abi96062/smolLM2-135-implementation

SmolLM2-135M: Complete From-Scratch Implementation

This Space demonstrates a complete reverse-engineered implementation of SmolLM2-135M.

Features

• —🔍 Reverse-engineered architecture
• —🏋️ Trained for 5,000+ steps
• —✅ Checkpoint validation
• —⚡ Optimized with Flash Attention & Mixed Precision

Links

• —GitHub Repository: abi2024/smollm2-135-implementation
• —Model Details: See the Model Info tab

Usage

Enter a prompt and adjust generation parameters to see the model in action!