profile
viewpoint
Ruotian(RT) Luo ruotianluo Toyota Technological Institute at Chicago Chicago http://ttic.uchicago.edu/~rluo Phd student at TTIC

ruotianluo/DiscCaptioning 91

Code for Discriminability objective for training descriptive captions(CVPR 2018)

ruotianluo/Faster-RCNN-Densecap-torch 88

Faster-RCNN based on Densecap(deprecated)

ruotianluo/Context-aware-ZSR 54

Official code for paper Context-aware Zero-shot Recognition (https://arxiv.org/abs/1904.09320 to appear at AAAI2020)

ruotianluo/cider 15

python codes for CIDEr - Consensus-based Image Caption Evaluation

ruotianluo/bert-convert-to-pth 5

TensorFlow code and pre-trained models for BERT

ruotianluo/awesome-torch 2

A curated list of awesome Torch tutorials, projects and communities

ruotianluo/coco-caption-old 2

Adds SPICE metric to coco-caption evaluation server codes

startedbytedance/piano_transcription

started time in 2 days

startedbytedance/GiantMIDI-Piano

started time in 2 days

startedeminorhan/baby-vision

started time in 2 days

startedBigRedT/info-ground

started time in 5 days

startedgoogle-research/vision_transformer

started time in 6 days

startedzomux/lanmt

started time in 7 days

startedzhegan27/VILLA

started time in 7 days

startedyumeng5/LOTClass

started time in 8 days

startedfacebookresearch/CollaQ

started time in 10 days

startedgoogle-research/xtreme

started time in 13 days

issue commentPyTorchLightning/pytorch-lightning

Resume training from a finished-training model will results in a new incorrect checkpoint

I add an if before https://github.com/PyTorchLightning/pytorch-lightning/blob/72f19768c828b734d8565ffef7b78fb9a57ba847/pytorch_lightning/trainer/training_loop.py#L178

         if self.trainer.current_epoch < self.trainer.max_epochs and not self.interrupted:
             self.check_checkpoint_callback(should_save=True, is_last=True)
ruotianluo

comment created time in 13 days

issue commentPyTorchLightning/pytorch-lightning

Resume training from a finished-training model will results in a new incorrect checkpoint

A simpler solution is adding an if before https://github.com/PyTorchLightning/pytorch-lightning/blob/b45b57cc587a11c3049fb5c605cf2d4b0689de2c/pytorch_lightning/trainer/trainer.py#L510

ruotianluo

comment created time in 13 days

issue commentPyTorchLightning/pytorch-lightning

Resume training from a finished-training model will results in a new incorrect checkpoint

I moved +1 from https://github.com/PyTorchLightning/pytorch-lightning/blob/dec31b3e761115b0a938caf35649acd814f557c9/pytorch_lightning/trainer/connectors/checkpoint_connector.py#L254 to https://github.com/PyTorchLightning/pytorch-lightning/blob/dec31b3e761115b0a938caf35649acd814f557c9/pytorch_lightning/trainer/connectors/checkpoint_connector.py#L150

ruotianluo

comment created time in 13 days

issue openedPyTorchLightning/pytorch-lightning

Resume training from a finished-training model will results in a new incorrect checkpoint

To verify a model if a model has finished training. I ran the training script again.

However, when I try to evaluate with the model, I got error that:

pytorch_lightning.utilities.exceptions.MisconfigurationException:
            you restored a checkpoint with current_epoch=11
            but the Trainer(max_epochs=10)

After digging a little bit, it seems that since the model has finished training, it will directly go to on_train_end, and save the model checkpoint with currect_epoch+1.

I am wondering if the design that saves the checkpoint with "next epoch" is reasonable? Why not add the epoch number after resuming?

created time in 13 days

push eventruotianluo/pytorch-lightning

Carlos Mocholí

commit sha eef1b5dbc862b0ed3b3974e524e6d1462afec935

Mention skipping steps in docs (#4108) * Mention skipping in docs * Use :class:

view details

William Falcon

commit sha 2d5a7f5e7dc686cfc8172101a81505bf421468af

Fixes #3276 (#4116)

view details

William Falcon

commit sha 710604838b2c7fc00b9f8a53561a79da0a741dd4

Fixes broken LM links (#4117)

view details

William Falcon

commit sha 85f1000ef0129371064618c9ae2542bebfff3e67

Update README.md

view details

William Falcon

commit sha 4ea31c2300d94c1bc3e2ed6db1ebc1af848d08c6

Update __init__.py

view details

Jirka Borovec

commit sha 19df091a3b29d941fa0fc7a2382263a5f63b973d

chlogs for 1.0 [skip ci] (#3978) * chlogs * logs * space * date * logs * logs * Apply suggestions from code review Co-authored-by: Adrian Wälchli <aedu.waelchli@gmail.com> * logs * logs Co-authored-by: Adrian Wälchli <aedu.waelchli@gmail.com>

view details

William Falcon

commit sha 09c2020a9325850bc159d2053b30c0bb627e5bbb

notices (#4118)

view details

William Falcon

commit sha 19a997870282874e97f6f06bcbc3db6ccc7c1df9

Update README.md

view details

Neven Miculinic

commit sha 01402e35948620e4ff77922ec30becf1f9b2d2fb

created minor doc fixes [ci skip] (#3958) * created minor fixes * adjusted the underline * Update docs/source/amp.rst * suggestion from code review Co-authored-by: Rohit Gupta <rohitgr1998@gmail.com>

view details

Sean Naren

commit sha 98eb7364962372f9c2020f400f1e013d45619af7

Added getstate/setstate method for torch.save serialization (#4127) * Added getstate/setstate method for torch.save serialization, added additional Optional Typing to results object * Added tests to ensure torch.save does not fail * Added flags to ensure compatible ddp cpu environment * Removed torch version check due to minimum already being 1.3, reduced epochs for speed * Moved tests to separate file * Update to accelerator, move to ddp_spawn to prevent hanging ddp

view details

William Falcon

commit sha bbbc111d520c220852298452a3b8ee1f7d4993d8

Update __init__.py

view details

William Falcon

commit sha 0474464c454a000e4cacfe188c86a0b8317288d5

Update README.md

view details

Stef | ステフ

commit sha fa737a5eb9c0313c662149f11f8ae66000011e8e

Add trace functionality to the function to_torchscript (#4142) * Add trace functionality to the function to_torchscript * used wrong parameter name in test * fix indentation to confirm to code style

view details

Jirka Borovec

commit sha 2c3b512c5066b39b419686a3de845a93e0d91e24

reverted "temporary drop metrics tests while speeding them up" and SKIP (#4115) * Revert "temporary drop metrics tests while speeding them up (#4071)" This reverts commit 86c70622fbae611dd45ccb104830e7e28639fe44. * skip metrics tests * skipping

