Skip to content

Conversation

cehongwang
Copy link
Collaborator

Description

Please include a summary of the change and which issue is fixed. Please also include relevant motivation and context. List any dependencies that are required for this change.

Fixes # (issue)

Type of change

Please delete options that are not relevant and/or add your own.

  • Bug fix (non-breaking change which fixes an issue)
  • New feature (non-breaking change which adds functionality)
  • Breaking change (fix or feature that would cause existing functionality to not work as expected)
  • This change requires a documentation update

Checklist:

  • My code follows the style guidelines of this project (You can use the linters)
  • I have performed a self-review of my own code
  • I have commented my code, particularly in hard-to-understand areas and hacks
  • I have made corresponding changes to the documentation
  • I have added tests to verify my fix or my feature
  • New and existing unit tests pass locally with my changes
  • I have added the relevant labels to my PR in so that relevant reviewers are notified

@meta-cla meta-cla bot added the cla signed label Sep 9, 2025
@github-actions github-actions bot added component: tests Issues re: Tests component: lowering Issues re: The lowering / preprocessing passes component: conversion Issues re: Conversion stage component: converters Issues re: Specific op converters component: api [Python] Issues re: Python API component: dynamo Issues relating to the `torch.compile` or `torch._dynamo.export` paths labels Sep 9, 2025
@github-actions github-actions bot requested a review from narendasan September 9, 2025 19:29
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Could you add detailed comments demonstrating what it does now that you've gone through the entire converter ? that would be helpful

model = model.to(torch.float32)

return model
return model.cuda()
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

model.cuda() is already done during initialization

# TODO: @Evan is waiting for TRT's feature to cache the weight-stripped engine
# if not self.compilation_settings.strip_engine_weights:
# # set EXCLUDE_WEIGHTS flag to strip weights
# runtime = trt.Runtime(TRT_LOGGER)
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

why are these being deleted?

Copy link
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Don't need to deserialize the engine anymore. The engine is already live

assert isinstance(
serialized_engine, bytes
), "Serialized engine must be a bytes object"
self.engine = serialized_engine
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

deserialize the engine here


return TRTInterpreterResult(
engine_str,
cuda_engine,
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

How will this work with the deferred engine setup?

Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Generally I think this is not clearly implemented. If we decided that PyRuntime will use initialized engines, then you should quickly reduce other cases to an initialized engine.

assert isinstance(
serialized_engine, bytes
), "Serialized engine must be a bytes object"
self.engine = serialized_engine
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Also do not overload the meaning of self.engine. Either it should be the serialized engine or the live engine not both


def setup_engine(self) -> None:

if isinstance(self.engine, trt.ICudaEngine):
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

We need to think about how this will interact with the greater system like the lazy_engine_init option.


if isinstance(self.engine, trt.ICudaEngine):
pass
elif isinstance(self.engine, bytes):
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

we should not check by type, have two members and a state flag


def _on_state_dict(self, state_dict: Dict[str, Any], prefix: str, _: Any) -> None:
state_dict[prefix + "engine"] = self.serialized_engine
state_dict[prefix + "engine"] = self.engine
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Same thing here, this should just be serialized engine

error_msgs: Any,
) -> None:
self.serialized_engine = state_dict[prefix + "engine"]
self.engine = state_dict[prefix + "engine"]
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

^^


def __init__(
self,
cuda_engine: Optional[trt.ICudaEngine | bytes] = None,
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Should not be a union. Its either you give a cuda_engine or a serialized engine

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
cla signed component: api [Python] Issues re: Python API component: conversion Issues re: Conversion stage component: converters Issues re: Specific op converters component: dynamo Issues relating to the `torch.compile` or `torch._dynamo.export` paths component: lowering Issues re: The lowering / preprocessing passes component: tests Issues re: Tests
Projects
None yet
Development

Successfully merging this pull request may close these issues.

3 participants