lovedheart/Qwen-AgentWorld-35B-A3B-FP8
Full fine-tune vs Qwen/Qwen-AgentWorld-35B-A3B. Fine-tune: 91% of the weights changed, across 39 of 40 layers.
Tensors changed: MLP 78/280, Attention 0/280, Norms 0/131, Embeddings 0/1, lm_head 0/1
Full fine-tune vs Qwen/Qwen-AgentWorld-35B-A3B. Fine-tune: 91% of the weights changed, across 39 of 40 layers.
Tensors changed: MLP 78/280, Attention 0/280, Norms 0/131, Embeddings 0/1, lm_head 0/1