view details

Jirka Borovec

commit sha 2232d372256801612ea6f64e968ab8005566c119

new log section [skip ci] (#4121) * new logs * formatting * 1.0.1

view details

Paul

commit sha 051d9b1d940f6c90aec128215ac142a35dbe02cf

Add trainer flag step [ci skip] (#4147) * Add trainer flag step Add step to disable automatic optimization in the trainer @justusschock * Apply suggestions from code review Co-authored-by: Nicki Skafte <skaftenicki@gmail.com> Co-authored-by: Jirka Borovec <Borda@users.noreply.github.com> Co-authored-by: Nicki Skafte <skaftenicki@gmail.com>

view details

Nrupatunga

commit sha d7f567126b5bfb8a61a08282a83ddf5d9a4141f4

[Doc] Lightning Module (#4123) * make current_epoch and global_step to be same as trainer, after model restore. * remove assignment here * test * minor modification * merge with parent's master * doc fix / improve * doc fix! Co-authored-by: Adrian Wälchli <aedu.waelchli@gmail.com>

view details

Nicki Skafte

commit sha edcb6e49b92ebcbf87fe55d59b34fa07b190c8c7

Speedup of metric tests (#4122) * speedup * something working * update the rest * more desc * recurse tests/metrics again * pep8 Co-authored-by: Teddy Koker <teddy.koker@gmail.com>

view details

Jirka Borovec

commit sha f967fbba436570adf78d9ff3ef2ede21212e346e

docs: fix fomatting

view details

Rohit Gupta

commit sha dec31b3e761115b0a938caf35649acd814f557c9

Call on_load_checkpoint before loading state_dict (#4057)

view details

push time in 13 days

startedneulab/REALSumm

started time in 13 days

startedVITA-Group/BERT-Tickets

started time in 13 days

startedairsplay/vokenization

started time in 14 days

startedclovaai/length-adaptive-transformer

started time in 14 days

issue commentruotianluo/ImageCaptioning.pytorch

What is the beam_size for?

Colab is using transformer for demonstration. Up down is a separate model

sbkim052

comment created time in 14 days

issue commentruotianluo/ImageCaptioning.pytorch

What is the beam_size for?

Transformer is transformer, it's a separate model

sbkim052

comment created time in 15 days

issue commentruotianluo/ImageCaptioning.pytorch

What is the beam_size for?

No, the uptown model is not using transformer.

sbkim052

comment created time in 15 days

starteddaniellevy/fast-dro

started time in 15 days

issue commentPyTorchLightning/pytorch-lightning

Graceful interrupt not graceful

Sure, so my proposal is maybe handling keyinterrupt a little bit different; for example, at least save the checkpoint with a different name. (There is a way to restore shuffled dataset when using ddp, because the distributed sampler is somewhat deterministic. In other cases, there are other nasty ways.)

ruotianluo

comment created time in 15 days

issue commentPyTorchLightning/pytorch-lightning

Graceful interrupt not graceful

It matters for research where the number of epochs or the number of iterations should be strictly controlled for fair comparison.

ruotianluo

comment created time in 15 days

issue commentruotianluo/self-critical.pytorch

The problem with loader.get_batch(split) when changing the image number of splits

It's just using the schedule. I tried to implement the instance normalization but got no better results, and I think the person in that thread also didn't get better results.

kaelsunkiller

comment created time in 16 days

startedtimoschick/pet

started time in 16 days

issue openedPyTorchLightning/pytorch-lightning

Graceful interrupt not graceful

The current key interrupt will call on_train_end which will save the checkpoint. However when resuming from the saved checkpoint, it starts a new epoch, which is not graceful: presumably it should restart from the middle of the epoch(ideally); otherwise it should not save the interrupted checkpoint with the same name as normal checkpoints (for example, save as a new name).

created time in 16 days

push eventruotianluo/pytorch-lightning

William Falcon

commit sha b34c7add23553f10f6f0d7caf4177c67ee213f3a

Fixes #3668, #3887 as a bonus (#3888) * Fixes #3668, #3887 as a bonus * Fixes #3668, #3887 as a bonus

view details

Adrian Wälchli

commit sha 0823cdd59ca6f13afac4217b49209b22b713c1b7

Mocking Loggers (part 4a, mlflow) (#3884) * extensive mlflow test * revert accidental commits

view details

edenlightning

commit sha 2119184801ef94117e52ddb6ed1056d0d9fb9103

Fix docs for auto_lr_find (#3883) * Fix docs for auto_lr_find * change testcode to codeblock we are not showing a complete example here

view details

Adrian Wälchli

commit sha 893bed741f32dbea2f5d10a18fa4cfcfda3376b2

Mocking Loggers (part 3b, comet) (#3853) * ref * Mocking Loggers (part 3c, comet) (#3859) * mock comet * new line

view details

William Falcon

commit sha cb2a3265e5eb329a48fb44df6ab8fd74df62b85a

Fixes #2936 (no fix needed) (#3892)

view details

Jeff Yang

commit sha 90929fa4333e5136020e9f9dcb7c1133e4c290f3

Fix apt repo issue for docker (#3823) * fix docker repo issue * docker * docker * docker * no cudnn * no cudnn * try 16.04 Co-authored-by: Jirka Borovec <jirka@pytorchlightning.ai>

view details

William Falcon

commit sha 04303b3a2efb3783fd3c3cbd14bd90dc9d1098b3

added bug report model (#3901)

view details

Nicki Skafte

commit sha 3ab43dd7797d3076d31b1679ac00bceeee35e033

Fix lr finder for optimizers with states (#3897) * fix lr finder * changelog * add test

view details

Nicki Skafte

commit sha f745c4a773fa9742a1c3cc9d051af0acbfe411ec

doc update (#3894)

view details

edenlightning

commit sha e6e5406511b1a499421b49faeddfede72851ee09

update bug template (#3902)

view details

Adrian Wälchli

commit sha 6272fd39b3d6ea5ad6318d73eeda0a37509be0cd

xfail if not installed (#3860) include mkpatch fix test

view details

Teddy Koker

commit sha 96009266192905093d9c2347f8f1ab08bf1e3df3

Rename log_save_interval, row_log_interval (#3748) * Rename row_log_interval -> log_every_n_steps log_save_interval -> flush_logs_every_n_steps * Changelog * fixed title underline length * typo * Update pytorch_lightning/trainer/trainer.py Co-authored-by: Jirka Borovec <Borda@users.noreply.github.com> * Update pytorch_lightning/trainer/trainer.py Co-authored-by: Jirka Borovec <Borda@users.noreply.github.com> * pep8 + deprecation test * 'todo: remove in 1.1 comment' * 1.1 -> 0.11 * log * docs * depr API * add depr tests * note * miss Co-authored-by: Jirka Borovec <Borda@users.noreply.github.com> Co-authored-by: Jirka Borovec <jirka@pytorchlightning.ai>

