MC7ever/spark-darwin-5way
Full fine-tune vs Takenoko12345678/Spark-X2.5-4B-Japanese. Fine-tune: 100% of the weights changed, across 36 of 36 layers.
Tensors changed: Embeddings 1/1, Norms 5/73, MLP 108/108, Attention 108/108
Full fine-tune vs Takenoko12345678/Spark-X2.5-4B-Japanese. Fine-tune: 100% of the weights changed, across 36 of 36 layers.
Tensors changed: Embeddings 1/1, Norms 5/73, MLP 108/108, Attention 108/108