Sync master/1.1.5 into release/1.2 [full merge, no squash]

.github/workflows/release-pypi.yml

@@ -10,9 +10,8 @@ on: # Trigger the workflow on push or pull request, but only for the master bra
 
 jobs:
  # based on https://github.com/pypa/gh-action-pypi-publish
- build-publish:
+ build-package:
  runs-on: ubuntu-20.04
-
  steps:
  - uses: actions/checkout@v2
  - uses: actions/setup-python@v2
@@ -28,6 +27,16 @@ jobs:
  python setup.py sdist bdist_wheel
  ls -lh dist/
 
+ - uses: actions/upload-artifact@v2
+ with:
+ name: pypi-packages
+ path: dist
+
+ publish-package:
+ runs-on: ubuntu-20.04
+ needs: build-package
+ steps:
+ - uses: actions/checkout@v2
  - name: Upload to release
  if: startsWith(github.event.ref, 'refs/tags') || github.event_name == 'release'
  uses: svenstaro/upload-release-action@v2
@@ -62,6 +71,14 @@ jobs:
  user: __token__
  password: ${{ secrets.pypi_password }}
 
+ create-legacy-ckpt:
+ runs-on: ubuntu-20.04
+ needs: [build-package, publish-package]
+ steps:
+ - uses: actions/checkout@v2
+ - uses: actions/setup-python@v2
+ with:
+ python-version: 3.7
  # Note: This uses an internal pip API and may not always work
  # https://github.com/actions/cache/blob/master/examples.md#multiple-oss-in-a-workflow
  - name: Cache pip
@@ -74,7 +91,6 @@ jobs:
  - name: Install dependencies
  run: |
  pip install -r requirements.txt --find-links https://download.pytorch.org/whl/cpu/torch_stable.html --quiet
- pip install virtualenv
  pip install awscli
 
  - name: Configure AWS credentials
@@ -84,25 +100,26 @@ jobs:
  aws-secret-access-key: ${{ secrets.AWS_SECRET_KEY_ID }}
  aws-region: us-east-1
 
+ - uses: actions/download-artifact@v2
+ with:
+ name: pypi-packages
+ path: dist
+
  - name: Pull files from S3
  run: |
  aws s3 cp --recursive s3://pl-public-data/legacy/checkpoints/ legacy/checkpoints/ # --acl public-read
  ls -l legacy/checkpoints/
 
  - name: Generate checkpoint
- if: startsWith(github.event.ref, 'refs/tags') || github.event_name == 'release'
+ # if: startsWith(github.event.ref, 'refs/tags') || github.event_name == 'release'
  run: |
- virtualenv vEnv --system-site-packages
- source vEnv/bin/activate
- pip install dist/*
+ ls -lh dist/
+ pip install dist/*.whl
 
  pl_ver=$(python -c "import pytorch_lightning as pl ; print(pl.__version__)" 2>&1)
  # generate checkpoint to this version
  bash legacy/generate_checkpoints.sh $pl_ver
 
- deactivate
- rm -rf vEnv
-
  - name: Push files to S3
  run: |
  aws s3 sync legacy/checkpoints/ s3://pl-public-data/legacy/checkpoints/

.mergify.yml

@@ -12,59 +12,45 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-#pull_request_rules:
-#
-# - name: Automatic merge on approval
-# conditions:
-# - base=master
-# # number of review approvals
-# - "#approved-reviews-by>=3"
-# # no waiting or assigned review
-# - "#review-requested=0"
-# # no requested chnages from any reviewer
-# - "#changes-requested-reviews-by=0"
-# # this serves as ALL check has to pass as we have actually around 40 tests in total
-# - "#status-success>=54"
-# # this is just in case since we rely on GPU tests (note: redundand to the above)
-# - status-success=continuous-integration/drone/pr
-# - "status-success=ci/circleci: TPU-tests"
-# # this is patter-like, unofrunatly serves as `any(...)` (note: redundand to the above)
-# #- "status-success~=^ci/circleci:"
-# # no conflict with master branch
-# - -conflict
-# # was not closed yet
-# - -closed
-# # filter-out GH draft PRs
-# - -draft
-# actions:
-# delete_head_branch: {}
-# merge:
-# # https://doc.mergify.io/merge-action.html#strict-merge
-# # (on head branch) $ git merge --no-ff base
-# # (on head branch) # Wait for CI to go green
-# # (on head branch) # Squash all commits
-# # (on base branch) $ git merge --ff head
-# strict: true
-# method: squash
-# comment:
-# message: Great job! =)
-#
-# - name: warn on conflicts
-# conditions:
-# - conflict
-# # filter-out GH draft PRs
-# - -draft
-# actions:
-# comment:
-# message: This pull request is now in conflict... :(
-#
-# - name: add core reviewer
-# conditions:
-# # filter-out GH draft PRs
-# - -draft
-# # number of review approvals
-# - "#approved-reviews-by<3"
-# actions:
-# request_reviews:
-# teams:
-# - core-contributors
+pull_request_rules:
+
+ - name: warn on conflicts
+ conditions:
+ - conflict
+ - -draft # filter-out GH draft PRs
+ - -label="has conflicts"
+ actions:
+ # comment:
+ # message: This pull request is now in conflict... :(
+ label:
+ add: [ "has conflicts" ]
+
+ - name: resolved conflicts
+ conditions:
+ - -conflict
+ - label="has conflicts"
+ - -draft # filter-out GH draft PRs
+ - -merged # not merged yet
+ - -closed
+ actions:
+ label:
+ remove: [ "has conflicts" ]
+
+ - name: update PR
+ conditions:
+ - conflict
+ - -draft # filter-out GH draft PRs
+ - label="0:] Ready-To-Go"
+ actions:
+ update: {}
+
+ - name: add core reviewer
+ conditions:
+ - -conflict # skip if conflict
+ - -draft # filter-out GH draft PRs
+ - label="0:] Ready-To-Go"
+ - "#approved-reviews-by<3" # number of review approvals
+ actions:
+ request_reviews:
+ teams:
+ - core-contributors

