fix model precision for qdq

[xieofxie]

No description provided.

[DingmaomaoBJTU]

Code review

Found 1 issue:

1. Stale _get_precision docstring: the PR changes _label() to always return w{w_bits}a{a_bits} format, so "int4", "int8", "int16" can no longer be returned for QDQ paths. However, the docstring still lists them as possible return values. The io_config property docstring also references "int8" as an example.

winml-cli/src/winml/modelkit/session/session.py

Lines 613 to 617 in 6d49466

	Returns one of: ``"fp32"``, ``"fp16"``, ``"bf16"``, ``"int4"``,
	``"int8"``, ``"int16"``, ``"w{w}a{a}"`` (mixed), or ``None`` when
	no signal can be derived.

🤖 Generated with Claude Code

_{- If this code review was useful, please react with 👍. Otherwise, react with 👎.}

[xieofxie]

Code review

Found 1 issue:

1. Stale _get_precision docstring: the PR changes _label() to always return w{w_bits}a{a_bits} format, so "int4", "int8", "int16" can no longer be returned for QDQ paths. However, the docstring still lists them as possible return values. The io_config property docstring also references "int8" as an example.

winml-cli/src/winml/modelkit/session/session.py

Lines 613 to 617 in 6d49466

	Returns one of: ``"fp32"``, ``"fp16"``, ``"bf16"``, ``"int4"``,
	``"int8"``, ``"int16"``, ``"w{w}a{a}"`` (mixed), or ``None`` when
	no signal can be derived.

🤖 Generated with Claude Code

• If this code review was useful, please react with 👍. Otherwise, react with 👎.

Will fix later