nineninesix/qwen3_5-full-attn-only-14

Partial fine-tune vs Qwen/Qwen3.5-0.8B-Base. Partial change: 41% of the weights changed (attention, mlp, norms), in 14 of 14 layers.

Tensors changed: Embeddings 0/1, Norms 35/35, MLP 42/42, Attention 12/12

1 derivatives checked

• nineninesix/gepard-1.0: Full fine-tune (fewer layers)

On the Hub · Downloads on Model Pulse