halt95/Swift1.5-Qwen3.8-Flash-Next-W4A16-Merlin

Partial fine-tune vs halt95/Qwen3.8-Flash-Next-W4A16-Merlin. Partial change: 6% of the weights changed (attention, mlp), in 48 of 48 layers.

Tensors changed: Embeddings 0/138, Attention 259/824, MLP 144/1727, Norms 0/302, other 0/8, lm_head 0/1

On the Hub · Downloads on Model Pulse