GitLab

Summary: Currently when launching a training flow, we read number of processes from resources.num_gpus. To be backward compatible with existing D2 (https://github.com/facebookresearch/d2go/commit/f82d44d3c33e6c781a3c6f2b27b376fdfbaeda53)Go training config, this diff changes to dist_config.num_processes_per_machine instead.

Reviewed By: wat3rBro

Differential Revision: D28630334

fbshipit-source-id: 3c684cd56e5d2e247c7b82e1d1eeff0f39e59ee4

Summary:
Pull Request resolved: https://github.com/facebookresearch/d2go/pull/76

Detectron2GoRunner doesn't contain configs about exporting RCNN models, use GeneralizedRCNNRunner instead

Reviewed By: zhanghang1989

Differential Revision: D28652627

fbshipit-source-id: 6f324f608d8b2abdf98179a36e4b79837f135340

Summary:
Pull Request resolved: https://github.com/facebookresearch/d2go/pull/70

DDP supports an fp16_compress_hook which compresses the gradient to FP16 before communication. This can result in a significant speed up.

Add one argument `_C.MODEL.DDP_FP16_GRAD_COMPRESS` to trigger it.

Reviewed By: zhanghang1989

Differential Revision: D28467701

fbshipit-source-id: 3c80865222f48eb8fe6947ea972448c445ee3ef3

Differential Revision:
D27881742 (https://github.com/facebookresearch/d2go/commit/90aff5daf608473dd312b300db8615326fa40a37)

Original commit changeset: 34a3ab7a88f4

fbshipit-source-id: 42c03b4f2b69c656b26774a4665b84b832262650

Summary: Pull Request resolved: https://github.com/facebookresearch/d2go/pull/73

Reviewed By: kazhang

Differential Revision: D28610607

Pulled By: zhanghang1989

fbshipit-source-id: c079bc6fff64cd452db6750b3e984546c47fdfe7

Summary:
- Enable sdk inference config specification in export step. This enables adding the sdk configuration as part of model file in the export step. The sdk config can be specified as infernece_config.yaml and is zipped together with torchscript model. The main goal of sdk configuration is to control the model inference behavior with model.
- SDK inference config design doc: https://docs.google.com/document/d/1j5qx8IrnFg1DJFzTnu4W8WmXFYJ-AgCDfSQHb2ACJsk/edit
- One click fblearner pipeline is in next diff on the stack

Differential Revision: D27881742

fbshipit-source-id: 34a3ab7a88f456b74841cf671ea1b3f678cdb733

Summary: Option to change only bounding boxes, others remain the same.

Differential Revision: D28339388

fbshipit-source-id: 7a6d4c5153cf10c473992119f4c684e0b9159b44

Summary: Add dataset visualization so that we could visualize test results in Tensorboard.

Reviewed By: zhanghang1989

Differential Revision: D28457363

fbshipit-source-id: 4c2fd9dce349c6fb9e1cec51c9138cf0abb45d7b

Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/58344

remove a helper function thats more trouble then its worth.

ghstack-source-id: 129131889

Reviewed By: dhruvbird

Differential Revision: D28460607

fbshipit-source-id: 31bd6c1cc169785bb360e3113d258b612cad47fc

Summary: Create new CfgNode that is consistent with the parent node.

Reviewed By: zhanghang1989

Differential Revision: D28318466

fbshipit-source-id: 38cb84de6bdfec2b283c4d9a1090cad47c118c9c

Summary: Pull Request resolved: https://github.com/facebookresearch/d2go/pull/68

Reviewed By: kazhang

Differential Revision: D28426745

Pulled By: zhanghang1989

fbshipit-source-id: 5b82b03847e5e1d06d1bfcdaf0192d65e9202e77

Summary:
Pull Request resolved: https://github.com/facebookresearch/d2go/pull/62

Lightning trainer set max step to cfg.SOLVER.MAX_ITER. However, this is the max iteration for all nodes, in multi-node training, we need to scale it down, as well as eval period and other configs.
This diff calls `auto_scale_world_size` before passing the config to trainer.

