manycore-research/SpatialLM1.1-Qwen-0.5B
Full fine-tune (fewer layers) vs Qwen/Qwen2.5-0.5B-Instruct. Qwen/Qwen2.5-0.5B-Instruct matches part of manycore-research/SpatialLM1.1-Qwen-0.5B (290 of its 748 tensors; the rest of manycore-research/SpatialLM1.1-Qwen-0.5B has no counterpart). On the shared part: Fine-tune: 100% of the weights changed, across 24 of 24 layers.
Tensors changed: Embeddings 1/1, Norms 49/49, MLP 72/72, Attention 168/168