CHANGELOG.md

@@ -169,8 +169,17 @@ The format is based on [Keep a Changelog](http://keepachangelog.com/en/1.0.0/).
 - Fixed loading yaml ([#5619](https://github.com/PyTorchLightning/pytorch-lightning/pull/5619))
 
 
+## [1.1.5] - 2021-01-19
 
-## [1.1.4] - YYYY-MM-DD
+### Fixed
+
+- Fixed a visual bug in the progress bar display initialization ([#4579](https://github.com/PyTorchLightning/pytorch-lightning/pull/4579))
+- Fixed logging `on_train_batch_end` in a callback with multiple optimizers ([#5521](https://github.com/PyTorchLightning/pytorch-lightning/pull/5521))
+- Fixed `reinit_scheduler_properties` with correct optimizer ([#5519](https://github.com/PyTorchLightning/pytorch-lightning/pull/5519))
+- Fixed `val_check_interval` with `fast_dev_run` ([#5540](https://github.com/PyTorchLightning/pytorch-lightning/pull/5540))
+
+
+## [1.1.4] - 2021-01-12
 
 ### Added
 
@@ -186,6 +195,7 @@ The format is based on [Keep a Changelog](http://keepachangelog.com/en/1.0.0/).
 - Logging only on `not should_accumulate()` during training ([#5417](https://github.com/PyTorchLightning/pytorch-lightning/pull/5417))
 - Resolve interpolation bug with Hydra ([#5406](https://github.com/PyTorchLightning/pytorch-lightning/pull/5406))
 - Check environ before selecting a seed to prevent warning message ([#4743](https://github.com/PyTorchLightning/pytorch-lightning/pull/4743))
+- Fixed signature mismatch in `model_to_device` of `DDPCPUHPCAccelerator` ([#5505](https://github.com/PyTorchLightning/pytorch-lightning/pull/5505))
 
 
 ## [1.1.3] - 2021-01-05

docs/source/advanced/multi_gpu.rst

@@ -362,7 +362,7 @@ project module) you can use the following method:
 .. code-block:: python
 
  # train on 8 GPUs (same machine (ie: node))
- trainer = Trainer(gpus=8, accelerator='ddp')
+ trainer = Trainer(gpus=8, accelerator='ddp_spawn')
 
 We STRONGLY discourage this use because it has limitations (due to Python and PyTorch):
 

pl_examples/domain_templates/reinforce_learn_Qnet.py

@@ -398,7 +398,7 @@ def add_model_specific_args(parent_parser): # pragma: no-cover
  parser.add_argument("--sync_rate", type=int, default=10, help="how many frames do we update the target network")
  parser.add_argument("--replay_size", type=int, default=1000, help="capacity of the replay buffer")
  parser.add_argument(
- "--warm_start_size",
+ "--warm_start_steps",
  type=int,
  default=1000,
  help="how many samples do we use to fill our buffer at the start of training"
@@ -407,8 +407,6 @@ def add_model_specific_args(parent_parser): # pragma: no-cover
  parser.add_argument("--eps_start", type=float, default=1.0, help="starting value of epsilon")
  parser.add_argument("--eps_end", type=float, default=0.01, help="final value of epsilon")
  parser.add_argument("--episode_length", type=int, default=200, help="max length of an episode")
- parser.add_argument("--max_episode_reward", type=int, default=200, help="max episode reward in the environment")
- parser.add_argument("--warm_start_steps", type=int, default=1000, help="max episode reward in the environment")
  return parser
 
 
@@ -429,7 +427,7 @@ def main(args) -> None:
  torch.manual_seed(0)
  np.random.seed(0)
 
