# PyTorch-Ignite internal meeting notes
## 2023-03-15
Attendees: Victor
### News
- GSoC: Student applications open soon
- Docker hub Free Team sunsetting -> Submit OSS application => DONE
### Work plan
- [x] (Victor) Fix failing pytorch version tests on CI
- [ ] (Sadra) Finish Pytest PR. My initial solution depended on pytest reordering algorithm (in the sense that I assumed items are reordered in a way that if the next item does not use a high-scope fixture and current item does, so this item is the last dependent item and we could tear down the fixture pre-emptively) which did not take tests using non-parametrized fixtures into account so I changed the code to encompass all tests. Meanwhile I came across a weakness in the reordering algorithm affecting my solution as well so I attempted to resolve it. After that I found a better solution to the issue which is independent of the reordering . Now I'd better break the PR into three PRs:
1. Pre-emptive fixture teardown
2. Improve and simplify reordering algo.
3. Improve reordering feature to consider items dependent on non-parametrized fixtures as well.
## 2023-03-01
Attendees: Sangho, Sadra, Victor
### News
- nothing
### Work plan
- [x] (Victor) Updated imagenet ref script
- [x] (Victor) Updated pascal voc 2012 ref script
- https://github.com/pytorch/ignite/tree/master/examples/references/segmentation/pascal_voc2012
- [ ] (Victor) reviewing mAP PR
- [ ] (Victor) Finalize object detection example : https://github.com/pytorch/ignite/pull/2520, https://github.com/louis-she/ignite/tree/detection-example
- https://github.com/louis-she/ignite/tree/detection-example/examples/contrib/detection
- [ ] (Sangho) Engine debug mode (#2851)
- [ ] Maybe add PROFILE flag as well: https://pytorch-ignite.ai/how-to-guides/03-time-profiling/
- [ ] (Sadra) WIP on test distributed scope
- [ ] PR for pytest: https://github.com/pytest-dev/pytest/pull/10771
## 2023-02-22
Attendees: Sangho, Sadra, Victor
### News
- v0.4.11 is released
- GSoC23 NumFocus is selected
- code-generator is again fully working
### Work plan
- [ ] (Victor) Finalize imagenet ref script
- [ ] (Victor) reviewing mAP PR
- [ ] (Victor) Finalize object detection example : https://github.com/pytorch/ignite/pull/2520, https://github.com/louis-she/ignite/tree/detection-example
- https://github.com/louis-she/ignite/tree/detection-example/examples/contrib/detection
- [ ] (Sangho) Engine debug mode (#2851)
- [ ] (Sadra) WIP on test distributed scope
## 2023-02-15
Attendees: Sangho, Sadra, Victor
### News
- WIP on v0.4.11 release
### Work plan
- [x] (Victor) Prepare for v0.4.11 release
- merge all open relevant PRs
- [x] create release branch
- WIP on release note
- [ ] (Victor) Finalize imagenet ref script
- [x] (Sangho) Sync website tutorial to script:
- https://github.com/pytorch/ignite/tree/master/examples/reinforcement_learning (master)
- https://pytorch-ignite.ai/tutorials/intermediate/03-reinforcement-learning/ (website)
- source: https://github.com/pytorch-ignite/examples/blob/main/tutorials/intermediate/03-reinforcement-learning.ipynb
- gym -> gymnasium
- [x] (Sadra) Finalize PRs: 2803, 2800, 2802
## 2023-02-01
Attendees: Sangho, Victor
### News
### Work plan
- [x] (Victor) Writing project ideas for GSoC 2023
- [ ] (Victor) Prepare for v0.4.11 release
- merge all open relevant PRs
- create release branch
- write release note
- [ ] (Victor) reviewing mAP PR
- [ ] (Victor) Finalize object detection example : https://github.com/pytorch/ignite/pull/2520, https://github.com/louis-she/ignite/tree/detection-example
- https://github.com/louis-she/ignite/tree/detection-example/examples/contrib/detection
- [x] (Sangho) Cross-post the recent blog's content
- Cross-post the content to https://pytorch-ignite.ai/blog/
- https://github.com/pytorch-ignite/pytorch-ignite.ai/tree/main/src/blog
- https://github.com/pytorch-ignite/pytorch-ignite.ai/tree/main/static/_images
- Sangho's blog post is published: https://labs.quansight.org/blog
```
https://github.com/pytorch-ignite/pytorch-ignite.ai/blob/main/src/blog/2021-06-28-pytorch-ignite-distributed.md
---
title: Distributed Training Made Easy with PyTorch-Ignite
slug: distributed-made-easy-with-ignite
description: Distributed code with PyTorch-Ignite
date: 2021-06-28
tags:
- Deep Learning
- Machine Learning
- PyTorch-Ignite
- PyTorch
- Horovod
- SLURM
- PyTorch XLA
- PyTorch DDP
- Distributed
---
# Distributed Training Made Easy with PyTorch-Ignite
Writing [agnostic](<https://en.wikipedia.org/wiki/Agnostic_(data)>) [distributed code](https://pytorch.org/tutorials/beginner/dist_overview.html) that supports different platforms, hardware configurations (GPUs, TPUs) and communication frameworks is tedious. In this blog, we will discuss how [PyTorch-Ignite](https://pytorch.org/ignite/) solves this problem with minimal code change.
<!--more-->
Crossposted from https://labs.quansight.org/blog/2021/06/distributed-made-easy-with-ignite/
```
## 2023-01-18
Attendees: Sangho, Victor
### News
- Prepare for v0.4.11 release, which date ? -> begining december
- create release branch
- write release note
- Mentoring GSoC 2023 ?
- Cross-post the content to https://pytorch-ignite.ai/blog/
- https://github.com/pytorch-ignite/pytorch-ignite.ai/tree/main/src/blog
- https://github.com/pytorch-ignite/pytorch-ignite.ai/tree/main/static/_images
```
https://github.com/pytorch-ignite/pytorch-ignite.ai/blob/main/src/blog/2021-06-28-pytorch-ignite-distributed.md
---
title: Distributed Training Made Easy with PyTorch-Ignite
slug: distributed-made-easy-with-ignite
description: Distributed code with PyTorch-Ignite
date: 2021-06-28
tags:
- Deep Learning
- Machine Learning
- PyTorch-Ignite
- PyTorch
- Horovod
- SLURM
- PyTorch XLA
- PyTorch DDP
- Distributed
---
# Distributed Training Made Easy with PyTorch-Ignite
Writing [agnostic](<https://en.wikipedia.org/wiki/Agnostic_(data)>) [distributed code](https://pytorch.org/tutorials/beginner/dist_overview.html) that supports different platforms, hardware configurations (GPUs, TPUs) and communication frameworks is tedious. In this blog, we will discuss how [PyTorch-Ignite](https://pytorch.org/ignite/) solves this problem with minimal code change.
