Merge pull request #510 from Zhufeng-Qiu/zephyr-week3_4_5
Add the community contribution for Week3/4/5
This commit is contained in:
Ed Donner
GitHub
@@ -0,0 +1,551 @@
|
||||
{
|
||||
"cells": [
|
||||
{
|
||||
"cell_type": "markdown",
|
||||
"metadata": {
|
||||
"id": "GD5Omr5EfWgb"
|
||||
},
|
||||
"source": [
|
||||
"# Date Generator\n",
|
||||
"\n",
|
||||
"generate synthetic data when given scheme, business problem description, model, number of records, file name, file type, and environment\n",
|
||||
"\n",
|
||||
"# Available models\n",
|
||||
" Model API:\n",
|
||||
"\n",
|
||||
" 1. gpt-4o-mini\n",
|
||||
" 2. claude-3-haiku-20240307\n",
|
||||
" 3. gemini-2.0-flash\n",
|
||||
" 4. deepseek-chat\"\n",
|
||||
"\n",
|
||||
" HuggingFace API:\n",
|
||||
"\n",
|
||||
" 5. meta-llama/Meta-Llama-3.1-8B-Instruct\n",
|
||||
"\n",
|
||||
"\n",
|
||||
"# Available environment\n",
|
||||
"\n",
|
||||
"Colab: set up HF token and API keys in Colab secret section\n",
|
||||
"\n",
|
||||
"Local: set up HF token and API keys in .env file\n",
|
||||
"\n",
|
||||
"\n",
|
||||
"\n",
|
||||
"### *** This project is developed based on the idea of 'week3/community-contributuins/Week3-Dataset_Generator-DP'. Really appreciate it! Then, the project is improved to run both on Colab or locally, and integrate HuggingFace API"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"metadata": {
|
||||
"id": "4FiCnE0MmU56"
|
||||
},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"!pip install -q --upgrade torch==2.5.1+cu124 torchvision==0.20.1+cu124 torchaudio==2.5.1+cu124 --index-url https://download.pytorch.org/whl/cu124\n",
|
||||
"!pip install -q requests bitsandbytes==0.46.0 transformers==4.48.3 accelerate==1.3.0\n",
|
||||
"!pip install anthropic dotenv pyarrow"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"metadata": {
|
||||
"id": "JeyKw5guoH3r"
|
||||
},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"# imports\n",
|
||||
"\n",
|
||||
"import os\n",
|
||||
"import requests\n",
|
||||
"from IPython.display import Markdown, display, update_display\n",
|
||||
"from openai import OpenAI\n",
|
||||
"from huggingface_hub import login\n",
|
||||
"from transformers import AutoTokenizer, AutoModelForCausalLM, TextStreamer, BitsAndBytesConfig\n",
|
||||
"from bs4 import BeautifulSoup\n",
|
||||
"from typing import List\n",
|
||||
"import google.generativeai\n",
|
||||
"import anthropic\n",
|
||||
"from itertools import chain\n",
|
||||
"from dotenv import load_dotenv\n",
|
||||
"import gradio as gr\n",
|
||||
"import json\n",
|
||||
"import pandas as pd\n",
|
||||
"import random\n",
|
||||
"import re\n",
|
||||
"import subprocess\n",
|
||||
"import pyarrow as pa\n",
|
||||
"import torch\n",
|
||||
"import gc"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"metadata": {
|
||||
"id": "7UyjFdRZoIAS"
|
||||
},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"# --- Schema Definition ---\n",
|
||||
"SCHEMA = [\n",
|
||||
" (\"Name\", \"TEXT\", '\"Northern Cafe\"'),\n",
|
||||
" (\"Location\", \"TEXT\", '\"2904 S Figueroa St, Los Angeles, CA 90007\"'),\n",
|
||||
" (\"Type\", \"TEXT\", 'One of [\"Chinese\",\"Mexico\",\"French\",\"Korean\",\"Italy\"] or other potential types'),\n",
|
||||
" (\"Average Price\", \"TEXT\", '\"$30\", or \"--\" if unkown'),\n",
|
||||
" (\"History/Age\", \"INT\", 'integer age of resturant, e.g., 7'),\n",
|
||||
" (\"Menu\", \"Array\", '[\"Beef Noodle\", \"Fried Rice\", \"Dumpling\", ...]'),\n",
|
||||
"]"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"metadata": {
|
||||
"id": "jXcTQATLoICV"
|
||||
},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"# Default schema text for the textbox\n",
|
||||
"DEFAULT_SCHEMA_TEXT = \"\\n\".join([f\"{i+1}. {col[0]} ({col[1]}) Example: {col[2]}\" for i, col in enumerate(SCHEMA)])"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"metadata": {
|
||||
"id": "4Irf5JV3oIEe"
|
||||
},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"# Available models\n",
|
||||
"MODELS = [\n",
|
||||
" \"gpt-4o-mini\",\n",
|
||||
" \"claude-3-haiku-20240307\",\n",
|
||||
" \"gemini-2.0-flash\",\n",
|
||||
" \"deepseek-chat\",\n",
|
||||
" \"meta-llama/Meta-Llama-3.1-8B-Instruct\"\n",
|
||||
"]"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"metadata": {
|
||||
"id": "JJ6r2SH9oIGf"
|
||||
},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"# Available file formats\n",
|
||||
