OpenGVLab/InternVL3_5-1B-MPO
Partial fine-tune vs OpenGVLab/InternVL3_5-1B-Instruct. Partial change: 85% of the weights changed (attention, embeddings, lm_head, mlp, norms), in 28 of 28 layers.
Tensors changed: lm_head 1/1, Embeddings 4/5, Norms 123/209, MLP 179/186, Attention 208/208, other 0/48
3 derivatives checked
- OpenGVLab/InternVL3_5-1B: Partial fine-tune
- OpenGVLab/InternVL3_5-1B-HF: Different architecture
- mlx-community/InternVL3_5-1B-4bit: Quantization