Commit History
26 days ago
-
Add zero-shot benchmark results (ARC, HellaSwag, SciQ, PIQA) and full-split val ppl (#1)
- Add zero-shot benchmark results (ARC, HellaSwag, SciQ, PIQA) and full-split val ppl (db5e6912758b7309daa4df401536e28af1f59daf)0f4e319 -
BPE-8k tokenizer (HF tokenizers format)
Verified signature6d8613e -
weights: 24.59M params, 75 tensors, float32
Verified signatured02797d -
TinyStories-24m: 24.59M BPE GPT trained from scratch on TinyStories (val ppl 8.76, coherent)
Verified signaturec6c9cdf -
initial commit
Verified signaturecbe0202