"FILE_FORMATS = [\".csv\", \".tsv\", \".jsonl\", \".parquet\", \".arrow\"]"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"metadata": {
|
||||
"id": "B98j45E3vq5g"
|
||||
},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"system_prompt = \"\"\"You are a helpful assistant whose main purpose is to generate datasets for a given business problem based on given schema.\"\"\""
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"metadata": {
|
||||
"id": "lsX16cWfwf6x"
|
||||
},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"def get_env_info(env):\n",
|
||||
" try:\n",
|
||||
" global hf_token, openai_api_key, anthropic_api_key, google_api_key, deepseek_api_key\n",
|
||||
" if env == \"Colab\":\n",
|
||||
" # Colab environment\n",
|
||||
" from google.colab import drive\n",
|
||||
" from google.colab import userdata\n",
|
||||
" hf_token = userdata.get('HF_TOKEN')\n",
|
||||
" openai_api_key = userdata.get('OPENAI_API_KEY')\n",
|
||||
" anthropic_api_key = userdata.get('ANTHROPIC_API_KEY')\n",
|
||||
" google_api_key = userdata.get('GOOGLE_API_KEY')\n",
|
||||
" deepseek_api_key = userdata.get('DEEPSEEK_API_KEY')\n",
|
||||
" elif env == \"Local\":\n",
|
||||
" # Local environment\n",
|
||||
" load_dotenv(override=True)\n",
|
||||
" hf_token = os.getenv('HF_TOKEN')\n",
|
||||
" openai_api_key = os.getenv('OPENAI_API_KEY')\n",
|
||||
" anthropic_api_key = os.getenv('ANTHROPIC_API_KEY')\n",
|
||||
" google_api_key = os.getenv('GOOGLE_API_KEY')\n",
|
||||
" deepseek_api_key = os.getenv('DEEPSEEK_API_KEY')\n",
|
||||
" except Exception as e:\n",
|
||||
" raise Exception(f\"Please check your environment: {str(e)}\")"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"metadata": {
|
||||
"id": "2gLUFAwGv29Q"
|
||||
},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"def get_prompt(schema_text, business_problem, nr_records):\n",
|
||||
" prompt = f\"\"\"\n",
|
||||
" The problem is: {business_problem}\n",
|
||||
"\n",
|
||||
" Generate {nr_records} rows data in JSONL format, each line a JSON object with the following fields:\n",
|
||||
"\n",
|
||||
" {schema_text}\n",
|
||||
"\n",
|
||||
" Do NOT repeat column values from one row to another.\n",
|
||||
"\n",
|
||||
" Only output valid JSONL.\n",
|
||||
" \"\"\"\n",
|
||||
" return prompt.strip()"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"metadata": {
|
||||
"id": "YZe1FVH8wf84"
|
||||
},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"# --- LLM Interface ---\n",
|
||||
"def query(user_prompt, model):\n",
|
||||
" try:\n",
|
||||
" if \"gpt\" in model.lower():\n",
|
||||
" client = OpenAI(api_key=openai_api_key)\n",
|
||||
" messages = [\n",
|
||||
" {\"role\": \"system\", \"content\": system_prompt},\n",
|
||||
" {\"role\": \"user\", \"content\": user_prompt}\n",
|
||||
" ]\n",
|
||||
" response = client.chat.completions.create(\n",
|
||||
" model=model,\n",
|
||||
" messages=messages,\n",
|
||||
" temperature=0.7\n",
|
||||
" )\n",
|
||||
" content = response.choices[0].message.content\n",
|
||||
"\n",
|
||||
" elif \"claude\" in model.lower():\n",
|
||||
" client = anthropic.Anthropic(api_key=anthropic_api_key)\n",
|
||||
" response = client.messages.create(\n",
|
||||
" model=model,\n",
|
||||
" messages=[{\"role\": \"user\", \"content\": user_prompt}],\n",
|
||||
" max_tokens=4000,\n",
|
||||
" temperature=0.7,\n",
|
||||
" system=system_prompt\n",
|
||||
" )\n",
|
||||
" content = response.content[0].text\n",
|
||||
" elif \"gemini\" in model.lower():\n",
|
||||
" client = OpenAI(\n",
|
||||
" api_key=google_api_key,\n",
|
||||
" base_url=\"https://generativelanguage.googleapis.com/v1beta/openai/\"\n",
|
||||
" )\n",
|
||||
" messages = [\n",
|
||||
" {\"role\": \"system\", \"content\": system_prompt},\n",
|
||||
" {\"role\": \"user\", \"content\": user_prompt}\n",
|
||||
" ]\n",
|
||||
" response = client.chat.completions.create(\n",
|
||||
" model=model,\n",
|
||||
" messages=messages,\n",
|
||||
" temperature=0.7\n",
|
||||
" )\n",
|
||||
" content = response.choices[0].message.content\n",
|
||||
"\n",
|
||||
" elif \"deepseek\" in model.lower():\n",
|
||||
" client = OpenAI(\n",
|
||||
" api_key=deepseek_api_key,\n",
|
||||
" base_url=\"https://api.deepseek.com\"\n",
|
||||
" )\n",
|
||||
" messages = [\n",
|
||||
" {\"role\": \"system\", \"content\": system_prompt},\n",
|
||||
" {\"role\": \"user\", \"content\": user_prompt}\n",
|
||||
" ]\n",
|
||||
" response = client.chat.completions.create(\n",
|
||||
" model=model,\n",
|
||||
" messages=messages,\n",
|
||||
" temperature=0.7\n",