<!--more-->
Crossposted from https://labs.quansight.org/blog/2021/06/distributed-made-easy-with-ignite/
```
### Work plan
- [ ] (Victor) reviewing mAP PR and check deepchecks implementation
- [ ] (Victor) Finalize object detection example : https://github.com/pytorch/ignite/pull/2520, https://github.com/louis-she/ignite/tree/detection-example
- https://github.com/louis-she/ignite/tree/detection-example/examples/contrib/detection
- [ ] (Victor) Setup CI for Horovod GPU
- [ ] (Victor?) WIP on Horovod all_reduce with groups https://github.com/pytorch/ignite/pull/2732
- [ ] try to repro the issue with pytorch and hvd only
- [ ] To check remove process set after all_reduce
- [ ] (Sangho)(4/5) Virtual try-on tutorial
- https://github.com/pytorch-ignite/examples/pull/84
- Dataset : VITON dataset (can be downloaded from Kaggle)
- Tutorial on website: https://github.com/pytorch-ignite/examples/tree/main/tutorials/intermediate
- https://pytorch-ignite.ai/tutorials/
- Figure out how to update the website:
- https://github.com/pytorch-ignite/pytorch-ignite.ai/tree/main/src/tutorials/intermediate
- Comparing to other tutorials, code is too long and model is a bit complicated
- How each part has background idea, what is virtual try-on task about
- Like, "This is general idea, ~~ "
- Remove `.cuda()`, simplify the code
- [ ] (Sangho)(3/5) Enable custom all_gather, all_reduce
- [ ] (Sadra) PR to fix RocCurve, https://github.com/pytorch/ignite/pull/2802
- [ ] (Sadra) Refactor EpochMetric and make it idempotent, https://github.com/pytorch/ignite/pull/2800
- [ ] (Sadra) Fix sync_all_reduce to consider update->compute->update case, https://github.com/pytorch/ignite/pull/2803
## 2022-12-21
Attendees: Sangho, Sadra, Victor
### News
- Prepare for v0.4.11 release, which date ? -> begining december
- create release branch
- write release note
- Sangho's blog post is published: https://labs.quansight.org/blog
- Cross-post the content to https://pytorch-ignite.ai/blog/
- https://github.com/pytorch-ignite/pytorch-ignite.ai/tree/main/src/blog
- https://github.com/pytorch-ignite/pytorch-ignite.ai/tree/main/static/_images
```
https://github.com/pytorch-ignite/pytorch-ignite.ai/blob/main/src/blog/2021-06-28-pytorch-ignite-distributed.md
---
title: Distributed Training Made Easy with PyTorch-Ignite
slug: distributed-made-easy-with-ignite
description: Distributed code with PyTorch-Ignite
date: 2021-06-28
tags:
- Deep Learning
- Machine Learning
- PyTorch-Ignite
- PyTorch
- Horovod
- SLURM
- PyTorch XLA
- PyTorch DDP
- Distributed
---
# Distributed Training Made Easy with PyTorch-Ignite
Writing [agnostic](<https://en.wikipedia.org/wiki/Agnostic_(data)>) [distributed code](https://pytorch.org/tutorials/beginner/dist_overview.html) that supports different platforms, hardware configurations (GPUs, TPUs) and communication frameworks is tedious. In this blog, we will discuss how [PyTorch-Ignite](https://pytorch.org/ignite/) solves this problem with minimal code change.
<!--more-->
Crossposted from https://labs.quansight.org/blog/2021/06/distributed-made-easy-with-ignite/
```
### Work plan
- [ ] (Victor) reviewing mAP PR and check deepchecks implementation
- [ ] (Victor) Finalize object detection example : https://github.com/pytorch/ignite/pull/2520, https://github.com/louis-she/ignite/tree/detection-example
- https://github.com/louis-she/ignite/tree/detection-example/examples/contrib/detection
- [ ] (Victor) Setup CI for Horovod GPU
- [ ] (Victor?) WIP on Horovod all_reduce with groups https://github.com/pytorch/ignite/pull/2732
- [ ] try to repro the issue with pytorch and hvd only
- [ ] To check remove process set after all_reduce
- [ ] (Sangho)(4/5) Virtual try-on tutorial
- https://github.com/pytorch-ignite/examples/pull/84
- Dataset : VITON dataset (can be downloaded from Kaggle)
- Tutorial on website: https://github.com/pytorch-ignite/examples/tree/main/tutorials/intermediate
- https://pytorch-ignite.ai/tutorials/
- Figure out how to update the website:
- https://github.com/pytorch-ignite/pytorch-ignite.ai/tree/main/src/tutorials/intermediate
- Comparing to other tutorials, code is too long and model is a bit complicated
- How each part has background idea, what is virtual tryon task about
- Like, "This is general idea, ~~ "
- Remove `cuda()`, simplify the code
- [ ] (Sangho)(3/5) Enable custom all_gather, all_reduce
- [ ] (Sadra) PR to fix RocCurve, https://github.com/pytorch/ignite/pull/2802
- [ ] (Sadra) Refactor EpochMetric and make it idempotent, https://github.com/pytorch/ignite/pull/2800
- [ ] (Sadra) Fix sync_all_reduce to consider update->compute->update case, https://github.com/pytorch/ignite/pull/2803
## 2022-12-14
Attendees: Sadra, Victor
### News
- Prepare for v0.4.11 release, which date ? -> begining december
- create release branch
- write release note
- Sangho's blog post PR was merged and will be soon published https://labs.quansight.org/blog
- Cross-post the content to https://pytorch-ignite.ai/blog/
- https://github.com/pytorch-ignite/pytorch-ignite.ai
### Ideas
- (Sadra) if you want to publish your pydata global content at https://pytorch-ignite.ai/blog/ ?
- Once mAP PR is merged, we can provide an object detection example using ignite, mAP and torchvision prototype transforms -> write a blog post and advertise on our website and torchvision repo.
```
acc = Accuracy()
acc.reset()
acc.update(o1)
acc.update(o2)
acc.update(o3)
v1 = acc.compute()
v1 = acc.compute()
acc.update(o4)
acc.update(o5)
v2 = acc.compute() # <--- possibly wrong value here
```
Check above snippet
### Work plan
- [ ] (Victor) reviewing mAP PR and check deepchecks implementation
- [ ] (Victor) Finalize object detection example : https://github.com/pytorch/ignite/pull/2520, https://github.com/louis-she/ignite/tree/detection-example
- https://github.com/louis-she/ignite/tree/detection-example/examples/contrib/detection
- [ ] (Victor) Setup CI for Horovod GPU
- [ ] (Victor?) WIP on Horovod all_reduce with groups https://github.com/pytorch/ignite/pull/2732
- [ ] try to repro the issue with pytorch and hvd only
- [ ] To check remove process set after all_reduce
- [ ] (Sangho)(4/5) Virtual try-on tutorial
- Dataset : VITON dataset (can be downloaded from Kaggle)
- Tutorial on website: https://github.com/pytorch-ignite/examples/tree/main/tutorials/intermediate
- https://pytorch-ignite.ai/tutorials/
- Figure out how to update the website:
- https://github.com/pytorch-ignite/pytorch-ignite.ai/tree/main/src/tutorials/intermediate
- Comparing to other tutorials, code is too long and model is a bit complicated
- [ ] (Sangho)(3/5) Enable custom all_gather, all_reduce
## 2022-12-07
Attendees: Sadra, Sangho, Victor
### News
- Prepare for v0.4.11 release, which date ? -> begining december
- create release branch
- write release note
```
t1 = torch.rand(1, 3, 32, 32)
print(id(t1), t1.data_ptr())
t2 = torch.Tensor(t1)
print(id(t2), t2.data_ptr())
l = [t1, ]
l += [1] * 100
# [t1, 1, 1, ..., 1]
assert l[0].data_ptr() == t1.data_ptr()
```
### Work plan
- [ ] (Victor) reviewing mAP PR and check deepchecks implementation
- [ ] (Victor) Setup CI for Horovod GPU
- [ ] (Victor) Finalize object detection example : https://github.com/pytorch/ignite/pull/2520, https://github.com/louis-she/ignite/tree/detection-example
- https://github.com/louis-she/ignite/tree/detection-example/examples/contrib/detection
- [ ] (Victor?) WIP on Horovod all_reduce with groups https://github.com/pytorch/ignite/pull/2732
- [ ] try to repro the issue with pytorch and hvd only
- [ ] To check remove process set after all_reduce
- [ ] (Sangho)(4/5) Virtual try-on tutorial
- Dataset : VITON dataset (can be downloaded from Kaggle)
- Tutorial on website: https://github.com/pytorch-ignite/examples/tree/main/tutorials/intermediate
- https://pytorch-ignite.ai/tutorials/
- Figure out how to update the website:
- https://github.com/pytorch-ignite/pytorch-ignite.ai/tree/main/src/tutorials/intermediate
- [ ] (Sangho)(3/5) Enable custom all_gather, all_reduce
- [x] (Sangho) Ping the developer on NDCG metric PR
- [ ] Once we figured new DDP test framework, port old tests to the new framework, https://github.com/pytorch/ignite/pull/2777
- There is an issue with pytest and fixture scope. Sadra would like to fix it for pytest. https://github.com/pytest-dev/pytest/issues/3806
- Workaround can be using `TestDistributed` class
- Port another metric to the new framework
- [ ] (Sadra) WIP on mAP: https://github.com/pytorch/ignite/pull/2546
- [ ] (Victor) https://github.com/deepchecks/deepchecks/blob/main/deepchecks/vision/metrics_utils/detection_precision_recall.py
- [ ] (low-prio) https://github.com/bes-dev/mean_average_precision/
## 2022-11-30
Attendees: Sangho, Victor
### News
- PyTorch Dev Conference this friday
- PyData Global starts this week
- Prepare for v0.4.11 release, which date ? -> begining december
- create release branch
- write release note
### Work plan
- [ ] (Victor) Setup CI for Horovod GPU
- [ ] (Victor) Finalize object detection example : https://github.com/pytorch/ignite/pull/2520, https://github.com/louis-she/ignite/tree/detection-example
- https://github.com/louis-she/ignite/tree/detection-example/examples/contrib/detection
- [ ] (Victor?) WIP on Horovod all_reduce with groups https://github.com/pytorch/ignite/pull/2732
- [ ] try to repro the issue with pytorch and hvd only
- [ ] To check remove process set after all_reduce
- [x] (Sangho) Update content for Quansight blog post:
- https://github.com/Quansight/Quansight-website/pull/497
- Link, typos, rewrite contents
- Will finalize by this week.