- parser = argparse.ArgumentParser()
+ parser = argparse.ArgumentParser(add_help=False)
  parser = DQNLightning.add_model_specific_args(parser)
  args = parser.parse_args()
 

pl_examples/domain_templates/semantic_segmentation.py

@@ -289,7 +289,7 @@ def main(hparams: Namespace):
 
 if __name__ == '__main__':
  cli_lightning_logo()
- parser = ArgumentParser()
+ parser = ArgumentParser(add_help=False)
  parser = SegModel.add_model_specific_args(parser)
  hparams = parser.parse_args()
 

pytorch_lightning/accelerators/legacy/accelerator.py

@@ -51,6 +51,10 @@ def __init__(self,
  def setup(self, model):
  pass
 
+ def train(self):
+ self.trainer.setup_trainer(self.trainer.model)
+ return self.train_or_test()
+
  def teardown(self):
  # Ensure if necessary all processes are finished
  self.barrier()
@@ -65,6 +69,7 @@ def train_or_test(self):
  if self.trainer.testing:
  results = self.trainer.run_test()
  else:
+ self.trainer.train_loop.setup_training()
  results = self.trainer.train()
  return results
 

pytorch_lightning/accelerators/legacy/accelerator_connector.py

@@ -364,8 +364,8 @@ def set_distributed_mode(self):
  _ddp = (DistributedType.DDP, DistributedType.DDP_SPAWN, DistributedType.DDP2)
  if (self.trainer.num_nodes > 1 and self.trainer._distrib_type not in _ddp):
  raise MisconfigurationException(
- 'DataParallel does not support num_nodes > 1. Switching to DistributedDataParallel for you. '
- 'To silence this warning set `accelerator="ddp"` or `accelerator="ddp2"`'
+ 'DataParallel does not support num_nodes > 1. '
+ 'To avoid this exception, set `accelerator="ddp"` or `accelerator="ddp2"`'
  )
 
  rank_zero_info(

pytorch_lightning/accelerators/legacy/cpu_accelerator.py

@@ -50,16 +50,6 @@ def setup(self, model):
 
  self.trainer.model = model
 
- def train(self):
- model = self.trainer.model
-
- # set up training routine
- self.trainer.train_loop.setup_training(model)
-
- # train or test
- results = self.train_or_test()
- return results
-
  def _step(self, model_step: Callable, args):
  if self.trainer.amp_backend == AMPType.NATIVE:
  with torch.cuda.amp.autocast():

pytorch_lightning/accelerators/legacy/ddp2_accelerator.py

@@ -186,9 +186,6 @@ def ddp_train(self, process_idx, mp_queue, model):
 
  self.ddp_plugin.on_after_setup_optimizers(self.trainer)
 
- # set model properties before going into wrapper
- self.trainer.model_connector.copy_trainer_model_properties(model)
-
  # 16-bit
  model = self.trainer.precision_connector.connect(model)
 
@@ -198,8 +195,7 @@ def ddp_train(self, process_idx, mp_queue, model):
  # allow user to configure ddp
  model = self.configure_ddp(model, device_ids)
 
- # set up training routine
- self.trainer.train_loop.setup_training(model)
+ self.trainer.setup_trainer(model)
 
  # train or test
  results = self.train_or_test()

pytorch_lightning/accelerators/legacy/ddp_accelerator.py

@@ -289,9 +289,6 @@ def ddp_train(self, process_idx, model):
  # allow for lr schedulers as well
  self.setup_optimizers(model)
 
- # set model properties before going into wrapper
- self.trainer.model_connector.copy_trainer_model_properties(model)
-
  # 16-bit
  model = self.trainer.precision_connector.connect(model)
 
@@ -301,9 +298,8 @@ def ddp_train(self, process_idx, model):
  # allow user to configure ddp
  model = self.configure_ddp(model, device_ids)
 
- # set up training routine
  self.barrier('ddp_setup')
- self.trainer.train_loop.setup_training(model)
+ self.trainer.setup_trainer(model)
 
  # train or test
  results = self.train_or_test()

pytorch_lightning/accelerators/legacy/ddp_cpu_hpc_accelerator.py

@@ -36,8 +36,7 @@ def __init__(self,
  super().__init__(trainer, cluster_environment, ddp_plugin)
  self.nickname = 'ddp_cpu'
 
- def model_to_device(self, model, process_idx):
- # Todo: required argument `process_idx` is not used
+ def model_to_device(self, model):
  model.cpu()
 
  def get_device_ids(self):

pytorch_lightning/accelerators/legacy/ddp_cpu_spawn_accelerator.py

@@ -142,9 +142,6 @@ def ddp_train(self, process_idx, mp_queue, model):
 
  self.ddp_plugin.on_after_setup_optimizers(self.trainer)
 
- # set model properties before going into wrapper
- self.trainer.model_connector.copy_trainer_model_properties(model)
-
  # 16-bit
  model = self.trainer.precision_connector.connect(model)
 
@@ -154,8 +151,7 @@ def ddp_train(self, process_idx, mp_queue, model):
  # allow user to configure ddp
  model = self.configure_ddp(model, device_ids)
 
- # set up training routine
- self.trainer.train_loop.setup_training(model)
+ self.trainer.setup_trainer(model)
 
  # train or test
  results = self.train_or_test()