view details

Jirka Borovec

commit sha f55a9cf63afe3c710ad0ab69c5e4444e44a4b0ea

fic CI parsing Horovod version (#3804)

view details

Sean Naren

commit sha e4a56fa5cfb5b67147c2013ae444ad0cd9a1b63a

Ensure global seed exists before passing into env subprocess.Popen call (#3904)

view details

William Falcon

commit sha 2cf17a3718f01b10fa2e9a14bad7dd6494965be1

Adds tests to make sure logging doesn't happen multiple times (#3899) * Makes sure logging doesn't ever happen from non-root zero * Makes sure logging doesn't ever happen from non-root zero * Makes sure logging doesn't ever happen from non-root zero * added bug report model * fix local model * fix local model * fix local model * fix local model

view details

Lezwon Castelino

commit sha 69833dad5b2a0e7e68ed60a91a5a8c32ae22f707

Added check to verify xla device is TPU (#3274) * tpu device check * replaced with xmp spawn * Revert "replaced with xmp spawn" This reverts commit 6835380f * replaced all instances of XLA_AVAILABLE * moved inner_f to global scope * made refactors * added changelog * added TPU_AVAILABLE variable * fix codefactor issues * removed form trainer and early stopping * add TORCHXLA_AVAILABLE check * added tests * refactoring * Update pytorch_lightning/utilities/xla_device_utils.py Co-authored-by: Adrian Wälchli <aedu.waelchli@gmail.com> * updated function names * fixed bug * updated CHANGELOG.md * added todo * added type hints * isort and black Co-authored-by: Adrian Wälchli <aedu.waelchli@gmail.com> Co-authored-by: William Falcon <waf2107@columbia.edu>

view details

maxjeblick

commit sha 39b3704285e40a29a5862c4d8145b68d3b35d45e

add current_epoch to dumped_params (#3261) * add current epoch to __dumped_params * log * reset * add to test * Update CHANGELOG.md Co-authored-by: Adrian Wälchli <aedu.waelchli@gmail.com> Co-authored-by: Jirka Borovec <jirka@pytorchlightning.ai> Co-authored-by: Nicki Skafte <skaftenicki@gmail.com> Co-authored-by: Adrian Wälchli <aedu.waelchli@gmail.com>

view details

Jirka Borovec

commit sha 064ae53d63389ed96e6ca4648eba55653aed0718

nb steps in early stop (#3909) * nb steps * if * skip * rev * seed * seed

view details

Nathan Painchaud

commit sha c510a7f90077140d60c47adf8e1e73638c2d1017

Additional test for logging during validation loop (#3907) * Added test for logging in validation step when using dict dataset with string value * fix recursive issue * fix recursive issue Co-authored-by: Nathan Painchaud <nathanpainchaud@gmail.com> Co-authored-by: William Falcon <waf2107@columbia.edu>

view details

Jirka Borovec

commit sha 4722cc0bf0fb8c5c902144f0d72dc55c3335e6d8

prune Results usage in notebooks (#3911) * notebooks * notebooks

view details

push time in 16 days

issue openedPyTorchLightning/pytorch-lightning

Slurm resubmit at the end of epoch.

From my understanding, the current resubmit will stop the model at the middle of epoch, which may have problem with dataloader resuming.

Is it possible that lightning automatically estimates that if a new epoch can be finished within the time limit, and decide if to halt or continue at the end of each epoch.

created time in 17 days

issue commentruotianluo/self-critical.pytorch

The problem with loader.get_batch(split) when changing the image number of splits

because it_max(which is the only place which is dependant on split in the collate_fn) is not used for training.

kaelsunkiller

comment created time in 17 days

issue commentruotianluo/self-critical.pytorch

The problem with loader.get_batch(split) when changing the image number of splits

I pushed some fix to the master, see if it helps.

kaelsunkiller

comment created time in 17 days

push eventruotianluo/ImageCaptioning.pytorch

Ruotian Luo

commit sha 5d1a632d721c5a533b721fbc85e52b0edfa2054f

update readme installing as a package

view details

Ruotian Luo

commit sha 02509ece2a00a7d59d5377a1888beb83e79d8d26

Use lmdbdict instead of original lmdb

view details

Ruotian Luo

commit sha c305eb009f3559da7041e57a4ff2296c35c71cce

bug fix: using lambda will make all colalte_fn to be collatefn(split='test'), because lambda is just a pointer to the local state, and the split will be the split at that local state, which is 'test'. (This is my guess, I don't know exactly the principle but this is my guess.) Use partial will return a new function so that the split will be remembered.

view details

Ruotian Luo

commit sha 67789e04340b8fa2e0a2e51e83cdb11e5b7a5b2c

Model: Add bart model.(The same as transformer, but less memory)

view details

Ruotian Luo

commit sha 8f908a043ad796401f49538aad849186e04e36d9

add tools for convert trans weights to bart weights.

view details

push time in 17 days

push eventruotianluo/self-critical.pytorch

Ruotian Luo

commit sha 5d1a632d721c5a533b721fbc85e52b0edfa2054f

update readme installing as a package

view details

Ruotian Luo

commit sha 02509ece2a00a7d59d5377a1888beb83e79d8d26

Use lmdbdict instead of original lmdb

view details

Ruotian Luo

commit sha c305eb009f3559da7041e57a4ff2296c35c71cce

bug fix: using lambda will make all colalte_fn to be collatefn(split='test'), because lambda is just a pointer to the local state, and the split will be the split at that local state, which is 'test'. (This is my guess, I don't know exactly the principle but this is my guess.) Use partial will return a new function so that the split will be remembered.

view details

push time in 17 days

push eventruotianluo/ImageCaptioning.pytorch

Ruotian Luo

commit sha 5d1a632d721c5a533b721fbc85e52b0edfa2054f

update readme installing as a package

view details

Ruotian Luo

commit sha 02509ece2a00a7d59d5377a1888beb83e79d8d26

Use lmdbdict instead of original lmdb

view details

Ruotian Luo

commit sha c305eb009f3559da7041e57a4ff2296c35c71cce

bug fix: using lambda will make all colalte_fn to be collatefn(split='test'), because lambda is just a pointer to the local state, and the split will be the split at that local state, which is 'test'. (This is my guess, I don't know exactly the principle but this is my guess.) Use partial will return a new function so that the split will be remembered.

view details

push time in 17 days

issue commentruotianluo/ImageCaptioning.pytorch

What is the beam_size for?

https://stackoverflow.com/questions/22273119/what-does-the-beam-size-represent-in-the-beam-search-algorithm Would this help?

