{
  "cells": [
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {},
      "outputs": [],
      "source": [
        "# @title Install dependencies\n",
        "!pip install -q -U autotrain-advanced huggingface_hub huggingface_hub[cli] optimum transformers accelerate peft bitsandbytes datasets"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {},
      "outputs": [],
      "source": [
        "# @title Login to Hugging Face\n",
        "from huggingface_hub import login\n",
        "import os\n",
        "token = os.getenv('HF_TOKEN')\n",
        "if not token:\n",
        "    raise ValueError('Set HF_TOKEN in Colab secrets (left sidebar 🔑)')\n",
        "login(token=token)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {},
      "outputs": [],
      "source": [
        "# @title Upload dataset to HF Hub\n",
        "from huggingface_hub import HfApi\n",
        "import json\n",
        "\n",
        "dataset = [\n",
        "    {\"text\": \"<|begin_of_text|><|start_header_id|>user<|end_header_id|>What is the current price of Bitcoin?<|eot_id|><|start_header_id|>assistant<|end_header_id|><call_tool>get_crypto_price({\\\"symbol\\\": \\\"BTC\\\"})</call_tool><|eot_id|>\"},\n",
        "    {\"text\": \"<|begin_of_text|><|start_header_id|>user<|end_header_id|>Multiply 42 by 58 for me.<|eot_id|><|start_header_id|>assistant<|end_header_id|><call_tool>multiply_numbers({\\\"a\\\": 42, \\\"b\\\": 58})</call_tool><|eot_id|>\"},\n",
        "    {\"text\": \"<|begin_of_text|><|start_header_id|>user<|end_header_id|>What is the capital of France?<|eot_id|><|start_header_id|>assistant<|end_header_id|>The capital of France is Paris.<|eot_id|>\"},\n",
        "    {\"text\": \"<|begin_of_text|><|start_header_id|>user<|end_header_id|>Calculate 144 multiplied by 37.<|eot_id|><|start_header_id|>assistant<|end_header_id|><call_tool>multiply_numbers({\\\"a\\\": 144, \\\"b\\\": 37})</call_tool><|eot_id|>\"},\n",
        "    {\"text\": \"<|begin_of_text|><|start_header_id|>user<|end_header_id|>What's the current Ethereum price?<|eot_id|><|start_header_id|>assistant<|end_header_id|><call_tool>get_crypto_price({\\\"symbol\\\": \\\"ETH\\\"})</call_tool><|eot_id|>\"}\n",
        "]\n",
        "\n",
        "with open('dataset.jsonl', 'w') as f:\n",
        "    for ex in dataset:\n",
        "        f.write(json.dumps(ex) + '\\n')\n",
        "\n",
        "api = HfApi()\n",
        "api.upload_file(\n",
        "    path_or_fileobj='dataset.jsonl',\n",
        "    path_in_repo='dataset.jsonl',\n",
        "    repo_id='yummyfiles/Bob',\n",
        "    repo_type='dataset',\n",
        "    token=os.getenv('HF_TOKEN')\n",
        ")\n",
        "print('Dataset uploaded to yummyfiles/Bob')"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {},
      "outputs": [],
      "source": [
        "# @title Fine-tune with AutoTrain (runs ~15 min on T4)\n",
        "!autotrain llm --train \\\n",
        "  --project-name Bob \\\n",
        "  --model meta-llama/Llama-3.2-1B-Instruct \\\n",
        "  --train-data yummyfiles/Bob \\\n",
        "  --data-path dataset.jsonl \\\n",
        "  --text-column text \\\n",
        "  --trainer peft \\\n",
        "  --peft-r 16 \\\n",
        "  --peft-alpha 32 \\\n",
        "  --peft-dropout 0.05 \\\n",
        "  --learning-rate 2e-4 \\\n",
        "  --batch-size 1 \\\n",
        "  --epochs 3 \\\n",
        "  --block-size 2048 \\\n",
        "  --warmup-ratio 0.03 \\\n",
        "  --optimizer paged_adamw_8bit \\\n",
        "  --scheduler cosine \\\n",
        "  --gradient-accumulation 4 \\\n",
        "  --mixed-precision bf16 \\\n",
        "  --quantization 4bit \\\n",
        "  --push-to-hub \\\n",
        "  --repo-id yummyfiles/Bob \\\n",
        "  --token {os.getenv('HF_TOKEN')}"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {},
      "outputs": [],
      "source": [
        "# @title Export to ONNX (4-bit) and upload\n",
        "!pip install -q optimum[onnxruntime]\n",
        "!optimum-cli export onnx \\\n",
        "  --model yummyfiles/Bob \\\n",
        "  --task text-generation \\\n",
        "  --quantize int4 \\\n",
        "  --output /content/web_model\n",
        "\n",
        "from huggingface_hub import HfApi\n",
        "api = HfApi()\n",
        "api.upload_folder(\n",
        "    folder_path='/content/web_model',\n",
        "    path_in_repo='web_model',\n",
        "    repo_id='yummyfiles/Bob',\n",
        "    repo_type='model',\n",
        "    token=os.getenv('HF_TOKEN')\n",
        ")\n",
        "print('ONNX model uploaded to yummyfiles/Bob/web_model')"
      ]
    }
  ],
  "metadata": {
    "colab": {
      "provenance": []
    },
    "kernelspec": {
      "display_name": "Python 3",
      "name": "python3"
    },
    "language_info": {
      "name": "python"
    }
  },
  "nbformat": 4,
  "nbformat_minor": 4
}