|
||||
" )\n",
|
||||
" content = response.choices[0].message.content\n",
|
||||
"\n",
|
||||
" elif \"llama\" in model.lower():\n",
|
||||
" global tokenizer, inputs, llama_model, outputs\n",
|
||||
" messages = [\n",
|
||||
" {\"role\": \"system\", \"content\": system_prompt},\n",
|
||||
" {\"role\": \"user\", \"content\": user_prompt}\n",
|
||||
" ]\n",
|
||||
"\n",
|
||||
" login(hf_token, add_to_git_credential=True)\n",
|
||||
" quant_config = BitsAndBytesConfig(\n",
|
||||
" load_in_4bit=True,\n",
|
||||
" bnb_4bit_use_double_quant=True,\n",
|
||||
" bnb_4bit_compute_dtype=torch.bfloat16,\n",
|
||||
" bnb_4bit_quant_type=\"nf4\"\n",
|
||||
" )\n",
|
||||
"\n",
|
||||
" tokenizer = AutoTokenizer.from_pretrained(model, trust_remote_code=True)\n",
|
||||
" tokenizer.pad_token = tokenizer.eos_token\n",
|
||||
" inputs = tokenizer.apply_chat_template(messages, return_tensors=\"pt\").to(\"cuda\")\n",
|
||||
" if llama_model == None:\n",
|
||||
" llama_model = AutoModelForCausalLM.from_pretrained(model, device_map=\"auto\", quantization_config=quant_config)\n",
|
||||
" outputs = llama_model.generate(inputs, max_new_tokens=4000)\n",
|
||||
"\n",
|
||||
" _, _, after = tokenizer.decode(outputs[0]).partition(\"assistant<|end_header_id|>\")\n",
|
||||
" content = after.strip()\n",
|
||||
" else:\n",
|
||||
" raise ValueError(f\"Unsupported model. Use one of {MODELS}\")\n",
|
||||
"\n",
|
||||
" # Parse JSONL output\n",
|
||||
" lines = [line.strip() for line in content.strip().splitlines() if line.strip().startswith(\"{\")]\n",
|
||||
" return [json.loads(line) for line in lines]\n",
|
||||
"\n",
|
||||
" except Exception as e:\n",
|
||||
" raise Exception(f\"Model query failed: {str(e)}\")"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"metadata": {
|
||||
"id": "4WUj-XqM5IYT"
|
||||
},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"# --- Output Formatter ---\n",
|
||||
"def save_dataset(records, file_format, filename):\n",
|
||||
" df = pd.DataFrame(records)\n",
|
||||
" if file_format == \".csv\":\n",
|
||||
" df.to_csv(filename, index=False)\n",
|
||||
" elif file_format == \".tsv\":\n",
|
||||
" df.to_csv(filename, sep=\"\\t\", index=False)\n",
|
||||
" elif file_format == \".jsonl\":\n",
|
||||
" with open(filename, \"w\") as f:\n",
|
||||
" for record in records:\n",
|
||||
" f.write(json.dumps(record) + \"\\n\")\n",
|
||||
" elif file_format == \".parquet\":\n",
|
||||
" df.to_parquet(filename, engine=\"pyarrow\", index=False)\n",
|
||||
" elif file_format == \".arrow\":\n",
|
||||
" table = pa.Table.from_pandas(df)\n",
|
||||
" with pa.OSFile(filename, \"wb\") as sink:\n",
|
||||
" with pa.ipc.new_file(sink, table.schema) as writer:\n",
|
||||
" writer.write(table)\n",
|
||||
" else:\n",
|
||||
" raise ValueError(\"Unsupported file format\")"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"metadata": {
|
||||
"id": "WenbNqrpwf-_"
|
||||
},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"# --- Main Generation Function ---\n",
|
||||
"def generate_dataset(schema_text, business_problem, model, nr_records, file_format, save_as, env):\n",
|
||||
" try:\n",
|
||||
" # Validation\n",
|
||||
" if nr_records <= 10:\n",
|
||||
" return \"❌ Error: Number of records must be greater than 10.\", None\n",
|
||||
" if nr_records > 1000:\n",
|
||||
" return \"❌ Error: Number of records must be less than or equal to 1000.\", None\n",
|
||||
"\n",
|
||||
" if file_format not in FILE_FORMATS:\n",
|
||||
" return \"❌ Error: Invalid file format.\", None\n",
|
||||
"\n",
|
||||
" if not (save_as or save_as.strip() == \"\"):\n",
|
||||
" save_as = f\"default{file_format}\"\n",
|
||||
" elif not save_as.endswith(file_format):\n",
|
||||
" save_as = save_as + file_format\n",
|
||||
"\n",
|
||||
" # Load env\n",
|
||||
" get_env_info(env)\n",
|
||||
"\n",
|
||||
" # Generate prompt\n",
|
||||
" user_prompt = get_prompt(schema_text, business_problem, nr_records)\n",
|
||||
"\n",
|
||||
" # Query model\n",
|
||||
" records = query(user_prompt, model)\n",
|
||||
"\n",
|
||||
" if not records:\n",
|
||||
" return \"❌ Error: No valid records generated from the model.\", None\n",
|
||||
"\n",
|
||||
" # Save dataset\n",
|
||||
" save_dataset(records, file_format, save_as)\n",
|
||||
"\n",
|
||||
" # Create preview\n",
|
||||
" df = pd.DataFrame(records)\n",
|
||||
" preview = df.head(10) # Show first 10 rows\n",
|
||||
"\n",
|
||||
" success_message = f\"✅ Generated {len(records)} records successfully!\\n📁 Saved to: {save_as}\\n📊 \"\n",