sbkim052

comment created time in 17 days

startedMCG-NKU/CVPR_Template

started time in 17 days

startedgpeyre/SinkhornAutoDiff

started time in 17 days

push eventruotianluo/lmdbdict

Ruotian Luo

commit sha e485825420579f107dd48136e1c4eb590be08f2e

make _keys a list instead of set

view details

Ruotian Luo

commit sha 65e9a2eb5288e30dca8e12ad91970bec9234b9a1

safer delete

view details

Ruotian Luo

commit sha 0a2dbaccdacada50d20edca1659d2e2a3692b5f5

bump version

view details

push time in 20 days

startedfacebookresearch/sound-spaces

started time in 21 days

startedintersun/CoDIR

started time in 21 days

startedvoxel51/fiftyone

started time in 21 days

startedallenai/scitldr

started time in 21 days

startedttengwang/dense-video-captioning-pytorch

started time in 23 days

push eventruotianluo/pytorch-lightning

Nathan Raw

commit sha ba01ec9dbf3bdac1a26cfbbee67631c3ff1fadad

Edited using Colaboratory (#3601)

view details

William Falcon

commit sha 2a10cfaf3d4e0e8cec77e1e774040947aa0e53e9

clarify forward (#3609) * clarify forward * clarify forward

view details

William Falcon

commit sha f53e73963751413e7c1b4545c51bd722e26318d0

clarify forward (#3611) * clarify forward * clarify forward * clarify forward * clarify forward

view details

edenlightning

commit sha b4c553b9e6634b7194087be52012ff2c5733c326

update readme with new notebooks (#3608)

view details

Adrian Wälchli

commit sha d2a3d6aa8e8b69e6f373243bd25165a0963d7a53

update ambiguous info (#3613)

view details

William Falcon

commit sha c591013708cb7b049e8f517ae39e21331882b45a

enable any logged metric to be accessible in callbacks (#3598) * enable any logged or written metric to be accessible in callbacks * enable any logged or written metric to be accessible in callbacks * enable any logged or written metric to be accessible in callbacks * enable any logged or written metric to be accessible in callbacks * enable any logged or written metric to be accessible in callbacks * enable any logged or written metric to be accessible in callbacks * enable any logged or written metric to be accessible in callbacks * enable any logged or written metric to be accessible in callbacks * enable any logged or written metric to be accessible in callbacks * enable any logged or written metric to be accessible in callbacks * enable any logged or written metric to be accessible in callbacks * enable any logged or written metric to be accessible in callbacks * enable any logged or written metric to be accessible in callbacks * enable any logged or written metric to be accessible in callbacks * enable any logged or written metric to be accessible in callbacks * enable any logged or written metric to be accessible in callbacks * enable any logged or written metric to be accessible in callbacks * enable any logged or written metric to be accessible in callbacks * enable any logged or written metric to be accessible in callbacks * enable any logged or written metric to be accessible in callbacks * enable any logged or written metric to be accessible in callbacks * clarify forward * clarify forward * clarify forward * clarify forward

view details

Jirka Borovec

commit sha 37a59be21b73baffd68d0cc16bc31dd508d2b4c8

build more docker configs (#3533) * update build cases * list * matrix * matrix * builds * docker * -j1 * -q * -q * sep * docker * docker * mergify * -j1 * -j1 * horovod * copy

view details

William Falcon

commit sha 031274c25dedc92e383d2715e283a55a2b102d29

fix dp issues + update examples and test examples (#3618) * fix dp * fix dp * fix dp * fix dp * fix examples * fix examples * fix examples * fix examples * fix examples * fix examples * fix examples * fix examples * fix examples * fix examples * fix examples * fix examples * fix examples * fix examples * fix examples * fix examples * fix examples * fix examples

view details

William Falcon

commit sha 12184854f97f3d0ef8d72aaa801e661dc10d7058

Update README.md

view details

s-rog

commit sha 49767e424f2a2ba2893d27ed0c757815e355c86e

fix on_fit_start (#3616) * init * fix call_hook args

view details

Jamie Morton

commit sha a2574d7dd21604d499586ee6db7b2a8c37ac26ae

Adding clarifying documentation on the usage of second_order_closure (#3551) * Adding clarifying documentation on the usage of second_order_closure * oops typo * making functions more sane * fixing spacing issues - I think * Apply suggestions from code review * suggestions Co-authored-by: Jirka Borovec <Borda@users.noreply.github.com> Co-authored-by: Rohit Gupta <rohitgr1998@gmail.com>

view details

William Falcon

commit sha c0e26b87661d37f8fee07a443c63bdd68384f12b

fix examples (#3623) * fix examples * fix examples

view details

William Falcon

commit sha 2db86381c6e9468828418cc77ef203486ae8f72f

Update README.md

view details

Adrian Wälchli

commit sha 3affa0e49aac25ed0e02bea777d995c5219e5719

use tmpdir in tests when writing predictions to disk (#3561) * save to tmpdir * path

view details

ananthsub

commit sha c61e1e697d2e4ed5cc97c187e6704a44f2f77aff

Add stronger typing to gradient accumulation scheduler callback (#3558) * Update gradient_accumulation_scheduler.py add types for gradient accumulation scheduler callback * Update gradient_accumulation_scheduler.py

view details

William Falcon

commit sha c94c0a2b1ee6b444ab1ecf58059e922229d44436

fix examples (#3631) * fix examples * fix examples

view details

Jirka Borovec

commit sha aa52c930f422106f643efbb2b5057a79f3d26911

test examples (#3643) * test examples * testing * testing * typo * req * exception Co-authored-by: Adrian Wälchli <aedu.waelchli@gmail.com> Co-authored-by: Adrian Wälchli <aedu.waelchli@gmail.com>

view details

JackCaster

commit sha 618eb913da3cb4e1d2afc8b0870f7de182d70d5c

Update docstring for early_stop_callback default Trainer argument (#3641)

view details

Jeff Yang

commit sha a2120130edb49bed2d38a034a7bf5cdfa31a736e

Lightning docker image based on base-cuda (#3637) * use lightning CI docker * exclude py3.8 and torch1.3 * torch 1.7 * mergify * Apply suggestions from code review Co-authored-by: Jirka Borovec <jirka@pytorchlightning.ai> Co-authored-by: Jirka Borovec <Borda@users.noreply.github.com>

view details

William Falcon

commit sha b5f0af182d08c398803c2cb751e13951c1a58c0a

Update __init__.py

view details

push time in 23 days

startedresume/resume.github.com

started time in 24 days

startedshtoshni92/long-doc-coref

started time in 24 days

startedfacebookresearch/Mephisto

started time in 24 days

startedKaihuaTang/Long-Tailed-Recognition.pytorch

started time in 24 days

startedjoannezhouyi/Control_3D_Objects_with_Camera

started time in 24 days

startedXuezheMax/flowseq

started time in 24 days

startedNVlabs/imaginaire

started time in 24 days

startedlucidrains/vit-pytorch

started time in 24 days

startedZitongYu/3DCDC-NAS

started time in 25 days

issue commentruotianluo/ImageCaptioning.pytorch

ModuleNotFoundError: No module named 'captioning.modules'

You can try add sorting here https://github.com/ruotianluo/ImageCaptioning.pytorch/blob/master/captioning/data/dataloaderraw.py#L79

Tahiya31

comment created time in a month

issue commentruotianluo/ImageCaptioning.pytorch

ModuleNotFoundError: No module named 'captioning.modules'

add --force --language_eval 0 may solve it.

