🚀 Add Multi-GPU Training Support

[ashwinvaidya17]

📝 Description

Testing script

#!/bin/bash

models=("Cfa" "Cflow" "Csflow" "Dfkde" "Dfm" "Draem" "Dsr" "EfficientAd" "Fastflow" "Fre" "Ganomaly" "Padim" "Patchcore" "ReverseDistillation" "Rkde" "Stfpm" "Uflow" "VlmAd" "WinClip" "AiVad")

# Loop through each model and run the anomalib train command
for model in "${models[@]}"; do
    anomalib train --model "$model" --data MVTec --trainer.max_epochs 2 --trainer.devices 2 --trainer.strategy='ddp_find_unused_parameters_true'
done

Works

1. CFA
2. CFlow
3. CSFlow
4. Dfkde
5. Dfm
6. Dsr
7. Fastflow
8. Ganomaly
9. Padim
10. Patchcore
11. ReverseDistillation
12. Stfpm
13. Uflow
14. WinCLIP
15. EfficientAd
16. VlmAd
17. AiVad // visualization stage does not work
18. Draem
19. Fre

Not Working

✨ Changes

Select what type of change your PR is:

• 🐞 Bug fix (non-breaking change which fixes an issue)
• 🔨 Refactor (non-breaking change which refactors the code base)
• 🚀 New feature (non-breaking change which adds functionality)
• 💥 Breaking change (fix or feature that would cause existing functionality to not work as expected)
• 📚 Documentation update
• 🔒 Security update

✅ Checklist

Before you submit your pull request, please make sure you have completed the following steps:

• 📋 I have summarized my changes in the CHANGELOG and followed the guidelines for my type of change (skip for minor changes, documentation updates, and test enhancements).
• 📚 I have made the necessary updates to the documentation (if applicable).
• 🧪 I have written tests that support my changes and prove that my fix is effective or my feature works (if applicable).

For more information about code review checklists, see the Code Review Checklist.

[ashwinvaidya17]

We need to revisit the docstrings for this method

[codecov]

Codecov Report

Attention: Patch coverage is 70.58824% with 10 lines in your changes missing coverage. Please review.

Please upload report for BASE (feature/v2@c73e411). Learn more about missing BASE report.

Files with missing lines	Patch %	Lines
src/anomalib/data/validators/torch/video.py	77.77%	2 Missing ⚠️
src/anomalib/metrics/evaluator.py	77.77%	2 Missing ⚠️
...rc/anomalib/models/video/ai_vad/lightning_model.py	33.33%	2 Missing ⚠️
...malib/models/components/base/memory_bank_module.py	66.66%	1 Missing ⚠️
...models/components/classification/kde_classifier.py	0.00%	1 Missing ⚠️
src/anomalib/models/image/dfm/torch_model.py	0.00%	1 Missing ⚠️
src/anomalib/models/image/dsr/anomaly_generator.py	0.00%	1 Missing ⚠️

Additional details and impacted files

@@              Coverage Diff              @@
##             feature/v2    #2435   +/-   ##
=============================================
  Coverage              ?   78.38%           
=============================================
  Files                 ?      302           
  Lines                 ?    12940           
  Branches              ?        0           
=============================================
  Hits                  ?    10143           
  Misses                ?     2797           
  Partials              ?        0           

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[samet-akcay]

looking good, thanks 🔥

[haimat]

Awesome, these are great news, thanks a lot!
When do you plan to release an official update with this change?

[samet-akcay]

Our plan is to pre-release before the 20th

[haimat]

Hello, we are so much looking forward to multi-GPU training :)
Do you know when you can release this change?

[samet-akcay]

Hi, as soon as passing this CI :)
#2465

We didn't release it on the 20th of December mainly because we thought the documentation is not sufficient. We worked on the documentation during Christmas.

@blaz-r kindly added a new algorithm to v2 as well, but one of the tests are failing now. As soon as we fix the test, we'll release, which we hope to sort out by the end of today.

[haimat]

Hi, as soon as passing this CI :) #2465

Awesome, thanks!

[blaz-r]

Hi, as soon as passing this CI :) #2465

We didn't release it on the 20th of December mainly because we thought the documentation is not sufficient. We worked on the documentation during Christmas.

@blaz-r kindly added a new algorithm to v2 as well, but one of the tests are failing now. As soon as we fix the test, we'll release, which we hope to sort out by the end of today.

I think I found the issue, I'm going to open a PR in a few mins. Apologies about that.

[samet-akcay]

Hi, as soon as passing this CI :) #2465
We didn't release it on the 20th of December mainly because we thought the documentation is not sufficient. We worked on the documentation during Christmas.
@blaz-r kindly added a new algorithm to v2 as well, but one of the tests are failing now. As soon as we fix the test, we'll release, which we hope to sort out by the end of today.

I think I found the issue, I'm going to open a PR in a few mins. Apologies about that.

no worries @blaz-r, I think I've fixed it already, let's wait for the test results
76dd186

[blaz-r]

Good @samet-akcay, that might fix it, but in case it doesn't I also opened a PR that also adds device= to that same lines in #2490 .

[samet-akcay]

Good @samet-akcay, that might fix it, but in case it doesn't I also opened a PR that also adds device= to that same lines in #2490 .

Looks like it is passing
https://github.com/openvinotoolkit/anomalib/actions/runs/12670086114/job/35308989508?pr=2465

[blaz-r]

Great! 😄

[haimat]

@samet-akcay Heyho, just wanted to ask for the current status of multi-GPU training?
Still working on it, any planned release date?

[samet-akcay]

You could try it with pip install anomalib==2.0.0b2

[haimat]

Thanks, that is great news - I will give it a try 👍
Do you have any docs on the breaking API changes?

[haimat]

@samet-akcay I tried to get beta 2 up and running, however, I am not sure how to update our workflow.
For example, neither the Folder nor the Engine classes accept the task argument any more.
How else do I specify the task?
Could you provide a short example of how a basic classificiation training with a given folder would look like in version 2?