Table of Contents
  • Docs >
  • Upgrade from 1.7 to the 2.0
Shortcuts

Upgrade from 1.7 to the 2.0

Regular User

reg. user 1.7

If

Then

Ref

have wrapped your loggers with LoggerCollection

directly pass a list of loggers to the Trainer and access the list via the trainer.loggers attribute.

#12147

used Trainer.lr_schedulers

access trainer.lr_scheduler_configs instead, which contains dataclasses instead of dictionaries.

#11443

used neptune-client API in the NeptuneLogger

upgrade to the latest API

#14727

used LightningDataModule.on_save hook

use LightningDataModule.on_save_checkpoint instead

#11887

used LightningDataModule.on_load_checkpoint hook

use LightningDataModule.on_load_checkpoint hook instead

#11887

used LightningModule.on_hpc_load hook

switch to general purpose hook LightningModule.on_load_checkpoint

#14315

used LightningModule.on_hpc_save hook

switch to general purpose hook LightningModule.on_save_checkpoint

#14315

used Trainer’s flag weights_save_path

use directly dirpath argument in the ModelCheckpoint callback.

#14424

used Trainer’s property Trainer.weights_save_path is dropped

#14424
reg. user 1.8

If

Then

Ref

used seed_everything_default=None in LightningCLI

set seed_everything_default=False instead

#12804

used Trainer.reset_train_val_dataloaders()

call Trainer.reset_train_dataloaders() and Trainer.reset_val_dataloaders() separately

#12184

imported pl.core.lightning

import pl.core.module instead

#12740
reg. user 1.9

If

Then

Ref

used Python 3.7

upgrade to Python 3.8 or higher

#16579

used PyTorch 1.10

upgrade to PyTorch 1.11 or higher

#16492

used Trainer’s flag gpus

use devices with the same number

#16171

used Trainer’s flag tpu_cores

use devices with the same number

#16171

used Trainer’s flag ipus

use devices with the same number

#16171

used Trainer’s flag num_processes

use devices with the same number

#16171

used Trainer’s flag resume_from_checkpoint

pass the path to the Trainer.fit(ckpt_path="...") method,

#10061

used Trainer’s flag auto_select_gpus

use devices="auto"

#16184

called the pl.tuner.auto_gpu_select.pick_single_gpu function

use Trainer’s flag``devices=”auto”``

#16184

called the pl.tuner.auto_gpu_select.pick_multiple_gpus functions

use Trainer’s flag``devices=”auto”``

#16184

used Trainer’s flag accumulate_grad_batches with a scheduling dictionary value

use the GradientAccumulationScheduler callback and configure it

#16729

imported profiles from pl.profiler

import from pl.profilers

#16359

used Tuner as part of Trainer in any form

move to a standalone Tuner object or use particular callbacks LearningRateFinder and BatchSizeFinder

https://lightning.ai/docs/pytorch/latest/advanced/training_tricks.html#batch-size-finder

used Trainer’s flag auto_scale_batch_size

use BatchSizeFinder callback instead and the Trainer.tune() method was removed

used Trainer’s flag auto_lr_find

use callbacks LearningRateFinder callback instead and the Trainer.tune() method was removed

Advanced User

adv. user 1.7

If

Then

Ref

used DDP2Strategy

switch to DDPStrategy

#14026

used Trainer.training_type_plugin property

now use Trainer.strategy and update the references

#11141

used any TrainingTypePluginsn

rename them to Strategy

#11120

used DistributedType

rely on protected _StrategyType

#10505

used DeviceType

rely on protected _AcceleratorType

#10503

used pl.utiltiies.meta functions

switch to built-in https://github.com/pytorch/torchdistx support

#13868

have implemented Callback.on_configure_sharded_model hook

move your implementation to Callback.setup

#14834

have implemented the Callback.on_before_accelerator_backend_setup hook

move your implementation to Callback.setup

#14834

have implemented the Callback.on_batch_start hook

move your implementation to Callback.on_train_batch_start

#14834

have implemented the Callback.on_batch_end hook

move your implementation to Callback.on_train_batch_end

#14834

have implemented the Callback.on_epoch_start hook

move your implementation to Callback.on_train_epoch_start , to Callback.on_validation_epoch_start , to Callback.on_test_epoch_start

