Upload quantize_fp8.ipynb

quantize_fp8.ipynb

@@ -0,0 +1,141 @@
+{
+  "nbformat": 4,
+  "nbformat_minor": 0,
+  "metadata": {
+    "colab": {
+      "provenance": []
+    },
+    "kernelspec": {
+      "name": "python3",
+      "display_name": "Python 3"
+    },
+    "language_info": {
+      "name": "python"
+    }
+  },
+  "cells": [
+    {
+      "cell_type": "code",
+      "execution_count": null,
+      "metadata": {
+        "id": "3eDO2Vp7y94X"
+      },
+      "outputs": [],
+      "source": [
+        "!pip install huggingface_hub hf_transfer"
+      ]
+    },
+    {
+      "cell_type": "code",
+      "source": [
+        "!export HF_HUB_ENABLE_HF_TRANSFER=1"
+      ],
+      "metadata": {
+        "id": "j4X2v5yh0qJC"
+      },
+      "execution_count": null,
+      "outputs": []
+    },
+    {
+      "cell_type": "code",
+      "source": [
+        "import torch\n",
+        "from huggingface_hub import notebook_login, HfApi\n",
+        "from safetensors.torch import load_file, save_file"
+      ],
+      "metadata": {
+        "id": "kPX5fBF5zHYN"
+      },
+      "execution_count": null,
+      "outputs": []
+    },
+    {
+      "cell_type": "code",
+      "source": [
+        "notebook_login() # WRITE token"
+      ],
+      "metadata": {
+        "id": "50ZzcvkI1F0W"
+      },
+      "execution_count": null,
+      "outputs": []
+    },
+    {
+      "cell_type": "code",
+      "source": [
+        "api = HfApi()"
+      ],
+      "metadata": {
+        "id": "zTFJTeRHzJ6a"
+      },
+      "execution_count": null,
+      "outputs": []
+    },
+    {
+      "cell_type": "code",
+      "source": [
+        "def download(repo_id: str, filename: str, save_name: str, dtype = torch.float8_e4m3fn):\n",
+        "  ckpt_path = api.hf_hub_download(\n",
+        "      repo_id,\n",
+        "      filename=filename,\n",
+        "  )\n",
+        "\n",
+        "  state_dict = load_file(ckpt_path)\n",
+        "  for key, value in state_dict.items():\n",
+        "      state_dict[key] = value.to(dtype)\n",
+        "\n",
+        "  save_file(state_dict, f\"./{save_name}\")\n",
+        "  return f\"./{save_name}\""
+      ],
+      "metadata": {
+        "id": "IDgia3bzzJBJ"
+      },
+      "execution_count": null,
+      "outputs": []
+    },
+    {
+      "cell_type": "code",
+      "source": [
+        "def push(repo_id: str, local_file: str):\n",
+        "  api.upload_file(\n",
+        "      path_or_fileobj=local_file,\n",
+        "      path_in_repo=local_file.split(\"/\")[-1],\n",
+        "      repo_id=repo_id,\n",
+        "      repo_type=\"model\",\n",
+        "  )"
+      ],
+      "metadata": {
+        "id": "o1-kDCqY1yZY"
+      },
+      "execution_count": null,
+      "outputs": []
+    },
+    {
+      "cell_type": "code",
+      "source": [
+        "model_path = download(\n",
+        "    \"OnomaAIResearch/Illustrious-xl-early-release-v0\",\n",
+        "    \"Illustrious-XL-v0.1.safetensors\",\n",
+        "    \"Illustrious-XL-v0.1.float8_e4m3fn.safetensors\",\n",
+        "    torch.float8_e4m3fn, # float8_e5m2\n",
+        ")"
+      ],
+      "metadata": {
+        "id": "dOBdDRTpzLcV"
+      },
+      "execution_count": null,
+      "outputs": []
+    },
+    {
+      "cell_type": "code",
+      "source": [
+        "push(YOUR_REPO_NAME, model_path)"
+      ],
+      "metadata": {
+        "id": "jYGONemw2Lj_"
+      },
+      "execution_count": null,
+      "outputs": []
+    }
+  ]
+}