- [ ] (Sangho)(3/5) Enable custom all_gather, all_reduce
- [ ] (Sangho)(4/5) Virtual try-on tutorial
- Dataset : VITON dataset (can be downloaded from Kaggle)
- Tutorial on website: https://github.com/pytorch-ignite/examples/tree/main/tutorials/intermediate
- https://pytorch-ignite.ai/tutorials/
- Figure out how to update the website:
- https://github.com/pytorch-ignite/pytorch-ignite.ai/tree/main/src/tutorials/intermediate
- [ ] Once we figured new DDP test framework, port old tests to the new framework, https://github.com/pytorch/ignite/pull/2777
- [ ] (Sadra) WIP on mAP: https://github.com/pytorch/ignite/pull/2546
- [x] Reference for tests is pycocotools
- [ ] https://github.com/deepchecks/deepchecks/blob/main/deepchecks/vision/metrics_utils/detection_precision_recall.py
- [ ] (low-prio) https://github.com/bes-dev/mean_average_precision/
- [x] Added some tests using randomly regenerated data
- [x] (Sadra) WIP on unified distributed tests framework
- [x] Bug fixes (https://github.com/pytorch/ignite/pull/2788)
## 2022-11-23
Attendees: Sadra, Victor
### News
- Sangho will not attend next meeting
### Work plan
- [x] (Victor)(5/5) Create new docker images: 1.13.0-0.4.10
- [ ] (Victor) Setup CI for Horovod GPU
- [ ] (Victor) Finalize object detection example : https://github.com/pytorch/ignite/pull/2520, https://github.com/louis-she/ignite/tree/detection-example
- https://github.com/louis-she/ignite/tree/detection-example/examples/contrib/detection
- [ ] (Victor?) WIP on Horovod all_reduce with groups https://github.com/pytorch/ignite/pull/2732
- [ ] try to repro the issue with pytorch and hvd only
- [ ] To check remove process set after all_reduce
- [ ] (Sangho)(3/5) Enable custom all_gather, all_reduce
- [ ] (Sangho)(4/5) Virtual try-on tutorial
- Dataset : VITON dataset (can be downloaded from Kaggle)
- Tutorial on website: https://github.com/pytorch-ignite/examples/tree/main/tutorials/intermediate
- https://pytorch-ignite.ai/tutorials/
- Figure out how to update the website:
- https://github.com/pytorch-ignite/pytorch-ignite.ai/tree/main/src/tutorials/intermediate
- [ ] (Sangho) Update content for Quansight blog post:
- https://github.com/Quansight/Quansight-website/pull/497
- Link, typos, rewrite contents
- Will finalize by this week.
- [ ] (Sadra) WIP on mAP: https://github.com/pytorch/ignite/pull/2546
- [ ] Reference for tests is pycocotools
- [ ] https://github.com/deepchecks/deepchecks/blob/main/deepchecks/vision/metrics_utils/detection_precision_recall.py
- [ ] (low-prio) https://github.com/bes-dev/mean_average_precision/
- [x] Added some tests using randomly regenerated data
- [ ] (Sadra) WIP on unified distributed tests framework
- [x] Feature implemented (https://github.com/pytorch/ignite/pull/2777)
- [ ] Bug fixes (https://github.com/pytorch/ignite/pull/2788)
## 2022-11-17
Attendees: Sadra, Sangho
### News
- Sangho will not attend next meeting
### Work plan
- [ ] (Victor)(5/5) Create new docker images: 1.13.0-0.4.10
- [ ] (Victor) Setup CI for Horovod GPU
- [ ] (Victor) Finalize object detection example : https://github.com/pytorch/ignite/pull/2520, https://github.com/louis-she/ignite/tree/detection-example
- https://github.com/louis-she/ignite/tree/detection-example/examples/contrib/detection
- [ ] (Victor?) WIP on Horovod all_reduce with groups https://github.com/pytorch/ignite/pull/2732
- [ ] try to repro the issue with pytorch and hvd only
- [ ] To check remove process set after all_reduce
- [ ] (Sangho)(3/5) Enable custom all_gather, all_reduce
- [ ] (Sangho)(4/5) Virtual try-on tutorial
- Dataset : VITON dataset (can be downloaded from Kaggle)
- Tutorial on website: https://github.com/pytorch-ignite/examples/tree/main/tutorials/intermediate
- https://pytorch-ignite.ai/tutorials/
- Figure out how to update the website:
- https://github.com/pytorch-ignite/pytorch-ignite.ai/tree/main/src/tutorials/intermediate
- [ ] (Sangho) Update content for Quansight blog post:
- https://github.com/Quansight/Quansight-website/pull/497
- Link, typos, rewrite contents
- Will finalize by this week.
- [ ] (Sadra) WIP on mAP: https://github.com/pytorch/ignite/pull/2546
- [ ] Reference for tests is pycocotools
- [ ] https://github.com/deepchecks/deepchecks/blob/main/deepchecks/vision/metrics_utils/detection_precision_recall.py
- [ ] (low-prio) https://github.com/bes-dev/mean_average_precision/
- [x] Added some tests using randomly regenerated data
- [ ] (Sadra) WIP on unified distributed tests framework
- [x] Feature implemented
- [ ] Bug fixes
## 2022-11-10
Attendees: Victor, Sadra, Sangho
### News
- Improving search feature on the website by Jeff Yang
- examples: https://github.com/pytorch-ignite/examples
- website: https://github.com/pytorch-ignite/pytorch-ignite.ai
- problem with auto-pr gh action: https://github.com/pytorch-ignite/pytorch-ignite.ai/actions/workflows/submodule.yml
- Questions from GH issues, forum: https://discuss.pytorch.org/c/ignite/15 and discord
- Blog post by Sangho for QS: https://github.com/Quansight/Quansight-website/pull/497
### Work plan
- [ ] Next stable PyTorch release: end of the month
- [ ] (Victor)(5/5) Create new docker images: 1.13.0-0.4.10
- [ ] (Victor) Setup CI for Horovod GPU
- [ ] (Victor) Finalize object detection example : https://github.com/pytorch/ignite/pull/2520, https://github.com/louis-she/ignite/tree/detection-example
- https://github.com/louis-she/ignite/tree/detection-example/examples/contrib/detection
- [ ] (Victor?) WIP on Horovod all_reduce with groups https://github.com/pytorch/ignite/pull/2732
- [ ] try to repro the issue with pytorch and hvd only
- [ ] To check remove process set after all_reduce
- [ ] (Sangho)(3/5) Enable custom all_gather, all_reduce
- [ ] (Sangho)(4/5) Virtual try-on tutorial
- Dataset : VITON dataset (can be downloaded from Kaggle)
- Tutorial on website: https://github.com/pytorch-ignite/examples/tree/main/tutorials/intermediate
- https://pytorch-ignite.ai/tutorials/
- Figure out how to update the website:
- https://github.com/pytorch-ignite/pytorch-ignite.ai/tree/main/src/tutorials/intermediate
- [ ](Sangho) Update content for Quansight blog post:
- https://github.com/Quansight/Quansight-website/pull/497
- Link, typos, rewrite contents
- [ ] (Sadra) WIP on mAP: https://github.com/pytorch/ignite/pull/2546
- [ ] Reference for tests is pycocotools
- [ ] https://github.com/deepchecks/deepchecks/blob/main/deepchecks/vision/metrics_utils/detection_precision_recall.py
- [ ] (low-prio) https://github.com/bes-dev/mean_average_precision/
- [ ] (Sadra) WIP on unified distributed tests framework
---
## 2022-11-03
Attendees: Victor, Sangho
### News
- PyData Global 2022 tutorial talk submission failed
### Work plan
- [ ] Next stable PyTorch release: end of the month
- [ ] (Victor)(5/5) Create new docker images: 1.13.0-0.4.10
- [x] (Sangho)(5/5) CI: Enable on mypy checks on stable pth
