tanabe1478/gyaim-lm-small-public-v1
Full fine-tune vs ku-nlp/gpt2-small-japanese-char. Fine-tune: 99% of the weights changed, across 12 of 12 layers.
Tensors changed: Attention 48/48, Norms 50/50, MLP 48/48, Embeddings 1/2
Full fine-tune vs ku-nlp/gpt2-small-japanese-char. Fine-tune: 99% of the weights changed, across 12 of 12 layers.
Tensors changed: Attention 48/48, Norms 50/50, MLP 48/48, Embeddings 1/2