|
||||
"\n",
|
||||
" return success_message, preview\n",
|
||||
"\n",
|
||||
" except Exception as e:\n",
|
||||
" return f\"❌ Error: {str(e)}\", None"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"metadata": {
|
||||
"id": "pHiP8ky8wgEb"
|
||||
},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"# --- Gradio Interface ---\n",
|
||||
"\n",
|
||||
"with gr.Blocks(title=\"Dataset Generator\", theme=gr.themes.Citrus()) as interface:\n",
|
||||
" hf_token = None\n",
|
||||
" openai_api_key = None\n",
|
||||
" anthropic_api_key = None\n",
|
||||
" google_api_key = None\n",
|
||||
" deepseek_api_key = None\n",
|
||||
" tokenizer = None\n",
|
||||
" inputs = None\n",
|
||||
" llama_model = None\n",
|
||||
" outputs = None\n",
|
||||
"\n",
|
||||
" gr.Markdown(\"# Dataset Generator\")\n",
|
||||
" gr.Markdown(\"Generate synthetic datasets using AI models\")\n",
|
||||
"\n",
|
||||
" with gr.Row():\n",
|
||||
" with gr.Column(scale=2):\n",
|
||||
" schema_input = gr.Textbox(\n",
|
||||
" label=\"Schema\",\n",
|
||||
" value=DEFAULT_SCHEMA_TEXT,\n",
|
||||
" lines=15,\n",
|
||||
" placeholder=\"Define your dataset schema here... Please follow this format: Field_Name, Field_Type, Field Example\"\n",
|
||||
" )\n",
|
||||
"\n",
|
||||
" business_problem_input = gr.Textbox(\n",
|
||||
" label=\"Business Problem\",\n",
|
||||
" value=\"I want to generate restuant records\",\n",
|
||||
" lines=1,\n",
|
||||
" placeholder=\"Enter business problem desciption for the model...\"\n",
|
||||
" )\n",
|
||||
"\n",
|
||||
" with gr.Row():\n",
|
||||
" model_dropdown = gr.Dropdown(\n",
|
||||
" label=\"Model\",\n",
|
||||
" choices=MODELS,\n",
|
||||
" value=MODELS[0],\n",
|
||||
" interactive=True\n",
|
||||
" )\n",
|
||||
"\n",
|
||||
" nr_records_input = gr.Number(\n",
|
||||
" label=\"Number of records\",\n",
|
||||
" value=27,\n",
|
||||
" minimum=11,\n",
|
||||
" maximum=1000,\n",
|
||||
" step=1\n",
|
||||
" )\n",
|
||||
"\n",
|
||||
" with gr.Row():\n",
|
||||
" save_as_input = gr.Textbox(\n",
|
||||
" label=\"Save as\",\n",
|
||||
" value=\"restaurant_dataset\",\n",
|
||||
" placeholder=\"Enter filename (extension will be added automatically)\"\n",
|
||||
" )\n",
|
||||
"\n",
|
||||
" file_format_dropdown = gr.Dropdown(\n",
|
||||
" label=\"File format\",\n",
|
||||
" choices=FILE_FORMATS,\n",
|
||||
" value=FILE_FORMATS[0],\n",
|
||||
" interactive=True\n",
|
||||
" )\n",
|
||||
"\n",
|
||||
" env_dropdown = gr.Dropdown(\n",
|
||||
" label=\"Environment\",\n",
|
||||
" choices=[\"Colab\", \"Local\"],\n",
|
||||
" value=\"Colab\",\n",
|
||||
" interactive=True\n",
|
||||
" )\n",
|
||||
"\n",
|
||||
"\n",
|
||||
"\n",
|
||||
" generate_btn = gr.Button(\"🚀 Generate\", variant=\"secondary\", size=\"lg\")\n",
|
||||
"\n",
|
||||
" with gr.Column(scale=1):\n",
|
||||
" output_status = gr.Textbox(\n",
|
||||
" label=\"Status\",\n",
|
||||
" lines=4,\n",
|
||||
" interactive=False\n",
|
||||
" )\n",
|
||||
"\n",
|
||||
" output_preview = gr.Dataframe(\n",
|
||||
" label=\"Preview (First 10 rows)\",\n",
|
||||
" interactive=False,\n",
|
||||
" wrap=True\n",
|
||||
" )\n",
|
||||
"\n",
|
||||
" # Connect the generate button\n",
|
||||
" generate_btn.click(\n",
|
||||
" fn=generate_dataset,\n",
|
||||
" inputs=[\n",
|
||||
" schema_input,\n",
|
||||
" business_problem_input,\n",
|
||||
" model_dropdown,\n",
|
||||
" nr_records_input,\n",
|
||||
" file_format_dropdown,\n",
|
||||
" save_as_input,\n",
|
||||
" env_dropdown\n",
|
||||
" ],\n",
|
||||
" outputs=[output_status, output_preview]\n",
|
||||
" )\n",
|
||||
"\n",
|
||||
" gr.Markdown(\"\"\"\n",
|
||||
" ### 📝 Instructions:\n",
|
||||
" 1. **Schema**: Define the structure of your dataset (pre-filled with restaurant schema)\n",
|
||||
" 2. **Business problem**: User prompt to guide the AI model\n",
|
||||
" 3. **Model**: Choose between GPT, Claude, Gemini, DeepSeek or Llama models\n",
|
||||
" 4. **Number of records**: Number of records to generate (minimum 11)\n",
|
||||
" 5. **File format**: Choose output format (.csv, .tsv, .jsonl, .parquet, .arrow)\n",
|
||||
" 6. **Save as**: Filename (extension added automatically)\n",
|
||||
" 7. Click **Generate** to create your dataset\n",
|
||||
"\n",
|
||||
" ### 🔧 Requirements:\n",
|
||||
" - For local mode, set up HF token and API keys in `.env` file (`OPENAI_API_KEY`, `ANTHROPIC_API_KEY`, `GOOGLE_API_KEY`, `DEEPSEEK_API_KEY`, `HF_TOKEN`)\n",
|
||||
