kulibinai/cadevolve-rl1

Full fine-tune vs Qwen/Qwen2-VL-2B-Instruct. Fine-tune: 100% of the weights changed, across 32 of 32 layers.

Tensors changed: Embeddings 2/2, Norms 144/187, MLP 216/216, Attention 324/324

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

On the Hub · Downloads on Model Pulse