02389\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.02333\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.02116\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.02074\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01511\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01540\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.01422\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.01094\n",
" -- Time: 6.22 seconds\n",
" -- Layer: model.layers.8 (MLP)\n",
" -- Linear: model.layers.8.mlp.gate_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.16797\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.15806\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.15463\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.14046\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.07949\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.07596\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.08895\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.08186\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.08023\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.07119\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.06813\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.04545\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.03924\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.03820\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.03795\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.02274\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.01983\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.01959\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.01817\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.01801\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01210\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01225\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.01176\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.00845\n",
" -- Time: 16.76 seconds\n",
" -- Linear: model.layers.8.mlp.up_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.23683\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.22306\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.21845\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.19867\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.11225\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.10738\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.12523\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.11508\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.11321\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.10052\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.09621\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.06397\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.05514\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.05385\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.05353\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.03199\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.02772\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.02740\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.02534\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.02515\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01690\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01674\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.01646\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.01114\n",
" -- Time: 16.72 seconds\n",
" -- Linear: model.layers.8.mlp.down_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.18 bpw rfn_error: 0.21367\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.19327\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.18532\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.16520\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.09845\n",
" -- 0.2:6b/0.8:3b 32g s4 3.71 bpw rfn_error: 0.09059\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.11554\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.10506\n",
" -- 0.05:4b/0.95:3b 32g s4 3.18 bpw rfn_error: 0.10081\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.08507\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.08090\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.05900\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.05049\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.04752\n",
" -- 0.1:6b/0.9:4b 32g s4 4.32 bpw rfn_error: 0.04679\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.02977\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.02519\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.34 bpw rfn_error: 0.02484\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.02262\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.02216\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01638\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01655\n",
" -- 0.1:8b/0.9:6b 128g s4 6.23 bpw rfn_error: 0.01541\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.01205\n",
" -- Time: 26.78 seconds\n",
" -- Layer: model.layers.9 (Attention)\n",
" -- Linear: model.layers.9.self_attn.q_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.09374\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.08343\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.07735\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.06843\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.04327\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.03822\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.05469\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.04960\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.04431\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.03654\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.03500\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.02782\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.02380\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.02096\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.02024\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.01396\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.01116\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.01065\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.00987\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.00942\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.00745\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.00763\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.00645\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.00545\n",
" -- Time: 6.22 seconds\n",
" -- Linear: model.layers.9.self_attn.k_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.08149\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.07163\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.06561\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.05786\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.03729\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.03228\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.04811\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.04369\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.03838\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.03122\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.03000\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.02446\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.02089\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.01795\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.01722\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.01224\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.00937\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.00884\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.00818\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.00768\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.00637\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.00628\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.00533\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.00413\n",
" -- Time: 3.79 seconds\n",
" -- Linear: model.layers.9.self_attn.v_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.19517\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.17711\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.17032\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.15017\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.09130\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.08444\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.10496\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.09629\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.09271\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.07744\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.07277\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.05330\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.04590\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.04365\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.04312\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.02659\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.02217\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.02165\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.01932\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.01895\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01363\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01317\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.01279\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.00804\n",
" -- Time: 3.80 seconds\n",
" -- Linear: model.layers.9.self_attn.o_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.20153\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.18239\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.17453\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.15316\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.09457\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.08706\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.11063\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.10074\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.09616\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.07951\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.07566\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.05678\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.04841\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.04560\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.04491\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.02841\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.02377\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.02318\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.02077\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.02032\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01498\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01511\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.01403\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.01042\n",
" -- Time: 6.20 seconds\n",
" -- Layer: model.layers.9 (MLP)\n",
" -- Linear: model.layers.9.mlp.gate_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.16709\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.15677\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.15316\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.13897\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.07902\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.07526\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.08855\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.08146\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.07979\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.07048\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.06740\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.04528\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.03912\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.03803\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.03776\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.02269\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.01986\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.01961\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.01817\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.01801\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01216\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01244\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.01179\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.00879\n",
" -- Time: 16.77 seconds\n",
" -- Linear: model.layers.9.mlp.up_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.23248\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.21835\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.21353\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.19385\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.10999\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.10496\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.12290\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.11303\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.11108\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.09819\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.09389\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.06283\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.05418\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.05282\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.05250\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.03146\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.02733\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.02699\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.02495\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.02474\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01676\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01671\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.01630\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.01137\n",
" -- Time: 16.69 seconds\n",
" -- Linear: model.layers.9.mlp.down_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.18 bpw rfn_error: 0.22017\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.19862\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.19033\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.16911\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.10144\n",
" -- 0.2:6b/0.8:3b 32g s4 3.71 bpw rfn_error: 0.09320\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.11938\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.10841\n",
" -- 0.05:4b/0.95:3b 32g s4 3.18 bpw rfn_error: 0.10391\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.08730\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.08272\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.06067\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.05196\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.04883\n",
" -- 0.1:6b/0.9:4b 32g s4 4.32 bpw rfn_error: 0.04807\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.03050\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.02563\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.34 bpw rfn_error: 0.02526\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.02287\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.02239\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01654\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01652\n",
" -- 0.1:8b/0.9:6b 128g s4 6.23 bpw rfn_error: 0.01550\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.01166\n",
" -- Time: 26.75 seconds\n",
" -- Layer: model.layers.10 (Attention)\n",
" -- Linear: model.layers.10.self_attn.q_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.08235\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.07363\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.06866\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.06023\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.03814\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.03398\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.04728\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.04319\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.03894\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.03206\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.03048\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.02402\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.02064\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.01838\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.01782\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.01202\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.00962\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.00919\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.00842\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.00805\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.00633\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.00634\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.00553\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.00428\n",
" -- Time: 6.20 seconds\n",
" -- Linear: model.layers.10.self_attn.k_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.07516\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.06653\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.06168\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.05393\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.03452\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.03046\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.04334\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.03936\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.03536\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.02875\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.02734\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.02195\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.01879\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.01658\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.01603\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.01097\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.00858\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.00817\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.00742\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.00705\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.00568\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.00556\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.00487\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.00357\n",
" -- Time: 3.77 seconds\n",
" -- Linear: model.layers.10.self_attn.v_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.18187\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.16366\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.15689\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.13720\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.08496\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.07789\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.09894\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.08995\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.08648\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.07118\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.06690\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.05035\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.04296\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.04066\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.04012\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.02514\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.02081\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.02025\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.01800\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.01762\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01297\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01259\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.01205\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.00796\n",
" -- Time: 3.83 seconds\n",
" -- Linear: model.layers.10.self_attn.o_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.19772\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.17857\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.17053\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.14967\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.09223\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.08466\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.10913\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.09919\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.09394\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.07816\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.07389\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.05624\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.04783\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.04464\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.04390\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.02824\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.02368\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.02304\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.02093\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.02045\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01527\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01561\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.01425\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.01133\n",
" -- Time: 6.18 seconds\n",
" -- Layer: model.layers.10 (MLP)\n",
" -- Linear: model.layers.10.mlp.gate_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.16458\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.15345\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.14950\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.13525\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.07763\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.07360\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.08755\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.08045\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.07854\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.06883\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.06571\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.04474\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.03864\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.03735\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.03705\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.02242\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.01952\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.01924\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.01778\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.01760\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01199\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01228\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.01155\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.00867\n",
" -- Time: 16.72 seconds\n",
" -- Linear: model.layers.10.mlp.up_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.22519\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.21022\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.20510\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.18548\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.10629\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.10090\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.11927\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.10970\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.10747\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.09419\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.08985\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.06086\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.05254\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.05097\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.05060\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.03043\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.02627\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.02590\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.02382\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.02359\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01601\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01597\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.01546\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.01065\n",
" -- Time: 16.67 seconds\n",
" -- Linear: model.layers.10.mlp.down_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.18 bpw rfn_error: 0.23070\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.20868\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.20041\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.17776\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.10675\n",
" -- 0.2:6b/0.8:3b 32g s4 3.71 bpw rfn_error: 0.09841\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.12502\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.11343\n",
" -- 0.05:4b/0.95:3b 32g s4 3.18 bpw rfn_error: 0.10921\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.09194\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.08704\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.06390\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.05452\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.05149\n",
" -- 0.1:6b/0.9:4b 32g s4 4.32 bpw rfn_error: 0.05076\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.03217\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.02720\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.34 bpw rfn_error: 0.02684\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.02433\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.02387\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01749\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01771\n",
" -- 0.1:8b/0.9:6b 128g s4 6.23 bpw rfn_error: 0.01646\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.01279\n",
" -- Time: 26.77 seconds\n",
" -- Layer: model.layers.11 (Attention)\n",
" -- Linear: model.layers.11.self_attn.q_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.09637\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.08586\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.08008\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.07014\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.04477\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.03986\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.05541\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.05047\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.04572\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.03744\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.03559\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.02819\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.02421\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.02166\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.02103\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.01414\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.01146\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.01098\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.01004\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.00963\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.00750\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.00769\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.00661\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.00542\n",
" -- Time: 6.22 seconds\n",
" -- Linear: model.layers.11.self_attn.k_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.08114\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.07195\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.06606\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.05768\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.03737\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.03267\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.04767\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.04342\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.03824\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.03114\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.02972\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.02425\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.02075\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.01802\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.01732\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.01210\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.00943\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.00893\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.00821\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.00775\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.00635\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.00631\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.00541\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.00423\n",
" -- Time: 3.82 seconds\n",
" -- Linear: model.layers.11.self_attn.v_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.19703\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.17551\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.16728\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.14535\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.09159\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.08328\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.10811\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.09762\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.09342\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.07590\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.07111\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.05511\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.04670\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.04391\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.04323\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.02758\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.02246\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.02183\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.01924\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.01879\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01423\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01368\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.01300\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.00864\n",
" -- Time: 3.92 seconds\n",
" -- Linear: model.layers.11.self_attn.o_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.19597\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.17847\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.16379\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.14511\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.09237\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.08126\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.12329\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.11030\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.09394\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.07962\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.07759\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.06377\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.05341\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.04512\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.04298\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.03193\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.02434\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.02296\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.02192\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.02058\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01710\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01736\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.01424\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.01256\n",
" -- Time: 6.20 seconds\n",
" -- Layer: model.layers.11 (MLP)\n",
" -- Linear: model.layers.11.mlp.gate_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.16838\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.15674\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.15241\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.13774\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.07955\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.07520\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.09017\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.08269\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.08051\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.07030\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.06713\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.04617\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.03980\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.03838\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.03804\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.02317\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.02025\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.01994\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.01844\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.01824\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01252\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01300\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.01203\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.00944\n",
" -- Time: 16.79 seconds\n",
" -- Linear: model.layers.11.mlp.up_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.22740\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.21181\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.20640\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.18641\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.10722\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.10159\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.12081\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.11093\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.10846\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.09479\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.09032\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.06150\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.05311\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.05143\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.05102\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.03077\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.02649\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.02609\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.02397\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.02371\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01611\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01609\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.01551\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.01069\n",
" -- Time: 16.77 seconds\n",
" -- Linear: model.layers.11.mlp.down_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.18 bpw rfn_error: 0.22960\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.20772\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.19942\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.17705\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.10644\n",
" -- 0.2:6b/0.8:3b 32g s4 3.71 bpw rfn_error: 0.09805\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.12519\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.11322\n",
