Fix --quantization fp8-scaled-mm CLI crash

[garrick99]

Summary

Fixes #146 — --quantization fp8-scaled-mm crashes with TypeError: QuantizationPolicy.fp8_scaled_mm() takes 1 positional argument but 2 were given.

Cause

args.py was passing amax_path to QuantizationPolicy.fp8_scaled_mm(), but the classmethod takes no arguments.

Fix

Match the pattern already used by fp8-cast: reject extra arguments and call without args.

 elif policy_name == "fp8-scaled-mm":
-    amax_path = resolve_path(values[1]) if len(values) > 1 else None
-    policy = QuantizationPolicy.fp8_scaled_mm(amax_path)
+    if len(values) > 1:
+        msg = f"{option_string} fp8-scaled-mm does not accept additional arguments"
+        raise argparse.ArgumentError(self, msg)
+    policy = QuantizationPolicy.fp8_scaled_mm()

🤖 Generated with Claude Code