NeuML/colbert-muvera-micro
Full fine-tune vs google/bert_uncased_L-2_H-128_A-2. Fine-tune: 97% of the weights changed, across 2 of 2 layers.
Tensors changed: Embeddings 4/4, Norms 8/8, Attention 16/16, other 8/10
Full fine-tune vs google/bert_uncased_L-2_H-128_A-2. Fine-tune: 97% of the weights changed, across 2 of 2 layers.
Tensors changed: Embeddings 4/4, Norms 8/8, Attention 16/16, other 8/10