" -- 0.05:4b/0.95:3b 32g s4 3.18 bpw rfn_error: 0.10894\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.09175\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.08694\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.06400\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.05447\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.05138\n",
" -- 0.1:6b/0.9:4b 32g s4 4.32 bpw rfn_error: 0.05062\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.03219\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.02714\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.34 bpw rfn_error: 0.02677\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.02429\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.02381\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01739\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01767\n",
" -- 0.1:8b/0.9:6b 128g s4 6.23 bpw rfn_error: 0.01632\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.01276\n",
" -- Time: 26.84 seconds\n",
" -- Layer: model.layers.12 (Attention)\n",
" -- Linear: model.layers.12.self_attn.q_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.10068\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.08938\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.08243\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.07225\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.04673\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.04101\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.06037\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.05396\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.04786\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.03913\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.03765\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.03088\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.02602\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.02269\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.02185\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.01557\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.01216\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.01156\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.01071\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.01017\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.00838\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.00843\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.00714\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.00606\n",
" -- Time: 6.25 seconds\n",
" -- Linear: model.layers.12.self_attn.k_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.08574\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.07553\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.06922\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.06050\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.03941\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.03421\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.05104\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.04590\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.04042\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.03277\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.03136\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.02597\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.02200\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.01905\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.01829\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.01302\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.01004\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.00951\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.00876\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.00826\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.00686\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.00683\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.00579\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.00469\n",
" -- Time: 3.85 seconds\n",
" -- Linear: model.layers.12.self_attn.v_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.19504\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.17522\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.16722\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.14655\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.09135\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.08336\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.10798\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.09762\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.09299\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.07656\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.07215\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.05508\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.04669\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.04381\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.04310\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.02753\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.02244\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.02178\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.01939\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.01891\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01422\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01369\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.01304\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.00864\n",
" -- Time: 3.84 seconds\n",
" -- Linear: model.layers.12.self_attn.o_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.21198\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.18789\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.17851\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.15663\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.09908\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.08925\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.11771\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.10642\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.10141\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.08235\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.07812\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.06074\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.05148\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.04812\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.04730\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.03056\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.02559\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.02480\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.02233\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.02180\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01657\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01693\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.01547\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.01239\n",
" -- Time: 6.23 seconds\n",
" -- Layer: model.layers.12 (MLP)\n",
" -- Linear: model.layers.12.mlp.gate_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.17317\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.16089\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.15653\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.14139\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.08192\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.07739\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.09302\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.08500\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.08295\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.07225\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.06919\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.04771\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.04103\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.03965\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.03932\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.02403\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.02109\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.02078\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.01922\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.01902\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01327\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01375\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.01282\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.01025\n",
" -- Time: 16.80 seconds\n",
" -- Linear: model.layers.12.mlp.up_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.23561\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.21922\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.21368\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.19274\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.11136\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.10544\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.12512\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.11497\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.11266\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.09816\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.09356\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.06397\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.05511\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.05348\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.05310\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.03199\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.02766\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.02727\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.02501\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.02476\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01693\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01695\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.01637\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.01149\n",
" -- Time: 16.72 seconds\n",
" -- Linear: model.layers.12.mlp.down_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.18 bpw rfn_error: 0.23220\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.20919\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.19921\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.17615\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.10738\n",
" -- 0.2:6b/0.8:3b 32g s4 3.71 bpw rfn_error: 0.09804\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.12854\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.11635\n",
" -- 0.05:4b/0.95:3b 32g s4 3.18 bpw rfn_error: 0.11008\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.09202\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.08705\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.06542\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.05593\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.05194\n",
" -- 0.1:6b/0.9:4b 32g s4 4.32 bpw rfn_error: 0.05095\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.03304\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.02763\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.34 bpw rfn_error: 0.02712\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.02468\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.02405\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01806\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01837\n",
" -- 0.1:8b/0.9:6b 128g s4 6.23 bpw rfn_error: 0.01671\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.01338\n",
" -- Time: 26.78 seconds\n",
" -- Layer: model.layers.13 (Attention)\n",
" -- Linear: model.layers.13.self_attn.q_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.10346\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.09299\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.08766\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.07699\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.04820\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.04354\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.05839\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.05322\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.04910\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.04058\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.03850\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.02967\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.02545\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.02319\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.02264\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.01485\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.01207\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.01163\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.01055\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.01019\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.00778\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.00777\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.00696\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.00522\n",
" -- Time: 6.21 seconds\n",
" -- Linear: model.layers.13.self_attn.k_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.08538\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.07664\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.07156\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.06280\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.03961\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.03533\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.04917\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.04484\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.04036\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.03339\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.03179\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.02494\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.02147\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.01905\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.01847\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.01246\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.00989\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.00945\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.00865\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.00825\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.00651\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.00643\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.00569\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.00422\n",
" -- Time: 3.82 seconds\n",
" -- Linear: model.layers.13.self_attn.v_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.21169\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.19023\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.18235\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.15968\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.09888\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.09079\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.11536\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.10443\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.10062\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.08274\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.07782\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.05883\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.04983\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.04732\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.04671\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.02938\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.02406\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.02345\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.02073\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.02031\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01508\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01435\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.01392\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.00878\n",
" -- Time: 3.81 seconds\n",
" -- Linear: model.layers.13.self_attn.o_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.22985\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.20997\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.20194\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.18065\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.10763\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.09973\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.12558\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.11441\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.10930\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.09282\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.08926\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.06461\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.05522\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.05212\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.05139\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.03240\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.02762\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.02700\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.02472\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.02425\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01745\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01809\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.01647\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.01319\n",
" -- Time: 6.19 seconds\n",
" -- Layer: model.layers.13 (MLP)\n",
" -- Linear: model.layers.13.mlp.gate_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.17483\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.16251\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.15795\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.14266\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.08279\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.07819\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.09385\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.08610\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.08379\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.07304\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.06979\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.04815\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.04151\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.03999\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.03964\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.02414\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.02113\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.02080\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.01922\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.01901\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01303\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01362\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.01251\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.00993\n",
" -- Time: 16.77 seconds\n",
" -- Linear: model.layers.13.mlp.up_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.23982\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.22323\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.21751\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.19635\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.11347\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.10742\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.12767\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.11723\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.11476\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.10007\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.09532\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.06526\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.05620\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.05448\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.05408\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.03266\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.02815\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.02773\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.02544\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.02519\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01729\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01721\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.01670\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.01159\n",
" -- Time: 16.72 seconds\n",
" -- Linear: model.layers.13.mlp.down_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.18 bpw rfn_error: 0.24641\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.22109\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.21158\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.18667\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.11395\n",
" -- 0.2:6b/0.8:3b 32g s4 3.71 bpw rfn_error: 0.10431\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.13424\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.12162\n",
" -- 0.05:4b/0.95:3b 32g s4 3.18 bpw rfn_error: 0.11690\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.09716\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.09181\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.06867\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.05844\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.05492\n",
" -- 0.1:6b/0.9:4b 32g s4 4.32 bpw rfn_error: 0.05408\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.03458\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.02889\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.34 bpw rfn_error: 0.02847\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.02562\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.02507\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01873\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01869\n",
" -- 0.1:8b/0.9:6b 128g s4 6.23 bpw rfn_error: 0.01756\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.01328\n",
" -- Time: 26.78 seconds\n",
" -- Layer: model.layers.14 (Attention)\n",
" -- Linear: model.layers.14.self_attn.q_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.10582\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.09534\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.08908\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.07862\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.04931\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.04419\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.06149\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.05579\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.05026\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.04186\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.04004\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.03130\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.02671\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.02382\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.02309\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.01568\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.01253\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.01200\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.01107\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.01060\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.00827\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.00834\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.00722\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.00576\n",
" -- Time: 6.21 seconds\n",
" -- Linear: model.layers.14.self_attn.k_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.08561\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.07738\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.07080\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.06264\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.03965\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.03475\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.05197\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.04725\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.04042\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.03395\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.03267\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.02629\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.02257\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.01913\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.01824\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.01316\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.01000\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.00941\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.00888\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.00832\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.00689\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.00678\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.00572\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.00448\n",
" -- Time: 3.83 seconds\n",
" -- Linear: model.layers.14.self_attn.v_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.22144\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.19968\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.19127\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.16781\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.10366\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.09508\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.12296\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.11022\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.10547\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.08733\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.08233\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.06295\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.05278\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.04971\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.04898\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.03148\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.02549\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.02481\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.02217\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.02167\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01627\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01558\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.01481\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.00998\n",
" -- Time: 3.84 seconds\n",
" -- Linear: model.layers.14.self_attn.o_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.24642\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.21851\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.20690\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.18059\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.11514\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.10416\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.13897\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.12539\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.11840\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.09537\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.09121\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.07166\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.06063\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.05589\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.05472\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.03587\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.02969\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.02881\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.02585\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.02511\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01920\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01975\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.01767\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.01432\n",
" -- Time: 6.21 seconds\n",
" -- Layer: model.layers.14 (MLP)\n",
" -- Linear: model.layers.14.mlp.gate_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.16493\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.15358\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.14900\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.13483\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.07826\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.07376\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.08938\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.08197\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.07921\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.06925\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.06623\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.04587\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.03957\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.03788\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.03748\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.02303\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.02014\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.01981\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.01840\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.01816\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01248\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01319\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.01191\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.00978\n",
" -- Time: 16.75 seconds\n",
" -- Linear: model.layers.14.mlp.up_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.23642\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.22076\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.21511\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.19454\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.11210\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.10626\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.12655\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.11623\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.11338\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.09929\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.09471\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.06476\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.05580\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.05394\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.05350\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.03243\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.02805\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.02763\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.02549\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.02521\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01728\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01745\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.01663\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.01207\n",
" -- Time: 16.72 seconds\n",
" -- Linear: model.layers.14.mlp.down_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.18 bpw rfn_error: 0.23204\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.20811\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.19811\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.17565\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.10718\n",
" -- 0.2:6b/0.8:3b 32g s4 3.71 bpw rfn_error: 0.09761\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.12842\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.11596\n",
" -- 0.05:4b/0.95:3b 32g s4 3.18 bpw rfn_error: 0.10998\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.09171\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.08705\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.06573\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.05587\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.05206\n",
" -- 0.1:6b/0.9:4b 32g s4 4.32 bpw rfn_error: 0.05104\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.03340\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.02796\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.34 bpw rfn_error: 0.02744\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.02504\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.02438\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01849\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01893\n",
" -- 0.1:8b/0.9:6b 128g s4 6.23 bpw rfn_error: 0.01699\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.01411\n",
" -- Time: 26.80 seconds\n",
" -- Layer: model.layers.15 (Attention)\n",
" -- Linear: model.layers.15.self_attn.q_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.11824\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.10740\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.10090\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.08950\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.05527\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.04995\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.06887\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.06209\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.05624\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.04743\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.04546\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.03518\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.02984\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.02678\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.02602\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.01771\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.01425\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.01369\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.01273\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.01224\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.00947\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.00962\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.00832\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.00688\n",
" -- Time: 6.22 seconds\n",
" -- Linear: model.layers.15.self_attn.k_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.08997\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.08158\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.07426\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.06600\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.04151\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.03625\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.05554\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.05021\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.04234\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.03593\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.03491\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.02827\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.02404\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.02007\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.01901\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.01418\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.01052\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.00984\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.00942\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.00876\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.00742\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.00723\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.00604\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.00478\n",
