complete PR

CHANGELOG.md

@@ -247,10 +247,10 @@ The format is based on [Keep a Changelog](http://keepachangelog.com/en/1.0.0/).
 - The `trainer.lightning_module` reference is now properly set at the very beginning of the run ([#8536](https://github.com/PyTorchLightning/pytorch-lightning/pull/8536))
 
 
-- The `Trainer` functions `reset_{train,val,test,predict}_dataloader`, `reset_train_val_dataloaders`, and `request_dataloader` `model` argument is now optional ([#8536](https://github.com/PyTorchLightning/pytorch-lightning/pull/8536))
+- Load ckpt path when model provided in validate/test/predict ([#8352](https://github.com/PyTorchLightning/pytorch-lightning/pull/8352))
 
 
-- Load ckpt path when model provided in validate/test/predict ([#8352](https://github.com/PyTorchLightning/pytorch-lightning/pull/8352)))
+- The `Trainer` functions `reset_{train,val,test,predict}_dataloader`, `reset_train_val_dataloaders`, and `request_dataloader` `model` argument is now optional ([#8536](https://github.com/PyTorchLightning/pytorch-lightning/pull/8536))
 
 
 - Saved checkpoints will no longer use the type of a `Callback` as the key to avoid issues with unpickling ([#6886](https://github.com/PyTorchLightning/pytorch-lightning/pull/6886))
@@ -281,9 +281,6 @@ The format is based on [Keep a Changelog](http://keepachangelog.com/en/1.0.0/).
 - Removed restrictions in the trainer that loggers can only log from rank 0. Existing logger behavior has not changed. ([#8608](https://github.com/PyTorchLightning/pytorch-lightning/pull/8608))
 
 
-- Disable quantization aware training observers by default during validating/testing/predicting stages ([#8540](https://github.com/PyTorchLightning/pytorch-lightning/pull/8540))
-
-
 - `Trainer.request_dataloader` now takes a `RunningStage` enum instance ([#8858](https://github.com/PyTorchLightning/pytorch-lightning/pull/8858))
 
 
@@ -330,13 +327,14 @@ The format is based on [Keep a Changelog](http://keepachangelog.com/en/1.0.0/).
 - `pytorch_lightning.utilities.grads.grad_norm` now raises an exception if parameter `norm_type <= 0` ([#9765](https://github.com/PyTorchLightning/pytorch-lightning/pull/9765))
 
 
-
 - Updated error message for interactive incompatible plugins ([#9896](https://github.com/PyTorchLightning/pytorch-lightning/pull/9896))
 
 
 - Updated several places in the loops and trainer to access `training_type_plugin` directly instead of `accelerator` ([#9901](https://github.com/PyTorchLightning/pytorch-lightning/pull/9901))
 
 
+- Disable quantization aware training observers by default during validating/testing/predicting stages ([#8540](https://github.com/PyTorchLightning/pytorch-lightning/pull/8540))
+
 
 ### Deprecated
 
@@ -411,6 +409,7 @@ The format is based on [Keep a Changelog](http://keepachangelog.com/en/1.0.0/).
 
 - Deprecated `GPUStatsMonitor` and `XLAStatsMonitor` in favor of `DeviceStatsMonitor` callback ([#9924](https://github.com/PyTorchLightning/pytorch-lightning/pull/9924))
 
+
 ### Removed
 
 - Removed deprecated `metrics` ([#8586](https://github.com/PyTorchLightning/pytorch-lightning/pull/8586/))
@@ -613,7 +612,6 @@ The format is based on [Keep a Changelog](http://keepachangelog.com/en/1.0.0/).
 - Fixed `LearningRateMonitor` logging with multiple param groups optimizer with no scheduler ([#10044](https://github.com/PyTorchLightning/pytorch-lightning/pull/10044))
 
 
-
 - Fixed undesired side effects being caused by `Trainer` patching dataloader methods on the `LightningModule` ([#9764](https://github.com/PyTorchLightning/pytorch-lightning/pull/9764))
 
 
@@ -935,9 +933,6 @@ The format is based on [Keep a Changelog](http://keepachangelog.com/en/1.0.0/).
 - Fixed `lr_scheduler` with metric (e.g. `torch.optim.lr_scheduler.ReduceLROnPlateau`) when using `automatic_optimization = False` ([#7643](https://github.com/PyTorchLightning/pytorch-lightning/pull/7643))
 - Fixed `DeepSpeed` breaking with no schedulers ([#8580](https://github.com/PyTorchLightning/pytorch-lightning/pull/8580))
 
-- Disable quantization aware training observers by default during validating/testing/predicting stages ([#8540](https://github.com/PyTorchLightning/pytorch-lightning/pull/8540))
-
-
 
 ## [1.3.8] - 2021-07-01
 

pytorch_lightning/callbacks/quantization.py

@@ -22,7 +22,6 @@
 
 import torch
 from torch import Tensor
-from torch.quantization import QConfig
 
 from pytorch_lightning.utilities.imports import _TORCH_GREATER_EQUAL_1_8
 
@@ -152,7 +151,6 @@ def custom_trigger_last(trainer):
 
     .. _PyTorch Quantization: https://pytorch.org/docs/stable/quantization.html#quantization-aware-training
     .. _torch.quantization.QConfig: https://pytorch.org/docs/stable/torch.quantization.html#torch.quantization.QConfig
-
     """
 
     OBSERVER_TYPES = ("histogram", "average")

tests/callbacks/test_quantization.py

@@ -17,12 +17,10 @@
 import pytest
 import torch
 from torchmetrics.functional import mean_absolute_percentage_error as mape
-from torch import Tensor
 
 from pytorch_lightning import seed_everything, Trainer
 from pytorch_lightning.callbacks import QuantizationAwareTraining
 from pytorch_lightning.utilities.exceptions import MisconfigurationException
-from pytorch_lightning.utilities.memory import get_model_size_mb
 from pytorch_lightning.utilities.imports import _TORCH_GREATER_EQUAL_1_8
 from pytorch_lightning.utilities.memory import get_model_size_mb
 from tests.helpers.boring_model import RandomDataset
@@ -171,7 +169,7 @@ def _get_observer_enabled(fake_quant: FakeQuantizeBase):
 )
 @RunIf(quantization=True)
 def test_quantization_disable_observers(tmpdir, observer_enabled_stages):
-    """Test disabling observers"""
+    """Test disabling observers."""
     qmodel = RegressionModel()
     qcb = QuantizationAwareTraining(observer_enabled_stages=observer_enabled_stages)
     trainer = Trainer(callbacks=[qcb], default_root_dir=tmpdir)
@@ -207,7 +205,7 @@ def test_quantization_disable_observers(tmpdir, observer_enabled_stages):
 
 @RunIf(quantization=True)
 def test_quantization_val_test_predict(tmpdir):
-    """Test the default quantization aware training not affected by validating, testing and predicting"""
+    """Test the default quantization aware training not affected by validating, testing and predicting."""
     seed_everything(42)
     num_features = 16
     dm = RegressDataModule(num_features=num_features)