" - For colab mode, set up HF token and API keys in Colab secret section (`OPENAI_API_KEY`, `ANTHROPIC_API_KEY`, `GOOGLE_API_KEY`, `DEEPSEEK_API_KEY`, `HF_TOKEN`)\n",
|
||||
" \"\"\")\n",
|
||||
"\n",
|
||||
"interface.launch(debug=True)\n",
|
||||
"\n",
|
||||
"del tokenizer, inputs, llama_model, outputs\n",
|
||||
"gc.collect()\n",
|
||||
"torch.cuda.empty_cache()"
|
||||
]
|
||||
}
|
||||
],
|
||||
"metadata": {
|
||||
"accelerator": "GPU",
|
||||
"colab": {
|
||||
"gpuType": "T4",
|
||||
"provenance": []
|
||||
},
|
||||
"kernelspec": {
|
||||
"display_name": "Python 3 (ipykernel)",
|
||||
"language": "python",
|
||||
"name": "python3"
|
||||
},
|
||||
"language_info": {
|
||||
"codemirror_mode": {
|
||||
"name": "ipython",
|
||||
"version": 3
|
||||
},
|
||||
"file_extension": ".py",
|
||||
"mimetype": "text/x-python",
|
||||
"name": "python",
|
||||
"nbconvert_exporter": "python",
|
||||
"pygments_lexer": "ipython3",
|
||||
"version": "3.11.13"
|
||||
}
|
||||
},
|
||||
"nbformat": 4,
|
||||
"nbformat_minor": 4
|
||||
}
|
||||
@@ -0,0 +1,523 @@
|
||||
{
|
||||
"cells": [
|
||||
{
|
||||
"cell_type": "markdown",
|
||||
"metadata": {
|
||||
"id": "It89APiAtTUF"
|
||||
},
|
||||
"source": [
|
||||
"# Create meeting minutes from an Audio file\n",
|
||||
"\n",
|
||||
"I downloaded some Denver City Council meeting minutes and selected a portion of the meeting for us to transcribe. You can download it here: \n",
|
||||
"https://drive.google.com/file/d/1N_kpSojRR5RYzupz6nqM8hMSoEF_R7pU/view?usp=sharing\n",
|
||||
"\n",
|
||||
"If you'd rather work with the original data, the HuggingFace dataset is [here](https://huggingface.co/datasets/huuuyeah/meetingbank) and the audio can be downloaded [here](https://huggingface.co/datasets/huuuyeah/MeetingBank_Audio/tree/main).\n",
|
||||
"\n",
|
||||
"The goal of this product is to use the Audio to generate meeting minutes, including actions.\n",
|
||||
"\n",
|
||||
"For this project, you can either use the Denver meeting minutes, or you can record something of your own!\n"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "markdown",
|
||||
"metadata": {
|
||||
"id": "sJPSCwPX3MOV"
|
||||
},
|
||||
"source": [
|
||||
"## Again - please note: 2 important pro-tips for using Colab:\n",
|
||||
"\n",
|
||||
"**Pro-tip 1:**\n",
|
||||
"\n",
|
||||
"The top of every colab has some pip installs. You may receive errors from pip when you run this, such as:\n",
|
||||
"\n",
|
||||
"> gcsfs 2025.3.2 requires fsspec==2025.3.2, but you have fsspec 2025.3.0 which is incompatible.\n",
|
||||
"\n",
|
||||
"These pip compatibility errors can be safely ignored; and while it's tempting to try to fix them by changing version numbers, that will actually introduce real problems!\n",
|
||||
"\n",
|
||||
"**Pro-tip 2:**\n",
|
||||
"\n",
|
||||
"In the middle of running a Colab, you might get an error like this:\n",
|
||||
"\n",
|
||||
"> Runtime error: CUDA is required but not available for bitsandbytes. Please consider installing [...]\n",
|
||||
"\n",
|
||||
"This is a super-misleading error message! Please don't try changing versions of packages...\n",
|
||||
"\n",
|
||||
"This actually happens because Google has switched out your Colab runtime, perhaps because Google Colab was too busy. The solution is:\n",
|
||||
"\n",
|
||||
"1. Kernel menu >> Disconnect and delete runtime\n",
|
||||
"2. Reload the colab from fresh and Edit menu >> Clear All Outputs\n",
|
||||
"3. Connect to a new T4 using the button at the top right\n",
|
||||
"4. Select \"View resources\" from the menu on the top right to confirm you have a GPU\n",
|
||||
"5. Rerun the cells in the colab, from the top down, starting with the pip installs\n",
|
||||
"\n",
|
||||
"And all should work great - otherwise, ask me!"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"metadata": {
|
||||
"id": "f2vvgnFpHpID"
|
||||
},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"!pip install -q --upgrade torch==2.5.1+cu124 torchvision==0.20.1+cu124 torchaudio==2.5.1+cu124 --index-url https://download.pytorch.org/whl/cu124\n",
|
||||
"!pip install -q requests bitsandbytes==0.46.0 transformers==4.48.3 accelerate==1.3.0 openai"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"metadata": {
|
||||
"id": "FW8nl3XRFrz0"
|
||||
},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"# imports\n",
|
||||
"\n",
|
||||
"import os\n",
|
||||
"import requests\n",
|
||||
"from IPython.display import Markdown, display, update_display\n",
|
||||
"from openai import OpenAI\n",