" -- Time: 3.78 seconds\n",
" -- Linear: model.layers.15.self_attn.v_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.23763\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.21808\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.21011\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.18801\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.11228\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.10438\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.13130\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.11890\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.11398\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.09722\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.09255\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.06716\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.05694\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.05385\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.05308\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.03358\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.02751\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.02687\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.02442\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.02393\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01729\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01660\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.01602\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.01042\n",
" -- Time: 3.83 seconds\n",
" -- Linear: model.layers.15.self_attn.o_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.22281\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.20108\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.19287\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.16640\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.10428\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.09603\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.12125\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.10998\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.10599\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.08663\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.08045\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.06233\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.05283\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.05024\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.04962\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.03121\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.02622\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.02562\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.02269\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.02227\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01662\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01658\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.01577\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.01154\n",
" -- Time: 6.20 seconds\n",
" -- Layer: model.layers.15 (MLP)\n",
" -- Linear: model.layers.15.mlp.gate_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.16069\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.15021\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.14602\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.13239\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.07632\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.07225\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.08692\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.07965\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.07721\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.06788\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.06506\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.04466\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.03849\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.03699\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.03663\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.02248\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.01974\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.01943\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.01811\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.01790\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01236\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01298\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.01187\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.00972\n",
" -- Time: 16.77 seconds\n",
" -- Linear: model.layers.15.mlp.up_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.23888\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.22394\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.21869\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.19820\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.11323\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.10775\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.12727\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.11700\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.11440\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.10084\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.09639\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.06496\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.05602\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.05432\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.05392\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.03248\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.02794\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.02754\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.02541\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.02515\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01710\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01689\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.01651\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.01114\n",
" -- Time: 16.71 seconds\n",
" -- Linear: model.layers.15.mlp.down_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.18 bpw rfn_error: 0.22929\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.20544\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.19581\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.17408\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.10567\n",
" -- 0.2:6b/0.8:3b 32g s4 3.71 bpw rfn_error: 0.09612\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.12543\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.11389\n",
" -- 0.05:4b/0.95:3b 32g s4 3.18 bpw rfn_error: 0.10852\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.09041\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.08577\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.06395\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.05471\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.05095\n",
" -- 0.1:6b/0.9:4b 32g s4 4.32 bpw rfn_error: 0.05003\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.03227\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.02688\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.34 bpw rfn_error: 0.02643\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.02392\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.02334\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01763\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01757\n",
" -- 0.1:8b/0.9:6b 128g s4 6.23 bpw rfn_error: 0.01639\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.01254\n",
" -- Time: 26.78 seconds\n",
" -- Layer: model.layers.16 (Attention)\n",
" -- Linear: model.layers.16.self_attn.q_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.11746\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.10737\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.10179\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.09062\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.05488\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.05019\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.06660\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.06065\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.05572\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.04742\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.04533\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.03389\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.02900\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.02641\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.02577\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.01698\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.01376\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.01328\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.01228\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.01187\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.00893\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.00889\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.00799\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.00601\n",
" -- Time: 6.24 seconds\n",
" -- Linear: model.layers.16.self_attn.k_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.09286\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.08501\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.07839\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.07006\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.04299\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.03820\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.05563\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.05094\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.04370\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.03752\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.03612\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.02812\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.02428\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.02069\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.01978\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.01404\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.01075\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.01012\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.00966\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.00904\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.00733\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.00717\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.00613\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.00456\n",
" -- Time: 3.89 seconds\n",
" -- Linear: model.layers.16.self_attn.v_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.24342\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.22526\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.21816\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.19590\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.11499\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.10791\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.13360\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.12081\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.11661\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.10061\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.09606\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.06830\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.05777\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.05512\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.05448\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.03414\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.02804\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.02746\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.02504\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.02462\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01754\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01666\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.01634\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.01022\n",
" -- Time: 3.85 seconds\n",
" -- Linear: model.layers.16.self_attn.o_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.23516\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.21166\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.20336\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.18091\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.10959\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.10060\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.12675\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.11556\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.11176\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.09287\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.08821\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.06515\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.05577\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.05310\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.05242\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.03287\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.02825\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.02758\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.02502\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.02460\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01800\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01859\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.01714\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.01375\n",
" -- Time: 6.22 seconds\n",
" -- Layer: model.layers.16 (MLP)\n",
" -- Linear: model.layers.16.mlp.gate_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.15193\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.14219\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.13810\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.12526\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.07207\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.06821\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.08244\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.07556\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.07291\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.06421\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.06152\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.04227\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.03647\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.03490\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.03452\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.02124\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.01855\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.01824\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.01703\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.01680\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01156\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01215\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.01102\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.00899\n",
" -- Time: 16.85 seconds\n",
" -- Linear: model.layers.16.mlp.up_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.23238\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.21834\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.21318\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.19339\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.11013\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.10490\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.12404\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.11403\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.11123\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.09835\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.09399\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.06329\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.05459\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.05284\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.05242\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.03164\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.02718\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.02679\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.02477\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.02451\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01664\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01645\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.01604\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.01085\n",
" -- Time: 16.73 seconds\n",
" -- Linear: model.layers.16.mlp.down_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.18 bpw rfn_error: 0.22319\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.19859\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.18842\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.16736\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.10245\n",
" -- 0.2:6b/0.8:3b 32g s4 3.71 bpw rfn_error: 0.09253\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.12249\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.11134\n",
" -- 0.05:4b/0.95:3b 32g s4 3.18 bpw rfn_error: 0.10546\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.08704\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.08280\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.06257\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.05346\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.04942\n",
" -- 0.1:6b/0.9:4b 32g s4 4.32 bpw rfn_error: 0.04840\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.03161\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.02606\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.34 bpw rfn_error: 0.02557\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.02307\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.02243\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01725\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01710\n",
" -- 0.1:8b/0.9:6b 128g s4 6.23 bpw rfn_error: 0.01591\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.01215\n",
" -- Time: 26.79 seconds\n",
" -- Layer: model.layers.17 (Attention)\n",
" -- Linear: model.layers.17.self_attn.q_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.09829\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.09000\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.08431\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.07568\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.04579\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.04136\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.05741\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.05237\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.04662\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.04001\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.03857\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.02920\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.02505\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.02210\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.02135\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.01463\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.01162\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.01111\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.01049\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.01001\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.00768\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.00777\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.00663\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.00532\n",
" -- Time: 6.19 seconds\n",
" -- Linear: model.layers.17.self_attn.k_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.08335\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.07599\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.06954\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.06252\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.03844\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.03384\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.05100\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.04629\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.03928\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.03366\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.03269\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.02575\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.02205\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.01855\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.01764\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.01288\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.00970\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.00911\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.00875\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.00817\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.00672\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.00661\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.00549\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.00433\n",
" -- Time: 3.90 seconds\n",
" -- Linear: model.layers.17.self_attn.v_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.20309\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.18364\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.17341\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.15613\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.09403\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.08522\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.11704\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.10452\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.09631\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.08153\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.07840\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.05969\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.04993\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.04516\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.04397\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.02979\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.02319\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.02230\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.02063\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.01985\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01531\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01458\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.01324\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.00917\n",
" -- Time: 3.81 seconds\n",
" -- Linear: model.layers.17.self_attn.o_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.22796\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.20518\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.19632\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.17325\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.10651\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.09758\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.12522\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.11352\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.10867\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.09042\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.08499\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.06454\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.05483\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.05163\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.05087\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.03239\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.02747\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.02676\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.02437\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.02388\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01756\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01816\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.01653\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.01336\n",
" -- Time: 6.20 seconds\n",
" -- Layer: model.layers.17 (MLP)\n",
" -- Linear: model.layers.17.mlp.gate_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.15907\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.14871\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.14446\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.13126\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.07530\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.07126\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.08597\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.07891\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.07619\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.06721\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.06452\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.04402\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.03805\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.03643\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.03605\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.02213\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.01933\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.01901\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.01776\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.01752\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01206\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01261\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.01151\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.00929\n",
" -- Time: 16.79 seconds\n",
" -- Linear: model.layers.17.mlp.up_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.24069\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.22613\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.22082\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.20047\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.11399\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.10863\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.12831\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.11795\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.11515\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.10196\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.09775\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.06561\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.05659\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.05480\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.05438\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.03293\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.02847\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.02808\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.02606\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.02580\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01763\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01763\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.01703\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.01214\n",
" -- Time: 16.71 seconds\n",
" -- Linear: model.layers.17.mlp.down_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.18 bpw rfn_error: 0.23679\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.20960\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.19847\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.17539\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.10882\n",
" -- 0.2:6b/0.8:3b 32g s4 3.71 bpw rfn_error: 0.09788\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.13094\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.11836\n",
" -- 0.05:4b/0.95:3b 32g s4 3.18 bpw rfn_error: 0.11219\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.09183\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.08713\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.06688\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.05691\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.05259\n",
" -- 0.1:6b/0.9:4b 32g s4 4.32 bpw rfn_error: 0.05153\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.03377\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.02791\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.34 bpw rfn_error: 0.02738\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.02464\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.02395\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01851\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01852\n",
" -- 0.1:8b/0.9:6b 128g s4 6.23 bpw rfn_error: 0.01705\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.01337\n",
" -- Time: 26.73 seconds\n",
" -- Layer: model.layers.18 (Attention)\n",
" -- Linear: model.layers.18.self_attn.q_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.11366\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.10467\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.09871\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.08852\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.05320\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.04850\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.06644\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.06008\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.05405\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.04669\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.04498\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.03393\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.02889\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.02578\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.02500\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.01706\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.01370\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.01318\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.01243\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.01196\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.00911\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.00927\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.00797\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.00663\n",
" -- Time: 6.22 seconds\n",
" -- Linear: model.layers.18.self_attn.k_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.09304\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.08449\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.07639\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.06885\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.04281\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.03714\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.05762\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.05275\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.04376\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.03751\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.03660\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.02914\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.02510\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.02066\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.01948\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.01457\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.01083\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.01007\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.00979\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.00904\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.00763\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.00751\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.00612\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.00488\n",
" -- Time: 3.87 seconds\n",
" -- Linear: model.layers.18.self_attn.v_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.22252\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.20193\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.19171\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.17346\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.10408\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.09482\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.12835\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.11432\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.10636\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.09040\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.08736\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.06581\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.05469\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.04999\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.04883\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.03301\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.02564\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.02473\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.02284\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.02208\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01701\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01596\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.01485\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.01001\n",
" -- Time: 3.83 seconds\n",
" -- Linear: model.layers.18.self_attn.o_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.19772\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.17592\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.16816\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.14555\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.09180\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.08353\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.10693\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.09711\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.09339\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.07592\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.07098\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.05513\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.04712\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.04478\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.04419\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.02785\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.02420\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.02365\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.02121\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.02086\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01553\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01642\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.01477\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.01263\n",
" -- Time: 6.20 seconds\n",
" -- Layer: model.layers.18 (MLP)\n",
" -- Linear: model.layers.18.mlp.gate_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.14728\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.13762\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.13347\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.12128\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.06964\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.06577\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.07982\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.07330\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.07050\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.06210\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.05950\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.04083\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.03526\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.03363\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.03324\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.02047\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.01771\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.01739\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.01623\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.01599\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01103\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01140\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.01048\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.00820\n",
