YuankaiLuo/SimVLA-LIBERO
Partial fine-tune (fewer layers) vs HuggingFaceTB/SmolVLM-500M-Instruct. HuggingFaceTB/SmolVLM-500M-Instruct matches part of YuankaiLuo/SimVLA-LIBERO (489 of its 786 tensors; the rest of YuankaiLuo/SimVLA-LIBERO has no counterpart). On the shared part: Partial change: 81% of the weights changed (attention, embeddings, mlp, norms, other), in 32 of 32 layers.
Tensors changed: lm_head 0/1, other 1/1, Embeddings 2/4, Norms 115/115, MLP 144/144, Attention 224/224