V0.2: r=128 injected graph (non-destructive, reuses model_q4.onnx_data)

V0.2a: injected r=128 graph under fresh name (cache-bust the wrong pristine push)

V0.2 FIX: push the INJECTED r=128 graph (adapter.data slots) — the previous push was the pristine base (LoRA never applied)

V0.2: r=128 injected graph (non-destructive, reuses model_q4.onnx_data)

Upload onnx/noop.data with huggingface_hub

Upload folder using huggingface_hub

initial commit