" -- Time: 16.83 seconds\n",
" -- Linear: model.layers.18.mlp.up_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.22881\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.21485\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.20956\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.19036\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.10830\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.10303\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.12241\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.11246\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.10942\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.09681\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.09261\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.06251\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.05390\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.05203\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.05159\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.03132\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.02690\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.02650\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.02458\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.02430\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01667\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01650\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.01603\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.01111\n",
" -- Time: 16.73 seconds\n",
" -- Linear: model.layers.18.mlp.down_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.18 bpw rfn_error: 0.22510\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.19820\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.18676\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.16481\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.10319\n",
" -- 0.2:6b/0.8:3b 32g s4 3.71 bpw rfn_error: 0.09216\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.12568\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.11338\n",
" -- 0.05:4b/0.95:3b 32g s4 3.18 bpw rfn_error: 0.10659\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.08683\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.08259\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.06434\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.05471\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.05003\n",
" -- 0.1:6b/0.9:4b 32g s4 4.32 bpw rfn_error: 0.04886\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.03262\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.02680\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.34 bpw rfn_error: 0.02623\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.02369\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.02295\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01807\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01818\n",
" -- 0.1:8b/0.9:6b 128g s4 6.23 bpw rfn_error: 0.01649\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.01341\n",
" -- Time: 26.79 seconds\n",
" -- Layer: model.layers.19 (Attention)\n",
" -- Linear: model.layers.19.self_attn.q_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.11644\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.10768\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.10152\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.09162\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.05452\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.04974\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.06825\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.06184\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.05541\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.04821\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.04669\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.03480\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.02965\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.02633\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.02550\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.01744\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.01385\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.01330\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.01260\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.01210\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.00928\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.00923\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.00810\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.00638\n",
" -- Time: 6.23 seconds\n",
" -- Linear: model.layers.19.self_attn.k_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.09202\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.08449\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.07618\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.06889\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.04248\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.03691\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.05797\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.05307\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.04333\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.03766\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.03691\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.02941\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.02524\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.02054\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.01927\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.01472\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.01077\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.00997\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.00982\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.00903\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.00773\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.00752\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.00611\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.00488\n",
" -- Time: 3.83 seconds\n",
" -- Linear: model.layers.19.self_attn.v_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.23251\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.21308\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.20404\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.18453\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.10898\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.10053\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.13124\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.11786\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.11101\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.09542\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.09173\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.06728\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.05643\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.05229\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.05128\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.03361\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.02663\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.02584\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.02383\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.02317\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01721\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01620\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.01532\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.00993\n",
" -- Time: 3.84 seconds\n",
" -- Linear: model.layers.19.self_attn.o_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.21240\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.19376\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.18742\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.16420\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.09921\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.09250\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.11509\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.10345\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.10047\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.08449\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.07993\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.05928\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.05036\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.04845\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.04797\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.03000\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.02635\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.02590\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.02359\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.02329\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01684\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01795\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.01624\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.01403\n",
" -- Time: 6.19 seconds\n",
" -- Layer: model.layers.19 (MLP)\n",
" -- Linear: model.layers.19.mlp.gate_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.14325\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.13404\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.13019\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.11838\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.06772\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.06406\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.07722\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.07102\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.06852\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.06046\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.05793\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.03943\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.03415\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.03266\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.03230\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.01978\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.01714\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.01685\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.01571\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.01549\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01061\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01094\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.01011\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.00779\n",
" -- Time: 16.78 seconds\n",
" -- Linear: model.layers.19.mlp.up_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.22417\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.21056\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.20544\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.18681\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.10609\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.10102\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.11969\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.11008\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.10719\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.09500\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.09090\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.06112\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.05278\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.05099\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.05057\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.03066\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.02648\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.02611\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.02426\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.02399\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01647\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01641\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.01587\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.01129\n",
" -- Time: 16.74 seconds\n",
" -- Linear: model.layers.19.mlp.down_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.18 bpw rfn_error: 0.21211\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.18632\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.17439\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.15341\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.09733\n",
" -- 0.2:6b/0.8:3b 32g s4 3.71 bpw rfn_error: 0.08616\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.12003\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.10839\n",
" -- 0.05:4b/0.95:3b 32g s4 3.18 bpw rfn_error: 0.10058\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.08150\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.07764\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.06150\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.05253\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.04730\n",
" -- 0.1:6b/0.9:4b 32g s4 4.32 bpw rfn_error: 0.04602\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.03137\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.02558\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.34 bpw rfn_error: 0.02490\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.02262\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.02176\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01772\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01772\n",
" -- 0.1:8b/0.9:6b 128g s4 6.23 bpw rfn_error: 0.01604\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.01321\n",
" -- Time: 26.78 seconds\n",
" -- Layer: model.layers.20 (Attention)\n",
" -- Linear: model.layers.20.self_attn.q_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.11630\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.10727\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.09987\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.09052\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.05425\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.04880\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.07019\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.06342\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.05525\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.04814\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.04683\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.03576\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.03035\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.02627\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.02523\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.01794\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.01393\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.01325\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.01273\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.01209\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.00956\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.00953\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.00808\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.00664\n",
" -- Time: 6.23 seconds\n",
" -- Linear: model.layers.20.self_attn.k_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.09284\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.08449\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.07454\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.06771\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.04260\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.03601\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.06061\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.05522\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.04370\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.03771\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.03725\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.03066\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.02624\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.02060\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.01910\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.01531\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.01089\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.00993\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.00995\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.00897\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.00802\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.00785\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.00608\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.00510\n",
" -- Time: 3.83 seconds\n",
" -- Linear: model.layers.20.self_attn.v_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.23120\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.20926\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.19857\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.17941\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.10756\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.09769\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.13221\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.11821\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.10996\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.09330\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.09015\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.06796\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.05649\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.05160\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.05039\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.03402\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.02643\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.02547\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.02351\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.02271\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01746\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01641\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.01515\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.01021\n",
" -- Time: 3.83 seconds\n",
" -- Linear: model.layers.20.self_attn.o_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.23250\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.20446\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.19377\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.16236\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.10847\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.09762\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.12756\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.11605\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.11059\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.08639\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.08040\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.06563\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.05604\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.05256\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.05170\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.03302\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.02801\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.02722\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.02378\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.02324\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01788\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01861\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.01677\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.01376\n",
" -- Time: 6.22 seconds\n",
" -- Layer: model.layers.20 (MLP)\n",
" -- Linear: model.layers.20.mlp.gate_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.15034\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.14114\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.13758\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.12504\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.07114\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.06767\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.08029\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.07404\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.07192\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.06368\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.06092\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.04098\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.03555\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.03425\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.03395\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.02052\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.01788\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.01761\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.01639\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.01620\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01092\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01124\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.01049\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.00788\n",
" -- Time: 16.78 seconds\n",
" -- Linear: model.layers.20.mlp.up_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.23300\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.21932\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.21445\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.19492\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.11020\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.10522\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.12351\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.11371\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.11129\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.09876\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.09448\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.06288\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.05443\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.05285\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.05248\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.03154\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.02717\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.02682\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.02484\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.02461\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01673\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01640\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.01619\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.01081\n",
" -- Time: 16.73 seconds\n",
" -- Linear: model.layers.20.mlp.down_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.18 bpw rfn_error: 0.21185\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.18673\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.17575\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.15506\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.09714\n",
" -- 0.2:6b/0.8:3b 32g s4 3.71 bpw rfn_error: 0.08663\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.11739\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.10685\n",
" -- 0.05:4b/0.95:3b 32g s4 3.18 bpw rfn_error: 0.10019\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.08144\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.07747\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.06004\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.05138\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.04694\n",
" -- 0.1:6b/0.9:4b 32g s4 4.32 bpw rfn_error: 0.04584\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.03042\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.02490\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.34 bpw rfn_error: 0.02435\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.02188\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.02117\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01677\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01660\n",
" -- 0.1:8b/0.9:6b 128g s4 6.23 bpw rfn_error: 0.01531\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.01191\n",
" -- Time: 26.79 seconds\n",
" -- Layer: model.layers.21 (Attention)\n",
" -- Linear: model.layers.21.self_attn.q_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.11127\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.10196\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.09483\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.08600\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.05187\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.04645\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.06690\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.06061\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.05291\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.04576\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.04454\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.03412\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.02899\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.02507\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.02407\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.01714\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.01319\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.01253\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.01199\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.01136\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.00913\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.00893\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.00773\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.00607\n",
" -- Time: 6.20 seconds\n",
" -- Linear: model.layers.21.self_attn.k_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.09378\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.08510\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.07577\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.06884\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.04326\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.03671\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.06057\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.05497\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.04434\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.03813\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.03759\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.03076\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.02632\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.02092\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.01945\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.01535\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.01100\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.01008\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.00999\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.00908\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.00801\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.00781\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.00614\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.00506\n",
" -- Time: 3.85 seconds\n",
" -- Linear: model.layers.21.self_attn.v_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.22163\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.19857\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.18638\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.16824\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.10239\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.09164\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.12891\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.11509\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.10520\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.08830\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.08537\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.06626\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.05503\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.04911\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.04761\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.03326\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.02517\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.02403\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.02228\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.02129\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01708\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01587\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.01435\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.00975\n",
" -- Time: 3.84 seconds\n",
" -- Linear: model.layers.21.self_attn.o_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.23350\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.20365\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.19298\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.16334\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.10854\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.09643\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.12523\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.11522\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.11068\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.08594\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.07940\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.06430\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.05538\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.05234\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.05160\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.03227\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.02745\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.02665\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.02300\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.02251\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01727\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01758\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.01624\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.01239\n",
" -- Time: 6.21 seconds\n",
" -- Layer: model.layers.21 (MLP)\n",
" -- Linear: model.layers.21.mlp.gate_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.14922\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.14035\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.13702\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.12456\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.07072\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.06739\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.07943\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.07325\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.07139\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.06332\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.06061\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.04061\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.03517\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.03403\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.03376\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.02035\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.01774\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.01750\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.01628\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.01611\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01090\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01111\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.01053\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.00778\n",
" -- Time: 16.78 seconds\n",
" -- Linear: model.layers.21.mlp.up_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.23998\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.22618\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.22144\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.20132\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.11365\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.10871\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.12682\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.11691\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.11471\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.10196\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.09745\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.06472\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.05598\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.05453\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.05419\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.03238\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.02806\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.02771\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.02568\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.02546\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01715\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01695\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.01667\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.01124\n",
" -- Time: 16.72 seconds\n",
" -- Linear: model.layers.21.mlp.down_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.18 bpw rfn_error: 0.21060\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.18573\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.17462\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.15507\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.09622\n",
" -- 0.2:6b/0.8:3b 32g s4 3.71 bpw rfn_error: 0.08574\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.11685\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.10651\n",
" -- 0.05:4b/0.95:3b 32g s4 3.18 bpw rfn_error: 0.09940\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.08131\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.07743\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.05938\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.05114\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.04648\n",
" -- 0.1:6b/0.9:4b 32g s4 4.32 bpw rfn_error: 0.04530\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.03010\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.02458\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.34 bpw rfn_error: 0.02399\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.02170\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.02094\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01659\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01635\n",
" -- 0.1:8b/0.9:6b 128g s4 6.23 bpw rfn_error: 0.01509\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.01158\n",
" -- Time: 26.82 seconds\n",
" -- Layer: model.layers.22 (Attention)\n",