#14834

have implemented the Callback.on_pretrain_routine_{start,end} hook

move your implementation to Callback.on_fit_start

#14834

used Callback.on_init_start hook

use Callback.on_train_start instead

#10940

used Callback.on_init_end hook

use Callback.on_train_start instead

#10940

used Trainer’s attribute Trainer.num_processes

it was replaced by Trainer.num_devices

#12388

used Trainer’s attribute Trainer.gpus

it was replaced by Trainer.num_devices

#12436

used Trainer’s attribute Trainer.num_gpus

use Trainer.num_devices instead

#12384

used Trainer’s attribute Trainer.ipus

use Trainer.num_devices instead

#12386

used Trainer’s attribute Trainer.tpu_cores

use Trainer.num_devices instead

#12437

used Trainer.num_processes attribute

switch to using Trainer.num_devices

#12388

used LightningIPUModule

it was removed

#14830

logged with LightningLoggerBase.agg_and_log_metrics

switch to LightningLoggerBase.log_metrics

#11832

used agg_key_funcs parameter from LightningLoggerBase

log metrics explicitly

#11871

used agg_default_func parameters in LightningLoggerBase

log metrics explicitly

#11871

used Trainer.validated_ckpt_path attribute

rely on generic read-only property Trainer.ckpt_path which is set when checkpoints are loaded via Trainer.validate(````ckpt_path=...)

#11696

used Trainer.tested_ckpt_path attribute

rely on generic read-only property Trainer.ckpt_path which is set when checkpoints are loaded via Trainer.test(````ckpt_path=...)

#11696

used Trainer.predicted_ckpt_path attribute

rely on generic read-only property Trainer.ckpt_path, which is set when checkpoints are loaded via Trainer.predict(````ckpt_path=...)

#11696

rely on the returned dictionary from Callback.on_save_checkpoint

call directly Callback.state_dict instead

#11887
adv. user 1.8

If

Then

Ref

imported pl.callbacks.base

import pl.callbacks.callback

#13031

imported pl.loops.base

import pl.loops.loop instead

#13043

imported pl.utilities.cli

import pl.cli instead

#13767

imported profiler classes from pl.profiler.*

import pl.profilers instead

#12308

used pl.accelerators.GPUAccelerator

use pl.accelerators.CUDAAccelerator

#13636

used LightningDeepSpeedModule

use strategy="deepspeed" or strategy=DeepSpeedStrategy(...)

https://pytorch-lightning.readthedocs.io/en/stable/api/pytorch_lightning.strategies.DeepSpeedStrategy.html

used the with init_meta_context() context manager from import pl.utilities.meta

switch to deepspeed-zero-stage-3

https://pytorch-lightning.readthedocs.io/en/stable/advanced/model_parallel.html#deepspeed-zero-stage-3

used the Lightning Hydra multi-run integration

removed support for it as it caused issues with processes hanging

#15689

used pl.utilities.memory.get_gpu_memory_map

use pl.accelerators.cuda.get_nvidia_gpu_stats

#9921
adv. user 1.9

If

Then

Ref

used the pl.lite module

switch to lightning_fabric

#15953

used Trainer’s flag strategy='dp'

use DDP with strategy='ddp' or DeepSpeed instead

#16748

implemented LightningModule.training_epoch_end hooks

port your logic to LightningModule.on_training_epoch_end hook

#16520

implemented LightningModule.validation_epoch_end hook

port your logic to LightningModule.on_validation_epoch_end hook

#16520

implemented LightningModule.test_epoch_end hooks

port your logic to LightningModule.on_test_epoch_end hook

#16520

used Trainer’s flag multiple_trainloader_mode

switch to CombinedLoader(..., mode=...) and set mode directly now

#16800

used Trainer’s flag move_metrics_to_cpu

implement particular offload logic in your custom metric or turn it on in torchmetrics

