Merge branch 'master' into bugfix/SWA_scheduler_step

Author: akihironitta

.github/workflows/ci_dockers.yml

@@ -75,8 +75,8 @@ jobs:
             pytorch_version: 1.6
           - python_version: 3.8
             pytorch_version: 1.7
-          #- python_version: 3.9
-          #  pytorch_version: 1.8
+          #  - python_version: 3.9
+          #    pytorch_version: 1.7
     steps:
       - name: Checkout
         uses: actions/checkout@v2
@@ -109,8 +109,10 @@ jobs:
             pytorch_version: 1.4
           - python_version: 3.7
             pytorch_version: 1.7
-          - python_version: 3.7
+          - python_version: 3.8
             pytorch_version: 1.8
+          #  - python_version: 3.9
+          #    pytorch_version: 1.8
     steps:
       - name: Checkout
         uses: actions/checkout@v2

.github/workflows/ci_pkg-install.yml

@@ -17,7 +17,7 @@ jobs:
       matrix:
         # PyTorch 1.5 is failing on Win and bolts requires torchvision>=0.5
         os: [ubuntu-20.04, macOS-10.15 , windows-2019]  #
-        python-version: [3.6, 3.8]
+        python-version: [3.6, 3.9]
 
     steps:
       - uses: actions/checkout@v2

.github/workflows/ci_test-base.yml

@@ -16,7 +16,7 @@ jobs:
       # max-parallel: 6
       matrix:
         os: [ubuntu-20.04, windows-2019, macOS-10.15]
-        python-version: [3.7]
+        python-version: [3.8]
 
     # Timeout: https://stackoverflow.com/a/59076067/4521646
     timeout-minutes: 20

.github/workflows/ci_test-full.yml

@@ -15,19 +15,24 @@ jobs:
       fail-fast: false
       matrix:
         os: [ubuntu-18.04, windows-2019, macOS-10.15]
-        python-version: [3.6, 3.7, 3.8]
+        python-version: [3.6, 3.7, 3.8, 3.9]
         requires: ['minimal', 'latest']
+        exclude:
+          - python-version: 3.9
+            requires: 'minimal'
 
     # Timeout: https://stackoverflow.com/a/59076067/4521646
-    timeout-minutes: 35  # TODO: the macOS is taking too long, probably caching did not work...
+    # TODO: the macOS is taking too long, probably caching did not work...
+    timeout-minutes: 40
+
     steps:
     - uses: actions/checkout@v2
     - name: Set up Python ${{ matrix.python-version }}
       uses: actions/setup-python@v2
       with:
         python-version: ${{ matrix.python-version }}
 
-    - name: Update Pip
+    - name: Update pip
       run: |
         # todo: unfreeze PIP after resolving minimal dependencies
         pip install --quiet "pip==20.1" --upgrade --user  # needed for get pip cacher folder
@@ -48,6 +53,19 @@ jobs:
         open(fname, 'w').writelines(lines)
       shell: python
 
+    # todo: re-enable when allow testing py 3.9 with min config, atm some Hydra issues
+    #- name: Adjust minimal for Python 3.9
+    #  if: matrix.requires == 'minimal' && matrix.python-version == 3.9
+    #  run: |
+    #    import re
+    #    def _req(fname, ptn, ver):
+    #        req = re.sub(ptn, ver, open(fname).read())
+    #        open(fname, 'w').write(req)
+    #
+    #    _req('requirements.txt', r'torch>=[\d\.]+', 'torch>=1.8.0')
+    #    _req('requirements/extra.txt', r'onnxruntime>=[\d\.]+', 'onnxruntime>=1.7.0')
+    #  shell: python
+
     - name: Set min. dependencies
       if: matrix.requires == 'minimal'
       run: |
@@ -84,7 +102,6 @@ jobs:
           ${{ runner.os }}-pip-py${{ matrix.python-version }}-${{ matrix.requires }}-
 
     - name: Pull checkpoints from S3
-      # todo: consider adding some caching, but ATM all models have less then 100KB
       run: |
         cd legacy
         # wget is simpler but does not work on Windows
@@ -93,6 +110,13 @@ jobs:
         unzip -o checkpoints.zip
         ls -l checkpoints/
 
+    # todo: re-enable testing with Horovod
+    - name: py3.9 - temp skip Horovod
+      if: matrix.python-version == 3.9
+      run: |
+        # pip uninstall -y horovod
+        python -c "fname = 'requirements/extra.txt' ; lines = [line for line in open(fname).readlines() if not line.startswith('horovod')] ; open(fname, 'w').writelines(lines)"
+
     - name: Install dependencies
       env:
         # MAKEFLAGS: "-j2"
@@ -112,7 +136,8 @@ jobs:
       shell: bash
 
     - name: Reinstall Horovod if necessary
-      if: runner.os != 'windows'
+      # todo: re-enable horovod on py3.9 when it will be supported
+      if: runner.os != 'windows' && matrix.python-version != 3.9
       env:
         HOROVOD_BUILD_ARCH_FLAGS: "-mfma"
       run: |

.github/workflows/release-docker.yml

@@ -14,7 +14,7 @@ jobs:
       fail-fast: false
       matrix:
         python_version: [3.6, 3.7, 3.8]
-        pytorch_version: [1.4, 1.5, 1.6, 1.7]
+        pytorch_version: [1.4, 1.5, 1.6, 1.7, 1.8]
     steps:
       - name: Checkout
         uses: actions/checkout@v2

CHANGELOG.md

@@ -9,6 +9,10 @@ The format is based on [Keep a Changelog](http://keepachangelog.com/en/1.0.0/).
 