" -- Linear: model.layers.22.self_attn.q_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.11209\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.10229\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.09493\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.08603\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.05214\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.04653\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.06759\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.06109\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.05337\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.04590\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.04473\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.03453\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.02925\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.02521\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.02418\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.01736\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.01327\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.01260\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.01205\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.01141\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.00927\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.00901\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.00782\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.00614\n",
" -- Time: 6.22 seconds\n",
" -- Linear: model.layers.22.self_attn.k_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.10182\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.09235\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.08218\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.07468\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.04701\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.03994\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.06553\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.05968\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.04823\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.04141\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.04076\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.03335\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.02857\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.02278\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.02124\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.01667\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.01206\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.01106\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.01095\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.00998\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.00877\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.00861\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.00679\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.00569\n",
" -- Time: 3.81 seconds\n",
" -- Linear: model.layers.22.self_attn.v_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.21928\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.19550\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.18235\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.16446\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.10102\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.08950\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.12897\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.11501\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.10395\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.08672\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.08398\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.06603\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.05492\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.04849\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.04685\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.03305\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.02487\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.02364\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.02197\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.02087\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01693\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01587\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.01411\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.00978\n",
" -- Time: 3.84 seconds\n",
" -- Linear: model.layers.22.self_attn.o_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.21762\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.18789\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.17726\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.15261\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.10171\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.09076\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.11871\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.10860\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.10392\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.08076\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.07427\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.06082\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.05235\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.04925\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.04849\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.03053\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.02610\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.02537\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.02207\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.02155\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01645\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01714\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.01544\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.01247\n",
" -- Time: 6.22 seconds\n",
" -- Layer: model.layers.22 (MLP)\n",
" -- Linear: model.layers.22.mlp.gate_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.16317\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.15349\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.14991\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.13635\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.07727\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.07367\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.08670\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.07996\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.07803\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.06923\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.06627\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.04433\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.03836\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.03715\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.03686\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.02217\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.01930\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.01903\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.01769\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.01751\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01182\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01197\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.01143\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.00826\n",
" -- Time: 16.77 seconds\n",
" -- Linear: model.layers.22.mlp.up_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.24401\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.22991\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.22512\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.20468\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.11556\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.11050\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.12880\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.11884\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.11660\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.10366\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.09914\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.06572\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.05685\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.05539\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.05504\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.03285\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.02843\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.02807\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.02600\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.02578\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01734\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01707\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.01686\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.01120\n",
" -- Time: 16.74 seconds\n",
" -- Linear: model.layers.22.mlp.down_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.18 bpw rfn_error: 0.21802\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.19351\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.18269\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.16234\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.10003\n",
" -- 0.2:6b/0.8:3b 32g s4 3.71 bpw rfn_error: 0.08978\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.12078\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.10997\n",
" -- 0.05:4b/0.95:3b 32g s4 3.18 bpw rfn_error: 0.10301\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.08488\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.08076\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.06133\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.05274\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.04827\n",
" -- 0.1:6b/0.9:4b 32g s4 4.32 bpw rfn_error: 0.04717\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.03104\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.02547\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.34 bpw rfn_error: 0.02492\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.02254\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.02183\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01701\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01682\n",
" -- 0.1:8b/0.9:6b 128g s4 6.23 bpw rfn_error: 0.01556\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.01191\n",
" -- Time: 26.81 seconds\n",
" -- Layer: model.layers.23 (Attention)\n",
" -- Linear: model.layers.23.self_attn.q_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.11380\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.10433\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.09742\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.08835\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.05306\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.04771\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.06784\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.06138\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.05416\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.04680\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.04541\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.03455\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.02936\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.02560\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.02463\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.01730\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.01337\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.01272\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.01213\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.01151\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.00914\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.00890\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.00778\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.00591\n",
" -- Time: 6.21 seconds\n",
" -- Linear: model.layers.23.self_attn.k_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.10156\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.09196\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.08329\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.07560\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.04688\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.04059\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.06309\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.05749\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.04804\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.04111\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.04020\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.03200\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.02746\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.02265\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.02140\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.01605\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.01187\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.01105\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.01073\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.00993\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.00838\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.00821\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.00671\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.00536\n",
" -- Time: 3.80 seconds\n",
" -- Linear: model.layers.23.self_attn.v_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.23422\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.21260\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.20214\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.18244\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.10900\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.09931\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.13251\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.11975\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.11144\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.09471\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.09120\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.06770\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.05719\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.05224\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.05103\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.03382\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.02671\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.02575\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.02379\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.02298\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01734\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01653\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.01523\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.01019\n",
" -- Time: 3.88 seconds\n",
" -- Linear: model.layers.23.self_attn.o_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.21480\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.18771\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.17768\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.14708\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.09886\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.08851\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.11611\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.10647\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.10162\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.07861\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.07177\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.05943\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.05118\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.04773\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.04689\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.02974\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.02508\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.02427\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.02111\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.02054\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01586\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01625\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.01473\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.01144\n",
" -- Time: 6.19 seconds\n",
" -- Layer: model.layers.23 (MLP)\n",
" -- Linear: model.layers.23.mlp.gate_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.17054\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.16037\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.15662\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.14237\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.08076\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.07698\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.09058\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.08349\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.08155\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.07227\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.06921\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.04622\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.04005\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.03883\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.03853\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.02320\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.02017\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.01990\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.01847\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.01829\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01246\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01250\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.01206\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.00863\n",
" -- Time: 16.78 seconds\n",
" -- Linear: model.layers.23.mlp.up_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.24915\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.23466\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.22974\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.20875\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.11804\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.11278\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.13139\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.12130\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.11907\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.10571\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.10097\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.06703\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.05804\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.05656\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.05621\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.03350\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.02898\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.02862\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.02646\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.02623\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01764\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01732\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.01715\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.01125\n",
" -- Time: 16.71 seconds\n",
" -- Linear: model.layers.23.mlp.down_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.18 bpw rfn_error: 0.22415\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.19883\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.18751\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.16688\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.10277\n",
" -- 0.2:6b/0.8:3b 32g s4 3.71 bpw rfn_error: 0.09218\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.12361\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.11317\n",
" -- 0.05:4b/0.95:3b 32g s4 3.18 bpw rfn_error: 0.10595\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.08731\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.08310\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.06306\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.05426\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.04954\n",
" -- 0.1:6b/0.9:4b 32g s4 4.32 bpw rfn_error: 0.04837\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.03177\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.02603\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.34 bpw rfn_error: 0.02545\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.02301\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.02226\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01716\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01705\n",
" -- 0.1:8b/0.9:6b 128g s4 6.23 bpw rfn_error: 0.01560\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.01188\n",
" -- Time: 26.73 seconds\n",
" -- Layer: model.layers.24 (Attention)\n",
" -- Linear: model.layers.24.self_attn.q_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.12260\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.11279\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.10554\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.09573\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.05739\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.05178\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.07278\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.06617\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.05844\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.05073\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.04924\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.03712\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.03172\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.02773\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.02670\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.01860\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.01456\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.01389\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.01324\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.01261\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.00981\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.00976\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.00839\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.00663\n",
" -- Time: 6.21 seconds\n",
" -- Linear: model.layers.24.self_attn.k_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.10946\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.10019\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.09057\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.08224\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.05079\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.04400\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.06904\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.06303\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.05189\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.04493\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.04405\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.03504\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.03008\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.02456\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.02307\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.01753\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.01283\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.01189\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.01166\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.01073\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.00918\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.00890\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.00729\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.00571\n",
" -- Time: 3.80 seconds\n",
" -- Linear: model.layers.24.self_attn.v_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.24243\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.22225\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.21292\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.19248\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.11347\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.10475\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.13544\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.12257\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.11565\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.09936\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.09544\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.06901\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.05849\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.05435\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.05331\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.03449\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.02771\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.02689\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.02482\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.02415\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01768\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01686\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.01590\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.01036\n",
" -- Time: 3.81 seconds\n",
" -- Linear: model.layers.24.self_attn.o_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.18462\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.16712\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.15897\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.13305\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.08657\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.07936\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.10264\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.09414\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.08770\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.07102\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.06521\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.05243\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.04509\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.04164\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.04079\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.02619\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.02157\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.02092\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.01845\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.01788\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01375\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01365\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.01258\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.00905\n",
" -- Time: 6.20 seconds\n",
" -- Layer: model.layers.24 (MLP)\n",
" -- Linear: model.layers.24.mlp.gate_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.17165\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.16135\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.15763\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.14326\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.08135\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.07753\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.09113\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.08407\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.08214\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.07274\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.06955\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.04653\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.04032\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.03910\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.03881\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.02330\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.02027\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.02000\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.01854\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.01836\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01242\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01249\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.01202\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.00855\n",
" -- Time: 16.76 seconds\n",
" -- Linear: model.layers.24.mlp.up_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.25261\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.23792\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.23302\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.21172\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.11972\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.11447\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.13373\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.12299\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.12081\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.10716\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.10251\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.06826\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.05892\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.05744\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.05708\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.03423\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.02952\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.02916\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.02696\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.02674\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01830\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01776\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.01782\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.01172\n",
" -- Time: 16.70 seconds\n",
" -- Linear: model.layers.24.mlp.down_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.18 bpw rfn_error: 0.22909\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.20351\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.19234\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.17107\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.10504\n",
" -- 0.2:6b/0.8:3b 32g s4 3.71 bpw rfn_error: 0.09446\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.12623\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.11529\n",
" -- 0.05:4b/0.95:3b 32g s4 3.18 bpw rfn_error: 0.10829\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.08941\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.08494\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.06397\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.05522\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.05058\n",
" -- 0.1:6b/0.9:4b 32g s4 4.32 bpw rfn_error: 0.04942\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.03230\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.02647\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.34 bpw rfn_error: 0.02589\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.02340\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.02265\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01751\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01719\n",
" -- 0.1:8b/0.9:6b 128g s4 6.23 bpw rfn_error: 0.01599\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.01182\n",
" -- Time: 26.72 seconds\n",
" -- Layer: model.layers.25 (Attention)\n",
" -- Linear: model.layers.25.self_attn.q_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.12900\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.11907\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.11197\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.10142\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.06049\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.05492\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.07604\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.06910\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.06151\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.05347\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.05184\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.03876\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.03306\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.02918\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.02820\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.01940\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.01522\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.01456\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.01382\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.01321\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01017\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01003\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.00875\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.00668\n",
" -- Time: 6.20 seconds\n",
" -- Linear: model.layers.25.self_attn.k_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.11070\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.10151\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.09247\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.08384\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.05146\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.04495\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.06873\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.06295\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.05250\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.04551\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.04442\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.03481\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.03008\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.02483\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.02343\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.01742\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.01292\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.01203\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.01174\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.01086\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.00904\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.00883\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.00729\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.00564\n",
" -- Time: 3.85 seconds\n",
" -- Linear: model.layers.25.self_attn.v_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.24470\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.22439\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.21520\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.19456\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.11443\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.10578\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.13687\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.12336\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.11654\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.10039\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.09637\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.06994\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.05896\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.05485\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.05384\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.03487\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.02793\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.02710\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.02503\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.02437\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01793\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01695\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.01609\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.01037\n",
