tchaton Follow

tchaton Follow

👻

Always up for it !

thomas chaton tchaton

👻

Always up for it !

Follow

Research Engineer - Grid.ai | Pytorch Lightning

80 followers · 84 following · 51

Grid.ai | Pytorch Lightning
London
Sign in to view email
https://www.linkedin.com/in/thomas-chaton-1529ba128/

Highlights

Arctic Code Vault Contributor

Popular repositories

lightning-geometric

Integrate pytorch

Python 24 3
lightning-hydra-seed

Pytorch Lightning seed project with hydra

Python 14 1
The-Nature-Conservancy-Fisheries-Monitoring-Challenge

Jupyter Notebook 8 3
sagemaker-pytorch-boilerplate

Jupyter Notebook 8
Self-Driving-Car-on-GTAV

Jupyter Notebook 6
DeepLDA

DeepLDA in tensorflow

Python 4 3

2,861 contributions in the last year

Learn how we count contributions.

Less More

Activity overview

Contributed to PyTorchLightning/pytorch-lightning, nicolas-chaulet/torch-points3d, tchaton/lightning-geometric and 5 other repositories

Contribution activity

March 2021

Created 31 commits in 3 repositories

Created a pull request in PyTorchLightning/pytorch-lightning that received 9 comments

[refactor] Move save_function to accelerator 1/n [DeepSpeed]

What does this PR do? This PR refactor save_checkpoint to be trainer - checkpoint_connector - accelerator - training type plugin responsibility. Be…

+136 −123 • 9 comments

Opened 23 other pull requests in 2 repositories

PyTorchLightning/pytorch-lightning

18 merged 1 open 3 closed

[bugfix] Resolve schedule step bug for PyTorch Profiler Mar 25
[refactor] Add setup to profilers + _run_stage_setup to trainer 2/5 Mar 22
Add PyTorch 1.8 Profiler 5/5 Mar 21
[doc] Update Dict Train Loader doc. Mar 18
Flash predict step Mar 18
[doc] Add Zero Grad set_to_none=True trick Mar 16
Feat/ds update Mar 16
[changelog] Update Changelog on release v1.2.3 Mar 9
[release] Weekly Patch Release v.1.2.3 [full merge, no squash] Mar 9
[release] Weekly Patch Release v.1.2.3 [full merge, no squash] Mar 9
[warning] Add warning when values are not being reduced Mar 8
[bug] All_gather support tensor on cpu Mar 8
[bug] Update broadcast + reduce decision ModelCheckpoint] Mar 8
[bugfix] Resolve hanging in DDP Mar 4
[bugfix] Resolve memory leak for evaluation Mar 3
[bugfix] Perform reduction for dict in training_step and DP Mar 3
[bugfix] Check LightningOptimizer doesn't delete optimizer hooks Mar 2
[bugfix] TPU + all_gather + SingleTPU shouldn't call xm.all_gather Mar 2
[doc] Improve Manual Optimization Example Mar 2
[CI] Pytorch 1.8 testing Mar 2
[bugfix] TPU test hangs to barrier on 1 process Mar 1
[bug] Fix Pytorch profiler with emit_nvtx Mar 1

PyTorchLightning/lightning-flash

1 merged

[refactor] DataPipeline 1/n Mar 22

Reviewed 149 pull requests in 5 repositories

PyTorchLightning/pytorch-lightning 127 pull requests

Do not add return dict items to callback_metrics Mar 26
Profilers - do not write when there's no summary Mar 26
Add model parallel setup hook Mar 26
[bugfix] Resolve schedule step bug for PyTorch Profiler Mar 25
Fix checkpoint callback & Trainer.test(_) issue for TPUs Mar 25
add copyr Mar 24
fix: update example autoencoder.py to reflect args Mar 24
MetricsHolder clean-up + typing Mar 24
Weekly Patch Release v.1.2.5 [full merge, no squash] Mar 24
Add PyTorch 1.8 Profiler 5/5 Mar 23
Docs/robots Mar 23
Fix disabled grads after call to predict Mar 23
Refactor PyTorch profiler 4/5 Mar 23
Flash predict step Mar 23
Refactor base profilers 3/5 Mar 23
[refactor] Add setup to profilers + _run_stage_setup to trainer 2/5 Mar 22
Simplify deprecations Mar 22
hotfix: mock examples Mar 22
Allow training type plugin to delay optimizer creation (FSDP 2/n) Mar 22
Fix csv extension check Mar 22
[Horovod] Fix Reduce for Horovod Mar 22
Don't update checkpoint in case of using Hydra Mar 22
Feature/double precision Mar 22
[1/N] Define dataclasses for progress tracking Mar 22
Clean utilities/argparse and add missing tests Mar 22
Some pull request reviews not shown.

PyTorchLightning/metrics 11 pull requests

Add testing agains each feat PT version Mar 25
seed all Mar 25
Fix ddp example Mar 24
Metric bootstrapper Mar 24
Simplify metrics testing utils Mar 24
fix link in canonical Mar 24
Add half precision testing [1/n] Mar 22
Multilabel support in ROC Mar 22
Transfer PL master [full merge, no squash] Mar 22
refactor setup Mar 19
fuse compositional Mar 3

PyTorchLightning/lightning-flash 9 pull requests

DataPipeline PoC Mar 26
refactoring setup Mar 25
[refactor] DataPipeline 1/n Mar 25
expose scheduler to Task Mar 24
Prototype modifications made for RANZCR kaggle challenge Mar 22
switch to torchmetrics Mar 12
add multilabel option Mar 9
Add download support for tar.gz & don't download data if exists Mar 9
Update new models from new torchvision version Mar 9

facebookresearch/fairscale 1 pull request

[feat] FSDP: add auto_wrap_bn Mar 18

PyTorchLightning/lightning-bolts 1 pull request

Refactor: shorten bolts name Mar 9

Created an issue in PyTorchLightning/pytorch-lightning that received 4 comments

trainer.training_type_plugin.broadcast doesn't seem to work properly

🐛 Bug Please reproduce using the BoringModel To Reproduce Use following BoringModel and post here Expected behavior Environment Note: Bugs with code …

4 comments

Opened 5 other issues in 3 repositories

PyTorchLightning/pytorch-lightning

3 open

pytorch/xla

1 open

Using XLA with subprocesses.Popen instead of spawn Mar 24

microsoft/DeepSpeed

1 closed

allreduce_gradients breaks in the narrow operation Mar 16

3 contributions in private repositories Mar 23 – Mar 25