OpenGVLab/InternVL3_5-38B-MPO

Full fine-tune vs OpenGVLab/InternVL3_5-38B-Instruct. Fine-tune: 98% of the weights changed, across 64 of 64 layers.

Tensors changed: Embeddings 3/5, Norms 250/437, Attention 391/391, other 87/90, MLP 373/378, lm_head 1/1

2 derivatives checked

• OpenGVLab/InternVL3_5-38B: Partial fine-tune
• OpenGVLab/InternVL3_5-38B-HF: Full fine-tune

On the Hub · Downloads on Model Pulse