README.md

metadata

title: Veda Programming LLM
emoji: 🕉️
colorFrom: purple
colorTo: blue
sdk: gradio
sdk_version: 3.50.0
app_file: app.py
pinned: false
license: mit

🕉️ Veda Programming LLM

A TensorFlow-based Large Language Model for programming code generation.

Features

• Code Generation: Generate Python code from prompts
• Custom Training: Train on your own code samples
• Transformer Architecture: Uses modern transformer blocks
• Interactive Interface: Easy-to-use Gradio interface

Model Architecture

• Transformer-based decoder architecture
• Configurable model sizes (small/medium/large)
• Causal attention masking for autoregressive generation
• Custom tokenizer optimized for code

Usage

1. Generate Code: Enter a code prompt and adjust generation parameters
2. Train Model: Paste your code samples and train the model
3. View Model Info: Check model architecture and parameters

Parameters

• Temperature: Controls randomness (lower = more deterministic)
• Top-K: Limits sampling to top K tokens
• Top-P: Nucleus sampling threshold
• Max Tokens: Maximum number of tokens to generate

Training Data

The model can be trained on programming.txt containing Python code samples.

License

MIT License