lmms-lab/llava-critic-7b
Full fine-tune vs lmms-lab/llava-onevision-qwen2-7b-ov. Fine-tune: 93% of the weights changed, across 28 of 28 layers.
Tensors changed: Embeddings 3/4, other 5/5, Norms 138/163, MLP 185/188, Attention 404/404, lm_head 1/1
Full fine-tune vs lmms-lab/llava-onevision-qwen2-7b-ov. Fine-tune: 93% of the weights changed, across 28 of 28 layers.
Tensors changed: Embeddings 3/4, other 5/5, Norms 138/163, MLP 185/188, Attention 404/404, lm_head 1/1