Skip to content

Commit

Permalink
Synthetic Instruction Dataset Notebook (#86)
Browse files Browse the repository at this point in the history 
* Synthetic Instruction Dataset Notebook

* minor fix
  • Loading branch information
@aravindputrevu
aravindputrevu authored Aug 30, 2024
1 parent 869b910 commit 48d5d4b
Showing 1 changed file with 392 additions and 0 deletions.
392 changes: 392 additions & 0 deletions examples/synthetic_data/Llama_3_1_Synthetic_Data.ipynb
View file
Original file line number Diff line number Diff line change
@@ -0,0 +1,392 @@
{
"cells": [
{
"cell_type": "markdown",
"id": "71a43144",
"metadata": {
"id": "71a43144"
},
"source": [
"# Get Started with Llama 3.1 Models\n",
"\n",
"\n",
"Llama 3.1 release comes with three sizes of models 7B, 70B and 405B\n",
"\n",
"In this notebook, we will look at :\n",
"\n",
"* How to access the Llama 3.1 models over a API?\n",
"* Generate Structured Synthetic Instruction Dataset with Llama 3.1 405B\n"
]
},
{
"cell_type": "markdown",
"id": "TWZqnPr0pcv8",
"metadata": {
"id": "TWZqnPr0pcv8"
},
"source": [
"## Setup\n",
"\n",
"Install all the dependencies and import the required python modules."
]
},
{
"cell_type": "code",
"execution_count": null,
"id": "YYRdRzHuHkjz",
"metadata": {
"colab": {
"base_uri": "https://localhost:8080/"
},
"id": "YYRdRzHuHkjz",
"outputId": "c497ab98-5b58-44cc-e1ef-3a67c65a4e7e"
},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"Collecting fireworks-ai\n",
" Downloading fireworks_ai-0.15.0-py3-none-any.whl.metadata (5.3 kB)\n",
"Collecting httpx (from fireworks-ai)\n",
" Downloading httpx-0.27.0-py3-none-any.whl.metadata (7.2 kB)\n",
"Collecting httpx-sse (from fireworks-ai)\n",
" Downloading httpx_sse-0.4.0-py3-none-any.whl.metadata (9.0 kB)\n",
"Requirement already satisfied: pydantic in /usr/local/lib/python3.10/dist-packages (from fireworks-ai) (2.8.2)\n",
"Requirement already satisfied: Pillow in /usr/local/lib/python3.10/dist-packages (from fireworks-ai) (9.4.0)\n",
"Requirement already satisfied: anyio in /usr/local/lib/python3.10/dist-packages (from httpx->fireworks-ai) (3.7.1)\n",
"Requirement already satisfied: certifi in /usr/local/lib/python3.10/dist-packages (from httpx->fireworks-ai) (2024.7.4)\n",
"Collecting httpcore==1.* (from httpx->fireworks-ai)\n",
" Downloading httpcore-1.0.5-py3-none-any.whl.metadata (20 kB)\n",
"Requirement already satisfied: idna in /usr/local/lib/python3.10/dist-packages (from httpx->fireworks-ai) (3.7)\n",
"Requirement already satisfied: sniffio in /usr/local/lib/python3.10/dist-packages (from httpx->fireworks-ai) (1.3.1)\n",
"Collecting h11<0.15,>=0.13 (from httpcore==1.*->httpx->fireworks-ai)\n",
" Downloading h11-0.14.0-py3-none-any.whl.metadata (8.2 kB)\n",
"Requirement already satisfied: annotated-types>=0.4.0 in /usr/local/lib/python3.10/dist-packages (from pydantic->fireworks-ai) (0.7.0)\n",
"Requirement already satisfied: pydantic-core==2.20.1 in /usr/local/lib/python3.10/dist-packages (from pydantic->fireworks-ai) (2.20.1)\n",
"Requirement already satisfied: typing-extensions>=4.6.1 in /usr/local/lib/python3.10/dist-packages (from pydantic->fireworks-ai) (4.12.2)\n",
"Requirement already satisfied: exceptiongroup in /usr/local/lib/python3.10/dist-packages (from anyio->httpx->fireworks-ai) (1.2.2)\n",
"Downloading fireworks_ai-0.15.0-py3-none-any.whl (83 kB)\n",
"\u001b[2K \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m84.0/84.0 kB\u001b[0m \u001b[31m1.2 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m\n",
"\u001b[?25hDownloading httpx-0.27.0-py3-none-any.whl (75 kB)\n",
"\u001b[2K \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m75.6/75.6 kB\u001b[0m \u001b[31m3.7 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m\n",
"\u001b[?25hDownloading httpcore-1.0.5-py3-none-any.whl (77 kB)\n",
"\u001b[2K \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m77.9/77.9 kB\u001b[0m \u001b[31m3.2 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m\n",
"\u001b[?25hDownloading httpx_sse-0.4.0-py3-none-any.whl (7.8 kB)\n",
"Downloading h11-0.14.0-py3-none-any.whl (58 kB)\n",
"\u001b[2K \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m58.3/58.3 kB\u001b[0m \u001b[31m1.3 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m\n",
"\u001b[?25hInstalling collected packages: httpx-sse, h11, httpcore, httpx, fireworks-ai\n",
"Successfully installed fireworks-ai-0.15.0 h11-0.14.0 httpcore-1.0.5 httpx-0.27.0 httpx-sse-0.4.0\n"
]
}
],
"source": [
"pip install --upgrade fireworks-ai"
]
},
{
"cell_type": "markdown",
"id": "sV4u2hezqV4X",
"metadata": {
"id": "sV4u2hezqV4X"
},
"source": [
"## Setup your API Key\n",
"\n",
"In order to use the Llama 3.1, you must first obtain Fireworks API Keys. If you don't already have one, you can one by following the instructions [here](https://docs.fireworks.ai/getting-started/quickstart)."
