nvidia/EGM-8B-SFT

Partial fine-tune vs Qwen/Qwen3-VL-8B-Thinking. Partial change: 87% of the weights changed (attention, lm_head, mlp, norms), in 36 of 36 layers.

Tensors changed: Embeddings 0/4, Norms 127/261, MLP 112/232, Attention 144/252, lm_head 1/1

1 other repos hold byte-for-byte identical weight files.

On the Hub · Downloads on Model Pulse