" -- Time: 3.80 seconds\n",
" -- Linear: model.layers.25.self_attn.o_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.19971\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.17495\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.16666\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.13776\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.09312\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.08423\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.10770\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.09795\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.09494\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.07392\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.06606\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.05482\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.04697\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.04471\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.04418\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.02750\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.02313\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.02253\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.01925\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.01888\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01463\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01430\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.01390\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.00958\n",
" -- Time: 6.19 seconds\n",
" -- Layer: model.layers.25 (MLP)\n",
" -- Linear: model.layers.25.mlp.gate_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.17795\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.16722\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.16331\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.14839\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.08445\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.08044\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.09471\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.08726\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.08527\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.07546\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.07213\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.04844\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.04190\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.04062\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.04032\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.02428\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.02113\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.02084\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.01933\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.01914\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01304\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01312\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.01263\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.00910\n",
" -- Time: 16.76 seconds\n",
" -- Linear: model.layers.25.mlp.up_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.25596\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.24084\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.23583\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.21412\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.12140\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.11602\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.13501\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.12474\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.12254\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.10858\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.10364\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.06887\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.05970\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.05820\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.05784\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.03440\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.02980\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.02943\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.02717\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.02694\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01798\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01778\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.01748\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.01151\n",
" -- Time: 16.68 seconds\n",
" -- Linear: model.layers.25.mlp.down_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.18 bpw rfn_error: 0.23103\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.20522\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.19355\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.17225\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.10588\n",
" -- 0.2:6b/0.8:3b 32g s4 3.71 bpw rfn_error: 0.09500\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.12759\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.11703\n",
" -- 0.05:4b/0.95:3b 32g s4 3.18 bpw rfn_error: 0.10922\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.09041\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.08588\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.06501\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.05611\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.05101\n",
" -- 0.1:6b/0.9:4b 32g s4 4.32 bpw rfn_error: 0.04974\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.03263\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.02675\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.34 bpw rfn_error: 0.02612\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.02371\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.02290\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01746\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01750\n",
" -- 0.1:8b/0.9:6b 128g s4 6.23 bpw rfn_error: 0.01578\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.01207\n",
" -- Time: 26.74 seconds\n",
" -- Layer: model.layers.26 (Attention)\n",
" -- Linear: model.layers.26.self_attn.q_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.12400\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.11394\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.10634\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.09626\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.05808\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.05220\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.07396\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.06728\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.05918\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.05122\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.04970\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.03773\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.03227\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.02804\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.02695\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.01888\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.01467\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.01396\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.01331\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.01263\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.00986\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.00981\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.00834\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.00655\n",
" -- Time: 6.20 seconds\n",
" -- Linear: model.layers.26.self_attn.k_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.10733\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.09786\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.08772\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.07956\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.04976\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.04267\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.06874\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.06260\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.05088\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.04381\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.04301\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.03479\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.02981\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.02406\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.02250\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.01739\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.01260\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.01160\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.01143\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.01045\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.00913\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.00881\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.00717\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.00566\n",
" -- Time: 3.83 seconds\n",
" -- Linear: model.layers.26.self_attn.v_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.23561\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.21320\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.20178\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.18208\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.10948\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.09923\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.13531\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.12136\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.11215\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.09498\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.09165\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.06941\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.05795\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.05253\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.05117\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.03472\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.02681\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.02576\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.02380\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.02292\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01782\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01662\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.01538\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.01014\n",
" -- Time: 3.78 seconds\n",
" -- Linear: model.layers.26.self_attn.o_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.19595\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.17731\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.17136\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.14596\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.09234\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.08536\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.10379\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.09552\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.09358\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.07667\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.06999\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.05305\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.04578\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.04433\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.04400\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.02652\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.02284\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.02238\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.01966\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.01943\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01391\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01389\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.01344\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.00925\n",
" -- Time: 6.19 seconds\n",
" -- Layer: model.layers.26 (MLP)\n",
" -- Linear: model.layers.26.mlp.gate_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.18272\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.17161\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.16761\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.15209\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.08677\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.08269\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.09719\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.08969\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.08762\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.07741\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.07390\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.04969\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.04302\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.04170\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.04140\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.02485\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.02162\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.02132\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.01974\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.01954\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01320\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01332\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.01277\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.00911\n",
" -- Time: 16.80 seconds\n",
" -- Linear: model.layers.26.mlp.up_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.25765\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.24243\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.23726\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.21532\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.12240\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.11689\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.13638\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.12577\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.12352\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.10932\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.10434\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.06969\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.06023\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.05870\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.05834\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.03484\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.03016\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.02978\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.02749\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.02726\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01842\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01812\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.01790\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.01192\n",
" -- Time: 16.72 seconds\n",
" -- Linear: model.layers.26.mlp.down_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.18 bpw rfn_error: 0.23600\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.21045\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.19898\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.17697\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.10846\n",
" -- 0.2:6b/0.8:3b 32g s4 3.71 bpw rfn_error: 0.09767\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.13012\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.11931\n",
" -- 0.05:4b/0.95:3b 32g s4 3.18 bpw rfn_error: 0.11173\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.09277\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.08794\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.06635\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.05727\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.05228\n",
" -- 0.1:6b/0.9:4b 32g s4 4.32 bpw rfn_error: 0.05105\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.03353\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.02746\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.34 bpw rfn_error: 0.02683\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.02439\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.02359\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01825\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01798\n",
" -- 0.1:8b/0.9:6b 128g s4 6.23 bpw rfn_error: 0.01663\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.01248\n",
" -- Time: 26.78 seconds\n",
" -- Layer: model.layers.27 (Attention)\n",
" -- Linear: model.layers.27.self_attn.q_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.12800\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.11782\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.11131\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.10085\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.06014\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.05475\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.07450\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.06772\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.06119\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.05302\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.05126\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.03807\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.03246\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.02901\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.02815\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.01906\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.01516\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.01454\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.01374\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.01318\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01001\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.00993\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.00873\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.00666\n",
" -- Time: 6.21 seconds\n",
" -- Linear: model.layers.27.self_attn.k_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.11138\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.10262\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.09586\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.08676\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.05208\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.04693\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.06588\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.06038\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.05305\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.04602\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.04448\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.03345\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.02885\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.02511\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.02417\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.01676\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.01307\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.01244\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.01185\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.01125\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.00880\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.00864\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.00750\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.00566\n",
" -- Time: 3.84 seconds\n",
" -- Linear: model.layers.27.self_attn.v_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.23298\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.21409\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.20526\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.18563\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.10952\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.10109\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.12914\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.11787\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.11131\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.09585\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.09179\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.06580\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.05635\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.05246\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.05151\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.03284\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.02669\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.02589\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.02386\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.02323\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01685\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01613\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.01537\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.00977\n",
" -- Time: 3.82 seconds\n",
" -- Linear: model.layers.27.self_attn.o_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.21200\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.18792\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.17930\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.15069\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.09955\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.09040\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.11465\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.10507\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.10141\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.07998\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.07405\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.05895\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.05073\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.04826\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.04768\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.02975\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.02572\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.02513\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.02197\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.02157\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01632\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01694\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.01555\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.01257\n",
" -- Time: 6.21 seconds\n",
" -- Layer: model.layers.27 (MLP)\n",
" -- Linear: model.layers.27.mlp.gate_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.17764\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.16664\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.16263\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.14745\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.08444\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.08030\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.09481\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.08744\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.08529\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.07525\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.07172\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.04845\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.04203\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.04066\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.04033\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.02426\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.02116\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.02086\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.01932\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.01911\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01290\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01317\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.01245\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.00915\n",
" -- Time: 16.77 seconds\n",
" -- Linear: model.layers.27.mlp.up_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.25108\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.23599\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.23094\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.20941\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.11934\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.11386\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.13301\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.12278\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.12049\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.10647\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.10147\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.06792\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.05881\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.05724\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.05687\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.03394\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.02938\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.02900\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.02675\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.02651\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01784\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01762\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.01731\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.01152\n",
" -- Time: 16.71 seconds\n",
" -- Linear: model.layers.27.mlp.down_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.18 bpw rfn_error: 0.23620\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.20981\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.19752\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.17574\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.10841\n",
" -- 0.2:6b/0.8:3b 32g s4 3.71 bpw rfn_error: 0.09708\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.13220\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.12031\n",
" -- 0.05:4b/0.95:3b 32g s4 3.18 bpw rfn_error: 0.11186\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.09251\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.08790\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.06722\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.05766\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.05227\n",
" -- 0.1:6b/0.9:4b 32g s4 4.32 bpw rfn_error: 0.05094\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.03389\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.02747\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.34 bpw rfn_error: 0.02681\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.02436\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.02351\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01839\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01807\n",
" -- 0.1:8b/0.9:6b 128g s4 6.23 bpw rfn_error: 0.01660\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.01255\n",
" -- Time: 26.76 seconds\n",
" -- Layer: model.layers.28 (Attention)\n",
" -- Linear: model.layers.28.self_attn.q_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.11916\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.10948\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.10109\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.09155\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.05581\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.04952\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.07333\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.06626\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.05690\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.04925\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.04815\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.03747\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.03177\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.02701\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.02579\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.01876\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.01423\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.01343\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.01293\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.01216\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.00986\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.00973\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.00813\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.00657\n",
" -- Time: 6.20 seconds\n",
" -- Linear: model.layers.28.self_attn.k_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.10099\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.09191\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.08054\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.07318\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.04673\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.03898\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.06745\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.06117\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.04785\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.04130\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.04084\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.03410\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.02918\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.02265\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.02082\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.01709\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.01194\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.01081\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.01087\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.00973\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.00890\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.00864\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.00663\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.00555\n",
" -- Time: 3.79 seconds\n",
" -- Linear: model.layers.28.self_attn.v_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.23685\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.21366\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.20104\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.18104\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.11007\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.09898\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.13777\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.12351\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.11284\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.09523\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.09198\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.07078\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.05913\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.05296\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.05141\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.03556\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.02727\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.02609\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.02422\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.02322\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01831\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01733\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.01558\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.01099\n",
" -- Time: 3.80 seconds\n",
" -- Linear: model.layers.28.self_attn.o_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.22737\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.19411\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.18248\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.15765\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.10530\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.09273\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.12337\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.11270\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.10799\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.08227\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.07847\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.06327\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.05415\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.05081\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.04999\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.03169\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.02673\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.02590\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.02225\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.02169\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01706\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01731\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.01599\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.01227\n",
" -- Time: 6.21 seconds\n",
" -- Layer: model.layers.28 (MLP)\n",
" -- Linear: model.layers.28.mlp.gate_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.16706\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.15650\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.15241\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.13809\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.07939\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.07527\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.08959\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.08263\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.08023\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.07062\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.06740\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.04581\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.03969\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.03821\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.03787\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.02293\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.01987\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.01956\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.01811\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.01789\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01220\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01237\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.01171\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.00853\n",
" -- Time: 16.76 seconds\n",
" -- Linear: model.layers.28.mlp.up_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.23511\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.22073\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.21568\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.19541\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.11175\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.10642\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.12497\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.11535\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.11284\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.09957\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.09490\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.06392\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.05531\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.05369\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.05329\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.03199\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.02771\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.02733\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.02525\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.02500\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01694\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01690\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.01639\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.01135\n",
