1 7 Devel - Pytorch Lightning

.. list-table:: devel 1.7 :widths: 40 40 20 :header-rows: 1

- If
- Then
- Ref
- Removed the legacy Trainer.get_deprecated_arg_names()
- PR14415_
- used the generic method Trainer.run_stage
- switch to a specific one depending on your purpose Trainer.{fit,validate,test,predict} .
- PR11000_
- used rank_zero_only from pl.utilities.distributed
- import it from pl.utilities.rank_zero
- PR11747_
- used rank_zero_debug from pl.utilities.distributed
- import it from pl.utilities.rank_zero
- PR11747_
- used rank_zero_info from pl.utilities.distributed
- import it from pl.utilities.rank_zero
- PR11747_
- used rank_zero_warn from pl.utilities.warnings
- import it from pl.utilities.rank_zero
- PR11747_
- used rank_zero_deprecation from pl.utilities.warnings
- import it from pl.utilities.rank_zero
- PR11747_
- used LightningDeprecationWarning from pl.utilities.warnings
- import it from pl.utilities.rank_zero
- PR11747_
- used LightningDeprecationWarning from pl.utilities.warnings
- import it from pl.utilities.rank_zero
- PR11747_
- used Trainer.data_parallel_device_ids attribute
- switch it to Trainer.device_ids
- PR12072_
- derived it from TrainerCallbackHookMixin
- use Trainer base class
- PR14401_
- used base class pytorch_lightning.profiler.BaseProfilerto
- switch to use pytorch_lightning.profiler.Profiler instead
- PR12150_
- set distributed backend via the environment variable PL_TORCH_DISTRIBUTED_BACKEND
- use process_group_backend in the strategy constructor
- PR11745_
- used PrecisionPlugin.on_load_checkpoint hooks
- switch to PrecisionPlugin.load_state_dict
- PR11978_
- used PrecisionPlugin.on_save_checkpoint hooks
- switch to PrecisionPlugin.load_state_dict
- PR11978_
- used Trainer.root_gpu attribute
- use Trainer.strategy.root_device.index when GPU is used
- PR12262_
- used Trainer.use_amp attribute
- rely on Torch native AMP
- PR12312_
- used LightningModule.use_amp attribute
- rely on Torch native AMP
- PR12315_
- used Trainer’s attribute Trainer.verbose_evaluate
- rely on loop constructor EvaluationLoop(verbose=...)
- PR10931_
- used Trainer’s attribute Trainer.should_rank_save_checkpoint
- it was removed
- PR11068_
- derived from TrainerOptimizersMixin
- rely on core/optimizer.py
- PR11155_
- derived from TrainerDataLoadingMixin
- rely on methods from Trainer and DataConnector
- PR11282_
- used Trainer’s attribute Trainer.lightning_optimizers
- switch to the Strategy and its attributes.
- PR11444_
- used Trainer.call_hook
- it was set as a protected method Trainer._call_callback_hooks, Trainer._call_lightning_module_hook, Trainer._call_ttp_hook, Trainer._call_accelerator_hook and shall not be used.
- PR10979_
- used Profiler’s attribute SimpleProfiler.profile_iterable
- it was removed
- PR12102_
- used Profiler’s attribute AdvancedProfiler.profile_iterable
- it was removed
- PR12102_
- used the device_stats_monitor.prefix_metric_keys
- PR11254_
- used on_train_batch_end(outputs, ...) with 2d list with sizes (n_optimizers, tbptt_steps)
- chang it to (tbptt_steps, n_optimizers). You can update your code by adding the following parameter to your hook signature: on_train_batch_end(outputs, ..., new_format=True).
- PR12182_
- used training_epoch_end(outputs) with a 3d list with sizes (n_optimizers, n_batches, tbptt_steps)
- change it to (n_batches, tbptt_steps, n_optimizers). You can update your code by adding the following parameter to your hook signature: training_epoch_end(outputs, new_format=True).
- PR12182_