]
},
{
"cell_type": "code",
"execution_count": null,
"id": "M9EtJxjJrlyD",
"metadata": {
"id": "M9EtJxjJrlyD"
},
"outputs": [],
"source": [
"from fireworks.client import Fireworks\n",
"\n",
"#replace the FIREWORKS_API_KEY with the key copied in the above step.\n",
"client = Fireworks(api_key=\"FIREWORKS_API_KEY\")"
]
},
{
"cell_type": "markdown",
"id": "FrFtrGQ6r-Mk",
"metadata": {
"id": "FrFtrGQ6r-Mk"
},
"source": [
"## Accessing Llama 3.1 Models using API\n",
"\n",
"We are sending a request to Llama 3.1 405B model, alternatively you can change the model string to access the otherm models.\n",
"\n",
"* accounts/fireworks/models/llama-v3p1-70b-instruct\n",
"* accounts/fireworks/models/llama-v3p1-8B-instruct"
]
},
{
"cell_type": "markdown",
"id": "TVGsp9wvtO2Q",
"metadata": {
"id": "TVGsp9wvtO2Q"
},
"source": [
"### Chat Completions API"
]
},
{
"cell_type": "code",
"execution_count": null,
"id": "DHAgBTk8Hpg_",
"metadata": {
"colab": {
"base_uri": "https://localhost:8080/"
},
"id": "DHAgBTk8Hpg_",
"outputId": "a2060088-0fac-4c29-c388-60c3760e8c0f"
},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"I'm an artificial intelligence model known as Llama. Llama stands for \"Large Language Model Meta AI.\"\n"
]
}
],
"source": [
"model_name = \"accounts/fireworks/models/llama-v3p1-405b-instruct\"\n",
"\n",
"response = client.chat.completions.create(\n",
"\tmodel=model_name,\n",
"\tmessages=[{\n",
"\t\t\"role\": \"user\",\n",
"\t\t\"content\": \"Who are you?\",\n",
"\t}],\n",
")\n",
"print(response.choices[0].message.content)"
]
},
{
"cell_type": "markdown",
"id": "O5i7rqkhJQJN",
"metadata": {
"id": "O5i7rqkhJQJN"
},
"source": [
"## Generate Synthetic Data\n",
"\n"
]
},
{
"cell_type": "code",
"execution_count": null,
"id": "yxVgS4xuBhFF",
"metadata": {
"colab": {
"base_uri": "https://localhost:8080/"
},
"id": "yxVgS4xuBhFF",
"outputId": "1ce8ebe2-c924-4065-d2c7-87e2a8d5e019"
},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"Requirement already satisfied: pydantic in /usr/local/lib/python3.10/dist-packages (2.8.2)\n",
"Requirement already satisfied: annotated-types>=0.4.0 in /usr/local/lib/python3.10/dist-packages (from pydantic) (0.7.0)\n",
"Requirement already satisfied: pydantic-core==2.20.1 in /usr/local/lib/python3.10/dist-packages (from pydantic) (2.20.1)\n",
"Requirement already satisfied: typing-extensions>=4.6.1 in /usr/local/lib/python3.10/dist-packages (from pydantic) (4.12.2)\n"
]
}
],
"source": [
"pip install pydantic"
]
},
{
"cell_type": "code",
"execution_count": null,
"id": "wcOA_lp1B9lD",
"metadata": {
"id": "wcOA_lp1B9lD"
},
"outputs": [],
"source": [
"from pydantic import BaseModel, Field"
]
},
{
"cell_type": "code",
"execution_count": null,
"id": "D9xgnLuGCAQi",
"metadata": {
"id": "D9xgnLuGCAQi"
},
"outputs": [],
"source": [
"from pydantic import BaseModel, Field\n",
"from typing import List, Optional\n",
"from enum import Enum\n",
"\n",
"\n",
"class Category(str, Enum):\n",
" COUNTRIES = \"Countries\"\n",
" CAPITALS = \"Capitals\"\n",
" RIVERS = \"Rivers\"\n",
" MOUNTAINS = \"Mountains\"\n",
" LANDMARKS = \"Landmarks\"\n",
" CLIMATE = \"Climate\"\n",
" CULTURE = \"Culture\"\n",
"\n",
"class Difficulty(str, Enum):\n",
" EASY = \"Easy\"\n",
" MEDIUM = \"Medium\"\n",
" HARD = \"Hard\"\n",
" EXPERT = \"Expert\"\n",
"\n",
"class QuestionType(str, Enum):\n",
" MULTIPLE_CHOICE = \"Multiple Choice\"\n",
" TRUE_FALSE = \"True/False\"\n",
" FILL_IN_THE_BLANK = \"Fill in the Blank\"\n",
" SHORT_ANSWER = \"Short Answer\"\n",
"\n",
"class Question(BaseModel):\n",
" instruction: str\n",
" context: str\n",
" response: str\n",
" question_type: QuestionType\n",
