antoinelouis/splade-max-camembert-base-mmarcoFR
Full fine-tune vs almanach/camembert-base. Fine-tune: 100% of the weights changed, across 12 of 12 layers.
Tensors changed: lm_head 2/5, Embeddings 3/5, Norms 0/48, Attention 48/96, other 24/48
Full fine-tune vs almanach/camembert-base. Fine-tune: 100% of the weights changed, across 12 of 12 layers.
Tensors changed: lm_head 2/5, Embeddings 3/5, Norms 0/48, Attention 48/96, other 24/48