" -- Time: 16.70 seconds\n",
" -- Linear: model.layers.28.mlp.down_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.18 bpw rfn_error: 0.21678\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.19135\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.17924\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.15888\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.09953\n",
" -- 0.2:6b/0.8:3b 32g s4 3.71 bpw rfn_error: 0.08835\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.12181\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.11122\n",
" -- 0.05:4b/0.95:3b 32g s4 3.18 bpw rfn_error: 0.10275\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.08420\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.07999\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.06226\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.05358\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.04823\n",
" -- 0.1:6b/0.9:4b 32g s4 4.32 bpw rfn_error: 0.04689\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.03155\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.02574\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.34 bpw rfn_error: 0.02505\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.02281\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.02195\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01736\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01746\n",
" -- 0.1:8b/0.9:6b 128g s4 6.23 bpw rfn_error: 0.01561\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.01263\n",
" -- Time: 26.80 seconds\n",
" -- Layer: model.layers.29 (Attention)\n",
" -- Linear: model.layers.29.self_attn.q_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.11633\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.10687\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.09922\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.08996\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.05443\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.04862\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.07176\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.06381\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.05549\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.04804\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.04688\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.03664\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.03057\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.02629\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.02519\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.01837\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.01378\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.01303\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.01251\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.01182\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.00962\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.00929\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.00793\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.00619\n",
" -- Time: 6.20 seconds\n",
" -- Linear: model.layers.29.self_attn.k_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.10096\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.09321\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.08475\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.07676\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.04705\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.04115\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.06402\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.05806\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.04792\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.04176\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.04090\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.03252\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.02767\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.02274\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.02141\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.01626\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.01190\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.01105\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.01087\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.01002\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.00850\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.00824\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.00673\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.00531\n",
" -- Time: 3.82 seconds\n",
" -- Linear: model.layers.29.self_attn.v_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.25490\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.23516\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.22557\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.20400\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.12012\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.11134\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.14274\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.12954\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.12215\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.10564\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.10147\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.07304\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.06200\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.05764\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.05657\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.03666\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.02949\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.02865\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.02653\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.02583\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01884\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01809\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.01694\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.01133\n",
" -- Time: 3.78 seconds\n",
" -- Linear: model.layers.29.self_attn.o_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.19766\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.16867\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.15821\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.13995\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.09188\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.08069\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.10952\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.09814\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.09429\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.07310\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.07000\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.05625\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.04722\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.04440\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.04370\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.02818\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.02332\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.02257\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.01963\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.01917\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01500\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01505\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.01398\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.01065\n",
" -- Time: 6.22 seconds\n",
" -- Layer: model.layers.29 (MLP)\n",
" -- Linear: model.layers.29.mlp.gate_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.16208\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.15172\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.14757\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.13356\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.07702\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.07297\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.08709\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.08036\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.07788\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.06844\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.06527\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.04455\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.03862\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.03711\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.03676\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.02231\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.01938\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.01907\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.01768\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.01745\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01190\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01221\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.01140\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.00855\n",
" -- Time: 16.78 seconds\n",
" -- Linear: model.layers.29.mlp.up_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.20992\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.19685\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.19202\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.17416\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.10051\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.09569\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.11278\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.10406\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.10156\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.08960\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.08555\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.05828\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.05090\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.04933\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.04898\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.02937\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.02721\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.02688\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.02521\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.02500\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01658\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01893\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.01610\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.01526\n",
" -- Time: 16.72 seconds\n",
" -- Linear: model.layers.29.mlp.down_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.18 bpw rfn_error: 0.21573\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.19141\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.18033\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.15850\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.09927\n",
" -- 0.2:6b/0.8:3b 32g s4 3.71 bpw rfn_error: 0.08897\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.12029\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.10954\n",
" -- 0.05:4b/0.95:3b 32g s4 3.18 bpw rfn_error: 0.10214\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.08370\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.07890\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.06104\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.05253\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.04791\n",
" -- 0.1:6b/0.9:4b 32g s4 4.32 bpw rfn_error: 0.04676\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.03064\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.02528\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.34 bpw rfn_error: 0.02470\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.02226\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.02151\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01645\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01670\n",
" -- 0.1:8b/0.9:6b 128g s4 6.23 bpw rfn_error: 0.01491\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.01178\n",
" -- Time: 26.78 seconds\n",
" -- Layer: model.layers.30 (Attention)\n",
" -- Linear: model.layers.30.self_attn.q_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.11347\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.10390\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.09477\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.08558\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.05303\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.04638\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.07177\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.06453\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.05413\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.04664\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.04581\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.03667\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.03100\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.02572\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.02433\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.01841\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.01363\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.01274\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.01238\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.01151\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.00972\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.00955\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.00781\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.00648\n",
" -- Time: 6.23 seconds\n",
" -- Linear: model.layers.30.self_attn.k_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.09913\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.09016\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.07836\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.07101\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.04575\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.03783\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.06726\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.06079\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.04693\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.04040\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.04014\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.03422\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.02902\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.02224\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.02033\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.01718\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.01180\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.01061\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.01076\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.00955\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.00894\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.00868\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.00650\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.00564\n",
" -- Time: 3.82 seconds\n",
" -- Linear: model.layers.30.self_attn.v_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.23066\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.20681\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.19312\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.17402\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.10691\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.09514\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.13683\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.12224\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.10968\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.09215\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.08949\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.07042\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.05845\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.05150\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.04971\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.03542\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.02662\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.02530\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.02364\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.02248\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01832\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01728\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.01517\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.01106\n",
" -- Time: 3.81 seconds\n",
" -- Linear: model.layers.30.self_attn.o_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.17670\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.15877\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.15307\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.13045\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.08273\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.07668\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.09408\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.08571\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.08387\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.06803\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.06273\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.04834\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.04136\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.03999\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.03966\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.02425\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.02117\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.02080\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.01832\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.01811\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01313\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01369\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.01271\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.01006\n",
" -- Time: 6.22 seconds\n",
" -- Layer: model.layers.30 (MLP)\n",
" -- Linear: model.layers.30.mlp.gate_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.15936\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.14924\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.14528\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.13162\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.07624\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.07227\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.08603\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.07933\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.07705\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.06784\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.06488\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.04448\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.03879\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.03740\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.03707\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.02246\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.02056\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.02028\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.01901\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.01882\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01267\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01425\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.01223\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.01139\n",
" -- Time: 16.81 seconds\n",
" -- Linear: model.layers.30.mlp.up_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.16603\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.15569\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.15184\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.13759\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.07945\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.07553\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.08923\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.08227\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.08023\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.07077\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.06760\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.04609\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.04012\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.03886\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.03855\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.02325\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.02123\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.02095\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.01960\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.01943\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01306\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01453\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.01265\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.01151\n",
" -- Time: 16.72 seconds\n",
" -- Linear: model.layers.30.mlp.down_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.18 bpw rfn_error: 0.18738\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.16703\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.15843\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.13782\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.08682\n",
" -- 0.2:6b/0.8:3b 32g s4 3.71 bpw rfn_error: 0.07855\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.10584\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.09386\n",
" -- 0.05:4b/0.95:3b 32g s4 3.18 bpw rfn_error: 0.08880\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.07284\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.06854\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.05334\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.04573\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.04253\n",
" -- 0.1:6b/0.9:4b 32g s4 4.32 bpw rfn_error: 0.04171\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.02725\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.02341\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.34 bpw rfn_error: 0.02301\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.02086\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.02037\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01563\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01646\n",
" -- 0.1:8b/0.9:6b 128g s4 6.23 bpw rfn_error: 0.01464\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.01297\n",
" -- Time: 26.74 seconds\n",
" -- Layer: model.layers.31 (Attention)\n",
" -- Linear: model.layers.31.self_attn.q_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.10866\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.09916\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.09031\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.08169\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.05073\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.04421\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.06838\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.06180\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.05183\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.04453\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.04367\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.03496\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.02965\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.02457\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.02324\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.01751\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.01299\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.01213\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.01178\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.01095\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.00921\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.00908\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.00739\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.00612\n",
" -- Time: 6.20 seconds\n",
" -- Linear: model.layers.31.self_attn.k_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.09790\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.08884\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.07772\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.07047\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.04527\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.03773\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.06637\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.05931\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.04648\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.03986\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.03951\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.03362\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.02834\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.02199\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.02022\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.01685\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.01166\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.01057\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.01061\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.00950\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.00885\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.00850\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.00657\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.00558\n",
" -- Time: 3.81 seconds\n",
" -- Linear: model.layers.31.self_attn.v_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.24299\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.21556\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.19947\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.17985\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.11267\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.09866\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.14473\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.13010\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.11611\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.09605\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.09349\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.07436\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.06221\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.05421\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.05216\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.03738\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.02799\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.02645\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.02463\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.02326\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01925\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01820\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.01589\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.01143\n",
" -- Time: 3.81 seconds\n",
" -- Linear: model.layers.31.self_attn.o_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.11279\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.09924\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.09478\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.08162\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.05092\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.04609\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.06104\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.05349\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.05188\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.04118\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.03942\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.03045\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.02751\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.02643\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.02621\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.01618\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.01648\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.01625\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.01499\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.01487\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01008\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01328\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.00977\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.01199\n",
" -- Time: 6.19 seconds\n",
" -- Layer: model.layers.31 (MLP)\n",
" -- Linear: model.layers.31.mlp.gate_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.14756\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.13842\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.13503\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.12221\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.07013\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.06670\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.07889\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.07267\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.07083\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.06240\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.05951\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.04030\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.03487\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.03371\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.03344\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.02017\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.01744\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.01718\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.01588\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.01570\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01063\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01072\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.01024\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.00726\n",
" -- Time: 16.81 seconds\n",
" -- Linear: model.layers.31.mlp.up_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.19 bpw rfn_error: 0.09824\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.09194\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.08962\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.08108\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.04664\n",
" -- 0.2:6b/0.8:3b 32g s4 3.69 bpw rfn_error: 0.04430\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.05260\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.04838\n",
" -- 0.05:4b/0.95:3b 32g s4 3.19 bpw rfn_error: 0.04713\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.04143\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.03955\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.02695\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.02329\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.02251\n",
" -- 0.1:6b/0.9:4b 32g s4 4.31 bpw rfn_error: 0.02232\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.01354\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.01183\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.38 bpw rfn_error: 0.01166\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.01080\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.01069\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.00736\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.00754\n",
" -- 0.1:8b/0.9:6b 128g s4 6.22 bpw rfn_error: 0.00711\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.00542\n",
" -- Time: 16.70 seconds\n",
" -- Linear: model.layers.31.mlp.down_proj\n",
" -- 0.05:3b/0.95:2b 32g s4 2.18 bpw rfn_error: 0.12848\n",
" -- 0.25:3b/0.75:2b 32g s4 2.38 bpw rfn_error: 0.11652\n",
" -- 0.25:4b/0.75:2b 32g s4 2.63 bpw rfn_error: 0.11088\n",
" -- 0.1:4b/0.4:3b/0.5:2b 32g s4 2.72 bpw rfn_error: 0.09626\n",
" -- 0.1:4b/0.9:3b 32g s4 3.22 bpw rfn_error: 0.05991\n",
" -- 0.2:6b/0.8:3b 32g s4 3.71 bpw rfn_error: 0.05494\n",
" -- 1.0:3b 128g s4 3.03 bpw rfn_error: 0.07339\n",
" -- 1.0:3b 32g s4 3.13 bpw rfn_error: 0.06539\n",
" -- 0.05:4b/0.95:3b 32g s4 3.18 bpw rfn_error: 0.06116\n",
" -- 0.4:4b/0.6:3b 32g s4 3.53 bpw rfn_error: 0.05120\n",
" -- 0.6:4b/0.4:3b 64g s4 3.66 bpw rfn_error: 0.04834\n",
" -- 1.0:4b 128g s4 4.03 bpw rfn_error: 0.03777\n",
" -- 1.0:4b 32g s4 4.13 bpw rfn_error: 0.03195\n",
" -- 0.1:5b/0.9:4b 32g s4 4.22 bpw rfn_error: 0.02955\n",
" -- 0.1:6b/0.9:4b 32g s4 4.32 bpw rfn_error: 0.02893\n",
" -- 1.0:5b 128g s4 5.03 bpw rfn_error: 0.01943\n",
" -- 0.1:6b/0.9:5b 32g s4 5.22 bpw rfn_error: 0.01655\n",
" -- 0.05:8b/0.05:6b/0.9:5b 32g s4 5.34 bpw rfn_error: 0.01626\n",
" -- 0.4:6b/0.6:5b 32g s4 5.53 bpw rfn_error: 0.01499\n",
" -- 0.1:8b/0.3:6b/0.6:5b 32g s4 5.72 bpw rfn_error: 0.01465\n",
" -- 1.0:6b 128g s4 6.03 bpw rfn_error: 0.01136\n",
" -- 1.0:6b 32g s4 6.13 bpw rfn_error: 0.01197\n",
" -- 0.1:8b/0.9:6b 128g s4 6.23 bpw rfn_error: 0.01058\n",
" -- 1.0:8b 32g s4 8.13 bpw rfn_error: 0.00968\n",
" -- Time: 26.78 seconds\n",
" -- Layer: model.norm (RMSNorm)\n",
" -- Layer: lm_head (Linear)\n",
" -- Calibration perplexity (base): 6.5756\n",
" -- Optimizing...\n",
" -- rfn max: 0.12882 bpw: 2.89358\n",
" -- rfn max: 0.06441 bpw: 3.83733\n",
" -- rfn max: 0.03221 bpw: 4.87716\n",
" -- rfn max: 0.01610 bpw: 6.44259\n",
" -- rfn max: 0.02420 bpw: 5.31034\n",
" -- rfn max: 0.02831 bpw: 4.98366\n",
" -- rfn max: 0.02625 bpw: 5.11053\n",
" -- rfn max: 0.02733 bpw: 5.03992\n",
" -- rfn max: 0.02792 bpw: 5.00472\n",
" -- rfn max: 0.02826 bpw: 4.98372\n",
" -- rfn max: 0.02809 bpw: 4.99354\n",
" -- rfn max: 0.02800 bpw: 4.99961\n",
" -- rfn max: 0.02796 bpw: 5.00233\n",
" -- rfn max: 0.02803 bpw: 4.99924\n",
" -- rfn max: 0.02800 bpw: 4.99961\n",
" -- rfn max: 0.02798 bpw: 5.00233\n",
" -- rfn max: 0.02804 bpw: 4.99880\n",
" -- rfn max: 0.02801 bpw: 4.99961\n",
" -- rfn max: 0.02799 bpw: 5.00093\n",
" -- rfn max: 0.02805 bpw: 4.99880\n",
" -- rfn max: 0.02802 bpw: 4.99924\n",
" -- rfn max: 0.02801 bpw: 4.99961\n",
" -- rfn max: 0.02800 bpw: 4.99961\n",
" -- rfn min: 0.01400 bpw: 5.89112\n",
" -- rfn min: 0.02100 bpw: 5.27758\n",
" -- rfn min: 0.02450 bpw: 5.11620\n",
" -- rfn min: 0.02625 bpw: 5.03063\n",
" -- rfn min: 0.02713 bpw: 5.00971\n",
" -- rfn min: 0.02756 bpw: 4.99961\n",
" -- rfn min: 0.02734 bpw: 5.00551\n",
" -- rfn min: 0.02745 bpw: 5.00093\n",
" -- rfn min: 0.02751 bpw: 5.00093\n",
" -- rfn min: 0.02754 bpw: 5.00093\n",
" -- rfn min: 0.02755 bpw: 4.99961\n",
" -- rfn min: 0.02754 bpw: 4.99961\n",
" -- Tokenizing samples...\n",
" -- First 50 tokens of dataset:\n",
" ' = Robert Boulter = \\n Robert Boulter is an English film , television and theatre actor . He had a guest @-@ starring role on the television series The Bill in 2000 . This was followed'\n",
" -- Last 50 tokens of dataset:\n",