- https://github.com/pytorch/ignite/blob/0f8e47bc43f0dbf4f36e43f50e3c62cee007d816/.github/workflows/unit-tests.yml#L98-L101
- Remove `&& matrix.pytorch-channel == 'pytorch-nightly'`
- Check other workflows if there are similar conditions
- [ ] (Victor) Setup CI for Horovod GPU
- [ ] (Sangho)(3/5) Enable custom all_gather, all_reduce
- [ ] (Sangho)(4/5) Virtual try-on tutorial
- Dataset : VITON dataset (can be downloaded from Kaggle)
- Tutorial on website: https://github.com/pytorch-ignite/examples/tree/main/tutorials/intermediate
- https://pytorch-ignite.ai/tutorials/
- Figure out how to update the website:
- https://github.com/pytorch-ignite/pytorch-ignite.ai/tree/main/src/tutorials/intermediate
- [ ] (Sadra?) Sentiment analysis with ignite tutorial
- Put in https://github.com/pytorch-ignite/examples/tree/main/tutorials/intermediate
- [ ] (Victor?) WIP on Horovod all_reduce with groups https://github.com/pytorch/ignite/pull/2732
- [ ] try to repro the issue with pytorch and hvd only
- [ ] To check remove process set after all_reduce
- [ ] (Sadra) WIP on `max_iters` args
- [ ] (Victor) review the latest comments and answer
- [ ] (Sadra) Review PR on mAP: https://github.com/pytorch/ignite/pull/2546
- [ ] Reference for tests is pycocotools
- [ ] https://github.com/deepchecks/deepchecks/blob/main/deepchecks/vision/metrics_utils/detection_precision_recall.py
- [ ] (low-prio) https://github.com/bes-dev/mean_average_precision/
- [ ] (Victor) Finalize object detection example : https://github.com/pytorch/ignite/pull/2520, https://github.com/louis-she/ignite/tree/detection-example
- https://github.com/louis-she/ignite/tree/detection-example/examples/contrib/detection
- [ ] (?, docs) https://github.com/pytorch/ignite/issues/2325
- [ ] (?, debug) https://github.com/pytorch/ignite/issues/1992
---
## 2022-10-20
Attendees: Sadra, Victor, Sangho
### News
- PyTorch Dev Conf, if we need invitations
- Hacktoberfest, create new good first issues, promote on twitter etc
### Work plan
- [ ] Next stable PyTorch release: end of the month
- (todo after pth release) Create new docker images: 1.13.0-0.4.10
- (todo after pth release) CI: Enable on mypy checks on stable pth
- [ ] (Sangho) WIP on Horovod all_reduce with groups https://github.com/pytorch/ignite/pull/2732
- [ ] try to repro the issue with pytorch and hvd only
- [ ] To check remove process set after all_reduce
- [ ] (Sadra) WIP on `max_iters` args
- [ ] (Victor) review the latest comments and answer
- [ ] (Sadra) Review PR on mAP: https://github.com/pytorch/ignite/pull/2546
- [ ] Reference for tests is pycocotools
- [ ] https://github.com/deepchecks/deepchecks/blob/main/deepchecks/vision/metrics_utils/detection_precision_recall.py
- [ ] (low-prio) https://github.com/bes-dev/mean_average_precision/
- [ ] (Victor) Finalize object detection example : https://github.com/pytorch/ignite/pull/2520, https://github.com/louis-she/ignite/tree/detection-example
- https://github.com/louis-she/ignite/tree/detection-example/examples/contrib/detection
- [ ] (docs) https://github.com/pytorch/ignite/issues/2325
- [ ] (debug) https://github.com/pytorch/ignite/issues/1992
---
## 2022-10-14
Attendees: Victor, Sangho
### News
- NF, monthly news update to send
- PyTorch Dev Conf, if we need invitations
- Hacktoberfest, create new good first issues, promote on twitter etc
- Next stable PyTorch release: end of the month
- (todo right now) CI: Update pytorch-versions workflow
- https://github.com/pytorch/ignite/blob/8d5d7ca58ff1e6e3ebc904ec7c7a38096b8be3c3/.github/workflows/pytorch-version-tests.yml#L17-L19
- Add 1.12.1 pth
- Add 3.10 python
- (todo after pth release) Create new docker images: 1.13.0-0.4.10
- (todo after pth release) CI: Enable on mypy checks on stable pth
- (todo right now) CI: Update cuda to 11.6, https://github.com/pytorch/ignite/blob/8d5d7ca58ff1e6e3ebc904ec7c7a38096b8be3c3/.github/workflows/gpu-tests.yml#L59
### Work plan
- (Victor) Finalize object detection example : https://github.com/pytorch/ignite/pull/2520, https://github.com/louis-she/ignite/tree/detection-example
- https://github.com/louis-she/ignite/tree/detection-example/examples/contrib/detection
- (Sangho) Finalize horovod issue
- Setup GPU CI Job:
- Draft: https://github.com/pytorch/ignite/pull/2619
- Problem: Could not find nvcc, please set CUDAToolkit_ROOT.
CMake Error at cmake/upstream/FindCUDAToolkit.cmake:649 (message):
Could not find nvcc, please set CUDAToolkit_ROOT.
- Conda env:
- conda create -n ignite python=3.8
- (?) conda install -y cmake nccl nvcc -c conda-forge
- `which nvcc` ensure that it is in conda path
- `HOROVOD_GPU_OPERATIONS=NCCL HOROVOD_WITH_PYTORCH=1 pip install horovod[pytorch]`
---
## 2022-03-15
Attendees: Victor, Taras
### News
- GSoC22: NF is accepted with all sub-orgs
- Plan to make new release
### Work plan
- Docker images: apex does not build on CircleCI -> manually ?
- Setup GHA self-hosted on a local server: ported 2/3 CircleCI tests: https://github.com/pytorch/ignite/issues/2466
- Finalize: https://github.com/pytorch/ignite/pull/2496
- https://github.com/pytorch/ignite/pull/2496#issuecomment-1065984440
- https://github.com/pytorch-ignite/pytorch-ignite.ai/pull/150
- Make v0.4.9 release once finished with #2496
---
## 2022-02-22
Attendees: Victor, Taras
### News
- MONAI/Aim integration: https://github.com/aimhubio/aim/blob/main/examples/pytorch_ignite_track.py
- https://github.com/osoblanco/tutorials/blob/master/3d_segmentation/spleen_segmentation_3d.ipynb
- 1-to-1 meetings with contributors
- Setup GHA self-hosted on a local server: ported 2/3 CircleCI tests
- CircleCI support responded with a config name for 2 GPUs, but runs are not successful
- Fixed bugs and improvements for ModelCheckpoint/Checkpoint
### Work plan
- Feature requests:
- https://github.com/pytorch/ignite/issues/2480
- https://github.com/pytorch/ignite/issues/2465
- https://github.com/pytorch/ignite/issues/2328
- Finalize migration to https://pytorch-ignite.ai
- Work on improving docs: https://github.com/pytorch/ignite/issues?q=is%3Aissue+is%3Aopen+label%3Adocs
- For Taras: https://github.com/pytorch-ignite/code-generator/issues/205 => under review
---
## 2022-02-15
Attendees: Victor, Taras
### News
- GSoC 2022 written proposal: https://github.com/pytorch/ignite/discussions/2456
- 1-to-1 meetings with contributors
- CircleCI tests on 2 GPUs configuration is failing due to removed configuration
### Work plan
- Finalize migration to https://pytorch-ignite.ai
- Work on improving docs: https://github.com/pytorch/ignite/issues?q=is%3Aissue+is%3Aopen+label%3Adocs
- For Taras: https://github.com/pytorch-ignite/code-generator/issues/205
---
## 2022-01-18
Attendees: Victor, Taras
### News
- GSoC 2022: ideas, proposals. Deadline: 07/02/2022
- Victor: have to figure out the ideas and write a proposal
- François may help
- Taras may help as well
### Work plan
- Finalize migration to https://pytorch-ignite.ai (by Priyansi?)
