nvidia/EGM-4B
Full fine-tune vs Qwen/Qwen3-VL-4B-Thinking. Fine-tune: 100% of the weights changed, across 36 of 36 layers.
Tensors changed: Embeddings 4/4, Norms 196/249, MLP 214/220, Attention 240/240
Full fine-tune vs Qwen/Qwen3-VL-4B-Thinking. Fine-tune: 100% of the weights changed, across 36 of 36 layers.
Tensors changed: Embeddings 4/4, Norms 196/249, MLP 214/220, Attention 240/240