hf_ptq.py: deprecated --auto_quantize_bits CLI flag silently no-ops instead of enabling AutoQuantize

[wyattearp]

Description

examples/hf_ptq/hf_ptq.py's deprecated --auto_quantize_bits CLI flag (and its siblings --auto_quantize_method, --auto_quantize_score_size, --auto_quantize_cost_model, --auto_quantize_active_moe_expert_ratio) is documented as still enabling the AutoQuantize CLI path via a conversion shim:

# Deprecated AutoQuantize CLI flags: kept as a backward-compat shim that converts them into an
# AutoQuantizeConfig on the fly (see _auto_quantize_config_from_cli). Prefer --recipe. The old
# CLI also lives on the 0.45 branch.
parser.add_argument(
    "--auto_quantize_bits",
    ...
)

_auto_quantize_config_from_cli does not exist anywhere in the file (or the repo). quantize_main() only enables AutoQuantize by checking whether --recipe resolves to a ModelOptAutoQuantizeRecipe; args.auto_quantize_bits is never read in the actual quantization path.

Effect: a legacy command still relying on --auto_quantize_bits (e.g. carried over from the 0.45 branch's CLI) silently runs plain PTQ instead of AutoQuantize, with no warning or error pointing back at the cause.

Fix

Fixed in #2129 by rejecting --auto_quantize_bits at parse_args() instead of letting it silently no-op, pointing users at an AutoQuantize --recipe.

Found via a CodeRabbit review comment on that PR.

[wyattearp]

Fixed by #2129