Merge branch 'master' into feat/lightning-cli-trainer-fn

CHANGELOG.md

@@ -19,7 +19,7 @@ The format is based on [Keep a Changelog](http://keepachangelog.com/en/1.0.0/).
 - Added a flavor of `training_step` that takes `dataloader_iter` as an argument ([#8807](https://github.com/PyTorchLightning/pytorch-lightning/pull/8807))
 
 
-- Added `state_id` property to the `Callback` base class ([#6886](https://github.com/PyTorchLightning/pytorch-lightning/pull/6886))
+- Added `state_key` property to the `Callback` base class ([#6886](https://github.com/PyTorchLightning/pytorch-lightning/pull/6886))
 
 
 - Progress tracking
@@ -56,12 +56,20 @@ The format is based on [Keep a Changelog](http://keepachangelog.com/en/1.0.0/).
     * Added `CaptureMapDataset` for state management in map-style datasets ([#8891](https://github.com/PyTorchLightning/pytorch-lightning/pull/8891))
     * Added Fault Tolerant Training to `DataFetcher` ([#8891](https://github.com/PyTorchLightning/pytorch-lightning/pull/8891))
     * Replaced old prefetch iterator with new `DataFetcher` in training loop ([#8953](https://github.com/PyTorchLightning/pytorch-lightning/pull/8953))
+    * Added partial support for global random state fault-tolerance in map-style datasets ([#8950](https://github.com/PyTorchLightning/pytorch-lightning/pull/8950))
 
 - Checkpoint saving & loading extensibility:
     * Added `CheckpointIO` to expose checkpoint IO from training type plugin ([#8743](https://github.com/PyTorchLightning/pytorch-lightning/pull/8743))
     * Refactored CheckpointConnector to offload validating logic to the checkpoitn IO plugin ([#9045](https://github.com/PyTorchLightning/pytorch-lightning/pull/9045))
 
 
+- Loop customization:
+    * Added `Closure` and `AbstractClosure` classes ([#8642](https://github.com/PyTorchLightning/pytorch-lightning/pull/8642))
+
+
+- Added support for saving and loading state of multiple callbacks of the same type ([#7187](https://github.com/PyTorchLightning/pytorch-lightning/pull/7187))
+
+
 - Added DeepSpeed Stage 1 support ([#8974](https://github.com/PyTorchLightning/pytorch-lightning/pull/8974))
 
 
@@ -86,6 +94,12 @@ The format is based on [Keep a Changelog](http://keepachangelog.com/en/1.0.0/).
 - Added Rich Progress Bar ([#8929](https://github.com/PyTorchLightning/pytorch-lightning/pull/8929))
 
 
+- Added validate logic for precision ([#9080](https://github.com/PyTorchLightning/pytorch-lightning/pull/9080))
+
+
+- Add support for CPU AMP autocast ([#9084](https://github.com/PyTorchLightning/pytorch-lightning/pull/9084))
+
+
 ### Changed
 
 - Parsing of the `gpus` Trainer argument has changed: `gpus="n"` (str) no longer selects the GPU index n and instead selects the first n devices. ([#8770](https://github.com/PyTorchLightning/pytorch-lightning/pull/8770))
@@ -140,10 +154,15 @@ The format is based on [Keep a Changelog](http://keepachangelog.com/en/1.0.0/).
 - Deprecated `DataModule` properties: `train_transforms`, `val_transforms`, `test_transforms`, `size`, `dims` ([#8851](https://github.com/PyTorchLightning/pytorch-lightning/pull/8851))
 
 
-- Deprecated `prepare_data_per_node` flag on Trainer and set it as a property of `DataHooks`, accessible in the `LightningModule` and `LightningDataModule` [#8958](https://github.com/PyTorchLightning/pytorch-lightning/pull/8958)
+- Deprecated `prepare_data_per_node` flag on Trainer and set it as a property of `DataHooks`, accessible in the `LightningModule` and `LightningDataModule` ([#8958](https://github.com/PyTorchLightning/pytorch-lightning/pull/8958))
+
+
+- Deprecated the `TestTubeLogger` ([#9065](https://github.com/PyTorchLightning/pytorch-lightning/pull/9065))
+
+
+- Updated deprecation of  `argparse_utils.py` from removal in 1.4 to 2.0 ([#9162](https://github.com/PyTorchLightning/pytorch-lightning/pull/9162))
 
 
--
 
 ### Removed
 
@@ -198,6 +217,15 @@ The format is based on [Keep a Changelog](http://keepachangelog.com/en/1.0.0/).
 - Removed `on_train_epoch_end` from `Accelerator` ([#9035](https://github.com/PyTorchLightning/pytorch-lightning/pull/9035))
 
 
+- Removed `InterBatchProcessor` in favor of `DataLoaderIterDataFetcher` ([#9052](https://github.com/PyTorchLightning/pytorch-lightning/pull/9052))
+
+
+- Removed `Plugin` in `base_plugin.py`, access `TrainingTypePlugin` and `PrecisionPlugin` directly instead ([#9066](https://github.com/PyTorchLightning/pytorch-lightning/pull/9066))
+
+
+- Removed `teardown` from `ParallelPlugin` ([#8943](https://github.com/PyTorchLightning/pytorch-lightning/pull/8943))
+
+
 ### Fixed
 
 - Fixed save/load/resume from checkpoint for DeepSpeed Plugin (
@@ -211,9 +239,13 @@ The format is based on [Keep a Changelog](http://keepachangelog.com/en/1.0.0/).
 
 - Fixed bug where data-loading functions where not getting the correct running stage passed ([#8858](https://github.com/PyTorchLightning/pytorch-lightning/pull/8858))
 
+
 - Fixed a bug in the binary search mode of auto batch size scaling where exception was thrown if the first trainer run resulted in OOM ([#8954](https://github.com/PyTorchLightning/pytorch-lightning/pull/8954))
 
 
+- Fixed not setting a default value for `max_epochs` if `max_time` was specified on the `Trainer` constructor ([#9072](https://github.com/PyTorchLightning/pytorch-lightning/pull/9072))
+
+
 ## [1.4.3] - 2021-08-17
 
 - Fixed plateau scheduler stepping on incomplete epoch ([#8861](https://github.com/PyTorchLightning/pytorch-lightning/pull/8861))