Skip to content

This issue was moved to a discussion.

You can continue the conversation there. Go to discussion →

✨[Feature] Add the ability to set breakpoints in the compilation process. #1256

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Closed
narendasan opened this issue Aug 12, 2022 · 2 comments
Closed
Assignees
Labels
component: conversion Issues re: Conversion stage component: partitioning feature request New feature or request

Comments

@narendasan
Copy link
Collaborator

Is your feature request related to a problem? Please describe.

Sometimes you want to pause or break compilation of a graph at a certain point to see what is happening at that operation. This can help figure out what is happening in the conversion process without having to recompile the entire library.

Describe the solution you'd like

At least for the conversion process it should be possible to specify a debugName of a Value and stop conversion and dump the state of the conversion context. This would include all arguments, the target function signature, and the node info.

Describe alternatives you've considered

People would need to add debug statements to the library themselves

Additional context

Related to #1235, #949

@narendasan narendasan added the feature request New feature or request label Aug 12, 2022
@narendasan narendasan self-assigned this Aug 12, 2022
@narendasan
Copy link
Collaborator Author

Similarly this could be helpful in partitioning @bowang007. Though we need to centralize the context first to do this I would think.

@narendasan
Copy link
Collaborator Author

narendasan commented Aug 12, 2022

I am thinking the way this would work is you would do something like:

torch_tensorrt.compile(module,
    inputs=...,
    breakpoints=["%32","@12"],
    dump_context=["%input.2", "@3"]
)

where breakpoints would interrupt conversion / partitioning and pause / maybe cancel conversion after that point. And context dumps would dump context at those ops onto the debug log.

@pytorch pytorch locked and limited conversation to collaborators Aug 12, 2022
@narendasan narendasan converted this issue into discussion #1257 Aug 12, 2022

This issue was moved to a discussion.

You can continue the conversation there. Go to discussion →

Labels
component: conversion Issues re: Conversion stage component: partitioning feature request New feature or request
Projects
None yet
Development

No branches or pull requests

3 participants