 ### Added
 
+
+- Added more explicit exception message when trying to execute `trainer.test()` or `trainer.validate()` with `fast_dev_run=True` ([#6667](https://github.com/PyTorchLightning/pytorch-lightning/pull/6667))
+
+
 - Trigger warning when non-metric logged value with multi processes hasn't been reduced ([#6417](https://github.com/PyTorchLightning/pytorch-lightning/pull/6417))
 
 
@@ -66,12 +70,18 @@ The format is based on [Keep a Changelog](http://keepachangelog.com/en/1.0.0/).
 - 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 `configure_sharded_model` hook ([#6679](https://github.com/PyTorchLightning/pytorch-lightning/pull/6679))
+
+
 - Added support for `precision=64`, enabling training with double precision ([#6595](https://github.com/PyTorchLightning/pytorch-lightning/pull/6595))
 
 
 - Added `artifact_location` argument to `MLFlowLogger` which will be passed to the `MlflowClient.create_experiment` call ([#6677](https://github.com/PyTorchLightning/pytorch-lightning/pull/6677))
 
 
+- Added `model` parameter to precision plugins' `clip_gradients` signature ([#6764](https://github.com/PyTorchLightning/pytorch-lightning/pull/6764))
+
+
 ### Changed
 
 - Renamed `pytorch_lightning.callbacks.swa` to `pytorch_lightning.callbacks.stochastic_weight_avg` ([#6259](https://github.com/PyTorchLightning/pytorch-lightning/pull/6259))
@@ -92,9 +102,6 @@ The format is based on [Keep a Changelog](http://keepachangelog.com/en/1.0.0/).
 - Changed `PyTorchProfiler` to use `torch.autograd.profiler.record_function` to record functions ([#6349](https://github.com/PyTorchLightning/pytorch-lightning/pull/6349))
 
 
-- Changed the behavior of `on_epoch_start` to run at the beginning of validation & test epoch ([#6498](https://github.com/PyTorchLightning/pytorch-lightning/pull/6498))
-
-
 ### 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))
@@ -149,20 +156,20 @@ The format is based on [Keep a Changelog](http://keepachangelog.com/en/1.0.0/).
 - Removed `mode='auto'` from `EarlyStopping` ([#6167](https://github.com/PyTorchLightning/pytorch-lightning/pull/6167))
 
 
+- Removed legacy references for magic keys in the `Result` object ([#6016](https://github.com/PyTorchLightning/pytorch-lightning/pull/6016))
+
+
 - Removed deprecated `LightningModule` `hparams` setter ([#6207](https://github.com/PyTorchLightning/pytorch-lightning/pull/6207))
 
 
-- Removed legacy code to include `step` dictionary returns in `callback_metrics`. Use `self.log_dict` instead. ([#6682](https://github.com/PyTorchLightning/pytorch-lightning/pull/6682))
+- Removed legacy code to log or include metrics in the progress bar by returning them in a dict with the `"log"/"progress_bar"` magic keys. Use `self.log` instead ([#6734](https://github.com/PyTorchLightning/pytorch-lightning/pull/6734))
 
 
 - Removed `optimizer_idx` argument from `training_step` in manual optimization ([#6093](https://github.com/PyTorchLightning/pytorch-lightning/pull/6093))
 
 
 ### Fixed
 
-- 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))
 
 
@@ -178,9 +185,6 @@ The format is based on [Keep a Changelog](http://keepachangelog.com/en/1.0.0/).
 - Fixed `ModelCheckpoint(save_top_k=0, save_last=True)` not saving the `last` checkpoint ([#6136](https://github.com/PyTorchLightning/pytorch-lightning/pull/6136))
 
 
-- Fixed duplicate logs appearing in console when using the python logging module ([#5509](https://github.com/PyTorchLightning/pytorch-lightning/pull/5509), [#6275](https://github.com/PyTorchLightning/pytorch-lightning/pull/6275))
-
-
 - Fixed `.teardown(stage='fit')` getting called during `trainer.test` ([#6386](https://github.com/PyTorchLightning/pytorch-lightning/pull/6386))
 
 
@@ -190,29 +194,47 @@ 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))
+- Fixed torch distributed not available in setup hook for DDP ([#6506](https://github.com/PyTorchLightning/pytorch-lightning/pull/6506))
 
