OpenGVLab/InternVL3_5-30B-A3B-Instruct

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

Tensors changed: Norms 289/289, MLP 2369/2369, Attention 288/288, lm_head 1/1, Embeddings 5/5, other 48/48

1 derivatives checked

• OpenGVLab/InternVL3_5-30B-A3B-MPO: Full fine-tune

On the Hub · Downloads on Model Pulse