albertge/llada-8b-dllm-registers-mix60k-r4-corrected
Full fine-tune vs GSAI-ML/LLaDA-8B-Base. Fine-tune: 100% of the weights changed, across 32 of 32 layers.
Tensors changed: Attention 128/128, Norms 20/65, Embeddings 1/1, other 65/65, MLP 32/32
Full fine-tune vs GSAI-ML/LLaDA-8B-Base. Fine-tune: 100% of the weights changed, across 32 of 32 layers.
Tensors changed: Attention 128/128, Norms 20/65, Embeddings 1/1, other 65/65, MLP 32/32