utahnlp/tevatron-elastic-bert-retriever-depth
Full fine-tune vs google-bert/bert-base-uncased. Fine-tune: 99% of the weights changed, across 12 of 12 layers.
Tensors changed: Embeddings 4/5, Norms 24/48, Attention 96/96, other 48/50
Full fine-tune vs google-bert/bert-base-uncased. Fine-tune: 99% of the weights changed, across 12 of 12 layers.
Tensors changed: Embeddings 4/5, Norms 24/48, Attention 96/96, other 48/50