lmms-lab/LLaVA-Video-7B-Qwen2

Full fine-tune vs lmms-lab/llava-onevision-qwen2-7b-si. Fine-tune: 100% of the weights changed, across 28 of 28 layers.

Tensors changed: Embeddings 4/4, other 5/5, Norms 139/163, MLP 188/188, Attention 404/404, lm_head 1/1

2 other repos hold byte-for-byte identical weight files.

On the Hub · Downloads on Model Pulse