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.

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

Zpx01/dit advanced training #11226

Open
wants to merge 4 commits into
base: main
Choose a base branch
from

Conversation

zpx01
Copy link
Contributor

@zpx01 zpx01 commented Nov 8, 2024

What does this PR do ?

Adds the following new features to the Diffusion Training Framework:

  • FSDP support
  • MovieGen 1B/5B/30B Stages 1-3 Training with FSDP + TP/SP + CP
  • EC-DiT model (https://arxiv.org/abs/2410.02098)
  • Adds mock video data module for mock training across stages 1-3
  • Mixed Image-Video Training with THD Packed Sequencing

Collection: diffusion

Changelog

  • Add specific line by line info of high level changes in this PR.

Usage

  • You can potentially add a usage example below
# Add a code snippet demonstrating how to use this 

GitHub Actions CI

The Jenkins CI system has been replaced by GitHub Actions self-hosted runners.

The GitHub Actions CI will run automatically when the "Run CICD" label is added to the PR.
To re-run CI remove and add the label again.
To run CI on an untrusted fork, a NeMo user with write access must first click "Approve and run".

Before your PR is "Ready for review"

Pre checks:

  • Make sure you read and followed Contributor guidelines
  • Did you write any new necessary tests?
  • Did you add or update any necessary documentation?
  • Does the PR affect components that are optional to install? (Ex: Numba, Pynini, Apex etc)
    • Reviewer: Does the PR have correct import guards for all optional libraries?

PR Type:

  • New Feature
  • Bugfix
  • Documentation

If you haven't finished some of the above items you can still open "Draft" PR.

Who can review?

Anyone in the community is free to review the PR once the checks have passed.
Contributor guidelines contains specific people who can review PRs to various areas.

Additional Information

  • Related to # (issue)

Zeeshan Patel added 2 commits November 7, 2024 23:49
Signed-off-by: Zeeshan Patel <[email protected]>
Signed-off-by: Zeeshan Patel <[email protected]>
ethanhe42
ethanhe42 previously approved these changes Nov 8, 2024
@@ -0,0 +1,166 @@
from random import randint

Check notice

Code scanning / CodeQL

Unused import Note

Import of 'randint' is not used.
from torch.utils.data import DataLoader

from nemo.collections.diffusion.models.model import DiTConfig
from nemo.collections.multimodal.data.energon.base import SimpleMultiModalDataModule

Check notice

Code scanning / CodeQL

Unused import Note

Import of 'SimpleMultiModalDataModule' is not used.

from nemo.collections.diffusion.models.model import DiTConfig
from nemo.collections.multimodal.data.energon.base import SimpleMultiModalDataModule
from nemo.lightning.io.mixin import IOMixin

Check notice

Code scanning / CodeQL

Unused import Note

Import of 'IOMixin' is not used.
from nemo.lightning.io.mixin import IOMixin
from nemo.lightning.pytorch.plugins import MegatronDataSampler

from .diffusion_energon_datamodule import DiffusionDataModule

Check notice

Code scanning / CodeQL

Unused import Note

Import of 'DiffusionDataModule' is not used.
@@ -19,24 +19,31 @@
import torch
from megatron.core.distributed import DistributedDataParallelConfig
from megatron.core.optimizer import OptimizerConfig
from megatron.core.transformer.enums import AttnMaskType

Check notice

Code scanning / CodeQL

Unused import Note

Import of 'AttnMaskType' is not used.
Comment on lines +471 to +477
torch.cuda.memory._record_memory_history(
True,
# Keep 100,000 alloc/free events from before the snapshot
trace_alloc_max_entries=100000,
# Record stack information for the trace events
trace_alloc_record_context=True,
)

Check warning

Code scanning / CodeQL

Unreachable code Warning

This statement is unreachable.
Signed-off-by: Zeeshan Patel <[email protected]>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

2 participants