Reviewed By: wat3rBro

Differential Revision: D28140877

fbshipit-source-id: 2639ae58773a4ec2a0cc59dfefb2f5d9b1afe1a8

Reviewed By: zhanghang1989

Differential Revision: D28346653

fbshipit-source-id: d80a1f824b097c05029edb171739a4928e47e4d8

Synchronize PyTorchLightning/pytorch-lightning (revision 7b283e3c@master) to github/third-party/PyTorchLightning/pytorch-lightning

Summary:
# Manual
 - remove fixme's in `model_checkpoint.py`, `parameter_monitor.py`, `test_quantization.py`, and `speed_monitor.py` now that `Trainer` is properly annotated.
- update `test_quantization.py` to `trainer.train_loop.global_step` instead of `trainer.global_step` which is a read-only.
- update `loop_callback.py` to read from `train_loop` for `batch_idx` (which is no longer available).

# Automatic
### New commit log messages
  7b283e3c Bugfix/Multiple dataloaders (#7433)
  d7c44cc6 Docs: sync chlog 1.3.1 (#7478)
  fdf50a5e Mark certain Trainer APIs as protected (#7420)
  ad9118f0 remove trainer hidden state | sanity refactor [1 / n] (#7437)
  4a1134db Log epoch metrics before firing the `on_evaluation_end` hook (#7272)
  b65ae794 Automatically check `DataModule.has_{setup,teardown,prepare_data}` [2/2] (#7238)
  8660d8cf [pre-commit.ci] pre-commit autoupdate (#7475)
  f6fe715e Fix Sphinx argument deprecation (#7464)

Reviewed By: shuyingsunshine21

Differential Revision: D28353491

fbshipit-source-id: 98b87d99e2f09b47b07270858fcbdb5d5299730b

Summary:
Pull Request resolved: https://github.com/facebookresearch/d2go/pull/67

When extending RCNN, sometimes we need to also modify `prepare_for_export/quant`. It might be inconvenient to create new meta-arch just for the purpose of modifying `prepare_for_export/quant`, therefore we create the registry so user can change them via config.

Reviewed By: zhanghang1989

Differential Revision: D28308056

fbshipit-source-id: 4f169eb38292a75d15d3b9f44694480eaa9244e0

Summary: Pull Request resolved: https://github.com/facebookresearch/d2go/pull/66

Reviewed By: zhanghang1989

Differential Revision: D28308297

fbshipit-source-id: 7f6effad7d289546b50f7b16739a04c358a22e21

Summary:
Pull Request resolved: https://github.com/facebookresearch/d2go/pull/59

* We have an internal dependency:
```
d2go/export/logfiledb.py", line 8, in <module>
    from mobile_cv.torch.utils_caffe2.ws_utils import ScopedWS
    ModuleNotFoundError: No module named 'mobile_cv.torch'
```
This cause the failure of unittest on GitHub
https://github.com/facebookresearch/d2go/pull/58/checks?check_run_id=2471727763

* use python 3.8 because another unittest failure on github ci
```
from typing import final
ImportError: cannot import name 'final' from 'typing' (/usr/share/miniconda/lib/python3.7/typing.py)
```

Reviewed By: wat3rBro

Differential Revision: D28109444

fbshipit-source-id: 95e9774bdaa94f622267aeaac06d7448f37a103f

Summary: Add upperbound-model based video evaluation dataset

Reviewed By: yc-fb

Differential Revision: D28122534

fbshipit-source-id: 37cf5ece1c6a5d537b91ae2f7db74a35b97c210d

Summary:
Pull Request resolved: https://github.com/facebookresearch/d2go/pull/63

As titled.

Reviewed By: zhanghang1989

Differential Revision: D27945627

fbshipit-source-id: 58180ace6ab327a5eb025f3227ae6c22601a3580

Summary: build_traceable_model is deprecated, remove it from default runner.

