Squeal-Studio/squeal_ai_20m-instruct
Full fine-tune vs Squeal-Studio/squeal_ai_20m-base. Fine-tune: 100% of the weights changed, across 8 of 8 layers.
Tensors changed: Embeddings 1/1, Norms 17/17, MLP 24/24, Attention 56/56
Full fine-tune vs Squeal-Studio/squeal_ai_20m-base. Fine-tune: 100% of the weights changed, across 8 of 8 layers.
Tensors changed: Embeddings 1/1, Norms 17/17, MLP 24/24, Attention 56/56