- Work on improving docs: https://github.com/pytorch/ignite/issues?q=is%3Aissue+is%3Aopen+label%3Adocs
---
## 2022-01-18
Attendees: Victor, François
### News
- Done mentored sprint for PyData Riyadh 2022
- Released v0.4.8
- GSoC 2022: ideas, proposals. Deadline: 07/02/2022
- François may help
- Victor: have to figure out the ideas and write a proposal
- Large-scale training with ignite ? 1GPU results ~ NGPUs results
### Work plan
- Figure out HVD CI issues
- Merge `idist.barrier` kwargs PR
- WIP on mAP by Victor
---
## 2022-01-11
Attendees: Victor, François, Priyansi
### Work plan
- Run a mentored sprint for PyData Riyadh
- finalizing dates: friday Jan 14 or next week
- who is available ?
- Prepare v0.4.8 release : branch cut + remaining fixes
- Finalize migration to pytorch-ignite.ai
---
## 2022-01-04
Attendees: Victor, Priyansi, Taras
### News and thoughts
- All PyData Global t-shirts are sent
- **Plan new release mid-January 2022**
- Planning a meet about Engine refactoring
### Topics for 2022
> In 2022 we also have big plans on our Roadmap. Few items:
> - Engine refactoring and finally v0.5.X release
> - GSoC 2022 (writing a proposal, apply, candidate selection, mentroing)
> - GSoD 2022 (writing a proposal, apply, candidate selection, mentroing) ?
> - Maybe, apply to NumFOCUS Sponsored project
> - Larger maintainers team
> - Domain toolkits (e.g. self-supervised learning) using PyTorch-Ignite
> - Better communications
#### What are your interests for the project ?
Victor:
- managing the team
- coding features for ignite
- coding domain toolkits
- helping writing proposals
- (if needed) managing the contributors
- (if needed and no one is doing) mentoring GSoC
- (if needed and no one is doing) communications
Sylvain:
- following a trend in the communauty
- coding
- pushing domain features
François:
- coding new features, fixing bugs
- if needed communications (like blog post showing new features usage)
Taras:
- Coding new features, bug fixing
- Coding domain toolkits/frameworks
- Dev advocacy
- Writing proposals and grant applications
- Showcasing: writing blog posts, tutorials; participation in meetups
- if needed mentoring GSoC (note: little/no prior experience)
Another name:
- topic of interest
---
## 2021-12-21
Attendees: Victor, Priyansi, Sylvain
### News
- Fix PyPi nightly releases
- Plan new release mid-January 2022
### Standup
Priyansi
- WIP on Concept page
- WIP on redrawing the diagram
Victor
- Reviewing/merging PRs
- WIP on fixing broken PyPi nightly releases
---
## 2021-12-17
Attendees: Victor, François, Priyansi, Taras
### Things to discuss
- Revamping Community
### News
- GSoD final stipend is paid on open-collective
- T-shirts for PyData Global are sent to me
### Standup
Priyansi
* WIP on finalizing GSoD tasks
Taras
* Nothing special
François
* Fixed TPU CI issue
* Tryied to fix GPU HVD CI issue
* WIP on idist.barrier
Victor
* Tryied to fix GPU HVD CI issue
* Project management
---
## 2021-12-10
Attendees: Victor, François, Priyansi
### News
- Ordered 5 t-shirts "Ignite at PyData Global 2021"
- Christmas gift for maintainers
### Standup
François
* Worked on fixing EMA Handler
* Configuration for idist barrier
Priyansi
- [ ] Action - Plan how to proceed with https://github.com/pytorch/ignite/issues/2325
Victor
- Working on MAP Metric
- Helping Fareha with the Timer issue
---
## 2021-12-03
Attendees: François, Taras, Victor, Priyansi
### News
- PTD2, poster was there. Any feedback?
- Discussed with a guy from W&B
- Interesting posters for medical domain
- End of GSoD 2021 program
- Case-study: https://github.com/pytorch/ignite/wiki/Google-Season-of-Docs-2021---Case-Study
- What's next ?
### Things to discuss
- moving fb page
- tensorly, data-api, data pipes, cuda graphs
### Standup
Priyansi:
* Finalizing GSoD 2021 open issues
* Participated at PTD2
Taras:
* WIP on fixing Metrics reset/update/compute API
* Helped writing GSoD case-study
François:
* Worked on state param scheduler
* Action: check on EMA Handler PR
Victor:
* [x] Action: send Moses' email to Priyansi to promote a tutorial from ClearML's side
* Helped writing GSoD case-study
---
## 2021-11-26
Attendees: Sylvain, Priyansi, Taras
### News
- GSoD: Final review to do (timeline, etc.) Reminder deadline 29th November.
- SSL meeting canceled. Suggestion : next Wenesday 1st December 2.00pm ? To be confirmed.
- PyTorch Dev Day soon, no information received yet on the schedule. Taras will attend.
### Standup
Sylvain:
* WIP on doctest for metrics.
Priyansi:
* Finalized PRs on tutorials and RLs.
Taras:
* WIP on https://github.com/pytorch/ignite/issues/2158
* WIP on fixing Metrics reset/update/compute API
* Will finished this week
---
## 2021-11-19
Attendees: Sylvain, Taras, Priyansi, Victor
### News
- PTD2 Poster sent and Victor reviewed and approved the latest edits by Helen
- Sylvain will make a presentation for mAP the next week (next friday)
- GSoD program finalization: final project evaluation and case study to do
-
### Standup
Victor:
* Reviewing issues/PRs and finalized PTD2 poster
Sylvain:
* PRs on doctests
Taras:
* WIP on https://github.com/pytorch/ignite/issues/2158
* WIP on fixing Metrics reset/update/compute API
Priyansi:
* WIP on how-to guides load checkpoint
* WIP on how-to guides Multi-output
* Open question on how to migrate from pytorch.org/ignite to pytorch-ignite.ai, https://github.com/pytorch/ignite/issues/2325
### Work on Mean Average Precision for detection task
Sylvain's presentation:
- https://docs.google.com/presentation/d/1G2nTPKq3kNi7xBhorAbv1vbJn_0dsww8
Refs:
- https://github.com/bes-dev/mean_average_precision/
- https://github.com/airbusgeo/playground-metrics
Points to decide:
- pycocotools is a ref for tests ? -> yes
- data format for update function
---
## 2021-11-12
Attendees: Sylvain, François, Taras, Victor, Ahmed, Priyansi
### News
- Hard deadline today to send the poster
- Did a kick-off meeting for SSL toolkit
- Sylvain will make a presentation for mAP the next week (next friday)
### Standup
Victor:
* Reviewing issues/PRs
Sylvain:
* WIP on doctest PRs
François:
* WIP on idist.barrier + kwargs
* WIP on fixing state param scheduler, 2 PRs
Ahmed:
* Reviewed PRs
* Finalizing ConvAI PR
* Reading papers for SSL
Priyansi:
* WIP on PR for HP notebook
Taras:
* WIP on https://github.com/pytorch/ignite/issues/2158
* WIP on fixing Metrics reset/update/compute API
---
## 2021-11-02
Attendees: Priyansi, Victor, Taras
### News
- Did mentored Sprint for PyData Global 2021
- Have to create ignite poster for PTD2
- Prioritize model parallelism support
- Victor: proposing to shift the kick-off meeting project for Model Parallism to the next week = Nov 7
- Victor: kick-off meeting for SSL project could be done Nov 5 depending on our workloads -> create a shared doc with papers, algorithms, ref tools etc
### Standup
Victor:
* Prepared slides for the sprint
* Run the sprint
* TODO: explain steps to do for Timer issue
* TODO: wip on the poster
Priyansi:
* Prepared slides for the sprint
* Run the sprint
* TODO: wip on the poster
Taras:
* PR with Sphinx version update
* TODO: wip on the poster
---
## 2021-10-26
Attendees: Ahmed, Priyansi, Victor, François
### News
- Release v0.4.7 is up and announced
- Mentored Sprint for PyData Global this friday (12h-16h UTC+2)
- Windows GPU CI is broken with release PyTorch 1.10
### Standup
Victor:
* Updating and fixing CI for pth 1.10
* Reviewing PRs for website
Priyansi:
* Finalized https://github.com/pytorch-ignite/examples/pull/48, to review
* Improvements for the website
* WIP on Hyperparams PR
Ahmed:
* Reviewed PR on NMT
François:
* WIP on adding kwargs to idist.barrier
* WIP on 2258 (improving codecov)
* WIP on https://github.com/pytorch/ignite/issues/2291
---
## 2021-10-19
Attendees: Ahmed, Priyansi, Taras, Victor, François
### News
- Release v0.4.7 is on Pipy, not yet anaconda
- New docker images with v0.4.7 and PyTorch v1.9.0
- Soon release PyTorch 1.10
### Standup
Victor:
* Reviewing and merging PRs
* Reviewed Release note
Ahmed:
* Reviewing PRs on Ignite, Examples
* PR reenabling DeterministicEngine tests
* Written with the release note
Taras:
* Fixed issue with the CI on Windows
* Built docker images
* Written with the release note
* WIP on Sphinx update -> update to v3.5.4
Priyansi:
* Finalizing open PR and addressed reviews
* We have to renew FB page
* Interested project (next week): available from Nov 1
* Model parallelism
* Self-supervised learning toolkit
François:
* Some activity on issue: https://github.com/pytorch/ignite/issues/2258
---
## 2021-10-12
Attendees: Ahmed, François, Taras, Victor
### News
- We did a presentation at Python Bucaramanga Meetup: https://www.youtube.com/watch?v=X9RJxmcvrJE
- Upcoming release v0.4.7 tomorrow
### Ideas and thoughts
Ahmed:
- make Public meeting once per two week
- make kick-off meeting for the projects
François:
- make live-coding sessions and specific topic meetings
### Standup
Victor:
* Fixed idist for the release
* Schedule GPU tests
* Preps for saturday's Meetup
* Interested project (next week): available from ~Nov 3~ Nov 5
* Model parallelism support (see roadmap and NF subproject)
* Self-supervised learning toolkit
Ahmed:
* WIP on Conv-AI code -> will send a PR
* Interested project (can start from the next week): availabile from Nov 1
* Self-supervised learning toolkit
* Improve distributed tests
François:
* Finalized long-standing PR on params scheduling
* WIP on scaling experiments
* Interested project: available from Nov 3
* Thoughts on integrating some NLP models from HuggingFace
* Something close to Ref Examples Showcase
Taras:
* WIP on release v0.4.7
* Fixed a bug with conda/pip migration
* WIP on sphinx version update
* Preps for saturday's Meetup
* Interested project: available from ~Nov 1~ Nov 7
* Self-supervised learning toolkit
* Code-Generator
---
## 2021-10-05
Attendees: Ahmed, Priyansi, Taras, Victor
### Topics to discuss
> By the way, for all of us who are less senior members (and very busy) as Sylvain ;), let's do the following exercise:
please pick from the list below or propose your own project that you could work on for 1-2 months and let's talk about them on the next tuesday. Everyone could present it and we can exchange on the pertinence of it. For example, for Priyansi it is writing how-to guides and tutorials.