" category: Category\n",
" difficulty: Difficulty\n",
"\n",
"class GeographyQuizDataset(BaseModel):\n",
" title: str = \"World Geography Challenge Dataset\"\n",
" description: str = \"Dataset for geography quiz questions and answers\"\n",
" questions: List[Question]"
]
},
{
"cell_type": "code",
"execution_count": null,
"id": "5fdLrBLx9vwL",
"metadata": {
"colab": {
"base_uri": "https://localhost:8080/"
},
"id": "5fdLrBLx9vwL",
"outputId": "f4e84aff-af6c-4075-e48f-7674d764ab06"
},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"{'instruction': \"What is the world's largest desert, covering over 9,000,000 square kilometers?\", 'context': 'Deserts are large areas of land that receive very little precipitation. The largest desert in the world is also the driest and hottest, covering much of North Africa.', 'response': 'Sahara', 'question_type': 'Multiple Choice', 'category': 'Landmarks', 'difficulty': 'Easy'}\n",
"Generated question 1/10: What is the world's largest desert, covering over 9,000,000 square kilometers?\n",
"{'instruction': 'Which African country has a coastline on the Mediterranean Sea and is home to the ancient city of Carthage?', 'context': 'Carthage was a major trading hub in the ancient world and is now a UNESCO World Heritage Site.', 'response': 'Tunisia', 'question_type': 'Multiple Choice', 'category': 'Countries', 'difficulty': 'Medium'}\n",
"Generated question 2/10: Which African country has a coastline on the Mediterranean Sea and is home to the ancient city of Carthage?\n",
"{'instruction': \"What is the world's longest river, which flows through 10 countries in northeastern Africa?\", 'context': 'This river is a primary source of water, transportation, and livelihood for millions of people in the region.', 'response': 'Nile', 'question_type': 'Multiple Choice', 'category': 'Rivers', 'difficulty': 'Easy'}\n",
"Generated question 3/10: What is the world's longest river, which flows through 10 countries in northeastern Africa?\n",
"{'instruction': 'Which river is the longest in South America?', 'context': 'The river in question is the largest river in the world by discharge volume, and flows through Brazil, Peru, and Colombia.', 'response': 'Amazon River', 'question_type': 'Multiple Choice', 'category': 'Rivers', 'difficulty': 'Medium'}\n",
"Generated question 4/10: Which river is the longest in South America?\n",
"{'instruction': 'Which river is the longest in South America and flows through Brazil, Peru, and Colombia before emptying into the Pacific Ocean?', 'context': 'Rivers of South America', 'response': 'Amazon River', 'question_type': 'Multiple Choice', 'category': 'Rivers', 'difficulty': 'Medium'}\n",
"Generated question 5/10: Which river is the longest in South America and flows through Brazil, Peru, and Colombia before emptying into the Pacific Ocean?\n",
"{'instruction': 'Which of the following rivers is the longest in South America?', 'context': 'The question requires the test-taker to identify the longest river in South America, which is a key geographical feature of the continent.', 'response': 'Amazon River', 'question_type': 'Multiple Choice', 'category': 'Rivers', 'difficulty': 'Medium'}\n",
"Generated question 6/10: Which of the following rivers is the longest in South America?\n",
"{'instruction': 'What is the longest river in South America?', 'context': 'This river flows through several countries in South America, including Brazil, Colombia, and Peru, before emptying into the Pacific Ocean.', 'response': 'Amazon River', 'question_type': 'Multiple Choice', 'category': 'Rivers', 'difficulty': 'Medium'}\n",