 
-- Update Gradient Clipping for the TPU Accelerator ([#6576](https://github.com/PyTorchLightning/pytorch-lightning/pull/6576))
+- Enforce an epoch scheduler interval when using SWA ([#6588](https://github.com/PyTorchLightning/pytorch-lightning/pull/6588))
 
 
-- Fixed torch distributed not available in setup hook for DDP ([#6506](https://github.com/PyTorchLightning/pytorch-lightning/pull/6506))
+## [1.2.6] - 2021-03-30
 
+### Changed
 
-- Fixed comparing required versions ([#6434](https://github.com/PyTorchLightning/pytorch-lightning/pull/6434))
+- Changed the behavior of `on_epoch_start` to run at the beginning of validation & test epoch ([#6498](https://github.com/PyTorchLightning/pytorch-lightning/pull/6498))
 
+### Removed
 
-- Enforce an epoch scheduler interval when using SWA ([#6588](https://github.com/PyTorchLightning/pytorch-lightning/pull/6588))
+- Removed legacy code to include `step` dictionary returns in `callback_metrics`. Use `self.log_dict` instead. ([#6682](https://github.com/PyTorchLightning/pytorch-lightning/pull/6682))
 
+### Fixed
 
+- Fixed `DummyLogger.log_hyperparams` raising a `TypeError` when running with `fast_dev_run=True` ([#6398](https://github.com/PyTorchLightning/pytorch-lightning/pull/6398))
 - Fixed error on TPUs when there was no `ModelCheckpoint` ([#6654](https://github.com/PyTorchLightning/pytorch-lightning/pull/6654))
+- Fixed `trainer.test` freeze on TPUs ([#6654](https://github.com/PyTorchLightning/pytorch-lightning/pull/6654))
+- Fixed a bug where gradients were disabled after calling `Trainer.predict` ([#6657](https://github.com/PyTorchLightning/pytorch-lightning/pull/6657))
+- Fixed bug where no TPUs were detected in a TPU pod env ([#6719](https://github.com/PyTorchLightning/pytorch-lightning/pull/6719))
 
 
-- Fixed `trainer.test` freeze on TPUs ([#6654](https://github.com/PyTorchLightning/pytorch-lightning/pull/6654))
+## [1.2.5] - 2021-03-23
 
+### Changed
 
-- Fixed a bug where gradients were disabled after calling `Trainer.predict` ([#6657](https://github.com/PyTorchLightning/pytorch-lightning/pull/6657))
+- Update Gradient Clipping for the TPU Accelerator ([#6576](https://github.com/PyTorchLightning/pytorch-lightning/pull/6576))
+- Refactored setup for typing friendly ([#6590](https://github.com/PyTorchLightning/pytorch-lightning/pull/6590))
 
+### Fixed
+
+- Fixed a bug where `all_gather` would not work correctly with `tpu_cores=8` ([#6587](https://github.com/PyTorchLightning/pytorch-lightning/pull/6587))
+- Fixed comparing required versions ([#6434](https://github.com/PyTorchLightning/pytorch-lightning/pull/6434))
+- Fixed duplicate logs appearing in console when using the python logging module ([#6275](https://github.com/PyTorchLightning/pytorch-lightning/pull/6275))
+- Added Autocast in validation, test and predict modes for Native AMP ([#6565](https://github.com/PyTorchLightning/pytorch-lightning/pull/6565))
+
+
+- Fixed resolve a bug with omegaconf and xm.save ([#6741](https://github.com/PyTorchLightning/pytorch-lightning/pull/6741))
 
 ## [1.2.4] - 2021-03-16
 
@@ -228,7 +250,6 @@ The format is based on [Keep a Changelog](http://keepachangelog.com/en/1.0.0/).
 - Fixed broadcast to use PyTorch `broadcast_object_list` and add `reduce_decision` ([#6410](https://github.com/PyTorchLightning/pytorch-lightning/pull/6410))
 - Fixed logger creating directory structure too early in DDP ([#6380](https://github.com/PyTorchLightning/pytorch-lightning/pull/6380))
 - Fixed DeepSpeed additional memory use on rank 0 when default device not set early enough ([#6460](https://github.com/PyTorchLightning/pytorch-lightning/pull/6460))
-- Fixed `DummyLogger.log_hyperparams` raising a `TypeError` when running with `fast_dev_run=True` ([#6398](https://github.com/PyTorchLightning/pytorch-lightning/pull/6398))
 - Fixed an issue with `Tuner.scale_batch_size` not finding the batch size attribute in the datamodule ([#5968](https://github.com/PyTorchLightning/pytorch-lightning/pull/5968))
 - Fixed an exception in the layer summary when the model contains torch.jit scripted submodules ([#6511](https://github.com/PyTorchLightning/pytorch-lightning/pull/6511))
 - Fixed when Train loop config was run during `Trainer.predict` ([#6541](https://github.com/PyTorchLightning/pytorch-lightning/pull/6541))