GPT-51M_step_6668
A 51M-parameter GPT-style model trained from scratch by Tralalabs.
Training Data
- TinyStories (50%)
- OpenWebText (30%)
- Wikipedia 20231101.en (20%)
Details
- Context: 768 tokens
- Tokenizer: GPT-2 BPE
- Training steps: 6668
- ~60M tokens seen
Use
Best for storytelling, simple explanations, and structured text generation.
- Downloads last month
- 1
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support