Tahiya31

comment created time in a month

startedyaodongyu/TRADES

started time in a month

startedkeitakurita/contextual_embedding_bias_measure

started time in a month

issue commentruotianluo/ImageCaptioning.pytorch

ModuleNotFoundError: No module named 'captioning.modules'

you can run pip install -e . under the root directory.

Tahiya31

comment created time in a month

startedZeweiChu/NatCat

started time in a month

startedEssayKillerBrain/EssayKiller_V2

started time in a month

startedfacebookresearch/frankmocap

started time in a month

startedgenforce/genforce

started time in a month

startedkrrish94/nerf-pytorch

started time in a month

startedEpiphqny/CondInst

started time in a month

startedtsujuifu/pytorch_sscr

started time in a month

issue openedhuggingface/tokenizers

thread '<unnamed>' panicked at 'The global thread pool has not been initialized.:

thread '<unnamed>' panicked at 'The global thread pool has not been initialized.: ThreadPoolBuildError { kind: IOError(Os { code: 11, kind: WouldBlock, message: "R esource temporarily unavailable" }) }', /github/home/.cargo/registry/src/github.com-1ecc6299db9ec823/rayon-core-1.7.0/src/registry.rs:168:5 Has anyone seen this error.

created time in a month

startedYyzHarry/imbalanced-semi-self

started time in a month

startedfacebookresearch/ELI5

started time in a month

startedlinjieli222/HERO

started time in a month

issue commentruotianluo/self-critical.pytorch

Train using self critical

Are you using the cider submodule in the repo?

alibabadoufu

comment created time in a month

startedgoogle/spectral-density

started time in a month

startedgoogle-research-datasets/paws

started time in a month

issue openedNVIDIA/apex

Why lazy init, and how to make it not lazy

Hi,

I met this problem https://github.com/NVIDIA/apex/issues/480#issuecomment-698696982

It seems the reason is when you load the first time, the saved states will be cast to fp16, while at this time the states are not properly initialized because of the lazy_init. After the first time to call amp.scale_loss, the states of optimizer are properly initialized and the states will be recast to fp32, and the precision difference here will cause the problem.

Then I attempted to move lazy_init right after the amp.intialize, but funnily the loss will still spike but less severe. I am wondering: 1 why do you want to use lazy init 2 where should put lazy_init if I want to manually init.

created time in a month

issue commentpytorch/pytorch

libgcc_s.so.1 must be installed for pthread_cancel to work

This problem disappear after I reinstall 1.6

ruotianluo

comment created time in a month

issue commentruotianluo/ImageCaptioning.pytorch

Eval runs ok first time but then throw an "ZeroDivisionError: division by zero"

Replace line 74,75 with if opt.dump_json == 1: # dump the json json.dump(predictions, open('vis/vis.json', 'w'))

Tetsujinfr

comment created time in a month

issue commentruotianluo/ImageCaptioning.pytorch

Eval runs ok first time but then throw an "ZeroDivisionError: division by zero"

Can you show the full error traceback?

Tetsujinfr

comment created time in a month

issue commentruotianluo/ImageCaptioning.pytorch

Eval runs ok first time but then throw an "ZeroDivisionError: division by zero"

I am in the middle of something, but I believe language_eval is an argument, would you mind try it?

Tetsujinfr

comment created time in a month

issue commentruotianluo/ImageCaptioning.pytorch

Eval runs ok first time but then throw an "ZeroDivisionError: division by zero"

remove what I asked you to add. That was wrong.

Tetsujinfr

comment created time in a month

issue commentruotianluo/ImageCaptioning.pytorch

Eval runs ok first time but then throw an "ZeroDivisionError: division by zero"

Do you want to turn --language_eval to 0?

Tetsujinfr

comment created time in a month

issue commentruotianluo/ImageCaptioning.pytorch

Eval runs ok first time but then throw an "ZeroDivisionError: division by zero"

Sorry, the fix in #100 is already there. Interesting, are you using the master?

Tetsujinfr

comment created time in a month

issue commentruotianluo/ImageCaptioning.pytorch

Eval runs ok first time but then throw an "ZeroDivisionError: division by zero"

tell me if the solution in https://github.com/ruotianluo/ImageCaptioning.pytorch/issues/100 fix it.

If so, I will push a bug fix.

Tetsujinfr

comment created time in a month

issue commentNVIDIA/apex

loss spike after checkpoint reload

Add optimizer.load_state_dict right before the first optimizer.step works for me too. (I basically add a manual checkpoint loading in optimizer_step of LightningModule)

For general pytorch user, this is what I have:

state_dict = torch.load(optimizer_ckpt)
optimizer.load_state_dict(state_dict)
....
if_equal(optimizer.state_dict(), state_dict) # return True
optimizer.load_state_dict(state_dict)
if_equal(optimizer.state_dict(), state_dict) # return True
optimizer.step()
optimizer.zero_grad()

I also check if the optimizer states match the checkpoint; it does match before my manual loading. This bug is so weird.

williamFalcon

comment created time in a month

issue commenthuggingface/transformers

Bart encoder with add_final_layer_norm

Any idea?

ruotianluo

comment created time in a month

issue commentPyTorchLightning/pytorch-lightning

Unlink 16-bit Precision and AMP

I believe

        precision=16,
        amp_backend='apex',
        amp_level='O2',

gives you precision 16.