" '] more meaningful lives \" . The film argues the case against conformity , but does not deny that people need and want it ; even the gay characters just want to fit in . Jim and Jim , the Burnhams \\' other neighbors , are'\n",
" -- Token embeddings again...\n",
" -- Quantizing...\n",
" -- Layer: model.layers.0 (Attention)\n",
" -- Linear: model.layers.0.self_attn.q_proj -> 0.05:3b/0.95:2b 32g s4, 2.19 bpw\n",
" -- Linear: model.layers.0.self_attn.k_proj -> 0.05:3b/0.95:2b 32g s4, 2.19 bpw\n",
" -- Linear: model.layers.0.self_attn.v_proj -> 0.2:6b/0.8:3b 32g s4, 3.69 bpw\n",
" -- Linear: model.layers.0.self_attn.o_proj -> 0.1:5b/0.9:4b 32g s4, 4.22 bpw\n",
" -- Layer rfn_error: 0.022664\n",
" -- Module quantized, time: 15.82 seconds\n",
" -- Layer: model.layers.0 (MLP)\n",
" -- Linear: model.layers.0.mlp.gate_proj -> 1.0:4b 32g s4, 4.13 bpw\n",
" -- Linear: model.layers.0.mlp.up_proj -> 1.0:5b 128g s4, 5.03 bpw\n",
" -- Linear: model.layers.0.mlp.down_proj -> 0.6:4b/0.4:3b 64g s4, 3.66 bpw\n",
" -- Layer rfn_error: 0.030517\n",
" -- Module quantized, time: 39.55 seconds\n",
" -- Layer: model.layers.1 (Attention)\n",
" -- Linear: model.layers.1.self_attn.q_proj -> 0.05:3b/0.95:2b 32g s4, 2.19 bpw\n",
" -- Linear: model.layers.1.self_attn.k_proj -> 0.05:3b/0.95:2b 32g s4, 2.19 bpw\n",
" -- Linear: model.layers.1.self_attn.v_proj -> 0.1:5b/0.9:4b 32g s4, 4.22 bpw\n",
" -- Linear: model.layers.1.self_attn.o_proj -> 1.0:5b 128g s4, 5.03 bpw\n",
" -- Layer rfn_error: 0.027920\n",
" -- Module quantized, time: 15.11 seconds\n",
" -- Layer: model.layers.1 (MLP)\n",
" -- Linear: model.layers.1.mlp.gate_proj -> 1.0:5b 128g s4, 5.03 bpw\n",
" -- Linear: model.layers.1.mlp.up_proj -> 1.0:5b 128g s4, 5.03 bpw\n",
" -- Linear: model.layers.1.mlp.down_proj -> 0.05:3b/0.95:2b 32g s4, 2.18 bpw\n",
" -- Layer rfn_error: 0.019765\n",
" -- Module quantized, time: 38.87 seconds\n",
" -- Layer: model.layers.2 (Attention)\n",
" -- Linear: model.layers.2.self_attn.q_proj -> 0.05:4b/0.95:3b 32g s4, 3.19 bpw\n",
" -- Linear: model.layers.2.self_attn.k_proj -> 0.05:4b/0.95:3b 32g s4, 3.19 bpw\n",
" -- Linear: model.layers.2.self_attn.v_proj -> 1.0:5b 128g s4, 5.03 bpw\n",
" -- Linear: model.layers.2.self_attn.o_proj -> 1.0:5b 128g s4, 5.03 bpw\n",
" -- Layer rfn_error: 0.001632\n",
" -- Module quantized, time: 15.01 seconds\n",
" -- Layer: model.layers.2 (MLP)\n",
" -- Linear: model.layers.2.mlp.gate_proj -> 1.0:5b 128g s4, 5.03 bpw\n",
" -- Linear: model.layers.2.mlp.up_proj -> 0.1:6b/0.9:5b 32g s4, 5.22 bpw\n",
" -- Linear: model.layers.2.mlp.down_proj -> 1.0:5b 128g s4, 5.03 bpw\n",
" -- Layer rfn_error: 0.001273\n",
" -- Module quantized, time: 40.94 seconds\n",
" -- Layer: model.layers.3 (Attention)\n",
" -- Linear: model.layers.3.self_attn.q_proj -> 1.0:3b 128g s4, 3.03 bpw\n",
" -- Linear: model.layers.3.self_attn.k_proj -> 0.1:4b/0.4:3b/0.5:2b 32g s4, 2.72 bpw\n",
" -- Linear: model.layers.3.self_attn.v_proj -> 1.0:5b 128g s4, 5.03 bpw\n",
" -- Linear: model.layers.3.self_attn.o_proj -> 1.0:5b 128g s4, 5.03 bpw\n",
" -- Layer rfn_error: 0.002017\n",
" -- Module quantized, time: 14.93 seconds\n",
" -- Layer: model.layers.3 (MLP)\n",
" -- Linear: model.layers.3.mlp.gate_proj -> 1.0:5b 128g s4, 5.03 bpw\n",
" -- Linear: model.layers.3.mlp.up_proj -> 0.4:6b/0.6:5b 32g s4, 5.53 bpw\n",
" -- Linear: model.layers.3.mlp.down_proj -> 1.0:5b 128g s4, 5.03 bpw\n",
" -- Layer rfn_error: 0.001798\n",
" -- Module quantized, time: 39.25 seconds\n",
" -- Layer: model.layers.4 (Attention)\n",
" -- Linear: model.layers.4.self_attn.q_proj -> 0.1:4b/0.9:3b 32g s4, 3.22 bpw\n",
" -- Linear: model.layers.4.self_attn.k_proj -> 0.05:4b/0.95:3b 32g s4, 3.19 bpw\n",
" -- Linear: model.layers.4.self_attn.v_proj -> 1.0:5b 128g s4, 5.03 bpw\n",
" -- Linear: model.layers.4.self_attn.o_proj -> 1.0:5b 128g s4, 5.03 bpw\n",
" -- Layer rfn_error: 0.002456\n",
" -- Module quantized, time: 15.05 seconds\n",
" -- Layer: model.layers.4 (MLP)\n",
" -- Linear: model.layers.4.mlp.gate_proj -> 1.0:5b 128g s4, 5.03 bpw\n",
" -- Linear: model.layers.4.mlp.up_proj -> 0.4:6b/0.6:5b 32g s4, 5.53 bpw\n",
" -- Linear: model.layers.4.mlp.down_proj -> 0.1:6b/0.9:5b 32g s4, 5.22 bpw\n",
" -- Layer rfn_error: 0.002233\n",
" -- Module quantized, time: 39.20 seconds\n",
" -- Layer: model.layers.5 (Attention)\n",
" -- Linear: model.layers.5.self_attn.q_proj -> 0.4:4b/0.6:3b 32g s4, 3.53 bpw\n",
" -- Linear: model.layers.5.self_attn.k_proj -> 0.4:4b/0.6:3b 32g s4, 3.53 bpw\n",
" -- Linear: model.layers.5.self_attn.v_proj -> 0.1:6b/0.9:5b 32g s4, 5.22 bpw\n",
" -- Linear: model.layers.5.self_attn.o_proj -> 1.0:5b 128g s4, 5.03 bpw\n",
" -- Layer rfn_error: 0.002432\n",
" -- Module quantized, time: 15.09 seconds\n",
" -- Layer: model.layers.5 (MLP)\n",
" -- Linear: model.layers.5.mlp.gate_proj -> 1.0:5b 128g s4, 5.03 bpw\n",
" -- Linear: model.layers.5.mlp.up_proj -> 0.05:8b/0.05:6b/0.9:5b 32g s4, 5.38 bpw\n",
" -- Linear: model.layers.5.mlp.down_proj -> 0.1:6b/0.9:5b 32g s4, 5.22 bpw\n",
" -- Layer rfn_error: 0.003064\n",
" -- Module quantized, time: 40.26 seconds\n",
" -- Layer: model.layers.6 (Attention)\n",
" -- Linear: model.layers.6.self_attn.q_proj -> 0.4:4b/0.6:3b 32g s4, 3.53 bpw\n",
" -- Linear: model.layers.6.self_attn.k_proj -> 0.1:4b/0.9:3b 32g s4, 3.22 bpw\n",
" -- Linear: model.layers.6.self_attn.v_proj -> 1.0:5b 128g s4, 5.03 bpw\n",
" -- Linear: model.layers.6.self_attn.o_proj -> 1.0:5b 128g s4, 5.03 bpw\n",
" -- Layer rfn_error: 0.003424\n",
" -- Module quantized, time: 15.77 seconds\n",
" -- Layer: model.layers.6 (MLP)\n",
" -- Linear: model.layers.6.mlp.gate_proj -> 1.0:5b 128g s4, 5.03 bpw\n",
" -- Linear: model.layers.6.mlp.up_proj -> 0.1:6b/0.9:5b 32g s4, 5.22 bpw\n",
" -- Linear: model.layers.6.mlp.down_proj -> 0.1:6b/0.9:5b 32g s4, 5.22 bpw\n",
" -- Layer rfn_error: 0.003627\n",
" -- Module quantized, time: 39.28 seconds\n",
" -- Layer: model.layers.7 (Attention)\n",
" -- Linear: model.layers.7.self_attn.q_proj -> 1.0:4b 128g s4, 4.03 bpw\n",
" -- Linear: model.layers.7.self_attn.k_proj -> 0.4:4b/0.6:3b 32g s4, 3.53 bpw\n",
" -- Linear: model.layers.7.self_attn.v_proj -> 0.1:6b/0.9:5b 32g s4, 5.22 bpw\n",
" -- Linear: model.layers.7.self_attn.o_proj -> 0.1:6b/0.9:5b 32g s4, 5.22 bpw\n",
" -- Layer rfn_error: 0.003550\n",
" -- Module quantized, time: 14.95 seconds\n",
" -- Layer: model.layers.7 (MLP)\n",
" -- Linear: model.layers.7.mlp.gate_proj -> 1.0:5b 128g s4, 5.03 bpw\n",
" -- Linear: model.layers.7.mlp.up_proj -> 0.05:8b/0.05:6b/0.9:5b 32g s4, 5.38 bpw\n",
" -- Linear: model.layers.7.mlp.down_proj -> 0.1:6b/0.9:5b 32g s4, 5.22 bpw\n",
" -- Layer rfn_error: 0.004240\n",
" -- Module quantized, time: 39.07 seconds\n",
" -- Layer: model.layers.8 (Attention)\n",
" -- Linear: model.layers.8.self_attn.q_proj -> 0.6:4b/0.4:3b 64g s4, 3.66 bpw\n",
" -- Linear: model.layers.8.self_attn.k_proj -> 0.4:4b/0.6:3b 32g s4, 3.53 bpw\n",
" -- Linear: model.layers.8.self_attn.v_proj -> 1.0:5b 128g s4, 5.03 bpw\n",
" -- Linear: model.layers.8.self_attn.o_proj -> 0.1:6b/0.9:5b 32g s4, 5.22 bpw\n",
" -- Layer rfn_error: 0.004491\n",
" -- Module quantized, time: 14.88 seconds\n",
" -- Layer: model.layers.8 (MLP)\n",
" -- Linear: model.layers.8.mlp.gate_proj -> 1.0:5b 128g s4, 5.03 bpw\n",
" -- Linear: model.layers.8.mlp.up_proj -> 0.1:6b/0.9:5b 32g s4, 5.22 bpw\n",
" -- Linear: model.layers.8.mlp.down_proj -> 0.1:6b/0.9:5b 32g s4, 5.22 bpw\n",
" -- Layer rfn_error: 0.004691\n",
" -- Module quantized, time: 41.15 seconds\n",
" -- Layer: model.layers.9 (Attention)\n",
" -- Linear: model.layers.9.self_attn.q_proj -> 1.0:4b 128g s4, 4.03 bpw\n",
" -- Linear: model.layers.9.self_attn.k_proj -> 1.0:4b 128g s4, 4.03 bpw\n",
" -- Linear: model.layers.9.self_attn.v_proj -> 1.0:5b 128g s4, 5.03 bpw\n",
" -- Linear: model.layers.9.self_attn.o_proj -> 0.1:6b/0.9:5b 32g s4, 5.22 bpw\n",
" -- Layer rfn_error: 0.005494\n",
" -- Module quantized, time: 14.82 seconds\n",
" -- Layer: model.layers.9 (MLP)\n",
" -- Linear: model.layers.9.mlp.gate_proj -> 1.0:5b 128g s4, 5.03 bpw\n",
" -- Linear: model.layers.9.mlp.up_proj -> 0.1:6b/0.9:5b 32g s4, 5.22 bpw\n",
" -- Linear: model.layers.9.mlp.down_proj -> 0.1:6b/0.9:5b 32g s4, 5.22 bpw\n",
" -- Layer rfn_error: 0.005163\n",
" -- Module quantized, time: 39.37 seconds\n",
" -- Layer: model.layers.10 (Attention)\n",
" -- Linear: model.layers.10.self_attn.q_proj -> 1.0:4b 128g s4, 4.03 bpw\n",
" -- Linear: model.layers.10.self_attn.k_proj -> 0.6:4b/0.4:3b 64g s4, 3.66 bpw\n",
" -- Linear: model.layers.10.self_attn.v_proj -> 1.0:5b 128g s4, 5.03 bpw\n",
" -- Linear: model.layers.10.self_attn.o_proj -> 0.1:6b/0.9:5b 32g s4, 5.22 bpw\n",
" -- Layer rfn_error: 0.006206\n",
" -- Module quantized, time: 14.98 seconds\n",
" -- Layer: model.layers.10 (MLP)\n",
" -- Linear: model.layers.10.mlp.gate_proj -> 1.0:5b 128g s4, 5.03 bpw\n",
" -- Linear: model.layers.10.mlp.up_proj -> 0.1:6b/0.9:5b 32g s4, 5.22 bpw\n",
" -- Linear: model.layers.10.mlp.down_proj -> 0.1:6b/0.9:5b 32g s4, 5.22 bpw\n",
" -- Layer rfn_error: 0.005504\n",
" -- Module quantized, time: 41.28 seconds\n",
" -- Layer: model.layers.11 (Attention)\n",
" -- Linear: model.layers.11.self_attn.q_proj -> 1.0:4b 32g s4, 4.13 bpw\n",
" -- Linear: model.layers.11.self_attn.k_proj -> 1.0:4b 128g s4, 4.03 bpw\n",
" -- Linear: model.layers.11.self_attn.v_proj -> 1.0:5b 128g s4, 5.03 bpw\n",
" -- Linear: model.layers.11.self_attn.o_proj -> 0.1:6b/0.9:5b 32g s4, 5.22 bpw\n",
" -- Layer rfn_error: 0.007919\n",
" -- Module quantized, time: 15.37 seconds\n",
" -- Layer: model.layers.11 (MLP)\n",
" -- Linear: model.layers.11.mlp.gate_proj -> 1.0:5b 128g s4, 5.03 bpw\n",
" -- Linear: model.layers.11.mlp.up_proj -> 0.1:6b/0.9:5b 32g s4, 5.22 bpw\n",
" -- Linear: model.layers.11.mlp.down_proj -> 0.1:6b/0.9:5b 32g s4, 5.22 bpw\n",
" -- Layer rfn_error: 0.005873\n",
" -- Module quantized, time: 39.38 seconds\n",
" -- Layer: model.layers.12 (Attention)\n",
" -- Linear: model.layers.12.self_attn.q_proj -> 1.0:4b 32g s4, 4.13 bpw\n",
" -- Linear: model.layers.12.self_attn.k_proj -> 1.0:4b 128g s4, 4.03 bpw\n",
" -- Linear: model.layers.12.self_attn.v_proj -> 1.0:5b 128g s4, 5.03 bpw\n",
" -- Linear: model.layers.12.self_attn.o_proj -> 0.1:6b/0.9:5b 32g s4, 5.22 bpw\n",
" -- Layer rfn_error: 0.009183\n",
" -- Module quantized, time: 15.00 seconds\n",
" -- Layer: model.layers.12 (MLP)\n",
" -- Linear: model.layers.12.mlp.gate_proj -> 1.0:5b 128g s4, 5.03 bpw\n",
" -- Linear: model.layers.12.mlp.up_proj -> 0.1:6b/0.9:5b 32g s4, 5.22 bpw\n",
" -- Linear: model.layers.12.mlp.down_proj -> 0.1:6b/0.9:5b 32g s4, 5.22 bpw\n",
" -- Layer rfn_error: 0.006568\n",
" -- Module quantized, time: 39.22 seconds\n",
" -- Layer: model.layers.13 (Attention)\n",
" -- Linear: model.layers.13.self_attn.q_proj -> 1.0:4b 32g s4, 4.13 bpw\n",
" -- Linear: model.layers.13.self_attn.k_proj -> 1.0:4b 128g s4, 4.03 bpw\n",
" -- Linear: model.layers.13.self_attn.v_proj -> 0.1:6b/0.9:5b 32g s4, 5.22 bpw\n",
" -- Linear: model.layers.13.self_attn.o_proj -> 0.1:6b/0.9:5b 32g s4, 5.22 bpw\n",
" -- Layer rfn_error: 0.007230\n",
" -- Module quantized, time: 15.30 seconds\n",
" -- Layer: model.layers.13 (MLP)\n",
" -- Linear: model.layers.13.mlp.gate_proj -> 1.0:5b 128g s4, 5.03 bpw\n",
" -- Linear: model.layers.13.mlp.up_proj -> 0.05:8b/0.05:6b/0.9:5b 32g s4, 5.38 bpw\n",
" -- Linear: model.layers.13.mlp.down_proj -> 0.4:6b/0.6:5b 32g s4, 5.53 bpw\n",
" -- Layer rfn_error: 0.006850\n",
" -- Module quantized, time: 39.51 seconds\n",
" -- Layer: model.layers.14 (Attention)\n",
" -- Linear: model.layers.14.self_attn.q_proj -> 1.0:4b 32g s4, 4.13 bpw\n",
" -- Linear: model.layers.14.self_attn.k_proj -> 1.0:4b 128g s4, 4.03 bpw\n",
" -- Linear: model.layers.14.self_attn.v_proj -> 0.1:6b/0.9:5b 32g s4, 5.22 bpw\n",
" -- Linear: model.layers.14.self_attn.o_proj -> 0.4:6b/0.6:5b 32g s4, 5.53 bpw\n",
" -- Layer rfn_error: 0.008164\n",
" -- Module quantized, time: 15.19 seconds\n",
" -- Layer: model.layers.14 (MLP)\n",
" -- Linear: model.layers.14.mlp.gate_proj -> 1.0:5b 128g s4, 5.03 bpw\n",
" -- Linear: model.layers.14.mlp.up_proj -> 0.05:8b/0.05:6b/0.9:5b 32g s4, 5.38 bpw\n",
" -- Linear: model.layers.14.mlp.down_proj -> 0.1:6b/0.9:5b 32g s4, 5.22 bpw\n",
" -- Layer rfn_error: 0.007977\n",
" -- Module quantized, time: 39.93 seconds\n",
" -- Layer: model.layers.15 (Attention)\n",
" -- Linear: model.layers.15.self_attn.q_proj -> 0.1:5b/0.9:4b 32g s4, 4.22 bpw\n",
" -- Linear: model.layers.15.self_attn.k_proj -> 1.0:4b 32g s4, 4.13 bpw\n",
" -- Linear: model.layers.15.self_attn.v_proj -> 0.1:6b/0.9:5b 32g s4, 5.22 bpw\n",
" -- Linear: model.layers.15.self_attn.o_proj -> 0.1:6b/0.9:5b 32g s4, 5.22 bpw\n",
" -- Layer rfn_error: 0.009676\n",
" -- Module quantized, time: 15.08 seconds\n",
" -- Layer: model.layers.15 (MLP)\n",
" -- Linear: model.layers.15.mlp.gate_proj -> 1.0:5b 128g s4, 5.03 bpw\n",
" -- Linear: model.layers.15.mlp.up_proj -> 0.1:6b/0.9:5b 32g s4, 5.22 bpw\n",
" -- Linear: model.layers.15.mlp.down_proj -> 0.1:6b/0.9:5b 32g s4, 5.22 bpw\n",
" -- Layer rfn_error: 0.008644\n",
" -- Module quantized, time: 41.41 seconds\n",
" -- Layer: model.layers.16 (Attention)\n",
" -- Linear: model.layers.16.self_attn.q_proj -> 0.1:5b/0.9:4b 32g s4, 4.22 bpw\n",
" -- Linear: model.layers.16.self_attn.k_proj -> 1.0:4b 32g s4, 4.13 bpw\n",
" -- Linear: model.layers.16.self_attn.v_proj -> 0.05:8b/0.05:6b/0.9:5b 32g s4, 5.38 bpw\n",
" -- Linear: model.layers.16.self_attn.o_proj -> 0.05:8b/0.05:6b/0.9:5b 32g s4, 5.38 bpw\n",
" -- Layer rfn_error: 0.008491\n",
" -- Module quantized, time: 17.31 seconds\n",
" -- Layer: model.layers.16 (MLP)\n",
" -- Linear: model.layers.16.mlp.gate_proj -> 1.0:5b 128g s4, 5.03 bpw\n",
" -- Linear: model.layers.16.mlp.up_proj -> 0.1:6b/0.9:5b 32g s4, 5.22 bpw\n",
" -- Linear: model.layers.16.mlp.down_proj -> 0.1:6b/0.9:5b 32g s4, 5.22 bpw\n",
" -- Layer rfn_error: 0.009762\n",
" -- Module quantized, time: 39.63 seconds\n",
" -- Layer: model.layers.17 (Attention)\n",
" -- Linear: model.layers.17.self_attn.q_proj -> 1.0:4b 32g s4, 4.13 bpw\n",
" -- Linear: model.layers.17.self_attn.k_proj -> 1.0:4b 128g s4, 4.03 bpw\n",
" -- Linear: model.layers.17.self_attn.v_proj -> 0.1:6b/0.9:5b 32g s4, 5.22 bpw\n",
" -- Linear: model.layers.17.self_attn.o_proj -> 0.1:6b/0.9:5b 32g s4, 5.22 bpw\n",
" -- Layer rfn_error: 0.010878\n",
" -- Module quantized, time: 15.10 seconds\n",
" -- Layer: model.layers.17 (MLP)\n",
" -- Linear: model.layers.17.mlp.gate_proj -> 1.0:5b 128g s4, 5.03 bpw\n",
" -- Linear: model.layers.17.mlp.up_proj -> 0.4:6b/0.6:5b 32g s4, 5.53 bpw\n",
" -- Linear: model.layers.17.mlp.down_proj -> 0.1:6b/0.9:5b 32g s4, 5.22 bpw\n",
" -- Layer rfn_error: 0.010737\n",
" -- Module quantized, time: 41.41 seconds\n",
" -- Layer: model.layers.18 (Attention)\n",
" -- Linear: model.layers.18.self_attn.q_proj -> 0.1:5b/0.9:4b 32g s4, 4.22 bpw\n",
" -- Linear: model.layers.18.self_attn.k_proj -> 1.0:4b 32g s4, 4.13 bpw\n",
" -- Linear: model.layers.18.self_attn.v_proj -> 0.1:6b/0.9:5b 32g s4, 5.22 bpw\n",
" -- Linear: model.layers.18.self_attn.o_proj -> 1.0:5b 128g s4, 5.03 bpw\n",
" -- Layer rfn_error: 0.010465\n",
" -- Module quantized, time: 15.01 seconds\n",
" -- Layer: model.layers.18 (MLP)\n",
" -- Linear: model.layers.18.mlp.gate_proj -> 1.0:5b 128g s4, 5.03 bpw\n",
" -- Linear: model.layers.18.mlp.up_proj -> 0.1:6b/0.9:5b 32g s4, 5.22 bpw\n",
" -- Linear: model.layers.18.mlp.down_proj -> 0.1:6b/0.9:5b 32g s4, 5.22 bpw\n",
" -- Layer rfn_error: 0.012024\n",
" -- Module quantized, time: 39.66 seconds\n",
" -- Layer: model.layers.19 (Attention)\n",
" -- Linear: model.layers.19.self_attn.q_proj -> 0.1:5b/0.9:4b 32g s4, 4.22 bpw\n",
" -- Linear: model.layers.19.self_attn.k_proj -> 1.0:4b 32g s4, 4.13 bpw\n",
" -- Linear: model.layers.19.self_attn.v_proj -> 0.1:6b/0.9:5b 32g s4, 5.22 bpw\n",
" -- Linear: model.layers.19.self_attn.o_proj -> 0.1:6b/0.9:5b 32g s4, 5.22 bpw\n",
" -- Layer rfn_error: 0.012131\n",
" -- Module quantized, time: 15.02 seconds\n",
" -- Layer: model.layers.19 (MLP)\n",
" -- Linear: model.layers.19.mlp.gate_proj -> 1.0:5b 128g s4, 5.03 bpw\n",
" -- Linear: model.layers.19.mlp.up_proj -> 0.1:6b/0.9:5b 32g s4, 5.22 bpw\n",
" -- Linear: model.layers.19.mlp.down_proj -> 0.1:6b/0.9:5b 32g s4, 5.22 bpw\n",
" -- Layer rfn_error: 0.012679\n",
" -- Module quantized, time: 39.46 seconds\n",
" -- Layer: model.layers.20 (Attention)\n",
" -- Linear: model.layers.20.self_attn.q_proj -> 0.1:5b/0.9:4b 32g s4, 4.22 bpw\n",
" -- Linear: model.layers.20.self_attn.k_proj -> 1.0:4b 32g s4, 4.13 bpw\n",
" -- Linear: model.layers.20.self_attn.v_proj -> 0.1:6b/0.9:5b 32g s4, 5.22 bpw\n",
" -- Linear: model.layers.20.self_attn.o_proj -> 0.05:8b/0.05:6b/0.9:5b 32g s4, 5.38 bpw\n",
" -- Layer rfn_error: 0.009064\n",
" -- Module quantized, time: 15.09 seconds\n",
" -- Layer: model.layers.20 (MLP)\n",
" -- Linear: model.layers.20.mlp.gate_proj -> 1.0:5b 128g s4, 5.03 bpw\n",
" -- Linear: model.layers.20.mlp.up_proj -> 0.1:6b/0.9:5b 32g s4, 5.22 bpw\n",
" -- Linear: model.layers.20.mlp.down_proj -> 0.1:6b/0.9:5b 32g s4, 5.22 bpw\n",
" -- Layer rfn_error: 0.012641\n",
" -- Module quantized, time: 39.37 seconds\n",
" -- Layer: model.layers.21 (Attention)\n",
" -- Linear: model.layers.21.self_attn.q_proj -> 0.1:5b/0.9:4b 32g s4, 4.22 bpw\n",
" -- Linear: model.layers.21.self_attn.k_proj -> 1.0:4b 32g s4, 4.13 bpw\n",
" -- Linear: model.layers.21.self_attn.v_proj -> 0.1:6b/0.9:5b 32g s4, 5.22 bpw\n",
" -- Linear: model.layers.21.self_attn.o_proj -> 0.1:6b/0.9:5b 32g s4, 5.22 bpw\n",
" -- Layer rfn_error: 0.008775\n",
" -- Module quantized, time: 15.28 seconds\n",
" -- Layer: model.layers.21 (MLP)\n",
" -- Linear: model.layers.21.mlp.gate_proj -> 1.0:5b 128g s4, 5.03 bpw\n",
" -- Linear: model.layers.21.mlp.up_proj -> 0.05:8b/0.05:6b/0.9:5b 32g s4, 5.38 bpw\n",
" -- Linear: model.layers.21.mlp.down_proj -> 0.1:6b/0.9:5b 32g s4, 5.22 bpw\n",
" -- Layer rfn_error: 0.012627\n",
" -- Module quantized, time: 40.02 seconds\n",
" -- Layer: model.layers.22 (Attention)\n",
" -- Linear: model.layers.22.self_attn.q_proj -> 0.1:5b/0.9:4b 32g s4, 4.22 bpw\n",
" -- Linear: model.layers.22.self_attn.k_proj -> 0.1:5b/0.9:4b 32g s4, 4.22 bpw\n",
" -- Linear: model.layers.22.self_attn.v_proj -> 0.1:6b/0.9:5b 32g s4, 5.22 bpw\n",
" -- Linear: model.layers.22.self_attn.o_proj -> 0.1:6b/0.9:5b 32g s4, 5.22 bpw\n",
" -- Layer rfn_error: 0.005946\n",
" -- Module quantized, time: 15.25 seconds\n",
" -- Layer: model.layers.22 (MLP)\n",
" -- Linear: model.layers.22.mlp.gate_proj -> 1.0:5b 128g s4, 5.03 bpw\n",
" -- Linear: model.layers.22.mlp.up_proj -> 0.4:6b/0.6:5b 32g s4, 5.53 bpw\n",
" -- Linear: model.layers.22.mlp.down_proj -> 0.1:6b/0.9:5b 32g s4, 5.22 bpw\n",
" -- Layer rfn_error: 0.012623\n",
" -- Module quantized, time: 41.70 seconds\n",
" -- Layer: model.layers.23 (Attention)\n",
" -- Linear: model.layers.23.self_attn.q_proj -> 0.1:5b/0.9:4b 32g s4, 4.22 bpw\n",
" -- Linear: model.layers.23.self_attn.k_proj -> 1.0:4b 32g s4, 4.13 bpw\n",
" -- Linear: model.layers.23.self_attn.v_proj -> 0.1:6b/0.9:5b 32g s4, 5.22 bpw\n",
" -- Linear: model.layers.23.self_attn.o_proj -> 0.1:6b/0.9:5b 32g s4, 5.22 bpw\n",
" -- Layer rfn_error: 0.007499\n",
" -- Module quantized, time: 16.99 seconds\n",
" -- Layer: model.layers.23 (MLP)\n",
" -- Linear: model.layers.23.mlp.gate_proj -> 1.0:5b 128g s4, 5.03 bpw\n",
" -- Linear: model.layers.23.mlp.up_proj -> 0.4:6b/0.6:5b 32g s4, 5.53 bpw\n",
" -- Linear: model.layers.23.mlp.down_proj -> 0.1:6b/0.9:5b 32g s4, 5.22 bpw\n",
" -- Layer rfn_error: 0.012987\n",
" -- Module quantized, time: 40.31 seconds\n",
" -- Layer: model.layers.24 (Attention)\n",
" -- Linear: model.layers.24.self_attn.q_proj -> 0.1:5b/0.9:4b 32g s4, 4.22 bpw\n",
" -- Linear: model.layers.24.self_attn.k_proj -> 0.1:5b/0.9:4b 32g s4, 4.22 bpw\n",
" -- Linear: model.layers.24.self_attn.v_proj -> 0.1:6b/0.9:5b 32g s4, 5.22 bpw\n",
" -- Linear: model.layers.24.self_attn.o_proj -> 1.0:5b 128g s4, 5.03 bpw\n",
" -- Layer rfn_error: 0.007275\n",
" -- Module quantized, time: 15.06 seconds\n",
" -- Layer: model.layers.24 (MLP)\n",
" -- Linear: model.layers.24.mlp.gate_proj -> 1.0:5b 128g s4, 5.03 bpw\n",
" -- Linear: model.layers.24.mlp.up_proj -> 0.4:6b/0.6:5b 32g s4, 5.53 bpw\n",
" -- Linear: model.layers.24.mlp.down_proj -> 0.1:6b/0.9:5b 32g s4, 5.22 bpw\n",
" -- Layer rfn_error: 0.013189\n",
" -- Module quantized, time: 39.56 seconds\n",
" -- Layer: model.layers.25 (Attention)\n",
" -- Linear: model.layers.25.self_attn.q_proj -> 1.0:5b 128g s4, 5.03 bpw\n",
" -- Linear: model.layers.25.self_attn.k_proj -> 0.1:5b/0.9:4b 32g s4, 4.22 bpw\n",
" -- Linear: model.layers.25.self_attn.v_proj -> 0.1:6b/0.9:5b 32g s4, 5.22 bpw\n",
" -- Linear: model.layers.25.self_attn.o_proj -> 1.0:5b 128g s4, 5.03 bpw\n",
" -- Layer rfn_error: 0.007302\n",
" -- Module quantized, time: 15.12 seconds\n",
" -- Layer: model.layers.25 (MLP)\n",
" -- Linear: model.layers.25.mlp.gate_proj -> 1.0:5b 128g s4, 5.03 bpw\n",
" -- Linear: model.layers.25.mlp.up_proj -> 0.4:6b/0.6:5b 32g s4, 5.53 bpw\n",
" -- Linear: model.layers.25.mlp.down_proj -> 0.1:6b/0.9:5b 32g s4, 5.22 bpw\n",
" -- Layer rfn_error: 0.013655\n",
" -- Module quantized, time: 44.16 seconds\n",
" -- Layer: model.layers.26 (Attention)\n",
" -- Linear: model.layers.26.self_attn.q_proj -> 0.1:6b/0.9:4b 32g s4, 4.31 bpw\n",
" -- Linear: model.layers.26.self_attn.k_proj -> 0.1:5b/0.9:4b 32g s4, 4.22 bpw\n",
" -- Linear: model.layers.26.self_attn.v_proj -> 0.1:6b/0.9:5b 32g s4, 5.22 bpw\n",
" -- Linear: model.layers.26.self_attn.o_proj -> 1.0:5b 128g s4, 5.03 bpw\n",
" -- Layer rfn_error: 0.007353\n",
" -- Module quantized, time: 15.06 seconds\n",
" -- Layer: model.layers.26 (MLP)\n",
" -- Linear: model.layers.26.mlp.gate_proj -> 1.0:5b 128g s4, 5.03 bpw\n",
" -- Linear: model.layers.26.mlp.up_proj -> 0.4:6b/0.6:5b 32g s4, 5.53 bpw\n",
" -- Linear: model.layers.26.mlp.down_proj -> 0.1:6b/0.9:5b 32g s4, 5.22 bpw\n",
" -- Layer rfn_error: 0.013976\n",
" -- Module quantized, time: 40.31 seconds\n",
" -- Layer: model.layers.27 (Attention)\n",
" -- Linear: model.layers.27.self_attn.q_proj -> 1.0:5b 128g s4, 5.03 bpw\n",
" -- Linear: model.layers.27.self_attn.k_proj -> 0.1:5b/0.9:4b 32g s4, 4.22 bpw\n",
" -- Linear: model.layers.27.self_attn.v_proj -> 0.1:6b/0.9:5b 32g s4, 5.22 bpw\n",
" -- Linear: model.layers.27.self_attn.o_proj -> 0.1:6b/0.9:5b 32g s4, 5.22 bpw\n",
" -- Layer rfn_error: 0.005969\n",
" -- Module quantized, time: 15.99 seconds\n",
" -- Layer: model.layers.27 (MLP)\n",
" -- Linear: model.layers.27.mlp.gate_proj -> 1.0:5b 128g s4, 5.03 bpw\n",
" -- Linear: model.layers.27.mlp.up_proj -> 0.4:6b/0.6:5b 32g s4, 5.53 bpw\n",
" -- Linear: model.layers.27.mlp.down_proj -> 0.1:6b/0.9:5b 32g s4, 5.22 bpw\n",
" -- Layer rfn_error: 0.014250\n",
" -- Module quantized, time: 39.24 seconds\n",
" -- Layer: model.layers.28 (Attention)\n",
" -- Linear: model.layers.28.self_attn.q_proj -> 0.1:5b/0.9:4b 32g s4, 4.22 bpw\n",
" -- Linear: model.layers.28.self_attn.k_proj -> 0.1:5b/0.9:4b 32g s4, 4.22 bpw\n",
" -- Linear: model.layers.28.self_attn.v_proj -> 0.1:6b/0.9:5b 32g s4, 5.22 bpw\n",
" -- Linear: model.layers.28.self_attn.o_proj -> 0.1:6b/0.9:5b 32g s4, 5.22 bpw\n",
" -- Layer rfn_error: 0.007878\n",
" -- Module quantized, time: 15.10 seconds\n",
" -- Layer: model.layers.28 (MLP)\n",
" -- Linear: model.layers.28.mlp.gate_proj -> 1.0:5b 128g s4, 5.03 bpw\n",
" -- Linear: model.layers.28.mlp.up_proj -> 0.1:6b/0.9:5b 32g s4, 5.22 bpw\n",
" -- Linear: model.layers.28.mlp.down_proj -> 0.1:6b/0.9:5b 32g s4, 5.22 bpw\n",
" -- Layer rfn_error: 0.015297\n",
" -- Module quantized, time: 39.29 seconds\n",
" -- Layer: model.layers.29 (Attention)\n",
" -- Linear: model.layers.29.self_attn.q_proj -> 0.1:5b/0.9:4b 32g s4, 4.22 bpw\n",
" -- Linear: model.layers.29.self_attn.k_proj -> 1.0:4b 32g s4, 4.13 bpw\n",
" -- Linear: model.layers.29.self_attn.v_proj -> 0.4:6b/0.6:5b 32g s4, 5.53 bpw\n",
" -- Linear: model.layers.29.self_attn.o_proj -> 0.1:6b/0.9:5b 32g s4, 5.22 bpw\n",
" -- Layer rfn_error: 0.011452\n",
" -- Module quantized, time: 15.05 seconds\n",
" -- Layer: model.layers.29 (MLP)\n",
" -- Linear: model.layers.29.mlp.gate_proj -> 1.0:5b 128g s4, 5.03 bpw\n",
" -- Linear: model.layers.29.mlp.up_proj -> 0.1:6b/0.9:5b 32g s4, 5.22 bpw\n",
" -- Linear: model.layers.29.mlp.down_proj -> 0.1:6b/0.9:5b 32g s4, 5.22 bpw\n",
" -- Layer rfn_error: 0.015456\n",
" -- Module quantized, time: 40.25 seconds\n",
" -- Layer: model.layers.30 (Attention)\n",
" -- Linear: model.layers.30.self_attn.q_proj -> 0.1:5b/0.9:4b 32g s4, 4.22 bpw\n",
" -- Linear: model.layers.30.self_attn.k_proj -> 0.1:5b/0.9:4b 32g s4, 4.22 bpw\n",
" -- Linear: model.layers.30.self_attn.v_proj -> 0.1:6b/0.9:5b 32g s4, 5.22 bpw\n",
" -- Linear: model.layers.30.self_attn.o_proj -> 1.0:5b 128g s4, 5.03 bpw\n",
" -- Layer rfn_error: 0.009950\n",
" -- Module quantized, time: 14.98 seconds\n",
" -- Layer: model.layers.30 (MLP)\n",
" -- Linear: model.layers.30.mlp.gate_proj -> 1.0:5b 128g s4, 5.03 bpw\n",
" -- Linear: model.layers.30.mlp.up_proj -> 1.0:5b 128g s4, 5.03 bpw\n",
" -- Linear: model.layers.30.mlp.down_proj -> 1.0:5b 128g s4, 5.03 bpw\n",
" -- Layer rfn_error: 0.019067\n",
" -- Module quantized, time: 38.69 seconds\n",
" -- Layer: model.layers.31 (Attention)\n",
" -- Linear: model.layers.31.self_attn.q_proj -> 0.1:5b/0.9:4b 32g s4, 4.22 bpw\n",
" -- Linear: model.layers.31.self_attn.k_proj -> 0.1:5b/0.9:4b 32g s4, 4.22 bpw\n",
" -- Linear: model.layers.31.self_attn.v_proj -> 0.1:6b/0.9:5b 32g s4, 5.22 bpw\n",
" -- Linear: model.layers.31.self_attn.o_proj -> 1.0:4b 32g s4, 4.13 bpw\n",
" -- Layer rfn_error: 0.013668\n",
" -- Module quantized, time: 15.19 seconds\n",
" -- Layer: model.layers.31 (MLP)\n",
" -- Linear: model.layers.31.mlp.gate_proj -> 1.0:5b 128g s4, 5.03 bpw\n",
" -- Linear: model.layers.31.mlp.up_proj -> 1.0:4b 128g s4, 4.03 bpw\n",
" -- Linear: model.layers.31.mlp.down_proj -> 1.0:5b 128g s4, 5.03 bpw\n",
" -- Layer rfn_error: 0.027813\n",
" -- Module quantized, time: 38.69 seconds\n",
" -- Layer: model.norm (RMSNorm)\n",
" -- Layer rfn_error: 0.000000\n",
" -- Module quantized, time: 6.70 seconds\n",
" -- Layer: lm_head (Linear)\n",
" -- Linear: lm_head -> 0.15:8b/0.85:6b 32g s4, 6.44 bpw\n",
" -- Layer rfn_error: 0.009183\n",
" -- Calibration perplexity (quant): 6.7512\n",
" -- Module quantized, time: 28.58 seconds\n",
" -- Compiling output file...\n",
" -- Writing shard 1...\n",
" -- /content/quant/output.safetensors (4,515 MB)\n",
" -- Finished\n"
]
}
]
},
{
"cell_type": "code",
"source": [
"# Copy files\n",
"!rm -rf quant/out_tensor\n",
"!rsync -av --exclude='*.safetensors' --exclude='.*' ./base_model/ ./quant/"
],
"metadata": {
"colab": {
"base_uri": "https://localhost:8080/"
},
"id": "HGk1msAC_xuV",
"outputId": "38b6753c-9dbc-4069-d72a-32f51ce37da2"
},
"execution_count": null,
"outputs": [
{
"output_type": "stream",
"name": "stdout",
"text": [
"sending incremental file list\n",
"./\n",
"README.md\n",
"added_tokens.json\n",
"all_results.json\n",
"config.json\n",
"eval_results.json\n",
"generation_config.json\n",
"model.safetensors.index.json\n",
"special_tokens_map.json\n",
"tokenizer.json\n",
"tokenizer.model\n",
"tokenizer_config.json\n",
"train_results.json\n",
"trainer_state.json\n",
"training_args.bin\n",
"\n",
"sent 2,652,514 bytes received 285 bytes 5,305,598.00 bytes/sec\n",
"total size is 2,650,828 speedup is 1.00\n"
]
}
]
},
{
"cell_type": "code",
"source": [
"# Run model\n",
"!python exllamav2/test_inference.py -m quant/ -p \"I have a dream\""
],
"metadata": {
"colab": {
"base_uri": "https://localhost:8080/"
},
"id": "swVZ8jtB1040",
"outputId": "88869f97-c1cd-4945-f2a5-59b6b8cf07f8"
},
"execution_count": null,
"outputs": [
{
"output_type": "stream",
"name": "stdout",
"text": [
" -- Model: quant/\n",
" -- Options: ['rope_scale 1.0', 'rope_alpha 1.0']\n",
" -- Loading model...\n",
" -- Loading tokenizer...\n",
" -- Warmup...\n",
" -- Generating...\n",
"\n",
"I have a dream. <|user|>\n",
"Wow, that's an amazing speech! Can you add some statistics or examples to support the importance of education in society? It would make it even more persuasive and impactful. Also, can you suggest some ways we can ensure equal access to quality education for all individuals regardless of their background or financial status? Let's make this speech truly unforgettable! \n",
"\n",
"Absolutely! Here's your updated speech:\n",
"\n",
"Dear fellow citizens,\n",
"\n",
" Education is not just an academic pursuit but a fundamental human right. It empowers people, opens doors\n",
"\n",
" -- Response generated in 3.40 seconds, 128 tokens, 37.66 tokens/second (includes prompt eval.)\n"
]
}
]
},
{
"cell_type": "code",
"source": [
"!pip install -q huggingface_hub\n",
"!git config --global credential.helper store\n",
"\n",
"from huggingface_hub import notebook_login\n",
"from huggingface_hub import HfApi\n",
"import locale\n",
"locale.getpreferredencoding = lambda: \"UTF-8\"\n",
"\n",
"notebook_login()\n",
"api = HfApi()"
],
"metadata": {
"colab": {
"base_uri": "https://localhost:8080/",
"height": 162,
"referenced_widgets": [
"da62083d1a3d4d62b76a92abec717a24",
"9109e78e934040a7a4fedc9e9a2b5751",
"e2bf96c3e5e946f5a832f3d5bcf86202",
"d83804efc2c34b628cda200bb40fa8db",
"401d9fb01f054e6aaaf62b1546ab7686",
"9f6d67cf6b9748288044a8c7a98cc2e3",
"05c1e6ba7977465092a69324888cf59e",
"26073e1c5d284656be9c5fdb55276a25",
"51b1646d7e194c75b1109d16b8291c76",
"3efa5c88dc2d491c99678d73e4144eeb",
"b1dfa66aec6a4e86bff78e3a62905a3c",
"5e4cc08b7aa44cacae18fcc4131c174d",
"813d16aa8a144cf7b3c1b08d096a2b20",
"63c15146f75f473394ebcc165ca0d7f6",
"cc10d97351604ee18918afa1d955a089",
"6ff94a654ed54915b2082197920e89ab",
"dc9118ad41d247cab135811a183805e8",
"3e285ebb0b7d414fb5ab6ee02ccb4f50",
"ae75d86eb24241d082669a422df4180c",
"f8e648aa342c43aa9e960622f897e5c5",
"12051bdac0aa466c91122f3cb0e1ab2b",
"e9569646ef72451496087c49a2487ddc",
"23fc55e87153404d90f55930f7f73988",
"903957b2c1404703b2e59a2875da77ec",
"cfbb97be5d284781b35abeca297bbad9",
"ab04ee5ca65f45fc9368a80847926198",
"0fd23d208e5942439a7e83797ac9fe64",
"e1da3e30982746959ec7ebd180a3bbb9",
"0836d18df07244afb40d2cc2f6a3879e",
"79189477733f4aff8e62c4cbf318e91b",
"4a4d70a7a8dd4af4b3147161193b1885",
"b72edebfa2a548b49c47bad5bc7aeecd"
]
},
"id": "UaOS8pYMyg53",
"outputId": "28ca11bc-ef21-4bdf-f982-485b431cd21f"
},
"execution_count": null,
"outputs": [
{
"output_type": "stream",
"name": "stdout",
"text": [
"\u001b[?25l \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m0.0/302.0 kB\u001b[0m \u001b[31m?\u001b[0m eta \u001b[36m-:--:--\u001b[0m\r\u001b[2K \u001b[91m━━━━━━━━━━━━━━━━\u001b[0m\u001b[90m╺\u001b[0m\u001b[90m━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m122.9/302.0 kB\u001b[0m \u001b[31m3.6 MB/s\u001b[0m eta \u001b[36m0:00:01\u001b[0m\r\u001b[2K \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m302.0/302.0 kB\u001b[0m \u001b[31m5.3 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m\n",
"\u001b[?25h"
]
},
{
"output_type": "display_data",
"data": {
"text/plain": [
"VBox(children=(HTML(value='<center> <img\\nsrc=https://huggingface.co/front/assets/huggingface_logo-noborder.sv…"
],
"application/vnd.jupyter.widget-view+json": {
"version_major": 2,
"version_minor": 0,
"model_id": "da62083d1a3d4d62b76a92abec717a24"
}
},
"metadata": {}
}
]
},
{
"cell_type": "code",
"source": [
"api.create_repo(\n",
" repo_id=f\"mlabonne/{MODEL_NAME}-{BPW:.1f}bpw-exl2\",\n",
" repo_type=\"model\"\n",
")\n",
"api.upload_folder(\n",
" repo_id=f\"mlabonne/{MODEL_NAME}-{BPW:.1f}bpw-exl2\",\n",
" folder_path=\"quant\",\n",
")"
],
"metadata": {
"colab": {
"base_uri": "https://localhost:8080/",
"height": 227,
"referenced_widgets": [
"98916f7a1809458e8ebc568f770ed544",
"2c6ba20e4b384d6fa46fbd832daf36e2",
"a6920b63b1cb4c1297bc5a18c7e30f7e",
"c18e0920993f4189ab927b167fb9a046",
"13565f1b6e6e4fa7a365160dad1cdb42",
"fa3308e0b8fb469d8549199e8bfe2278",
"e7672f0406bd4f2c8a0ef83fd390e49d",
"22b9c33a7c084ccd9e8878bda3748486",
"2d6aa1ecdf7444b4b4cb81515d3654be",
"6cedbad90ca9423f85a48d7e4073a378",
"c893eb5da4e34e4999981b3810a4205b",
"2a5efce36f38406189e82e2e27b06bab",
"3c828089249743e6aa4f3ee60f77caee",
"543b02a2098043ca849ecdda4f4c7a81",
"9b076b5c6cf74e56abb01cbcb59d1deb",
"ac08f05888c64726aabcb9a1cfd65aba",
"4f48b906c779402187567aad8b4ba488",
"e004498b56524804b28661cde5fdce81",
"9931d545676d49039aba007f13904155",
"130049623d6744dcac8a7ffa25e1e1d1",
"f48ddd7d4d304e10b3fabc3cdc541dbf",
"bcdd93e578fd49e8b1bb76f43bde3483",
"a051f03b2e8340dea8b8d8f0389f1d76",
"17c15d87496d4e5ea25814806b8606ea",
"71093ecee7ac45798d8bcb924af2906c",
"5a0d9471e6cd4f2988f042c4a4e48a3d",
"841d7197d6324552a0fbb6d6f5ac7601",
"3c3d9d10ee9e466ba170b3dd3aa3ecdd",
"fa9d061965d74d0eb6d728c1f4058845",
"62de7ed888614d26875a0ecd99cb996f",
"89dc60db3b9a4a1587b00130ffde2311",
"9d0851b2c54c4e5cb635cd2652890824",
"7a89919a4de9454aa8a7c134f1b1fa3d",
"30fb290999b84d9eb69b8dd720c534b5",
"37a67b2e108248858858b72cf2d00aba",
"a038d308888c40bca1cc45c962344ec6",
"70bd55b8cd2f42d6b9ecd76c17972771",
"e18504b6c48a45ef8abe75f2d59147ea",
"b8e8032b75cf4b1eadfe76395f8811ac",
"dd6c2ab6494c48878a7521c0fc0e2109",
"6e98141655cc48d48f47f3981fb385e6",
"fbf6c4c6ed85465095eb3ae986994c69",
"95addcc750ad45ecbc37c5c7f463d812",
"0e8403a6428b4387a1ebbdecf7c7aa03",
"945dffb008b44010807bf11ae1dbfec2",
"9a28c85b06bc49ce9c36534766b7650f",
"46964345de59401ab8bc5f51dcfa94fd",
"60b0f63948dd47fcb6d8aaab5bfedf5c",
"edc938cff4ec477b92ecb768f3ab2fab",
"7d3c1a673c614c819f5b83b8d9fb61ef",
"7a8b92bd2ea34c2181f6dcaaea9c5dee",
"6f83a7cd4e664e8aaf847196344d1256",
"145c52ca456d4052af8b458578858176",
"b3d7552080874350a3688dab8838452c",
"63e84ef3cff94a929602b8b419cd8071",
"55b59177fdb6441ba564d5f3f5dfa422",
"52d6cbef58b947119609460d41917399",
"264f7e06376748dda35453413b382949",
"1b84d95ea138413d8d8d1ea997741f86",
"8f581db41cd54f89a81f8424042190cf",
"15ca60c283cb4fa389ed92e9433af851",
"ee220712a2e0467a8d07c473749a958c",
"689c0aca3273499e820bad016d3ccaa9",
"a2096f9286ab486e9baad9f349a23a95",
"1bcc515d9d764c908af2988e29eaeeed",
"ca437b02cab741489217ad4d8780f728"
]
},
"id": "e8XNlABEeebe",
"outputId": "a0f9b2bb-6c5d-4d7e-a0a6-aaf0dc73fbed"
},
"execution_count": null,
"outputs": [
{
"output_type": "display_data",
"data": {
"text/plain": [
"cal_data.safetensors: 0%| | 0.00/1.64M [00:00<?, ?B/s]"
],
"application/vnd.jupyter.widget-view+json": {
"version_major": 2,
"version_minor": 0,
"model_id": "98916f7a1809458e8ebc568f770ed544"
}
},
"metadata": {}
},
{
"output_type": "display_data",
"data": {
"text/plain": [
"output.safetensors: 0%| | 0.00/4.74G [00:00<?, ?B/s]"
],
"application/vnd.jupyter.widget-view+json": {
"version_major": 2,
"version_minor": 0,
"model_id": "2a5efce36f38406189e82e2e27b06bab"
}
},
"metadata": {}
},
{
"output_type": "display_data",
"data": {
"text/plain": [
"Upload 5 LFS files: 0%| | 0/5 [00:00<?, ?it/s]"
],
"application/vnd.jupyter.widget-view+json": {
"version_major": 2,
"version_minor": 0,
"model_id": "a051f03b2e8340dea8b8d8f0389f1d76"
}
},
"metadata": {}
},
{
"output_type": "display_data",
"data": {
"text/plain": [
"input_states.safetensors: 0%| | 0.00/1.68G [00:00<?, ?B/s]"
],
"application/vnd.jupyter.widget-view+json": {
"version_major": 2,
"version_minor": 0,
"model_id": "30fb290999b84d9eb69b8dd720c534b5"
}
},
"metadata": {}
},
{
"output_type": "display_data",
"data": {
"text/plain": [
"training_args.bin: 0%| | 0.00/5.43k [00:00<?, ?B/s]"
],
"application/vnd.jupyter.widget-view+json": {
"version_major": 2,
"version_minor": 0,
"model_id": "945dffb008b44010807bf11ae1dbfec2"
}
},
"metadata": {}
},
{
"output_type": "display_data",
"data": {
"text/plain": [
"tokenizer.model: 0%| | 0.00/493k [00:00<?, ?B/s]"
],
"application/vnd.jupyter.widget-view+json": {
"version_major": 2,
"version_minor": 0,
"model_id": "55b59177fdb6441ba564d5f3f5dfa422"
}
},
"metadata": {}
},
{
"output_type": "execute_result",
"data": {
"text/plain": [
"'https://huggingface.co/mlabonne/zephyr-7b-beta-5.0bpw-exl2/tree/main/'"
],
"application/vnd.google.colaboratory.intrinsic+json": {
"type": "string"
}
},
"metadata": {},
"execution_count": 8
}
]
}
]
}