Fix IntermediateTensors initialization and add type hints

[OthmanMohammad]

Purpose

Add missing type hints and fix field initialization bug in IntermediateTensors.__init__ method in vllm/sequence.py.

Changes:

1. Added type hint for tensors parameter: dict[str, torch.Tensor]
2. Added type hint for kv_connector_output parameter with default value None
3. Fixed bug where kv_connector_output field was not initialized in custom __init__, which could cause AttributeError at runtime

Changes

• File: vllm/sequence.py:63-68
• Added: Type hints for both tensors and kv_connector_output parameters
• Added: Return type annotation -> None
• Fixed: Proper initialization of kv_connector_output field in method body

Benefits

• Improves type safety and IDE autocomplete support
• Makes parameter types explicit and consistent with class definition
• Fixes potential runtime bug where kv_connector_output could be accessed before initialization
• Helps static type checkers catch potential bugs

Test Plan

# Verify Python syntax is valid
python -m py_compile vllm/sequence.py

# Verify the class can be instantiated correctly
python -c "from vllm.sequence import IntermediateTensors; import torch; obj = IntermediateTensors({}); print(obj.kv_connector_output)"

Test Result

✅ File compiles successfully with correct type annotations. ✅ Class instantiates properly with all fields initialized. ✅ kv_connector_output defaults to None when not provided.

Essential Elements of an Effective PR Description Checklist

• The purpose of the PR, such as "Fix some issue (link existing issues this PR will resolve)".
• The test plan, such as providing test command.
• The test results, such as pasting the results comparison before and after, or e2e results
• (Optional) The necessary documentation update, such as updating supported_models.md and examples for a new model.
• (Optional) Release notes update. If your change is user facing, please update the release notes draft in the Google Doc.

[gemini-code-assist]

Code Review

This pull request adds type hints to the __init__ method of the IntermediateTensors dataclass, which is a good improvement for type safety and code clarity. However, I noticed that the custom __init__ method doesn't initialize the kv_connector_output field defined in the class. This could lead to AttributeError exceptions at runtime. I've provided a suggestion to update the __init__ method to ensure all class fields are properly initialized, making the class more robust.

[gemini-code-assist]

The custom __init__ method does not initialize the kv_connector_output field. Since IntermediateTensors is a dataclass with a custom __init__, it is responsible for initializing all fields that do not have a default value. Currently, kv_connector_output is not initialized upon object creation, which will cause an AttributeError if it's accessed before being explicitly set elsewhere.

To fix this, you should initialize it within __init__, likely with a default value of None. This requires updating both the method signature and its body.

I've suggested a change for the signature below. You will also need to add self.kv_connector_output = kv_connector_output inside the method body to complete the fix.

Suggested change

	def __init__(self, tensors: dict[str, torch.Tensor]) -> None:
	def __init__(self, tensors: dict[str, torch.Tensor], kv_connector_output: KVConnectorOutput | None = None) -> None:

[OthmanMohammad]

@gemini-code-assist I've updated the __init__ method to initialize the kv_connector_output field. The signature now includes the parameter with a default value of None, and the field is properly initialized in the method body.

[gemini-code-assist]

Thanks for the update, @OthmanMohammad! I've reviewed the changes to the __init__ method in vllm/sequence.py, and they correctly include the kv_connector_output parameter with a default value of None and proper initialization. This looks good!