mlabonne/chesspythia-70m
Partial fine-tune vs EleutherAI/pythia-70m-deduped. Partial change: 63% of the weights changed (attention, embeddings, mlp, norms), in 6 of 6 layers.
Tensors changed: Embeddings 1/2, Norms 25/26, Attention 24/24, MLP 24/24
Partial fine-tune vs EleutherAI/pythia-70m-deduped. Partial change: 63% of the weights changed (attention, embeddings, mlp, norms), in 6 of 6 layers.
Tensors changed: Embeddings 1/2, Norms 25/26, Attention 24/24, MLP 24/24