Onely7/llava-1.5-llm-jp-3.1-1.8b-instruct4
Partial fine-tune (fewer layers) vs llm-jp/llm-jp-3.1-1.8b-instruct4. llm-jp/llm-jp-3.1-1.8b-instruct4 matches part of Onely7/llava-1.5-llm-jp-3.1-1.8b-instruct4 (219 of its 614 tensors; the rest of Onely7/llava-1.5-llm-jp-3.1-1.8b-instruct4 has no counterpart). On the shared part: Partial change: 89% of the weights changed (attention, lm_head, mlp, norms), in 24 of 24 layers.
Tensors changed: lm_head 1/1, Embeddings 0/1, Norms 28/49, MLP 72/72, Attention 96/96