Merge branch 'master' into refactor/deprecate

.github/workflows/ci_test-base.yml

@@ -68,7 +68,7 @@ jobs:
     - name: Test Package [only]
       run: |
         # NOTE: run coverage on tests does not propagare faler status for Win, https://github.com/nedbat/coveragepy/issues/1003
-        python -m pytest pytorch_lightning -v --cov=pytorch_lightning --junitxml=junit/test-results-${{ runner.os }}-${{ matrix.python-version }}-${{ matrix.requires }}.xml
+        coverage run --source pytorch_lightning -m pytest pytorch_lightning -v --junitxml=junit/test-results-${{ runner.os }}-${{ matrix.python-version }}-${{ matrix.requires }}.xml
 
     - name: Upload pytest test results
       uses: actions/upload-artifact@v2

.github/workflows/ci_test-conda.yml

@@ -44,7 +44,7 @@ jobs:
     - name: Tests
       run: |
         # NOTE: run coverage on tests does not propagare faler status for Win, https://github.com/nedbat/coveragepy/issues/1003
-        python -m pytest pytorch_lightning tests --cov=pytorch_lightning -v --durations=50 --junitxml=junit/test-results-${{ runner.os }}-torch${{ matrix.pytorch-version }}.xml
+        coverage run --source pytorch_lightning -m pytest pytorch_lightning tests -v --durations=50 --junitxml=junit/test-results-${{ runner.os }}-torch${{ matrix.pytorch-version }}.xml
       shell: bash -l {0}
 
     - name: Upload pytest results

.github/workflows/ci_test-full.yml

@@ -134,7 +134,7 @@ jobs:
     - name: Tests
       run: |
         # NOTE: do not include coverage report here, see: https://github.com/nedbat/coveragepy/issues/1003
-        python -m pytest pytorch_lightning tests --cov=pytorch_lightning -v --durations=50 --junitxml=junit/test-results-${{ runner.os }}-py${{ matrix.python-version }}-${{ matrix.requires }}.xml
+        coverage run --source pytorch_lightning -m pytest pytorch_lightning tests -v --durations=50 --junitxml=junit/test-results-${{ runner.os }}-py${{ matrix.python-version }}-${{ matrix.requires }}.xml
 
     - name: Examples
       run: |

.github/workflows/docs-checks.yml

@@ -98,7 +98,7 @@ jobs:
           # First run the same pipeline as Read-The-Docs
           cd docs
           make clean
-          make html --debug --jobs $(nproc) SPHINXOPTS="-W"
+          make html --debug --jobs $(nproc) SPHINXOPTS="-W --keep-going"
 
       - name: Upload built docs
         uses: actions/upload-artifact@v2

.gitignore

@@ -157,3 +157,4 @@ tags
 data
 MNIST
 runs
+*trace*

.pre-commit-config.yaml

@@ -33,8 +33,3 @@ repos:
     hooks:
       - id: yapf
         args: [--parallel, --in-place]
-
-  - repo: https://github.com/pre-commit/mirrors-mypy
-    rev: v0.790
-    hooks:
-      - id: mypy

CHANGELOG.md

@@ -9,13 +9,13 @@ The format is based on [Keep a Changelog](http://keepachangelog.com/en/1.0.0/).
 
 ### Added
 
-- Added `RetrievalMAP` metric, the corresponding functional version `retrieval_average_precision` and a generic superclass for retrieval metrics `RetrievalMetric` ([#5032](https://github.com/PyTorchLightning/pytorch-lightning/pull/5032))
-
 
 - Added a way to print to terminal without breaking up the progress bar ([#5470](https://github.com/PyTorchLightning/pytorch-lightning/pull/5470))
 
+
 - Added support to checkpoint after training steps in `ModelCheckpoint` callback ([#6146](https://github.com/PyTorchLightning/pytorch-lightning/pull/6146))
 
+
 - Added `checkpoint` parameter to callback's `on_save_checkpoint` hook ([#6072](https://github.com/PyTorchLightning/pytorch-lightning/pull/6072))
 
 
@@ -37,11 +37,28 @@ The format is based on [Keep a Changelog](http://keepachangelog.com/en/1.0.0/).
 - Added arg to `self.log` that enables users to give custom names when dealing with multiple dataloaders ([#6274](https://github.com/PyTorchLightning/pytorch-lightning/pull/6274))
 
