DAMO-NLP-SG/VideoLLaMA3-2B-Image
Full fine-tune (fewer layers) vs Qwen/Qwen2.5-1.5B-Instruct. Qwen/Qwen2.5-1.5B-Instruct matches part of DAMO-NLP-SG/VideoLLaMA3-2B-Image (338 of its 778 tensors; the rest of DAMO-NLP-SG/VideoLLaMA3-2B-Image has no counterpart). On the shared part: Fine-tune: 100% of the weights changed, across 28 of 28 layers.
Tensors changed: Embeddings 1/1, Norms 57/57, MLP 84/84, Attention 196/196
1 derivatives checked
- DAMO-NLP-SG/VideoLLaMA3-2B: Full fine-tune