Available high-prio projects are the following:
> - Model parallelism support (see roadmap and NF subproject)
> - Ref examples showcase (training models on our slurm infra + log on clearml)
> - Improve distributed tests (avoid writing manually all test_something_nccl, test_something_gloo etc, it should be code once and internally generate others)
> - Self-supervised learning toolkit (NF subproject)
> - Work on Code-Generator
### News
- We did 2nd public meeting
- Ahmed: once per 2 weeks
### Standup
* Taras:
- was busy with PhD
- WIP release v0.4.7
* Ahmed:
- reviewed PRs
- WIP on Conv-AI
- Can help release v0.4.7
- Prefers project "Self-supervised learning toolkit"
* Priyansi:
- WIP on Cross-Validation notebook
- Reviewed Ishan's PR
- WIP on HP notebook
- Finalized PR on save_handler
* Victor:
- Fixed a bug with CI
- Reviewed some PRs
---
## 2021-09-28
Attendees: Priyansi, Ahmed, François, Victor
### News
- We did our 1st public meeting
- For the next one: create breaking sessions to talk about Hfest issues
- Label repositories as HFest friendly
- Shift to 14h30UTC
- https://pytorchdeveloperday.fbreg.com/ , submitted poster proposal
### Standup
* Victor:
* Planning v0.4.7 bugfix release + features
* Preparing presentation for Python Columbia meetup
* PyTorch-Ignite presentation to Agenium
* Ahmed:
* Helping with release
* Create a PR to remove commented code for LRFinder tests
* Talk on Conv-AI
* Priyansi:
* Contacted Livia to work on Cross-Validation tutorial
* WIP on HP tutorial
* cc Ishan
* Finalize: https://github.com/pytorch/ignite/pull/2198
* François:
* Finalize : [PR2090](https://github.com/pytorch/ignite/pull/2090) : fix conflicts, get Sylvain's review, I will address its comments this week
---
## 2021-09-21
Attendees: Victor, Priyansi
### News
- Public meeting open to everyone (max 20-30 minutes)
- Project presentation: team etc
- Q&A about PyTorch-Ignite
- how to start contributing
### Standup
* Priyansi:
* Finalized CIFAR10 example
* WIP on ClearML links
* WIP on HP tuning
* Victor:
* AWS PCluster
* Help debug SLURM issue
---
## 2021-09-16
Attendees: Sylvain, Victor, Priyansi, Ahmed
### News
- Presented ignite on PyTorch Community Voices
- Cool NLP internship by Ahmed
- Public meeting open to everyone (max 20-30 minutes)
- Ideas on the topics:
- how to start contributing
- Q&A about Ignite
- ...
### Standup
* Victor:
* Issues/PR/Support, PTCV
* WIP on AWS PCluster
* Sylvain:
* Help with PTCV
* WIP on SLURM issues
* Priyansi:
* Finalizing CIFAR10 tutorial
* Content on Transformers: https://github.com/pytorch-ignite/examples/pull/32
* Added Issue and a PR for Ignite
* Ahmed:
* WIP on LR finder
* Reviewing PR by Ishan on transformers example
---
## 2021-09-07
Attendees: Sylvain, Victor, Taras
### News
- PyTorch Community Voices
- https://gpuhackathons.org/event/monai-miccai-bootcamp-2021
### Standup
* Victor:
* slides for PTCV
* devs on AWS PCluster
* issue https://github.com/pytorch/ignite/issues/2185
* Sylvain:
* Finished helping and reviewing with micro/macro Bleu: https://github.com/pytorch/ignite/pull/2179
* Taras:
* Busy with PhD preps
* Priyansi:
* CIFAR10 distributed tutorials
* PyTorch 2 Ignite GIF and slides
* NLP beginner tutorial
* Communicated about end of GSoC on FB
Actions:
- Ask Priyansi to communicate about FID/IS on twitter and FB
- https://pytorch-ignite.ai/posts/gan-evaluation-with-fid-and-is/
---
## 2021-09-02
Attendees: Priyansi, Sylvain, Victor, Ahmed, Taras
### News
- End of GSoC:
- Model API (postponed), Engine refactoring (ongoing)
- Metrics: 2 FID/IS, PPL (postponed), mAP (ongoing)
- PyTorch Community Voices
- PyData Global: didn't apply -> AWS Parallel Cluster presentation
### Standup
* Victor: PTO
* Sylvain:
* Batch-wise compatible BLEU metric
* Micro average BLEU metric
* WIP same for ROUGE metric
* Taras:
* Reviewed GSoD PRs
* GSoD meeting
* Priyansi:
* PRs on tutorial: CIFAR10, NLP
* See yesterday's meeting notes
* Ahmed: PTO
### Actions:
- [ ] Victor: add access to 2 GPUs machine to Priyansi
- [ ] Victor: answer PyTorch Community Voices email
- [ ] Fix broken CI
- [ ] EMA on XLA, PR by SandyLaker
- [ ] Temporarily disable RandomSampler related tests in deterministic engine
---
## 2021-09-01 (GSoD and Website)
Attendees: Taras, Priyansi, Jeff
### What was done for the website
- [x] Search bar baked by algolia
- [x] Fine tune Features block for Mobile
- [x] Improve syntax with shiki
- [x] Show copy button on hover
- [x] Download/Open in colab bar for Blog posts
### Tasks for the next week for the Website:
- [ ] Fix missing image for Gan evaluation using FID and IS
- [ ] Discuss search feature: algolia etc
### Tasks for the next week, GSoD:
- [ ] Reviewing the PRs for tutorials
- [ ] `idist` tutorial
---
## 2021-08-24
Attendees: Priyansi, François, Sylvain
### News
- [PyTorch Developer Day 2021](https://pytorch.org/blog/pytorch-developer-day-2021/) : Deadline for submission is September 24, 2021.
### Standup
* Sylvain:
* PyTorch Dev Day : could we highlight idist ? Given the upcoming deadline, an update is relevant.
