metadata
tags:
- gpt
- from-scratch
- code-generation
gpt-python-124m
A ~124M parameter GPT-2-style decoder-only transformer, trained from scratch on the devoppro/Python dataset.
- Layers: 12
- Heads: 12
- Embedding dim: 768
- Context length: 512
- Trained to step 0, val loss 10.3904
- Tokenizer: GPT-2 BPE (tiktoken 'gpt2')