"Generated question 7/10: What is the longest river in South America?\n",
"{'instruction': \"What is the world's longest river, which flows through 11 countries in northeastern Africa?\", 'context': 'The river originates in Burundi and flows northwards through Rwanda and Tanzania before emptying into the Mediterranean Sea.', 'response': 'Nile', 'question_type': 'Multiple Choice', 'category': 'Rivers', 'difficulty': 'Easy'}\n",
"Generated question 8/10: What is the world's longest river, which flows through 11 countries in northeastern Africa?\n",
"{'instruction': 'What is the longest river in South America?', 'context': 'South America is home to many significant rivers, but one stands out as the longest.', 'response': 'Amazon River', 'question_type': 'Multiple Choice', 'category': 'Rivers', 'difficulty': 'Medium'}\n",
"Generated question 9/10: What is the longest river in South America?\n",
"{'instruction': 'Which river is the longest in South America?', 'context': 'The question requires the test-taker to identify the longest river in South America.', 'response': 'Amazon River', 'question_type': 'Multiple Choice', 'category': 'Rivers', 'difficulty': 'Medium'}\n",
"Generated question 10/10: Which river is the longest in South America?\n",
"Generated and saved 10 questions to geography_quiz_dataset.jsonl\n"
]
}
],
"source": [
"import json\n",
"def generate_question():\n",
" prompt = \"\"\"Generate a geography quiz question. Format your response as a JSON object with the following structure:\n",
" {\n",
" \"instruction\": \"The full question text\",\n",
" \"context\": \"Provide context about the question\",\n",
" \"response\": \"The correct answer\",\n",
" \"question_type\": \"The type of question (e.g., 'Multiple Choice')\",\n",
" \"category\": \"The category should be marked as one of these: Countries, Capitals, Rivers, Mountains, Landmarks, Climate, Culture\",\n",
" \"difficulty\": \"The difficulty level of the question (e.g., 'Easy')\"\n",
" }\"\"\"\n",
"\n",
" response = client.chat.completions.create(\n",
" model=\"accounts/fireworks/models/llama-v3p1-405b-instruct\",\n",
" response_format={\"type\": \"json_object\"},\n",
" messages=[\n",
" {\"role\": \"system\", \"content\": \"You are a geography expert creating quiz questions.\"},\n",
" {\"role\": \"user\", \"content\": prompt}\n",
" ]\n",
" )\n",
"\n",
" question_data = json.loads(response.choices[0].message.content)\n",
" print(question_data)\n",
" return Question(**question_data)\n",
"\n",
"def main(num_questions=10):\n",
" with open(\"geography_quiz_dataset.jsonl\", \"w\") as f:\n",
" for i in range(num_questions):\n",
" question = generate_question()\n",
" json.dump(question.dict(), f)\n",
" f.write(\"\\n\")\n",
" print(f\"Generated question {i+1}/{num_questions}: {question.instruction}\")\n",
"\n",
" print(f\"Generated and saved {num_questions} questions to geography_quiz_dataset.jsonl\")\n",
"\n",
"if __name__ == \"__main__\":\n",
" main()"
]
},
{
"cell_type": "markdown",
"id": "sDoxKlYwDORC",
"metadata": {
"id": "sDoxKlYwDORC"
},
"source": [
"## Conclusion\n",
"\n",
"We’re excited to see how the community leverages Llama 3.1 API to create interesting applications.\n",
"\n",
"\n",
"For more information and to get started with Llama 3.1, visit [docs.fireworks.ai](https://docs.fireworks.ai) or join our [discord community](https://discord.gg/fireworks-ai)"
]
}
],
"metadata": {
"colab": {
"provenance": [],
"toc_visible": true
},
"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": "3.11.3"
}
},
"nbformat": 4,
"nbformat_minor": 5
}

0 comments on commit 48d5d4b

Please sign in to comment.