v0.3.0.dev20260715011344
The GGUF load-source probe added in #467 reads model_config.quantization, but the AWQ lifecycle stubs predate that field. Set quantization=None so the dtype-alignment tests reach the AWQ checkpoints again. Tests: pytest tests/quant/ -m slow -v --tb=short (5 passed, 39 deselected) Fixes #507 Signed-off-by: PinJhih <isaacwang29@gmail.com>