tsteffek

comment created time in a month

push eventruotianluo/pytorch-lightning

Abe Botros

commit sha 76c4afb840b0ae5fcafee07d527c58e9245d099d

Fix IoU score for classes not present in target or pred (#3098) * Fix IoU score for classes not present in target or pred Fixes #3097 - Allow configurable not_present_score for IoU for classes not present in target or pred. Defaults to 1.0. - Also allow passing `num_classes` parameter through from iou metric class down to its underlying functional iou call. * Changelog: move IoU not-present score fix to [unreleased] * IoU: avoid recomputing class presence in target and pred Use already-computed support, true positives, and false positives to determine if a class is not present in either target or pred. * Test IoU against sklearn jaccard_score Also add TODO to test our IoU's not_present_score against sklearn's jaccard_score's zero_division when it beecomes available. * IoU: remove_bg -> ignore_index Fixes #2736 - Rename IoU metric argument from `remove_bg` -> `ignore_index`. - Accept an optional int class index to ignore, instead of a bool and instead of always assuming the background class has index 0. - If given, ignore the class index when computing the IoU output, regardless of reduction method. * Improve documentation for IoU not_present_score * Update default IoU not_present_score to 0.0 * Add note about IoU division by zero * Rename IoU not_present_score -> absent_score * Update IoU absent score changelog wording * Condense IoU absent_score argument docstring * Remove unnecessary IoU ignore_index comment * docstrings * isort * flake8 * Fix test of IoU against sklearn jaccard Use macro instead of micro averaging in sklearn's jaccard score, to match multi-class IoU, which conventionally takes per-class scores before averaging. Co-authored-by: rohitgr7 <rohitgr1998@gmail.com>

view details

Jirka Borovec

commit sha 7b64472ced732c6762978f2a3a844f6ef5261157

fix lib paths after Wandb 0.10 (#3520) * try * try * drop 0.20 * drop 0.19.5 * -U * Fixed Horovod in CI due to wandb==0.10.0 sys.path modifications (#3525) Co-authored-by: Jirka Borovec <Borda@users.noreply.github.com> * format * wb freeze * types Co-authored-by: Travis Addair <taddair@uber.com>

view details

Jeff Yang

commit sha 8be79a9a967347427d2b376707ba7a560507948b

stable, dev PyTorch in Dockerfile and conda gh actions (#3074) * dockerfile and actions file * dockerfile and actions file * added pytorch conda cpu nightly * added pytorch conda cpu nightly * recopy base reqs * gh action `include` torch nightly * add pytorch nightly & conda gh badge * rebase * fix horovod * proposal refactor * Update .github/workflows/ci_pt-conda.yml Co-authored-by: Jirka Borovec <Borda@users.noreply.github.com> * Update .github/workflows/ci_pt-conda.yml Co-authored-by: Jirka Borovec <Borda@users.noreply.github.com> * update * update * fix cmd * filled && * fix * add -y * torchvision >0.7 allowed * explicitly install torchvision * use HOROVOD_GPU_OPERATIONS env variable * CI * skip 1.7 * table Co-authored-by: Jirka Borovec <jirka@pytorchlightning.ai> Co-authored-by: Jirka Borovec <Borda@users.noreply.github.com>

view details

Carlos Mocholí

commit sha a9c0ed920a5a6d0c2bb92f2c61046aeb5070db4f

Fix log debug call (#3528)

view details

Nathan Raw

commit sha c46de8a3d4a18e5030dc9f07e96bb6abc7643945

Fix misuse of transforms in docs (#3546) * :pencil: docs * :pencil: docs * :pencil: docs * :pencil: docs * Apply suggestions from code review Co-authored-by: Adrian Wälchli <aedu.waelchli@gmail.com> Co-authored-by: Jirka Borovec <Borda@users.noreply.github.com> Co-authored-by: Adrian Wälchli <aedu.waelchli@gmail.com>

view details

Lucas Steinmann

commit sha 197acd535fee5e79dafeeff14cc742095c77bd70

Fix early stopping with training step's return dict (#3347) * Fixes the test for early stopping without val step. The expression which checked, if early stopping was triggered, had an off-by-one error and hence was true even if early stopping was not triggered. Furthermore set patience to 0 and max epochs to 10, to ensure loss has enough time to flatten. * Fixes early stopping without val step. The issue has been, that only `early_stop_on` key was checked and not an arbitrary monitor key. * Fixes branch, which checks whether early stopping is done during validation. Before only `val_early_stop_on` was checked. Since arbitrary keys can be used, the set of possible validation keys cannot be exhaustive. Hence this disables "early stopping on_train_epoch_end" via an instance attribute if early stopping was executed in on_validation_epoch_end. Furthermore adds a test, which ensures arbitrary keys work. * Improve check whether eval results are used. Only disable early checking with train results if eval results are actually used. Before they were always disabled in ``on_validation_epoch_end``. Rename and document instance variable, to make it more clear. * Remove wrong documentation on behaviour of early stopping with train result' dict. * Apply suggestions from code review * Apply suggestions from code review Co-authored-by: Jirka Borovec <Borda@users.noreply.github.com>

view details

Carlos Mocholí

commit sha 580b04b490d4d6819133a5604ea0ef82e2a21727

Fix ModelCheckpoints name formatting (#3163) * Fix ModelCheckpoint's name formatting * Fix failing tests * Add dot to CHECKPOINT_SUFFIX * Set variables to their default values at the end of tests * Fix logic for filepath='' and filename=None. Add test * Fix Windows tests * Fix typo. Remove leading line break and zeroes * Remove CHECKPOINT_SUFFIX * Fix typos. Use appropriate f-string format * Apply suggestions from code review * Fix broken tests after #3320 * Finish changes suggested by Borda * Use explicit test var names * Apply suggestions Co-authored-by: Rohit Gupta <rohitgr1998@gmail.com> * Apply suggestions Co-authored-by: Rohit Gupta <rohitgr1998@gmail.com> * Update CHANGELOG * Apply suggestions from code review * for * prepend whitespace in warn msg Co-authored-by: Jirka Borovec <Borda@users.noreply.github.com> Co-authored-by: Rohit Gupta <rohitgr1998@gmail.com> Co-authored-by: Adrian Wälchli <aedu.waelchli@gmail.com>

view details

Boris Feld

commit sha e2af4f120e300ccf41f26a52b80be038b2e87a97

Improve Comet Logger pickled behavior (#2553) * Improve Comet Logger pickled behavior * Delay the creation of the actual experiment object for as long as we can. * Save the experiment id in case an Experiment object is created so we can continue the same experiment in the sub-processes. * Run pre-commit on the comet file. * Handle review comment Make most Comet Logger attribute protected as they might not reflect the final Experiment attributes. Also fix the typo in the test name. * Ensure that CometLogger.name and CometLogger.version always returns str * Add new test for CometLogger.version behavior * Add new tests for CometLogger.name and CometLogger.version * Apply review suggestions * Apply suggestions from code review Co-authored-by: Jirka Borovec <Borda@users.noreply.github.com> * Remove extraneous comments in Comet logger tests * Fix lint issues * Apply suggestions from code review Co-authored-by: Adrian Wälchli <aedu.waelchli@gmail.com> Co-authored-by: Jirka Borovec <Borda@users.noreply.github.com> Co-authored-by: Adrian Wälchli <aedu.waelchli@gmail.com>

view details

Lucas Steinmann

commit sha 07006238ea3a52843ecbd2c257ac725e7902b803

Early stopping doc (#3193) * Updated explanation to enabling early stopping via boolean flag. Now also includes case of returning Result objects. * Improved API documentation for checkpoint_on and early_stp_on in results. * Apply suggestions from code review * Fix terminology. * Fix wrong documentation. Strict checking is disabled when using structured results. * element typo * update remaining edits from code review Co-authored-by: Adrian Wälchli <aedu.waelchli@gmail.com>

