koyeb/Apple-MLX-QA

---
language:
- en
license: mit
size_categories:
- 1K<n<10K
task_categories:
- question-answering
pretty_name: Apple MLX Documentation Question/Answer
dataset_info:
features:
- name: question
dtype: string
- name: answer
dtype: string
- name: chunk
dtype: string
splits:
- name: train
num_bytes: 2599338.9379084967
num_examples: 1162
- name: test
num_bytes: 138691.06209150326
num_examples: 62
download_size: 1015973
dataset_size: 2738030.0
configs:
- config_name: default
data_files:
- split: train
path: data/train-*
- split: test
path: data/test-*
---

# Apple MLX Documentation QA Dataset

## Overview

This dataset contains question-answer pairs generated from the official documentation of Apple's latest machine learning framework, MLX. The dataset was created to facilitate the development and evaluation of models designed for natural language understanding, specifically in the context of technical documentation.

## Dataset Description

The dataset consists of three columns:
- **question**: A question generated from a specific chunk of the MLX documentation.
- **answer**: The corresponding answer to the question, also based on the same chunk of documentation.
- **chunk**: The original snippet (or chunk) from the MLX documentation that was used to generate both the question and the answer.

## Data Generation Process

The questions and answers in this dataset were generated using OpenAI's GPT-4o. The process involved two main steps:
1. **Question Generation**: GPT-4 was prompted to generate questions based on specific chunks of the MLX documentation.
2. **Answer Generation**: GPT-4 was then asked to answer the questions using the content of the same documentation chunk.

### System Prompts

Below are the system prompts used for question and answer generation.

- **Question Generation:**

System Prompt:

```plaintext
You are a helpful AI assistant. Your task is to help a user understand how to use functions and classes from Apple's Deep Learning framework, MLX. Carefully examine the function documentation snippet and generate 3 questions a medium to experienced MLX user could ask. Questions must be answerable from the information in the snippet. Do not assume anything about MLX's API that is not discussed in the snippet. If the snippet is too short or contains too little information, output an empty JSON array.
```

OpenAI Structured Outputs Schema:
```json
{
"name": "questions",
"strict": true,
"schema": {
"type": "object",
"properties": {
"questions": {
"type": "array",
"items": {
"type": "string"
}
}
},
"required": ["questions"],
"additionalProperties": false
}
}
```

- **Answer Generation:**

System Prompt:

```plaintext
You are a helpful AI assistant. Your task is to help a user understand how to use functions and classes from Apple's Deep Learning framework, MLX. Carefully examine the function documentation and generate an explanatory response based on the user's question which showcases usage and examples. Do not assume anything about MLX's API that is not discussed in the reference documentation snippet.
```

## Usage

This dataset is particularly useful for:
- Training question-answering models on technical documentation.
- Evaluating the ability to inject up-to-date information in the post-training phase of LLMs.

Please feel free to explore and make use of the dataset in your projects.