todiadiyatmo/Qwen3.8-Flash-Next-W4A16-Attn8-FP8PLE

Full fine-tune vs Intel/Qwen3.8-Flash-Next-W4A16-AutoRound. Fine-tune: 95% of the weights changed, across 1 of 48 layers.

Tensors changed: Embeddings 128/136, MLP 0/1440, Attention 0/447, Norms 0/302, other 0/7, lm_head 0/1

On the Hub · Downloads on Model Pulse