notebook.ipynb
| 1 | { |
| 2 | "nbformat": 4, |
| 3 | "nbformat_minor": 0, |
| 4 | "metadata": { |
| 5 | "colab": { |
| 6 | "provenance": [], |
| 7 | "gpuType": "A100" |
| 8 | }, |
| 9 | "kernelspec": { |
| 10 | "name": "python3", |
| 11 | "display_name": "Python 3" |
| 12 | }, |
| 13 | "language_info": { |
| 14 | "name": "python" |
| 15 | }, |
| 16 | "accelerator": "GPU" |
| 17 | }, |
| 18 | "cells": [ |
| 19 | { |
| 20 | "cell_type": "code", |
| 21 | "execution_count": null, |
| 22 | "metadata": { |
| 23 | "id": "aCl-IzLoDr2H" |
| 24 | }, |
| 25 | "outputs": [], |
| 26 | "source": [ |
| 27 | "!pip install -U transformers mamba-ssm" |
| 28 | ] |
| 29 | }, |
| 30 | { |
| 31 | "cell_type": "markdown", |
| 32 | "source": [ |
| 33 | "# Load Models" |
| 34 | ], |
| 35 | "metadata": { |
| 36 | "id": "SpRo_KJIRsxv" |
| 37 | } |
| 38 | }, |
| 39 | { |
| 40 | "cell_type": "code", |
| 41 | "source": [ |
| 42 | "import torch\n", |
| 43 | "from transformers import AutoTokenizer, AutoModelForCausalLM\n", |
| 44 | "\n", |
| 45 | "# Load tokenizer and model\n", |
| 46 | "tokenizer = AutoTokenizer.from_pretrained(\"nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B-BF16\")\n", |
| 47 | "model = AutoModelForCausalLM.from_pretrained(\n", |
| 48 | " \"nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B-BF16\",\n", |
| 49 | " torch_dtype=torch.bfloat16,\n", |
| 50 | " trust_remote_code=True,\n", |
| 51 | " device_map=\"auto\"\n", |
| 52 | ")\n" |
| 53 | ], |
| 54 | "metadata": { |
| 55 | "id": "waveliieEI1n" |
| 56 | }, |
| 57 | "execution_count": null, |
| 58 | "outputs": [] |
| 59 | }, |
| 60 | { |
| 61 | "cell_type": "markdown", |
| 62 | "source": [ |
| 63 | "# Define Input with Tools" |
| 64 | ], |
| 65 | "metadata": { |
| 66 | "id": "xjVkqaSdRx0_" |
| 67 | } |
| 68 | }, |
| 69 | { |
| 70 | "cell_type": "code", |
| 71 | "source": [ |
| 72 | "from transformers.utils import get_json_schema\n", |
| 73 | "\n", |
| 74 | "def multiply(a: float, b: float):\n", |
| 75 | " \"\"\"\n", |
| 76 | " A function that multiplies two numbers\n", |
| 77 | "\n", |
| 78 | " Args:\n", |
| 79 | " a: The first number to multiply\n", |
| 80 | " b: The second number to multiply\n", |
| 81 | " \"\"\"\n", |
| 82 | " return a * b\n", |
| 83 | "\n", |
| 84 | "messages = [\n", |
| 85 | " {\"role\": \"user\", \"content\": \"what is 2.0909090923 x 0.897987987\"},\n", |
| 86 | "]\n", |
| 87 | "\n", |
| 88 | "tokenized_chat = tokenizer.apply_chat_template(\n", |
| 89 | " messages,\n", |
| 90 | " tools=[\n", |
| 91 | " multiply\n", |
| 92 | " ],\n", |
| 93 | " tokenize=True,\n", |
| 94 | " add_generation_prompt=True,\n", |
| 95 | " return_tensors=\"pt\"\n", |
| 96 | ").to(model.device)\n" |
| 97 | ], |
| 98 | "metadata": { |
| 99 | "id": "zxZZ7iMZETsw" |
| 100 | }, |
| 101 | "execution_count": null, |
| 102 | "outputs": [] |
| 103 | }, |
| 104 | { |
| 105 | "cell_type": "markdown", |
| 106 | "source": [ |
| 107 | "# Inference" |
| 108 | ], |
| 109 | "metadata": { |
| 110 | "id": "SVBAG3dLRw4v" |
| 111 | } |
| 112 | }, |
| 113 | { |
| 114 | "cell_type": "code", |
| 115 | "source": [ |
| 116 | "outputs = model.generate(\n", |
| 117 | " tokenized_chat,\n", |
| 118 | " max_new_tokens=1024,\n", |
| 119 | " temperature=1.0,\n", |
| 120 | " top_p=1.0,\n", |
| 121 | " eos_token_id=tokenizer.eos_token_id\n", |
| 122 | ")\n", |
| 123 | "print(tokenizer.decode(outputs[0]))" |
| 124 | ], |
| 125 | "metadata": { |
| 126 | "id": "BKYqPT5ORDx3" |
| 127 | }, |
| 128 | "execution_count": null, |
| 129 | "outputs": [] |
| 130 | } |
| 131 | ] |
| 132 | } |