MicroLM Shakespeare 1.8M
A custom, character-level causal Transformer trained entirely from scratch on CPU using pure PyTorch.
Architecture
- Parameters: ~1.81M
- Context Length: 128 tokens
- Layers: 4
- Attention Heads: 4
- Embedding Dimension: 192
- Vocabulary: 65 characters (Tiny Shakespeare)
Usage
Download chat.py, vocab.json, and micro_lm_weights.pt to run text generation locally:
python chat.py
Inference Providers NEW
This model isn't deployed by any Inference Provider. ๐ Ask for provider support