Reviewed By: zhanghang1989

Differential Revision: D28243469

fbshipit-source-id: e09729dcc083dc273329e1c2f731065578d8ea42

Reviewed By: zhanghang1989

Differential Revision: D28243043

fbshipit-source-id: fae5491d8978bcdb096fcee69846c707cda0ac00

Summary: `augment_model_with_bundled_inputs` can compress the tensor when values are constant, however it requires contiguous layout and `zero_like` can return non-contiguous ones

Reviewed By: zhanghang1989

Differential Revision: D28224987

fbshipit-source-id: 32b13728ff8fadd53412dbf2d59c4b46e92af04a

Summary: Add a bounding manipulation tool to padding bounding box data.

Reviewed By: newstzpz

Differential Revision: D28082071

fbshipit-source-id: f168cae48672c4fa5c4ec98697c57ed7833787ab

Summary:
[WIP] Will add pretrained weights and update model url & scores

build mask head using fbnet builder and retrain weights

Reviewed By: wat3rBro

Differential Revision: D27992340

fbshipit-source-id: a216a99954eb3784438d595cd09cbb19e70ec3c3

Reviewed By: newstzpz

Differential Revision: D27747996

fbshipit-source-id: 6ae3b89c3944098828e246e5a4a89209b8e171a1

Summary:
1. Add a keypoint metadata registry for registering different keypoint metadata
2. Add option to inject_coco_dataset for adding keypoint metadata

Reviewed By: newstzpz

Differential Revision: D27730541

fbshipit-source-id: c6ba97f60664fce4dcbb0de80222df7490bc6d5d

Reviewed By: zhanghang1989

Differential Revision: D28083131

fbshipit-source-id: 8bad642800d3923db3f42047d1b1d85625c01bd9

Reviewed By: zhanghang1989

Differential Revision: D28081681

fbshipit-source-id: 3722f5db668c36c4f23c3fd0c10657a3cf14ad3c

Summary: PointRend mask doesn't work for quantization. Add a patch to disable it.

Reviewed By: wat3rBro

Differential Revision: D27800349

fbshipit-source-id: ae0268ee78b000245ebdb2edbfc679a62c85a59a

Synchronize PyTorchLightning/pytorch-lightning (revision 7fe8d184@master) to github/third-party/PyTorchLightning/pytorch-lightning

Summary:
### New commit log messages
  7fe8d184 Do not `shuffle` in `LightningDataModule.from_datasets` for `IterableDataset` (#7053)
  bab72255 [fix] Add barriers before and after setup hook is run (#7202)
  f920ba29 [bugfix] Metric not logged properly in manual optimization (#7228)
  e147127c [feat] Add better support for predict + ddp 2/3 (#7215)
  ca6c87ff Add back `clip_gradients(model)` (#7231)
  3b36d81c Fixed `num_sanity_val_steps` affecting reproducibility of training data shuffling (#7014)
  5cf9afa1 Add fairscale install msg for Sharded Plugins (#7213)
  52a5cee0 Set smarter default for DDP sharded for performance optimization (#6937)
  dd5ec75e Deprecate save_function from model checkpoint callback (#7201)
  ac7d6a35 Fix `NeptuneLogger.log_text(step=None)` (#7194)
  6be0a859 Update teardown for TPU acc (#7211)
  bc3f08b0 [fix] Add barrier to accelerator's teardown (#6814)
  68eac4d9 Enforce Lightning module as source of truth for automatic optimization (#7130)
  44d775fc Update Error message for ProfileConnector (#7204)
  31fcd7d0 Deprecate write_predictions on the LightningModule (#7066)
  591b9cee make bug_report_model minimal (#7191)
  b3fe8366 Move metrics_to_scalars to a dedicated utilities file (#7180)
  f58865aa Properly set `LightningModule.device` after model replacement (#7188)
  8439aead Update FairScale on CI (#7017)
  92af3632 Fix `lr_finder` suggesting too high learning rates (#7076)
  d534e53e add missing predict docs (#7150)

Reviewed By: kazhang

Differential Revision: D28032962

fbshipit-source-id: 18cd01e8ecc13fe25f0890ac0f4b20c3c3e1fed3

Summary:
Pull Request resolved: https://github.com/facebookresearch/d2go/pull/54

This arg is being deprecated, and its use case was really only for modules that use functions besides forward for inference.  The new plan is just to optimize every function. Since this script was just created I'm hoping I can edit this without throwing lots of stuff out of wack.