* Work with Arpan on mAP (OnGoing draft [#2130](https://github.com/pytorch/ignite/pull/2130)). Still a lot of work to do before a merge.
* Final evaluation of GSoC to do until August 28th.
* Contacted by Ishan about the implementation of Perplexity metric. Suggestion to work on a show case of BLEU and Rouge metrics to prepare this.
* François:
* PR [#2090](https://github.com/pytorch/ignite/pull/2090) on State Parameter Scheduler: core dev almost done (still minor conflicts to fix). Work left: coherent naming and adding more tests. It could be finished soon. Sylvain will help to review the PR if needed.
* IFPEN / PyTorch-Ignite collaboration project ? Scope: detection of the shape of microfossils / core grains (Research topic). All means of collaboration are opened (students, convention, expertise, etc.). Should be discussed in a few weeks. Meanwhile, ClearML was evaluated as very relevant feature for IFPEN, this is an opportunity for Ignite. Note that a support is needed for reducing the learning curve for the teams (code generator, low/no code approaches?).
* Priyansi:
* CIFAR10 Distributed Tutorial completed.
* Work on combining text CNN and Transformers notebook for text classification.
* Promoting in the community the blog article using PyTorch-Ignite: [Classifying Tiny ImageNet with EfficientNet](https://towardsdatascience.com/pytorch-ignite-classifying-tiny-imagenet-with-efficientnet-e5b1768e5e8f), to be discussed.
- [ ] Discuss the PyTorch Dev Day poster update
- [ ] Discuss the opportunity of collaboration IFPEN / PyTorch-Ignite
- [ ] Sylvain: review PR #2090
- [ ] Discuss the promotion of the blog article
---
## 2021-08-18 (GSoD and Website)
Attendees: Taras, Priyansi, Jeff
### Tasks for the next week for the Website:
### Tasks for the next week, GSoD:
---
## 2021-08-11 (GSoD and Website)
Attendees: Victor, Taras, Priyansi, Jeff
### Tasks for the next week for the Website:
- [x] Add dev.to link near twitter, fb logos (https://dev.to/pytorch-ignite)
- [x] Add pytorch like docs, tutorials, ressources block
- [x] Improve features representation block (like pytorch)
- [x] Put API Reference link to Docs drop-out menu
- [x] Fix API Ref link
- [ ] Blog posts, remove "next steps" and put crossposting to qs labs site
- [ ] Replace all other links to qs labs posts by our links
- [x] Fix mobile layout (ecosystem)
### Tasks for the next week, GSoD:
- [x] Finalize the beginner tutorial
- [ ] GIF
- [x] Finalize the time profiling notebook
- [x] Remove Timer from tutorial and create an issue to pytorch/ignite about its maybe misleading behaviour
- [ ] Create a draft notebook for CIFAR10 Distributed
- [x] Fix links in Getting Started Notebook
---
## 2021-08-10
Attendees: Victor, Taras, Ahmed, Arpan, Priyansi
### News:
- PyData Global 2021 - CFP, deadline Aug 15th
- Created Medium channel
- Publish the first post about PyTorch-Ignite
- No decent support of Markdown
### Standup
* Victor:
* Issues/PRs/Support/Reviews
* Taras:
* WIP on Code-Generator's CI
* Arpan:
* PR on mAP implementation: https://github.com/pytorch/ignite/pull/2130
* Finalizing on blog and example with FID/IS
* there are 2 versions: long and short
* Short version: PR to pytorch-ignite site as blog post
* Ping Jeff to get exact info on how to proceed with the PR
* Long version: Victor publish it on medium once the channel is set up
* Example of PPL: https://colab.research.google.com/drive/1uqGmxMiEO7iDorrsZd7ccQfwMoGsqmoZ
* Ahmed:
* WIP on refactoring Engine: https://github.com/pytorch/ignite/wiki/Roadmap#refactor-engine
* merged 1 pr with events refactoring
* PR on moving contrib code
* Finalizing https://github.com/pytorch/ignite/pull/2146
* Priyansi:
* WIP on Time Profiling
* PRs merged on FastaiLRFinder and ...
* Submitted new PRs to review with tutorials
* WIP on beginer tutorials
### Actions
- [ ] Victor: review long version FID/IS blog and publish it on medium once the channel is set up.
- [ ] Arpan: Ping Jeff to get exact info on how to proceed with the PR with the GAN evaluation (short version)
---
## 2021-08-03
Attendees: Victor, Taras, Ahmed, Arpan, François, Priyansi
### News:
- Release v0.4.6
- Docker images to update
- Created Medium channel
- Publish the first post about PyTorch-Ignite
- No decent support of Markdown
### Standup
* Victor:
* Issues/PRs/Support/Reviews
* Release v0.4.6
* Taras:
* WIP on fixed MNIST download with CircleCI
* Problem with VOC template and deeplabv3_resnet101
* Finalized the script for the Release Note (on Wiki, Maintainers Note)
* Release v0.4.6
* Arpan:
* Draft PR on mAP implementation
* Finalizing on blog and example with FID/IS
* there are 2 versions: long and short
* Short version: PR to pytorch-ignite site as blog post -> website
* Decide where to store ipynb: examples or website/content/posts -> website
* Long version: Victor publish it on medium once the channel is set up
* Example of PPL: https://colab.research.google.com/drive/1uqGmxMiEO7iDorrsZd7ccQfwMoGsqmoZ
* Priyansi:
* PRs merged on FastaiLRFinder and ...
* Submitted new PRs to review with tutorials
* WIP on beginer tutorials
* Ahmed:
* WIP on refactoring Engine: https://github.com/pytorch/ignite/wiki/Roadmap#refactor-engine
* merged 1 pr with events refactoring
### Actions
- [ ] Victor: review long version FID/IS blog and publish it on medium once the channel is set up.
---
## 2021-07-27
Attendees: Victor, Taras, Ahmed, Arpan
### News:
- Planning the next release v0.4.6 (30/07 - 02/08)
### Standup
* Victor:
* Issues/PRs/Support/Reviews
* Advanced and released playground Slurm Cluster
* Arpan:
* WIP on the design of mAP implementation
* Finalizing on blog and example with FID/IS
* there are 2 versions: long and short
* Short version: ping Priyansi and ask where to put (tutorial or blog)
* Long version: Victor publish it on medium once the channel is set up
* Taras:
* Review for the tests for playground Slurm Cluster
* Fixed MNIST download with CircleCI
* Finalize and make available the script for the Release Note
* Ahmed:
* WIP on refactoring Engine: https://github.com/pytorch/ignite/wiki/Roadmap#refactor-engine
* merged 1 pr with events refactoring
* Finalizing on CIFAR10 example with LRFinder
* WIP on EpochMetric issue
### Actions
- [x] Victor: finalize Medium channel for PyTorch-Ignite
- [ ] Victor: review long version FID/IS blog and publish it on medium once the channel is set up.
- [ ] Ahmed: Finalize EpochMetric issue
- [ ] Update DDP uncovered remaining metrics: ~AP, ~PRCurve
- [x] Taras: generate the release note
---
## 2021-07-20
Attendees: Victor, Taras, Ahmed, François, Arpan, Priyansi
### News:
- PyTorch-Ignite Code-generator blog is online and as a project sponsored by NumFOCUS is over.
- Accepted round 2 SDG 2021: Engine refactor, Pipeline Parallelism, tool for self-supervised learning.
