Moeblack/Qwen3.8-27B-chinese-taste-sft-lora

model card v11: training env note (Pro6000 1xGPU ~20h)

model card v10: pre-converted GGUF usage

add pre-converted GGUF (q8_0) for llama.cpp

model card v9: simplify sequence description

model card v8: drop style-assimilation limitation line

model card v7: sequence-level granularity wording

model card v6: drop rank limitation mention

model card v5: de-emphasize corpus privacy, prompt combination upper bound

model card v4: qwen-style wording + agent positioning + drop details

model card: agent model wording (no prompt recipe)

chinese-taste SFT LoRA v2: adapter + model card + sample comparison

initial commit