albertge/dream-7b-dllm-registers-mix60k-t4
Full fine-tune vs Dream-org/Dream-v0-Base-7B. Fine-tune: 93% of the weights changed, across 28 of 28 layers.
Tensors changed: Embeddings 0/1, Norms 33/57, MLP 84/84, Attention 195/196, lm_head 1/1
Full fine-tune vs Dream-org/Dream-v0-Base-7B. Fine-tune: 93% of the weights changed, across 28 of 28 layers.
Tensors changed: Embeddings 0/1, Norms 33/57, MLP 84/84, Attention 195/196, lm_head 1/1