Merge branch 'master' into fix_signal_connector_has_already_handler

CHANGELOG.md

@@ -68,10 +68,10 @@ The format is based on [Keep a Changelog](http://keepachangelog.com/en/1.0.0/).
 - Raised an error if the `batch_size` cannot be inferred from the current batch if it contained a string or was a custom batch object ([#10541](https://github.com/PyTorchLightning/pytorch-lightning/pull/10541))
 
 
--
+- Moved optimizer related logics from `Accelerator` to `TrainingTypePlugin` ([#10596](https://github.com/PyTorchLightning/pytorch-lightning/pull/10596))
 
 
--
+- Moved `batch_to_device` method from `Accelerator` to `TrainingTypePlugin` ([#10649](https://github.com/PyTorchLightning/pytorch-lightning/pull/10649))
 
 
 -
@@ -184,6 +184,9 @@ The format is based on [Keep a Changelog](http://keepachangelog.com/en/1.0.0/).
 
 ### Fixed
 
+- Fixed support for `--key.help=class` with the `LightningCLI` ([#10767](https://github.com/PyTorchLightning/pytorch-lightning/pull/10767))
+
+
 - Fixed `_compare_version` for python packages ([#10762](https://github.com/PyTorchLightning/pytorch-lightning/pull/10762))
 
 
@@ -193,6 +196,9 @@ The format is based on [Keep a Changelog](http://keepachangelog.com/en/1.0.0/).
 - Fixed a consolidation error in Lite when attempting to save the state dict of a sharded optimizer ([#10746](https://github.com/PyTorchLightning/pytorch-lightning/pull/10746))
 
 
+- Fixed the default logging level for batch hooks associated with training from `on_step=False, on_epoch=True` to `on_step=True, on_epoch=False` ([#10756](https://github.com/PyTorchLightning/pytorch-lightning/pull/10756))
+
+
 - Fixed `SignalConnector._has_already_handler` check for callable type ([#10483](https://github.com/PyTorchLightning/pytorch-lightning/pull/10483))
 
 

dockers/nvidia/Dockerfile

@@ -12,7 +12,7 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-ARG PYTORCH_VERSION=21.07
+ARG PYTORCH_VERSION=21.11
 
 # https://docs.nvidia.com/deeplearning/frameworks/pytorch-release-notes
 FROM nvcr.io/nvidia/pytorch:${PYTORCH_VERSION}-py3
@@ -30,18 +30,18 @@ RUN \
     # replace by specific version if asked
     if [ ! -z "$LIGHTNING_VERSION" ] ; then \
         rm -rf pytorch-lightning ; \
-        wget https://github.com/PyTorchLightning/pytorch-lightning/archive/${LIGHTNING_VERSION}.zip --progress=bar:force:noscroll ; \
-        unzip ${LIGHTNING_VERSION}.zip ; \
-        mv pytorch-lightning-*/* pytorch-lightning ; \
-        rm -rf pytorch-lightning-* ; \
-        rm *.zip ; \
+        git clone https://github.com/PyTorchLightning/pytorch-lightning.git ; \
+        cd pytorch-lightning ; \
+        git checkout ${LIGHTNING_VERSION} ; \
+        git submodule update --init --recursive ; \
+        cd .. ; \
     fi && \
 # save the examples
     mv pytorch-lightning/_notebooks notebooks && \
     mv pytorch-lightning/pl_examples . && \
 
 # Installations
-    python .github/prune-packages.py ./pytorch-lightning/requirements/extra.txt "horovod" && \
+    python ./pytorch-lightning/.github/prune-packages.py ./pytorch-lightning/requirements/extra.txt "horovod" && \
     pip install "Pillow>=8.2, !=8.3.0" "cryptography>=3.4" "py>=1.10" --no-cache-dir --upgrade-strategy only-if-needed && \
     pip install -r ./pytorch-lightning/requirements/extra.txt --no-cache-dir --upgrade-strategy only-if-needed && \
     pip install -r ./pytorch-lightning/requirements/examples.txt --no-cache-dir --upgrade-strategy only-if-needed && \

docs/source/common/weights_loading.rst

@@ -168,7 +168,7 @@ To load a model along with its weights, biases and hyperparameters use the follo
 
 .. code-block:: python
 
-    model = MyLightingModule.load_from_checkpoint(PATH)
+    model = MyLightningModule.load_from_checkpoint(PATH)
 
     print(model.learning_rate)
     # prints the learning_rate you used in this checkpoint

docs/source/extensions/accelerators.rst

@@ -3,7 +3,7 @@
 ############
 Accelerators
 ############
-Accelerators connect a Lightning Trainer to arbitrary accelerators (CPUs, GPUs, TPUs, etc). Accelerators
+Accelerators connect a Lightning Trainer to arbitrary accelerators (CPUs, GPUs, TPUs, IPUs). Accelerators
 also manage distributed communication through :ref:`Plugins` (like DP, DDP, HPC cluster) and
 can also be configured to run on arbitrary clusters or to link up to arbitrary
 computational strategies like 16-bit precision via AMP and Apex.
@@ -26,7 +26,7 @@ One to handle differences from the training routine and one to handle different
     from pytorch_lightning.plugins import NativeMixedPrecisionPlugin, DDPPlugin
 
     accelerator = GPUAccelerator(
-        precision_plugin=NativeMixedPrecisionPlugin(16, "cuda"),
+        precision_plugin=NativeMixedPrecisionPlugin(precision=16, device="cuda"),
         training_type_plugin=DDPPlugin(),
     )
     trainer = Trainer(accelerator=accelerator)