|
||||
"from google.colab import drive\n",
|
||||
"from huggingface_hub import login\n",
|
||||
"from google.colab import userdata\n",
|
||||
"from transformers import AutoTokenizer, AutoModelForCausalLM, TextStreamer, BitsAndBytesConfig\n",
|
||||
"import torch"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"metadata": {
|
||||
"id": "q3D1_T0uG_Qh"
|
||||
},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"# Constants\n",
|
||||
"\n",
|
||||
"AUDIO_MODEL = \"whisper-1\"\n",
|
||||
"LLAMA = \"meta-llama/Meta-Llama-3.1-8B-Instruct\""
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"metadata": {
|
||||
"id": "Es9GkQ0FGCMt"
|
||||
},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"# New capability - connect this Colab to my Google Drive\n",
|
||||
"# See immediately below this for instructions to obtain denver_extract.mp3\n",
|
||||
"\n",
|
||||
"drive.mount(\"/content/drive\")\n",
|
||||
"audio_filename = \"/content/drive/MyDrive/llms/denver_extract.mp3\""
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "markdown",
|
||||
"metadata": {
|
||||
"id": "HTl3mcjyzIEE"
|
||||
},
|
||||
"source": [
|
||||
"# Download denver_extract.mp3\n",
|
||||
"\n",
|
||||
"You can either use the same file as me, the extract from Denver city council minutes, or you can try your own..\n",
|
||||
"\n",
|
||||
"If you want to use the same as me, then please download my extract here, and put this on your Google Drive: \n",
|
||||
"https://drive.google.com/file/d/1N_kpSojRR5RYzupz6nqM8hMSoEF_R7pU/view?usp=sharing\n"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"metadata": {
|
||||
"id": "xYW8kQYtF-3L"
|
||||
},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"# Sign in to HuggingFace Hub\n",
|
||||
"\n",
|
||||
"hf_token = userdata.get('HF_TOKEN')\n",
|
||||
"login(hf_token, add_to_git_credential=True)"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"metadata": {
|
||||
"id": "qP6OB2OeGC2C"
|
||||
},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"# Sign in to OpenAI using Secrets in Colab\n",
|
||||
"\n",
|
||||
"openai_api_key = userdata.get('OPENAI_API_KEY')\n",
|
||||
"openai = OpenAI(api_key=openai_api_key)"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"metadata": {
|
||||
"id": "GMShdVGlGGr4"
|
||||
},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"# Use the Whisper OpenAI model to convert the Audio to Text\n",
|
||||
"# If you'd prefer to use an Open Source model, class student Youssef has contributed an open source version\n",
|
||||
"# which I've added to the bottom of this colab\n",
|
||||
"\n",
|
||||
"audio_file = open(audio_filename, \"rb\")\n",
|
||||
"transcription = openai.audio.transcriptions.create(model=AUDIO_MODEL, file=audio_file, response_format=\"text\")\n",
|
||||
"print(transcription)"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"metadata": {
|
||||
"id": "piEMmcSfMH-O"
|
||||
},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"system_message = \"You are an assistant that produces minutes of meetings from transcripts, with summary, key discussion points, takeaways and action items with owners, in markdown.\"\n",
|
||||
"user_prompt = f\"Below is an extract transcript of a Denver council meeting. Please write minutes in markdown, including a summary with attendees, location and date; discussion points; takeaways; and action items with owners.\\n{transcription}\"\n",
|
||||
"\n",
|
||||
"messages = [\n",
|
||||
" {\"role\": \"system\", \"content\": system_message},\n",
|
||||
" {\"role\": \"user\", \"content\": user_prompt}\n",
|
||||
" ]\n"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"metadata": {
|
||||
"id": "UcRKUgcxMew6"
|
||||
},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"quant_config = BitsAndBytesConfig(\n",
|
||||
" load_in_4bit=True,\n",
|
||||
" bnb_4bit_use_double_quant=True,\n",
|
||||
" bnb_4bit_compute_dtype=torch.bfloat16,\n",
|
||||
" bnb_4bit_quant_type=\"nf4\"\n",
|
||||
")"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"metadata": {
|
||||
"id": "6CujZRAgMimy"
|
||||
},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"tokenizer = AutoTokenizer.from_pretrained(LLAMA)\n",
|
||||
"tokenizer.pad_token = tokenizer.eos_token\n",
|
||||
"# inputs = tokenizer.apply_chat_template(messages, return_tensors=\"pt\").to(\"cuda\")\n",
|
||||
"streamer = TextStreamer(tokenizer)\n",
|
||||
"model = AutoModelForCausalLM.from_pretrained(LLAMA, device_map=\"auto\", quantization_config=quant_config, trust_remote_code=True)\n",