### Standup
* Victor:
* Issues/PRs/Support
* Advanced and released playground Slurm Cluster
* Arpan:
* Finalizing on blog and example with FID/IS
* WIP on mAP (colab notebook)
* PPL is on-hold
* François:
* WIP on State Parameter Scheduler PR
* Taras:
* Working on the website
* Work on MNIST download with CircleCI
* Ahmed:
* WIP on refactoring Engine: https://github.com/pytorch/ignite/wiki/Roadmap#refactor-engine
* merged 2 preliminary PRs
* WIP on CIFAR10 example with LRFinder
* remains a proper code update
* Integration of LR Finder to CodeGen templates
* Priyansi:
* PRs to pytorch-ignite/examples
* WIP on Notion page about future features
### Actions
- [ ] Elaborate a plan on how to engage with the community => WIP
- [x] Finalize and publish the blogpost on CodeGen => WIP
---
## 2021-07-13
Attendees: Victor, Taras, Ahmed, François, Arpan, Priyansi
* Victor:
* Issues/PRs/Support
* Helped to Merged [EMAHandler PR](https://github.com/pytorch/ignite/pull/2098)
* Arpan:
* WIP on blog and example with FID/IS
* Two versions: compact (our blog) and long (e.g. medium)
* Implement PPL as a loss and use as a metric
* François:
* WIP on State Parameter Scheduler PR
* Taras:
* Finalized blog post, submitted a PR to QS-Labs
* Ahmed:
* WIP on EpochMetric to support tuples output of user function
* WIP on refactoring Engine: https://github.com/pytorch/ignite/wiki/Roadmap#refactor-engine
* WIP on CIFAR10 example with LRFinder
* -> remains a proper code update
* Integration of LR Finder to CodeGen templates
* Priyansi:
* Reviewed CodeGen blog
* WIP on website, tutorials
Actions:
- [ ] Elaborate a plan on how to engage with the community => WIP
- [ ] Finalize and publish the blogpost on CodeGen => WIP
---
## 2021-07-06
Attendees: Victor, Taras, Ahmed, François, Arpan
* Victor:
* Issues/PRs/Support
* Reviewing and thinking about the API for [EMAHandler PR](https://github.com/pytorch/ignite/pull/2098)
* Taras:
* Worked on GSoD21
* "Elaborate a plan on how to engage with the community" ?
* More activity on Kaggle ? Providing competition baselines ?
* Hackathon ?
* François:
* Working on Any Parameter scheduler
* Ahmed:
* Updating CIFAR10 example with LRFinder
* Working on refactoring Engine: https://github.com/pytorch/ignite/wiki/Roadmap#refactor-engine
* Solving issue with EpochMetric to support any output of user function
* Arpan:
* Blog and example with FID/IS
* Implement PPL as a loss and use as a metric
* "Elaborate a plan on how to engage with the community" ?
* Write blog around DL techniques
* Convert examples and tutorials with Ignite
Actions:
- [ ] Elaborate a plan on how to engage with the community => WIP
- [ ] Finalize and publish the blogpost on CodeGen => WIP
---
## 2021-06-29
Attendees: Victor, Taras, Priyansi, Ahmed, François, Sylvain
* Victor:
* PyTorch-Ignite v0.4.5 release is done
* CodeGen v0.2.0 is released
* `idist` blogpost is published
* Taras:
* Fixed burning release issues
* Sylvain:
* PING project is finished (no significant results)
* FID/IS metrics are merged
* Next work priority:
* Examples with FID/IS
* Short blogpost about that
* PPL
* mAP
* François:
* Checked blogpost code snippets
* Blogpost text review
* Any param scheduler
* mypy issues
* Priyansi:
* Studied PyTorch tutorials -> too complicated
* Studied Numpy tutorials
* Issue with building numpy tutorials locally
* Updated friction log
*
* Ahmed:
* Model API meeting and a bit postponed
* Updating remaining metrics to work DDP
* reset/update/compute support for MetricsLambda
* Updating CIFAR10 example with LRFinder
Actions:
- [ ] Elaborate a plan on how to engage with the community
- [ ] Finalize and publish the blogpost on CodeGen
---
## 2021-06-23
Attendees: Victor, Taras, Priyansi, Ahmed, Arpan, François, Sylvain
* Victor:
* Delayed PyTorch-Ignite release v0.4.5
* Remove FID
* Fixing https://github.com/pytorch/ignite/issues/2076 ?
* Taras:
* Reviewing PR for CodeGenerator
* WIP on Blogpost
* Worked on CI issues
* Refactor conda/pypi workflow
* Fix Windows GPU CI with 1.9.0
* Sylvain:
* WIP on FID/IS code and PR
* François:
* Checked examples for the blogpost
* Ahmed:
* Another iteration on Model API
* WIP using our Model API with Dask
* Finalize DDP support and checks for all ignite metrics
* Arpan:
* Finalizing FID/IS PRs
* Example with GAN and metrics
* Priyansi:
* Finalizing blogpost comments and storing the code
* WIP on tutorials
Actions:
- [x] Ignite v0.4.5 Release before June 25th by the team
- [x] After the release create new docker images
- [x] Test blogpost codes upon v0.4.5 branch by François
---
## 2021-06-15
Attendees: Victor, Taras, Priyansi, Ahmed, Arpan, François
* Victor:
* Release on June 18th
* Implemented some of remaining features to be included to the release
* Taras:
* Worked on Codegenerator: open zip in Colab
* Still WIP as creating new colab with our content is tricky
* Arpan:
* PR on FID metric under review
* PR on IS metric under review
* WIP on an example with FID/IS metrics: DCGAN on CIFAR10
* François:
* Code for the blog post
* Check the code
* Added README.md
* Ahmed:
* WIP on another approach with Model API
* Help with integrating LR finder to CIFAR10 example
* Priyansi:
* WIP on renew README.md
* WIP on idist blogpost
* Numpy tutorials docs
Actions:
- [ ] Ignite v0.4.5 Release before June 18th by the team
- [ ] After the release create new docker images
- [x] Review FID/IS metric by Victor
- [x] Check the code for the blog post by François
- [ ] Experiment with CIFAR10 and LR finder by Ahmed
- [x] Blogpost layout fixes and helping with publishing by Priyansi
- [ ] Advertise FB page to have more followers and the goal is to change the username.
---
## 2021-06-08
Attendees: Victor, Taras, Priyansi, Ahmed, Arpan
* Victor:
* Postponed the release to the next week and will be fixing few bugs with idist.
* Taras:
* Worked Release note, investigated the source of numpy.org
* Worked on Codegenerator
* Started blog post for Codegenerator
* Ahmed:
* Team meeting and Ruslan from Argus library on Higher-level API
* Worked on other issues/PRs and the release note
* Updates for LR finder
* Arpan:
* Working on FID metric
* Soon will be working on IS metric
* Priyansi:
* WIP on restructuring our docs according to Diataxis format
* Numpy docs exploration
Actions:
- [ ] Ignite v0.4.5 Release before June 18th by the team
- [ ] After the release create new docker images
- [x] Review FID metric by Victor
- [x] Updates for LR finder by Ahmed
- [x] Send a PR for IS by Arpan
- [x] Summarize the results the survey and present it on the next meeting by Priyansi
- [ ] Advertise FB page to have more followers and the goal is to change the username.
---
## 2021-06-01
Attendees: Victor, Taras, Priyansi, Ahmed, Arpan
* Victor:
* Cut the branch 0.4.5 for the release
* Written a draft NF-SDG2021 with Taras
* Issues tracker
* WIP on state.data
* WIP on list of tensors support for metrics
* WIP on Porting Self-supervised algos: DINO
* Taras:
* Fixed Prec issue 2007
* Written a draft NF-SDG2021
* WIP: Metric reset-update-compute
* CodeGen: ported text-cls example with Jeff
* Draft release note
* Ahmed:
* Meeting with Ruslan from Argus library on Higher-level API
* Working on another prototype of Model API with multiple models/opts
* Arpan:
* WIP on FID and DCGAN cifar10 example
* Priyansi:
* Worked on blog post
Actions:
- [ ] Write the release note by Taras, Victor
- [ ] Create the release in the end of this week by Victor
- [x] Submit NF-SDG2021 proposal by Victor
- [x] Present new prototype for the next or later meeting by Ahmed
- [x] Send a draft PR with FID by Arpan
- [ ] Summarize the results the survey and present it on the next meeting by Priyansi
- [ ] Advertise FB page to have more followers and the goal is to change the username.
---
## 2021-05-25
Attendees: Victor, Taras, Priyansi, Ahmed, Arpan
* Victor:
* CZI full-app sent
* NF-SDG2021
* Issues tracker
* api (Nic)
* prec bug
* deepspeed
* Porting Self-supervised algos: DINO
* Taras:
* Prec issue investigation
* Metric reset-update-compute
* CodeGen: porting text cls example
* Priyansi:
* Sending email
* Survey exploration
* Setup fb page
* Investigation ignite docs, concepts, exampl
* Q:
* Ahmed:
* Experiments w/ api
* Arpan:
* Work on fid metric w/ Sylvain
Actions:
- [x] Victor: Write a draft NF-SDG2021 proposal before 31/05/2021
- [x] Prepare the release in juin, v0.4.5
- [x] Taras: Fix Precision issue 2007 before the release
- [ ] Advertise FB page to have more followers and the goal is to change the username.
Sign in with Wallet
Connect another wallet