 
+- Added `teardown` method to `BaseProfiler` to enable subclasses defining post-profiling steps outside of `__del__` ([#6370](https://github.com/PyTorchLightning/pytorch-lightning/pull/6370))
+
+
+- Added `setup` method to `BaseProfiler` to enable subclasses defining pre-profiling steps for every process ([#6633](https://github.com/PyTorchLightning/pytorch-lightning/pull/6633))
+
+
 - Added no return warning to predict ([#6139](https://github.com/PyTorchLightning/pytorch-lightning/pull/6139))
 
 
-- Added `outputs` parameter to callback's `on_validation_epoch_end` & `on_test_epoch_end` hooks ([#6120](https://github.com/PyTorchLightning/pytorch-lightning/pull/6120))
+- Added `Trainer.predict` config validation ([#6543](https://github.com/PyTorchLightning/pytorch-lightning/pull/6543))
+
+
+- Added `AbstractProfiler` interface ([#6621](https://github.com/PyTorchLightning/pytorch-lightning/pull/6621))
+
 
+- Added support for including module names for forward in the autograd trace of `PyTorchProfiler` ([#6349](https://github.com/PyTorchLightning/pytorch-lightning/pull/6349))
+
+
+- Added support for the PyTorch 1.8.1 autograd profiler ([#6618](https://github.com/PyTorchLightning/pytorch-lightning/pull/6618))
+
+
+- Added `outputs` parameter to callback's `on_validation_epoch_end` & `on_test_epoch_end` hooks ([#6120](https://github.com/PyTorchLightning/pytorch-lightning/pull/6120))
 
 
 ### Changed
@@ -58,6 +75,12 @@ The format is based on [Keep a Changelog](http://keepachangelog.com/en/1.0.0/).
 - Changed `setup()` and `teardown()` stage argument to take any of `{fit,validate,test,predict}` ([#6386](https://github.com/PyTorchLightning/pytorch-lightning/pull/6386))
 
 
+- Changed profilers to save separate report files per state and rank ([#6621](https://github.com/PyTorchLightning/pytorch-lightning/pull/6621))
+
+
+- Changed `PyTorchProfiler` to use `torch.autograd.profiler.record_function` to record functions ([#6349](https://github.com/PyTorchLightning/pytorch-lightning/pull/6349))
+
+
 ### Deprecated
 
 - `period` has been deprecated in favor of `every_n_val_epochs` in the `ModelCheckpoint` callback ([#6146](https://github.com/PyTorchLightning/pytorch-lightning/pull/6146))
@@ -66,6 +89,12 @@ The format is based on [Keep a Changelog](http://keepachangelog.com/en/1.0.0/).
 - Deprecated `trainer.running_sanity_check` in favor of `trainer.sanity_checking` ([#4945](https://github.com/PyTorchLightning/pytorch-lightning/pull/4945))
 
 
+- Deprecated `Profiler(output_filename)` in favor of `dirpath` and `filename` ([#6621](https://github.com/PyTorchLightning/pytorch-lightning/pull/6621))
+
+
+- Deprecated `PytorchProfiler(profiled_functions)` in favor of `record_functions` ([#6349](https://github.com/PyTorchLightning/pytorch-lightning/pull/6349))
+
+
 - Deprecated metrics in favor of `torchmetrics` ([#6505](https://github.com/PyTorchLightning/pytorch-lightning/pull/6505),
 
     [#6530](https://github.com/PyTorchLightning/pytorch-lightning/pull/6530),
@@ -80,6 +109,10 @@ The format is based on [Keep a Changelog](http://keepachangelog.com/en/1.0.0/).
 
     [#6584](https://github.com/PyTorchLightning/pytorch-lightning/pull/6584),
 
+    [#6636](https://github.com/PyTorchLightning/pytorch-lightning/pull/6636),
+
+    [#6637](https://github.com/PyTorchLightning/pytorch-lightning/pull/6637),
+
 )
 
 
@@ -118,6 +151,7 @@ The format is based on [Keep a Changelog](http://keepachangelog.com/en/1.0.0/).
 
