DAMO-NLP-SG/VideoLLaMA3-7B
Full fine-tune vs DAMO-NLP-SG/VideoLLaMA3-7B-Image. Fine-tune: 93% of the weights changed, across 28 of 28 layers.
Tensors changed: Embeddings 2/3, Norms 144/167, MLP 191/192, Attention 412/412, lm_head 1/1, other 4/4
Full fine-tune vs DAMO-NLP-SG/VideoLLaMA3-7B-Image. Fine-tune: 93% of the weights changed, across 28 of 28 layers.
Tensors changed: Embeddings 2/3, Norms 144/167, MLP 191/192, Attention 412/412, lm_head 1/1, other 4/4