Saren/Qwen3.8-Flash-Next-W4A16-AutoRound-hybrid
LoRA-style vs Intel/Qwen3.8-Flash-Next-W4A16-AutoRound. Looks like a merged LoRA (or a fine-tune with frozen norms): only the weight matrices of down_proj, gate_proj, in_proj_qkv, in_proj_z, k_proj, o_proj, out_proj, q_proj, up_proj, v_proj changed, in 48 of 48 layers; norms, biases and embeddings are untouched.
Tensors changed: Embeddings 0/9, MLP 144/1770, Attention 156/608, Norms 0/302, other 0/8
1 derivatives checked
- Saren/Qwen3.8-Flash-Next-W4A16-AutoRound-hybrid-MTP_int4RTN: Same weights, repacked