tensorrt_llm/_torch/models/modeling_qwen3_moe.py (1)

420-427: 🧹 Nitpick | 🔵 Trivial | ⚡ Quick win

Add an explicit None return annotation to the renamed hook.

setup_aliases is a state-mutating hook and does not return a value.

Proposed fix

-    def setup_aliases(self):
+    def setup_aliases(self) -> None:

As per coding guidelines, “Always annotate functions. Make the return type None if the function does not return anything.”

🤖 Prompt for AI Agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

In `@tensorrt_llm/_torch/models/modeling_qwen3_moe.py` around lines 420 - 427, The
setup_aliases method is missing an explicit return type annotation. Add `->
None` to the method signature of setup_aliases to indicate that this
state-mutating hook does not return a value, as per coding guidelines requiring
all functions to have return type annotations.

Source: Coding guidelines

tensorrt_llm/_torch/models/modeling_gpt_oss.py (1)

634-645: 🧹 Nitpick | 🔵 Trivial | ⚡ Quick win

Add an explicit None return annotation to the renamed hook.

setup_aliases mutates alias fields and returns nothing; annotating it keeps the new lifecycle hook contract clear.

Proposed fix

-    def setup_aliases(self):
+    def setup_aliases(self) -> None:

As per coding guidelines, “Always annotate functions. Make the return type None if the function does not return anything.”

🤖 Prompt for AI Agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

In `@tensorrt_llm/_torch/models/modeling_gpt_oss.py` around lines 634 - 645, The
setup_aliases method lacks an explicit return type annotation. Add `-> None` to
the method signature after the closing parenthesis in the setup_aliases method
definition to indicate that this method mutates state but does not return any
value, keeping the function contract clear and consistent with coding guidelines
that require all functions to be annotated with their return types.

Source: Coding guidelines

tensorrt_llm/_torch/models/modeling_qwen3_next.py (1)

983-990: 🧹 Nitpick | 🔵 Trivial | ⚡ Quick win

Add an explicit None return annotation to the renamed hook.

setup_aliases only wires aliases, so the new lifecycle method should be annotated as returning None.

Proposed fix

-    def setup_aliases(self):
+    def setup_aliases(self) -> None:

As per coding guidelines, “Always annotate functions. Make the return type None if the function does not return anything.”

🤖 Prompt for AI Agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

In `@tensorrt_llm/_torch/models/modeling_qwen3_next.py` around lines 983 - 990,
The setup_aliases method in the Qwen3 model class lacks an explicit return type
annotation. Add a `-> None` return type annotation to the method signature of
setup_aliases to explicitly indicate that this method does not return any value,
as per the coding guidelines requiring all functions to have return type
annotations.

Source: Coding guidelines

tensorrt_llm/_torch/models/checkpoints/mx/checkpoint_loader.py (1)

201-212: 🧹 Nitpick | 🔵 Trivial | ⚡ Quick win

Add return annotations to the new MX lifecycle methods.

These methods expose boolean/side-effect lifecycle contracts, so annotate them explicitly.

Proposed fix

-    def is_post_transform_weights_preloaded(self) -> bool:
+    def is_post_transform_weights_preloaded(self) -> bool:
         """Whether the last successful MX preload delivered transformed bytes.
@@
-    ) -> None:
+    ) -> None:
         """Publish this instance's weights so other ranks can pull via P2P.
@@
-    ) -> None:
+    ) -> None:
         """Publish locally loaded weights as an MX source when appropriate.

As per coding guidelines, “Always annotate functions. Make the return type None if the function does not return anything.”

Also applies to: 561-567, 664-670

🤖 Prompt for AI Agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

In `@tensorrt_llm/_torch/models/checkpoints/mx/checkpoint_loader.py` around lines
201 - 212, Add explicit return type annotations to all three MX lifecycle
methods to comply with coding guidelines. The method
is_post_transform_weights_preloaded already shows the required -> bool
annotation in the diff, but ensure the other two methods at lines 561-567 and
664-670 also have their appropriate return type annotations added (these likely
also return bool based on the lifecycle contract pattern). Verify each method
has its return type explicitly annotated rather than relying on implicit type
inference.

Source: Coding guidelines

tensorrt_llm/_torch/modules/fused_moe/interface.py (1)

830-841: 🧹 Nitpick | 🔵 Trivial | ⚡ Quick win

Add explicit -> None annotations to the new lifecycle hooks.

These hooks do not return values, so annotate them explicitly.

As per coding guidelines, “Always annotate functions. Make the return type None if the function does not return anything.”

Proposed fix

-    def transform_weights(self):
+    def transform_weights(self) -> None:
         if getattr(self, "_weights_transformed", False):
             return
         self.quant_method.transform_weights(self)
         self._weights_transformed = True
 
-    def cache_derived_state(self):
+    def cache_derived_state(self) -> None:
         self.quant_method.cache_derived_state(self)
 
-    def post_load_weights(self):
+    def post_load_weights(self) -> None:
         self.transform_weights()
         self.cache_derived_state()

🤖 Prompt for AI Agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

In `@tensorrt_llm/_torch/modules/fused_moe/interface.py` around lines 830 - 841,
The methods transform_weights, cache_derived_state, and post_load_weights are
missing explicit return type annotations. Add `-> None` to the method signature
of each of these three methods since they do not return any values. This follows
the coding guideline that all functions must be annotated with their return
type, using `None` when the function does not return anything.

Source: Coding guidelines

tensorrt_llm/_torch/modules/fused_moe/quantization.py (1)

562-580: 🧹 Nitpick | 🔵 Trivial | ⚡ Quick win

Annotate the changed lifecycle hook signatures.

The new/renamed hooks should explicitly return None; Line 1015 should also type module consistently with the other hooks.

As per coding guidelines, “Always annotate functions. Make the return type None if the function does not return anything.”

Proposed signature updates

-    def transform_weights(self, module: torch.nn.Module):
+    def transform_weights(self, module: torch.nn.Module) -> None:

-    def cache_derived_state(self, module: torch.nn.Module):
+    def cache_derived_state(self, module: torch.nn.Module) -> None:

-    def post_load_weights(self, module: torch.nn.Module):
+    def post_load_weights(self, module: torch.nn.Module) -> None:

-    def transform_weights(self, module):
+    def transform_weights(self, module: torch.nn.Module) -> None:

Also applies to: 784-787, 1015-1016, 1280-1300, 3106-3111, 5351-5352

🤖 Prompt for AI Agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

In `@tensorrt_llm/_torch/modules/fused_moe/quantization.py` around lines 562 -
580, Add explicit return type annotations to the lifecycle hook methods
transform_weights, cache_derived_state, and post_load_weights by appending ->
None to their signatures. Additionally, ensure that the module parameter is
consistently typed as torch.nn.Module across all these hook methods and at the
other locations mentioned (784-787, 1015-1016, 1280-1300, 3106-3111, 5351-5352).
This follows the coding guideline that all functions must be annotated with
their return types, using None when the function does not return a value.

Source: Coding guidelines

Fix all unresolved CodeRabbit comments on this PR:

• Push a commit to this branch (recommended)
• Create a new PR with the fixes