sewoong/korean-neural-sparse-encoder-base-klue-large
Full fine-tune vs klue/roberta-large. Fine-tune: 100% of the weights changed, across 24 of 24 layers.
Tensors changed: lm_head 5/5, Embeddings 5/5, Norms 96/96, Attention 192/192, other 96/96
Full fine-tune vs klue/roberta-large. Fine-tune: 100% of the weights changed, across 24 of 24 layers.
Tensors changed: lm_head 5/5, Embeddings 5/5, Norms 96/96, Attention 192/192, other 96/96