|
||||
"# outputs = model.generate(inputs, max_new_tokens=2000, streamer=streamer)"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"metadata": {
|
||||
"id": "MaLNmJ5PSqcH"
|
||||
},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"inputs = tokenizer.apply_chat_template(messages, return_tensors=\"pt\").to(\"cuda\")\n",
|
||||
"outputs = model.generate(inputs, max_new_tokens=2000, streamer=streamer)"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"metadata": {
|
||||
"id": "102tdU_3Peam"
|
||||
},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"response = tokenizer.decode(outputs[0])\n",
|
||||
"response"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"metadata": {
|
||||
"id": "KlomN6CwMdoN"
|
||||
},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"display(Markdown(response))"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "markdown",
|
||||
"metadata": {
|
||||
"id": "0jZElVOMSPAr"
|
||||
},
|
||||
"source": [
|
||||
"Day5 exercise - Gradio UI for meeting minutes"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"metadata": {
|
||||
"id": "5iiYYxQMHf0i"
|
||||
},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"import gradio as gr\n",
|
||||
"import tempfile\n",
|
||||
"import soundfile as sf"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"metadata": {
|
||||
"id": "aGwXW7BjPcTM"
|
||||
},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"# !pip install pydub\n",
|
||||
"# !apt-get install ffmpeg\n",
|
||||
"\n",
|
||||
"from pydub import AudioSegment"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"metadata": {
|
||||
"id": "RNu-reHuCYj_"
|
||||
},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"# Make sure that the tokenizeer and model is already generated\n",
|
||||
"\n",
|
||||
"# tokenizer = AutoTokenizer.from_pretrained(LLAMA)\n",
|
||||
"# tokenizer.pad_token = tokenizer.eos_token\n",
|
||||
"# streamer = TextStreamer(tokenizer)\n",
|
||||
"# model = AutoModelForCausalLM.from_pretrained(LLAMA, device_map=\"auto\", quantization_config=quant_config)"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"metadata": {
|
||||
"id": "KOuoH0YOPruE"
|
||||
},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"# def save_as_mp3(audio_np):\n",
|
||||
"# sr, data = audio_np\n",
|
||||
"# # Convert float32 or int16 to PCM wav and then mp3\n",
|
||||
"# wav_path = tempfile.NamedTemporaryFile(suffix=\".wav\", delete=False).name\n",
|
||||
"# mp3_path = tempfile.NamedTemporaryFile(suffix=\".mp3\", delete=False).name\n",
|
||||
"\n",
|
||||
"# sf.write(wav_path, data, sr)\n",
|
||||
"# audio_segment = AudioSegment.from_wav(wav_path)\n",
|
||||
"# audio_segment.export(mp3_path, format=\"mp3\", bitrate=\"64k\") # Low bitrate = small file\n",
|
||||
"# return mp3_path"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"metadata": {
|
||||
"id": "toBIPBJoSNw0"
|
||||
},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"# Handles audio input as numpy array and returns updated chat history\n",
|
||||
"def speak_send(audio_np):\n",
|
||||
"\n",
|
||||
" # If use numpy as input: audio_input = gr.Audio(sources=\"upload\", type=\"numpy\", label=\"Upload audio file to generate meeting minutes\")\n",
|
||||
" # mp3_path = save_as_mp3(audio_np)\n",
|
||||
"\n",
|
||||
" # with open(mp3_path, \"rb\") as audio_file:\n",
|
||||
" # transcription = openai.audio.transcriptions.create(\n",
|
||||
" # model=AUDIO_MODEL,\n",
|
||||
" # file=audio_file,\n",
|
||||
" # response_format=\"text\"\n",
|
||||
" # )\n",
|
||||
"\n",
|
||||
" audio = AudioSegment.from_file(audio_np)\n",
|
||||
" with tempfile.NamedTemporaryFile(suffix=\".mp3\", delete=False) as tmpfile:\n",
|
||||
" audio.export(tmpfile.name, format=\"mp3\")\n",
|
||||
" with open(tmpfile.name, \"rb\") as file:\n",
|
||||
" transcript = openai.audio.transcriptions.create(\n",
|
||||
" model=AUDIO_MODEL,\n",
|
||||
" file=file,\n",
|
||||
" response_format=\"text\"\n",
|
||||
" )\n",
|
||||
"\n",
|
||||
" system_message = \"You are an assistant that produces minutes of meetings from transcripts, with summary, key discussion points, takeaways and action items with owners, in markdown.\"\n",
|
||||
" user_prompt = f\"Below is an extract transcript of a Denver council meeting. Please write minutes in markdown, including a summary with attendees, location and date; discussion points; takeaways; and action items with owners.\\n{transcription}\"\n",
|
||||
"\n",
|
||||
" messages = [\n",
