amwangfan/privacy-gateway-v4-qwen2.5-0.5b

Clarify three-layer split: weak passwords are not hard-ruled

Bilingual zh/en model card (step-220, FPR-only probe)

Replace Q8 GGUF with step-220 merge

Replace F16 GGUF with step-220 merge

step-220 LoRA: FPR-only probe card, no training jsonl

Rename Hugging Face namespace venti1888 -> amwangfan

Combined v4 LoRA+GGUF model card

Add v4 Q8_0 GGUF

Add v4 F16 GGUF

v4 LoRA adapter + card

initial commit