 - Added Autocast in validation, test and predict modes for Native AMP ([#6565](https://github.com/PyTorchLightning/pytorch-lightning/pull/6565))
 
+
 - Made the `Plugin.reduce` method more consistent across all Plugins to reflect a mean-reduction by default ([#6011](https://github.com/PyTorchLightning/pytorch-lightning/pull/6011))
 
 
@@ -145,9 +179,21 @@ The format is based on [Keep a Changelog](http://keepachangelog.com/en/1.0.0/).
 - Fixed LightningModule `all_gather` on cpu tensors ([#6416](https://github.com/PyTorchLightning/pytorch-lightning/pull/6416))
 
 
+- Fixed a bug where `all_gather` would not work correctly with `tpu_cores=8` ([#6587](https://github.com/PyTorchLightning/pytorch-lightning/pull/6587))
+
+
+- Update Gradient Clipping for the TPU Accelerator ([#6576](https://github.com/PyTorchLightning/pytorch-lightning/pull/6576))
+
+
 - Fixed torch distributed not available in setup hook for DDP ([#6506](https://github.com/PyTorchLightning/pytorch-lightning/pull/6506))
 
 
+- Fixed comparing required versions ([#6434](https://github.com/PyTorchLightning/pytorch-lightning/pull/6434))
+
+
+- Fixed a bug where gradients were disabled after calling `Trainer.predict` ([#6657](https://github.com/PyTorchLightning/pytorch-lightning/pull/6657))
+
+
 ## [1.2.4] - 2021-03-16
 
 ### Changed
@@ -168,12 +214,6 @@ The format is based on [Keep a Changelog](http://keepachangelog.com/en/1.0.0/).
 - Fixed when Train loop config was run during `Trainer.predict` ([#6541](https://github.com/PyTorchLightning/pytorch-lightning/pull/6541))
 
 
-- Fixed a bug where `all_gather` would not work correctly with `tpu_cores=8` ([#6587](https://github.com/PyTorchLightning/pytorch-lightning/pull/6587))
-
-
-- Update Gradient Clipping for the TPU Accelerator ([#6576](https://github.com/PyTorchLightning/pytorch-lightning/pull/6576))
-
-
 ## [1.2.3] - 2021-03-09
 
 ### Fixed

Makefile

@@ -29,4 +29,4 @@ test: clean
 
 docs: clean
 	pip install --quiet -r requirements/docs.txt
-	python -m sphinx -b html -W docs/source docs/build
+	python -m sphinx -b html -W --keep-going docs/source docs/build

azure-pipelines.yml

@@ -78,7 +78,7 @@ jobs:
       displayName: 'Get legacy checkpoints'
 
     - bash: |
-        python -m pytest pytorch_lightning tests -v --cov=pytorch_lightning --junitxml=$(Build.StagingDirectory)/test-results.xml --durations=50
+        python -m coverage run --source pytorch_lightning -m pytest pytorch_lightning tests -v --junitxml=$(Build.StagingDirectory)/test-results.xml --durations=50
       displayName: 'Testing: standard'
 
     - bash: |
@@ -121,4 +121,6 @@ jobs:
         # cd pl_examples/basic_examples
         # bash submit_ddp_job.sh
         # bash submit_ddp2_job.sh
+      env:
+        PL_USE_MOCKED_MNIST: "1"
       displayName: 'Examples'

docs/source/advanced/multi_gpu.rst

@@ -267,7 +267,7 @@ Lightning allows multiple ways of training
 - TPUs (``tpu_cores=8|x``) (tpu or TPU pod)
 
 .. note::
-    If you request multiple GPUs or nodes without setting a mode, DDP will be automatically used.
+    If you request multiple GPUs or nodes without setting a mode, DDP Spawn will be automatically used.
 
 For a deeper understanding of what Lightning is doing, feel free to read this
 `guide <https://medium.com/@_willfalcon/9-tips-for-training-lightning-fast-neural-networks-in-pytorch-8e63a502f565>`_.

docs/source/conf.py

@@ -13,7 +13,6 @@
 # documentation root, use os.path.abspath to make it absolute, like shown here.
 
