Skip to content
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.

Sign up for GitHub

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

Jump to bottom

chore: [NCCL] reorg and better error messages #3338

Open
wants to merge 1 commit into
base: nccl_ops_multi_gpu
Choose a base branch
from
Open

chore: [NCCL] reorg and better error messages #3338

wants to merge 1 commit into from

Conversation

narendasan
Copy link
Collaborator

Description

Precommit was clearly not run, assorted fixes and reorg. Some AIs for @apbose re: custom ops and some hardcoded special cases

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

@github-actions github-actions bot added component: lowering Issues re: The lowering / preprocessing passes component: conversion Issues re: Conversion stage component: converters Issues re: Specific op converters component: build system Issues re: Build system component: api [Python] Issues re: Python API component: dynamo Issues relating to the `torch.compile` or `torch._dynamo.export` paths labels Dec 25, 2024
@github-actions github-actions bot requested a review from zewenli98 December 25, 2024 00:18
@narendasan narendasan changed the title chore: reorg and better error messages chore: [NCCL] reorg and better error messages Dec 25, 2024
apbose
apbose reviewed Dec 25, 2024
View reviewed changes
py/torch_tensorrt/dynamo/conversion/impl/select.py
@@ -25,6 +23,8 @@
)
from torch_tensorrt.fx.types import TRTTensor

import tensorrt as trt

Copy link
Collaborator

Choose a reason for hiding this comment

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

Why is this change here? Is this a pre-commit change?

Copy link
Collaborator Author

Choose a reason for hiding this comment

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

Precommit

py/torch_tensorrt/dynamo/lowering/passes/fuse_distributed_ops.py
@@ -12,16 +12,23 @@
logger = logging.getLogger(__name__)


def tensorrt_fused_nccl_all_gather_op(args0, args1, args2):
# TODO: @apbose make these actual torch custom ops, should allow aot_export
def tensorrt_fused_nccl_all_gather_op(
Copy link
Collaborator

Choose a reason for hiding this comment

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

Ok noted in the TODOs. Just to get clarity, what is meant by allowing aot_export? What is the difference in the present behavior?

Copy link
Collaborator Author

Choose a reason for hiding this comment

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

Sonthat you dont need to use the autograd workflow and could potentially support torch.export flows

pyproject.toml
prerelease = "if-necessary-or-explicit"

index-strategy = "unsafe-best-match" # Needed for TRT-LLM

Copy link
Collaborator

Choose a reason for hiding this comment

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

Why is this added? What is the significance?

Copy link
Collaborator Author

Choose a reason for hiding this comment

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

This is for uv to install Tensorrt llm

@apbose
Copy link
Collaborator

apbose commented Dec 25, 2024

Wanted to get clarity on- precommit was clearly not run. I generally run pre-commit run, and push in the modified changes and cross check if the CI Lint check passes. Should I be mindful of other things before pushing in?
Also what are the hardcoded changes?
Have some review comments/questions, else the PR changes look good to me.

apbose
apbose reviewed Dec 25, 2024
View reviewed changes
pyproject.toml
monitoring-tools = ["rich>=13.7.1"]
jupyter = ["rich[jupyter]>=13.7.1"]
distributed = ["tensorrt-llm>=0.16.0"]

Copy link
Collaborator

Choose a reason for hiding this comment

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

How do we ensure this goes through when we have a torchTRT container having python version different than 3.10. Won't this cause an issue here?

Copy link
Collaborator Author

Choose a reason for hiding this comment

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

Its an extra so it wont effect install

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers

@apbose apbose apbose approved these changes

@zewenli98 zewenli98 Awaiting requested review from zewenli98

Assignees
No one assigned
Labels
cla signed component: api [Python] Issues re: Python API component: build system Issues re: Build system 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
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.
3 participants
@narendasan @apbose @facebook-github-bot