OpenGVLab/Mono-InternVL-2B
Full fine-tune (fewer layers) vs internlm/internlm2-chat-1_8b. internlm/internlm2-chat-1_8b matches part of OpenGVLab/Mono-InternVL-2B (169 of its 253 tensors; the rest of OpenGVLab/Mono-InternVL-2B has no counterpart). On the shared part: Fine-tune: 100% of the weights changed, across 24 of 24 layers.
Tensors changed: Attention 48/48, Norms 49/49, MLP 72/72