OpenGVLab/InternVL3_5-30B-A3B-Flash

Full fine-tune vs OpenGVLab/InternVL3_5-30B-A3B. Fine-tune: 100% of the weights changed, across 48 of 48 layers.

Tensors changed: MLP 2308/2308, Norms 283/289, Attention 288/288, other 48/48, Embeddings 5/5, lm_head 1/1

On the Hub · Downloads on Model Pulse