metadata

title: Chat Generation
emoji: 🏃
colorFrom: gray
colorTo: yellow
sdk: gradio
sdk_version: 5.17.1
app_file: app.py
pinned: false

Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference

AI Kids' Story Generator - English & Arabic

Project Objectives

The AI Kids' Story Generator aims to create engaging and imaginative short stories for children. Users can input a character name and a setting, and the model will generate a fun story in either English or Arabic. This project leverages advanced natural language processing techniques to provide a creative tool for storytelling.

Implemented Pipelines

The project uses the following pipelines from the transformers library:

Text Generation Pipeline:
- Model: The model adapts based on the selected language. For English, it uses the microsoft/Phi-3-mini-4k-instruct model, while for Arabic, it employs the ALLaM-AI/ALLaM-7B-Instruct-preview model.
- The pipeline is configured for text generation with parameters optimized for creative outputs, including:
  - max_new_tokens: Limits the length of the generated story.
  - do_sample: Enables sampling to enhance creativity.
  - top_k and top_p: Control the diversity of generated text.

Instructions for Using the Interface

Input Fields:
- Character Name: Enter a name for the story's main character.
- Setting: Specify the story's setting (e.g., a magical forest).
- Choose Language: Select the desired language for the story (English or Arabic).
Generate Story: Click the "Submit" button to generate the story. The output will be displayed in the "Kids' Story" textbox.
Examples: The interface includes examples you can use to see how the app functions. Simply click on an example to populate the fields.

Justifications for Model and Pipeline Choices

Model Selection: The chosen models are specifically designed for instruction-following tasks, making them suitable for generating narratives. The English model (microsoft/Phi-3-mini-4k-instruct) has been optimized for various textual tasks, while the Arabic model (ALLaM-AI/ALLaM-7B-Instruct-preview) is tailored for Arabic language generation.
Pipeline Configuration: The parameters for the text generation pipeline were chosen to encourage creativity and engagement, essential for children's stories.

Bilingual Implementation

The project supports both English and Arabic languages, allowing users to receive stories in their preferred language. Each model is specifically chosen to handle the linguistic nuances of its respective language, ensuring that the output is both coherent and culturally relevant. The interface dynamically adjusts based on user input, providing a seamless bilingual experience.

Feel free to customize this README file further to meet your project's specific needs or to add any additional information you find relevant!