view details

Nathan Raw

commit sha 00ab67edc3c32ebc580f7c92da2b3ed489e8435b

Make dims a property in datamodule (#3547) * :bug: make dims a property * :bug: fix

view details

Jirka Borovec

commit sha 8eb77cd06aff18c14242e194f6959752a9b08817

drop v0.10 deprecated (#3454) * drop v0.10 deprecated * import * missed

view details

Rohit Gupta

commit sha 07b857769a3c9694791c34a46b4b25c5edf18a57

Allow kwargs in Wandb & Neptune + kwargs docstring (#3475) * Allow kwargs in WandbLogger * isort * kwargs docstring * typo * kwargs for other loggers * pep and isort * formatting * fix failing test Co-authored-by: Adrian Wälchli <aedu.waelchli@gmail.com>

view details

Yigit Ozen

commit sha 67fed01a16bb4026e2435bf112738302e1769654

Fixed failed import warning for torch.distributed.group (#3553)

view details

clement-f

commit sha e7ee473c90da76e0b406c7bfb6519540be10d517

Fix typo in loggers.rst (#3557)

view details

ananthsub

commit sha 31388eda1a16c128e7c85dc1f6e7f727565ff408

Add type hints to model checkpoint callback (#3560) * Update gradient_accumulation_scheduler.py add types for gradient accumulation scheduler callback * Apply black formatting to model checkpoint callback auto-format, no other changes * Update gradient_accumulation_scheduler.py drop other changes * Add type hints to model checkpoint callback * Update model_checkpoint.py remove trainer/lightning modules types to avoid circular import

view details

ananthsub

commit sha 79607e4b9497d7026e77705046746bb575e6ab45

Black format the model checkpoint callback (#3559) * Update gradient_accumulation_scheduler.py add types for gradient accumulation scheduler callback * Apply black formatting to model checkpoint callback auto-format, no other changes * Update gradient_accumulation_scheduler.py drop other changes Co-authored-by: William Falcon <waf2107@columbia.edu>

view details

Jirka Borovec

commit sha 0284f7ab5a9f1749f7904de5c27bd0f432f2db74

nightly releases (#3552) * nightly * nightly * ls

view details

Adrian Wälchli

commit sha 99f05ed23f818d4f21c2c6925a66e75df606c859

fix warnings on windows (#3555)

view details

Adrian Wälchli

commit sha e6c7548b306055e41552e23d57f0057e7f441256

Fix overfit_batches > 0 on distributed_backend = "ddp" (#3534) * example * ex * example * sampler * fix * fix * remove example * changelog

view details

William Falcon

commit sha 9acee67c31c84dac74cc6169561a483d3b9c9f9d

fixes 3549 (#3564)

view details

push time in a month

issue commentruotianluo/ImageCaptioning.pytorch

about model zoo

  1. https://drive.google.com/drive/folders/0B7fNdx_jAqhtQ1E0NUUyRjVtb0U?usp=sharing
  2. neuraltalk2 (specific to that repo) randomly select train/val/test from all the coco images, while in karparthy split, the val and test were chosen from coco val2014.
  3. I actually suggest running it by yourself, I think it's fairly easy to run, and should be able to get better results than what's in the paper. All the models I trained for that project are already archived, it's a little bit complicated to mine in that. You can try to run it first; if you have any problem, let me see how I can retrieve from the archived data.
wanboyang

comment created time in a month

issue commentruotianluo/ImageCaptioning.pytorch

about model zoo

1 == FC in [1] 2 is in my paper https://arxiv.org/abs/2003.09971. "att2in+self_critical" is Att2in in [2] (roughly speaking, there may be some difference in architecture detail). 3 correct.

wanboyang

comment created time in a month

issue commenthuggingface/transformers

Some weights of AlbertModel were not initialized ['albert.embeddings.position_ids']

Position_ids seems unnecessary to be saved? Why not use register_buffer with persistent==False

vgaraujov

comment created time in a month

startedpyutils/line_profiler

started time in a month

pull request commentPyTorchLightning/pytorch-lightning

Finish PR #2432: Imagenet example updates + basic testing

Doesn't that assume trainer already has the model? My understanding, if you want to use this, you have to run it after trainer.fit(model) otherwise trainer doesn't know what is the model.

awaelchli

comment created time in a month

pull request commentPyTorchLightning/pytorch-lightning

Finish PR #2432: Imagenet example updates + basic testing

https://github.com/PyTorchLightning/pytorch-lightning/blob/8eb77cd06aff18c14242e194f6959752a9b08817/pytorch_lightning/trainer/trainer.py#L538

It won't use the ckpt_path

awaelchli

comment created time in a month

pull request commentPyTorchLightning/pytorch-lightning

Finish PR #2432: Imagenet example updates + basic testing

I am not sure what is the most elegant solution: One possibility would just be:

model.load_state_dict(torch.load(args.resume_from_checkpoint, map_location=''cpu)['state_dict'])

But I think it's too ugly.

awaelchli

comment created time in a month

pull request commentPyTorchLightning/pytorch-lightning

Finish PR #2432: Imagenet example updates + basic testing

It seems the current --evaluate won't load the checkpoint from --resume_from_checkpoint, right?

I saw in the doc:

# (4) test with an explicit model (will use this model and not load a checkpoint)
traiener.test(model)
awaelchli

comment created time in a month

issue commentPyTorchLightning/pytorch-lightning

cuda runtime error (101) : invalid device ordinal

Training with: commit c64520.

I am still getting error....... I am using CUDA_VISIBLE_DEVICES=3 (set by slurm), and in the code I have Trainer(..., gpus=-1, ...)

ruotianluo

comment created time in a month

startedhendrycks/test

started time in a month

push eventruotianluo/pytorch-lightning

William Falcon

commit sha d13e5c9e53db72064900e3c17f08b779fc58a868

document lightiningmodule better (#2920) * updated docs

view details

Brendan Fahy

commit sha 56396abe9839fa075bcc087c32f098145b0bdc9f

fix checkpointing to remote file paths (#2925)

view details

Phil

commit sha e3528afae3f178cf9d5d8ea6bc3f8a876646054a

Move optimizer creation after device placement for ddp backends. (#2904)

view details

Nathan Raw

commit sha 118bd14d164f080cb664c9daf6e0c534ca6617d6

Update CONTRIBUTING.md (#2927) * Update CONTRIBUTING.md * Update CONTRIBUTING.md * Apply suggestions from code review Co-authored-by: Jirka Borovec <Borda@users.noreply.github.com>

