open-athena/Snowball-67B-A2B-5.7T-Mixed-RLVR-Step38
Full fine-tune vs laion/snowball-67b-a2b-sft-s3-nemotron-terminal-step1888. Fine-tune: 100% of the weights changed, across 26 of 26 layers.
Tensors changed: lm_head 1/1, Embeddings 3/4, Norms 106/159, MLP 182/208, Attention 130/130