random1st/secguard-models

Fix MODEL_CARD placeholders (val loss, duration)

Update model card for v4 (Qwen3.5-0.8B, ~800MB Q8)

Update to Qwen3.5-0.8B fine-tuned v4 (rank 16, 1000 iter, fixed MLX→GGUF pipeline)

v3.1: F16 GGUF (Q8 had weight overflow)

Update model card: Qwen3.5-2B v3

v3: Qwen3.5-2B fine-tuned, LoRA, Q8_0 (val loss 0.394)

Update model card: Qwen3-0.6B fine-tuned v2

v2: Qwen3-0.6B fine-tuned on 21K labeled commands (LoRA, Q8_0)

Upload README.md with huggingface_hub

Upload secguard-guard.gguf with huggingface_hub

initial commit