The Wayback Machine - https://web.archive.org/web/20200724031917/https://github.com/awaelchli

awaelchli Follow

awaelchli Follow

Adrian Wälchli awaelchli

PhD student at University of Bern

Follow

25 followers · 26 following · 37

University of Bern
Switzerland
@adrianwaelchli

Highlights

Arctic Code Vault Contributor
Pro

Pinned

PyTorchLightning/pytorch-lightning

The lightweight PyTorch wrapper for ML researchers. Scale your models. Write less boilerplate

Python 7k 789
GTA-V-Camera-Tracker

Grand Theft Auto V mod for recording and exporting camera pose

C++
ese_wordfinder

Forked from ese-unibe-ch/ese2013-team3

Java
notecards

Flip Card Android App

Java

592 contributions in the last year

Contribution activity

July 2020

awaelchli/stylegan2-pytorch-lightning Python Jul 23

Created a pull request in PyTorchLightning/pytorch-lightning that received 11 comments

save_dir fix for MLflowLogger + save_dir tests for others

What does this PR do? Fixes the error "Yaml file not found" when setting the default_root_dir in Trainer to something other than the default. Befor…

+219 −84 • 11 comments

Fix weights_save_path when logger is used + simplify path handling + better docs Jul 23
docs: clarify which shuffle argument (follow up to #2630) Jul 21
fix dtype/device property not getting updated in submodules Jul 21
update ddp docs about launching training script Jul 14
truncate long version number in progress bar Jul 13
update conda packages Jul 12
docs: add hint about register_buffer Jul 10
Fix CI crash on coverage upload timeout Jul 8
fix codecov command error in drone for master branch Jul 7
make loggers pickleable Jul 5
Fixed skipped horovod tests Jul 5
fix dtype conversion of example_input_array in model summary Jul 5
Warn user when IterableDataset has __len__ defined Jun 30

Datamodule Jul 23
EvalResult support for val loop (PR 3/5) Jul 22
truncate long version number in progress bar Jul 22
metrics: add SSIM Jul 22
Structured results (train loop only. val loop separate PR) (PR 2/5) Jul 20
Fix local rank zero casting Jul 18
[WIP] Improve Comet Logger pickled behavior Jul 18
metrics: add BLEU Jul 18
Add tracking of basic states in Trainer. Jul 18
Clarify limit_{train,val,test}_batches behaviour Jul 17
Fix typo of handling Jul 16
fix setup call while testing Jul 16
tests: add default_root_dir=tmpdir Jul 12
Fix max_batches with fast_dev_run. Jul 11
missing changelog Jul 10
Typo Jul 10
.fit() returns last not best weights in ddp_spawn Jul 9
Support returning python scalars in DP Jul 9
Fix ddp tests + .test() Jul 7
flatten Wandb hyperparameters dict Jul 6
fix worker warning Jul 6
[tiny] Fix training_dataloader usage to be train_dataloader instead. Jul 6
make loggers pickleable Jul 5
Start accumulate gradients schedule at epoch 0 (continued) Jul 5
Start accumulate gradients schedule at epoch 0 Jul 4
Some pull request reviews not shown.

Created an issue in wandb/client that received 10 comments

wandb in pytest causes OSError: [Errno 29] Illegal seek

Weights and Biases version: 0.9.2 Python version: 3.7.6 Operating System: Linux Description Tried to use wanb in pytest here in this PR PyTorchLi…

10 comments

Log debug information to a file Jul 12

You can’t perform that action at this time.