view details

William Falcon

commit sha a46130cdc19447ea86466bf4fcd0c7226e968a05

add weighted average to results obj (#2930) * track batch size in result obj

view details

Rosario Scalise

commit sha f9d88f8088bbc27341f9d19c4aaf27259d22e072

Support **DictConfig hparam serialization (#2519) * change to OmegaConf API Co-authored-by: Omry Yadan <omry@fb.com> * Swapped Container for OmegaConf sentinel; Limited ds copying * Add Namespace check. * Container removed. Pass local tests. Co-authored-by: Omry Yadan <omry@fb.com>

view details

Adrian Wälchli

commit sha 411914bd2bc225b0500f993d74f2f16d71f7ac6a

Fix hparams loading for model that accepts *args (#2911) * fix hparams loading for model that accepts *args * add test case * changelog * pep * fix test Co-authored-by: William Falcon <waf2107@columbia.edu>

view details

Jirka Borovec

commit sha fcf3c40172687b7d4a454cf4ea03419588c9e876

update changelogs

view details

Gerardo Roa Dabike

commit sha f6a3d8fd8da36a332d7bcf43a3967f3c5f10dbb4

GPU Usage Logger (#2932) * GPU utilisation Callback * GPU utilisation Callback * Fixing style * Fixing style * Fixing CodeFactor: partial executable path * Fix a misspelling in the Class name

view details

edenlightning

commit sha 2c31beccfbfe0752306122a2ba6f9822ec5cb6b8

Add magicleap/atlas to community examples (#2937)

view details

Jirka Borovec

commit sha 665c1507f035d7777a5ce9c414c0ba73e120fb74

deterministic=True (#2944)

view details

Jirka Borovec

commit sha 519b97effdeaccb0833ca20f1771e14c6e161f06

Clean save (#2933) * thr deterministic=True * clean * clean * Apply suggestions from code review Co-authored-by: Vadym Stupakov <vadim.stupakov@gmail.com> * Apply suggestions from code review Co-authored-by: Vadym Stupakov <vadim.stupakov@gmail.com>

view details

William Falcon

commit sha 6c5a0a172fbcde719cffd57e380f1db9478797a2

Resultd (#2947) * updated docs

view details

Jirka Borovec

commit sha 4354690e55704f0c1742e5a95411b63c6055ddc7

add apex test (#2921) * add apex test * rename * level * events * wrap * evt * miss * apex * apex * apex * apex * apex * apex * Update tests/models/test_amp.py Co-authored-by: William Falcon <waf2107@columbia.edu> * notes * notes Co-authored-by: William Falcon <waf2107@columbia.edu>

view details

William Falcon

commit sha 054ac94bd1da2c20b0f5b5ee2c3ec177469c3908

track batch size (#2950)

view details

William Falcon

commit sha 2c935d048e69a2890889dea768ecceb0252cf321

track batch size (#2954)

view details

shijianjian

commit sha 18d31a3b634d7dc6873bf918cbfb49efc8efc835

Added strict=False for load_from_checkpoint (#2819) * Added strict=False and hparams_file accepcts dict * Apply suggestions from code review Co-authored-by: Justus Schock <12886177+justusschock@users.noreply.github.com> * Type check fix * Added tests * Linting & test fix * Removed redundant code & test * Added strict=False and hparams_file accepcts dict * Apply suggestions from code review Co-authored-by: Justus Schock <12886177+justusschock@users.noreply.github.com> * Type check fix * Added tests * Linting & test fix * Removed redundant code & test * Apply suggestions from code review * tests * tests * chlog * Update tests/models/test_restore.py Co-authored-by: Rohit Gupta <rohitgr1998@gmail.com> * update test comments * Added docstring for the strict attribute * Added supplementary tests * Update saving.py * Apply suggestions from code review Co-authored-by: Rohit Gupta <rohitgr1998@gmail.com> * pep8, removed extra func Co-authored-by: Jirka Borovec <Borda@users.noreply.github.com> Co-authored-by: Justus Schock <12886177+justusschock@users.noreply.github.com> Co-authored-by: Jirka Borovec <jirka@pytorchlightning.ai> Co-authored-by: Rohit Gupta <rohitgr1998@gmail.com> Co-authored-by: Adrian Wälchli <aedu.waelchli@gmail.com> Co-authored-by: ananyahjha93 <ananya@pytorchlightning.ai>

view details

shijianjian

commit sha 53f855cdbf1214dbfae8c431749e07a8bdb44b60

Added strict=False for load_from_checkpoint (#2819) * Added strict=False and hparams_file accepcts dict * Apply suggestions from code review Co-authored-by: Justus Schock <12886177+justusschock@users.noreply.github.com> * Type check fix * Added tests * Linting & test fix * Removed redundant code & test * Added strict=False and hparams_file accepcts dict * Apply suggestions from code review Co-authored-by: Justus Schock <12886177+justusschock@users.noreply.github.com> * Type check fix * Added tests * Linting & test fix * Removed redundant code & test * Apply suggestions from code review * tests * tests * chlog * Update tests/models/test_restore.py Co-authored-by: Rohit Gupta <rohitgr1998@gmail.com> * update test comments * Added docstring for the strict attribute * Added supplementary tests * Update saving.py * Apply suggestions from code review Co-authored-by: Rohit Gupta <rohitgr1998@gmail.com> * pep8, removed extra func Co-authored-by: Jirka Borovec <Borda@users.noreply.github.com> Co-authored-by: Justus Schock <12886177+justusschock@users.noreply.github.com> Co-authored-by: Jirka Borovec <jirka@pytorchlightning.ai> Co-authored-by: Rohit Gupta <rohitgr1998@gmail.com> Co-authored-by: Adrian Wälchli <aedu.waelchli@gmail.com> Co-authored-by: ananyahjha93 <ananya@pytorchlightning.ai>

view details

SiddhantRanade

commit sha 88bfed371e9597e813384b3d951b0e5280be71bd

Fix enforce_datamodule_dataloader_override() for iterable datasets (#2957) This function has the if statement `if (train_dataloader or val_dataloaders) and datamodule:`. The issue is similar to that in https://github.com/PyTorchLightning/pytorch-lightning/pull/1560. The problem is that the `if(dl)` translates to `if(bool(dl))`, but there's no dataloader.__bool__ so bool() uses dataloader.__len__ > 0. But... dataloader.__len__ uses IterableDataset.__len__ for IterableDatasets for which __len__ is undefined. The fix is also the same, the `if dl` should be replaced by `if dl is not None`. Co-authored-by: mergify[bot] <37929162+mergify[bot]@users.noreply.github.com>

view details

Jeff Yang

commit sha 07c023c32f114925b6eef6f371143cf1f8ccc1e8

fix(docs): docstring for amp_backend (#2960) * fix(docs): docstring for amp_backend * fix(docs): early_stop_checkpoint -> early_stop_callback * docs Co-authored-by: ananyahjha93 <ananya@pytorchlightning.ai>

view details

push time in a month

more