Huangsssz/SmolLM3-3B-MOPD-Top64-DR-NoThinking-500

Full fine-tune vs BytedTsinghua-SIA/Open-MOPD-SmolLM3-3B-MixSFT. Fine-tune: 91% of the weights changed, across 36 of 36 layers.

Tensors changed: Embeddings 0/1, Norms 10/73, MLP 108/108, Attention 144/144

On the Hub · Downloads on Model Pulse