Reviewed By: wat3rBro

Differential Revision: D27954176

fbshipit-source-id: fbe178fcc0404e5d2524d8edb4052e2cd17f43ba

Summary: Customization via `export_predictor` is now deprecated, this diff move the functionality to `FacegenExportMethod`.

Reviewed By: danthe3rd

Differential Revision: D27935492

fbshipit-source-id: 4f35cff7f3709eff290edefce570cfeea47e687d

Summary:
This diff cleans up the process of exporting RCNN to predictor by tracing.
- Implement a new `D2 (https://github.com/facebookresearch/d2go/commit/d86ecc92eb97f14fcd97d626185f61c6817351e4)TorchscriptTracingExport` which utilizes D2 (https://github.com/facebookresearch/d2go/commit/d86ecc92eb97f14fcd97d626185f61c6817351e4)'s `TracingAdapter`. It's capable to handle more complicated input/output data structures, for example the `MultiDictInMultiDictOut` in unit test. Some duplicated code for serializing can also be removed.
- Later on we'll move `DefaultTorchscriptExport` to `mobile_cv.predictor` which doesn't have D2 (https://github.com/facebookresearch/d2go/commit/d86ecc92eb97f14fcd97d626185f61c6817351e4) dependency, while keep `D2 (https://github.com/facebookresearch/d2go/commit/d86ecc92eb97f14fcd97d626185f61c6817351e4)TorchscriptTracingExport` in D2 (https://github.com/facebookresearch/d2go/commit/d86ecc92eb97f14fcd97d626185f61c6817351e4)Go as a more advanced version.
- Using `D2 (https://github.com/facebookresearch/d2go/commit/d86ecc92eb97f14fcd97d626185f61c6817351e4)TorchscriptTracingExport` we can simply the `prepare_for_export` quite a bit and remove hacky code.

Reviewed By: zhanghang1989

Differential Revision: D27931029

fbshipit-source-id: 4a8d5e5ee3f10e29d98fca63e0e1c68bbda22745

Reviewed By: zhanghang1989

Differential Revision: D27916281

fbshipit-source-id: 7ea01e99e9c2a9b19992f458abc786713ba5a4cd

Reviewed By: zhanghang1989

Differential Revision: D27805428

fbshipit-source-id: c588bdb456e606ca333c2f99eb5c3668edddcbfa

Reviewed By: zhanghang1989

Differential Revision: D27898376

fbshipit-source-id: 87549b0cc24bd38f114977503f4eba97e9166ab8

Summary:
Pull Request resolved: https://github.com/facebookresearch/d2go/pull/46

As titled. The test is flaky because the tensorboard logger might still be writing to temporary folder when we tear down the folder.

Reviewed By: ananthsub

Differential Revision: D27844504

fbshipit-source-id: 3987f9ec3cd05b2f193e75cd4d85109a46f4ee71

Summary:
Pull Request resolved: https://github.com/facebookresearch/d2go/pull/51

As titled. Group the D2 (https://github.com/facebookresearch/d2go/commit/788cf41206fcd761da1974747f0c2c6c671ce871)go runner methods together.

Reviewed By: zhanghang1989, wat3rBro

Differential Revision: D27777726

fbshipit-source-id: f300bce444a401b61ff2adfb45b0c640b1f14855

Summary:
Pull Request resolved: https://github.com/facebookresearch/d2go/pull/50

Found a few custom runners override this method of default runner.

Reviewed By: zhanghang1989, wat3rBro

Differential Revision: D27777505

fbshipit-source-id: 0463cc36bad4af4cbfbe09ab46962cfc1dafbd5d

Summary: model to export is either `nn.Module` and or dict of `nn.Module`

Reviewed By: zhanghang1989

Differential Revision: D27835097

fbshipit-source-id: 869446b36d3e8cc30d6d947f1fc8970cc9ba6c12