 # import m2r
-import builtins
 import glob
 import os
 import shutil
@@ -27,10 +26,13 @@
 
 FOLDER_GENERATED = 'generated'
 SPHINX_MOCK_REQUIREMENTS = int(os.environ.get('SPHINX_MOCK_REQUIREMENTS', True))
-if SPHINX_MOCK_REQUIREMENTS:
-    builtins.__LIGHTNING_SETUP__ = True
 
-import pytorch_lightning  # noqa: E402
+try:
+    from pytorch_lightning import info
+except ImportError:
+    # alternative https://stackoverflow.com/a/67692/4521646
+    sys.path.append(os.path.join(PATH_ROOT, "pytorch_lightning"))
+    import info
 
 # -- Project documents -------------------------------------------------------
 
@@ -79,13 +81,13 @@ def _transform_changelog(path_in: str, path_out: str) -> None:
 # -- Project information -----------------------------------------------------
 
 project = 'PyTorch Lightning'
-copyright = pytorch_lightning.__copyright__
-author = pytorch_lightning.__author__
+copyright = info.__copyright__
+author = info.__author__
 
 # The short X.Y version
-version = pytorch_lightning.__version__
+version = info.__version__
 # The full version, including alpha/beta/rc tags
-release = pytorch_lightning.__version__
+release = info.__version__
 
 # -- General configuration ---------------------------------------------------
 
@@ -176,8 +178,8 @@ def _transform_changelog(path_in: str, path_out: str) -> None:
 # documentation.
 
 html_theme_options = {
-    'pytorch_project': pytorch_lightning.__homepage__,
-    'canonical_url': pytorch_lightning.__homepage__,
+    'pytorch_project': info.__homepage__,
+    'canonical_url': info.__homepage__,
     'collapse_navigation': False,
     'display_version': True,
     'logo_only': False,
@@ -279,6 +281,7 @@ def _transform_changelog(path_in: str, path_out: str) -> None:
     'torch': ('https://pytorch.org/docs/stable/', None),
     'numpy': ('https://numpy.org/doc/stable/', None),
     'PIL': ('https://pillow.readthedocs.io/en/stable/', None),
+    'torchmetrics': ('https://torchmetrics.readthedocs.io/en/stable/', None),
 }
 
 # -- Options for todo extension ----------------------------------------------
@@ -331,6 +334,7 @@ def package_list_from_file(file):
 }
 MOCK_PACKAGES = []
 if SPHINX_MOCK_REQUIREMENTS:
+    MOCK_PACKAGES += ['fairscale']
     # mock also base packages when we are on RTD since we don't install them there
     MOCK_PACKAGES += package_list_from_file(os.path.join(PATH_ROOT, 'requirements.txt'))
     MOCK_PACKAGES += package_list_from_file(os.path.join(PATH_ROOT, 'requirements', 'extra.txt'))

docs/source/starter/introduction_guide.rst

@@ -882,8 +882,8 @@ Or maybe we have a model that we use to do generation
     generated_imgs = model(z)
 
 
-To perform inference at scale, it is possible to use ``trainer.predict`` with LightningModule ``predict`` function
-By default, LightningModule ``predict`` calls forward, but it can be overriden to add any processing logic.
+To perform inference at scale, it is possible to use ``trainer.predict`` with LightningModule ``predict_step`` function
+By default, LightningModule ``predict_step`` calls forward, but it can be overriden to add any processing logic.
 
 .. code-block:: python
 
@@ -893,7 +893,7 @@ By default, LightningModule ``predict`` calls forward, but it can be overriden t
             imgs = self.decoder(z)
             return imgs
 
-        def predict(self, batch, batch_idx: int , dataloader_idx: int = None):
+        def predict_step(self, batch, batch_idx: int , dataloader_idx: int = None):
             return self(batch)
 
 

docs/source/starter/new-project.rst

@@ -83,7 +83,7 @@ Step 1: Define LightningModule
 
 .. testcode::
 
-    class LitAutoEncoder(LightningModule):
+    class LitAutoEncoder(pl.LightningModule):
 
         def __init__(self):
             super().__init__()

pl_examples/__init__.py

@@ -15,10 +15,10 @@
 _DATASETS_PATH = os.path.join(_PACKAGE_ROOT, 'Datasets')
 
 _TORCHVISION_AVAILABLE = _module_available("torchvision")
-_TORCHVISION_MNIST_AVAILABLE = True
+_TORCHVISION_MNIST_AVAILABLE = not bool(os.environ.get("PL_USE_MOCKED_MNIST", False))
 _DALI_AVAILABLE = _module_available("nvidia.dali")
 
-if _TORCHVISION_AVAILABLE:
+if _TORCHVISION_MNIST_AVAILABLE:
     try:
         from torchvision.datasets.mnist import MNIST
         MNIST(_DATASETS_PATH, download=True)