MashiroLn/Curious-VLA
Full fine-tune (fewer layers) vs Qwen/Qwen2.5-3B-Instruct. Qwen/Qwen2.5-3B-Instruct matches part of MashiroLn/Curious-VLA (433 of its 825 tensors; the rest of MashiroLn/Curious-VLA has no counterpart). On the shared part: Fine-tune: 100% of the weights changed, across 36 of 36 layers.
Tensors changed: Norms 73/73, MLP 108/108, Attention 252/252
2 derivatives checked
- mradermacher/Curious-VLA-i1-GGUF: Quantization (fewer layers)
- mradermacher/Curious-VLA-GGUF: Same weights, converted (fewer layers)