Qwen3.8-27B-MTP-Q4_0_ROCMFP4_STRIX_LEAN
Qwen3.8-27B — the dense hybrid-attention Qwen release (48 gated-delta-net layers + 16 full-attention layers, fullattentioninterval = 4, native...
Qwen3.8-27B — the dense hybrid-attention Qwen release (48 gated-delta-net layers + 16 full-attention layers, fullattentioninterval = 4, native...
Text-only ROCmFPX/ROCmFP4 GGUF builds of nightmedia/Qwen3.6-27B-Architect-Polaris2-Fable-B-F451, with the checkpoint’s embedded one-layer MTP draft model preserved. These are experimental...
> ## ⚠️ Эти файлы НЕ загружаются в стандартный llama.cpp > Они используют собственные тензорные типы AMD RCMFPX...