r0b0tlab/Qwen3.8-27B-DFlash2-EXL3-4.00bpw
Same weights, converted vs incoai/Qwen3.8-27B-DFlash2. Same weights stored in another dtype (BF16 → I16/F16), plus 146 new tensors.
Tensors changed: other 0/1, Norms 0/22, Attention 0/10, MLP 0/10
Same weights, converted vs incoai/Qwen3.8-27B-DFlash2. Same weights stored in another dtype (BF16 → I16/F16), plus 146 new tensors.
Tensors changed: other 0/1, Norms 0/22, Attention 0/10, MLP 0/10