|
||||
" {\"role\": \"system\", \"content\": system_message},\n",
|
||||
" {\"role\": \"user\", \"content\": user_prompt}\n",
|
||||
" ]\n",
|
||||
"\n",
|
||||
" inputs = tokenizer.apply_chat_template(messages, return_tensors=\"pt\").to(\"cuda\")\n",
|
||||
" outputs = model.generate(inputs, max_new_tokens=2000)\n",
|
||||
"\n",
|
||||
" _, _, after = tokenizer.decode(outputs[0]).partition(\"assistant<|end_header_id|>\")\n",
|
||||
" return after.strip()\n"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"metadata": {
|
||||
"id": "xXJfabpDSN5R"
|
||||
},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"with gr.Blocks() as demo:\n",
|
||||
"\n",
|
||||
" with gr.Row():\n",
|
||||
" audio_input = gr.Audio(sources=\"upload\", type=\"filepath\", label=\"Upload audio file to generate meeting minutes\")\n",
|
||||
" with gr.Row():\n",
|
||||
" audio_submit = gr.Button(\"Send\")\n",
|
||||
" with gr.Row():\n",
|
||||
" outputs = [gr.Markdown(label=\"Meeting minutes:\")]\n",
|
||||
"\n",
|
||||
" audio_submit.click(speak_send, inputs=audio_input, outputs=outputs)\n",
|
||||
"\n",
|
||||
"demo.launch(debug=True)\n"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "markdown",
|
||||
"metadata": {
|
||||
"id": "kuxYecT2QDQ9"
|
||||
},
|
||||
"source": [
|
||||
"# Student contribution\n",
|
||||
"\n",
|
||||
"Student Emad S. has made this powerful variation that uses `TextIteratorStreamer` to stream back results into a Gradio UI, and takes advantage of background threads for performance! I'm sharing it here if you'd like to take a look at some very interesting work. Thank you, Emad!\n",
|
||||
"\n",
|
||||
"https://colab.research.google.com/drive/1Ja5zyniyJo5y8s1LKeCTSkB2xyDPOt6D"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "markdown",
|
||||
"metadata": {
|
||||
"id": "AU3uAEyU3a-o"
|
||||
},
|
||||
"source": [
|
||||
"## Alternative implementation\n",
|
||||
"\n",
|
||||
"Class student Youssef has contributed this variation in which we use an open-source model to transcribe the meeting Audio.\n",
|
||||
"\n",
|
||||
"Thank you Youssef!"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"metadata": {
|
||||
"id": "phYYgAbBRvu5"
|
||||
},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"import torch\n",
|
||||
"from transformers import AutoModelForSpeechSeq2Seq, AutoProcessor, pipeline"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"metadata": {
|
||||
"id": "HdQnWEzW3lzP"
|
||||
},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"AUDIO_MODEL = \"openai/whisper-medium\"\n",
|
||||
"speech_model = AutoModelForSpeechSeq2Seq.from_pretrained(AUDIO_MODEL, torch_dtype=torch.float16, low_cpu_mem_usage=True, use_safetensors=True)\n",
|
||||
"speech_model.to('cuda')\n",
|
||||
"processor = AutoProcessor.from_pretrained(AUDIO_MODEL)"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"metadata": {
|
||||
"id": "ZhA_fbeCSAeZ"
|
||||
},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"pipe = pipeline(\n",
|
||||
" \"automatic-speech-recognition\",\n",
|
||||
" model=speech_model,\n",
|
||||
" tokenizer=processor.tokenizer,\n",
|
||||
" feature_extractor=processor.feature_extractor,\n",
|
||||
" torch_dtype=torch.float16,\n",
|
||||
" device='cuda',\n",
|
||||
" return_timestamps=True\n",
|
||||
")"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"metadata": {
|
||||
"id": "nrQjKtD53omJ"
|
||||
},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"# Use the Whisper OpenAI model to convert the Audio to Text\n",
|
||||
"result = pipe(audio_filename)"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"metadata": {
|
||||
"id": "G_XSljOY3tDf"
|
||||
},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"transcription = result[\"text\"]\n",
|
||||
"print(transcription)"
|
||||
]
|
||||
}
|
||||
],
|
||||
"metadata": {
|
||||
"accelerator": "GPU",
|
||||
"colab": {
|
||||
"gpuType": "T4",
|
||||
"provenance": []
|
||||
},
|
||||
"kernelspec": {
|
||||
"display_name": "Python 3 (ipykernel)",
|
||||
"language": "python",
|
||||
"name": "python3"
|
||||
},
|
||||
"language_info": {
|
||||
"codemirror_mode": {
|
||||
"name": "ipython",
|
||||
"version": 3
|
||||
},
|
||||
"file_extension": ".py",
|
||||
"mimetype": "text/x-python",
|
||||
"name": "python",
|
||||
"nbconvert_exporter": "python",
|
||||
"pygments_lexer": "ipython3",
|
||||
"version": "3.11.13"
|
||||
}
|
||||
},
|
||||
"nbformat": 4,
|
||||
"nbformat_minor": 4
|
||||
}
|
||||
Reference in New Issue
Block a user