#16358

used Trainer’s flag track_grad_norm

overwrite on_before_optimizer_step hook and pass the argument directly and LightningModule.log_grad_norm() hook

#16745 #16745

used Trainer’s flag replace_sampler_ddp

use use_distributed_sample; the sampler gets created not only for the DDP strategies

relied on the on_tpu argument in LightningModule.optimizer_step hook

switch to manual optimization

#16537 Manual Optimization

relied on the using_lbfgs argument in LightningModule.optimizer_step hook

switch to manual optimization

#16538 Manual Optimization

were using nvidia/apex in any form

switch to PyTorch native mixed precision torch.amp instead

#16039 Precision

used Trainer’s flag using_native_amp

use PyTorch native mixed precision

#16039 Precision

used Trainer’s flag amp_backend

use PyTorch native mixed precision

#16039 Precision

used Trainer’s flag amp_level

use PyTorch native mixed precision

#16039 Precision

used Trainer’s attribute using_native_amp

use PyTorch native mixed precision

#16039 Precision

used Trainer’s attribute amp_backend

use PyTorch native mixed precision

#16039 Precision

used Trainer’s attribute amp_level

use PyTorch native mixed precision

#16039 Precision

use the FairScale integration

consider using PyTorch’s native FSDP implementation or outsourced implementation into own project

https://github.com/Lightning-Sandbox/lightning-Fairscale

used pl.overrides.fairscale.LightningShardedDataParallel

use native FSDP instead

#16400 FSDP

used pl.plugins.precision.fully_sharded_native_amp.FullyShardedNativeMixedPrecisionPlugin

use native FSDP instead

#16400 FSDP

used pl.plugins.precision.sharded_native_amp.ShardedNativeMixedPrecisionPlugin

use native FSDP instead

#16400 FSDP

used pl.strategies.fully_sharded.DDPFullyShardedStrategy

use native FSDP instead

#16400 FSDP

used pl.strategies.sharded.DDPShardedStrategy

use native FSDP instead

#16400 FSDP

used pl.strategies.sharded_spawn.DDPSpawnShardedStrategy

use native FSDP instead

#16400 FSDP

used save_config_overwrite parameters in LightningCLI

pass this option and via dictionary of save_config_kwargs parameter

#14998

used save_config_multifile parameters in LightningCLI

pass this option and via dictionary of save_config_kwargs parameter

#14998

have customized loops Loop.replace()

implement your training loop with Fabric.

#14998 Fabric

have customized loops Loop.run()

implement your training loop with Fabric.

#14998 Fabric

have customized loops Loop.connect()

implement your training loop with Fabric.

#14998 Fabric

used the Trainer’s trainer.fit_loop property

implement your training loop with Fabric

#14998 Fabric

used the Trainer’s trainer.validate_loop property

implement your training loop with Fabric

#14998 Fabric

used the Trainer’s trainer.test_loop property

implement your training loop with Fabric

#14998 Fabric

used the Trainer’s trainer.predict_loop property

implement your training loop with Fabric

#14998 Fabric

used the Trainer.loop and fetching classes

being marked as protected

used opt_idx argument in BaseFinetuning.finetune_function

use manual optimization

#16539

used opt_idx argument in Callback.on_before_optimizer_step

use manual optimization

#16539 Manual Optimization

used optimizer_idx as an optional argument in LightningModule.training_step

use manual optimization

#16539 Manual Optimization

used optimizer_idx argument in LightningModule.on_before_optimizer_step

use manual optimization

#16539 Manual Optimization

used optimizer_idx argument in LightningModule.configure_gradient_clipping

use manual optimization

#16539 Manual Optimization

used optimizer_idx argument in LightningModule.optimizer_step

use manual optimization

#16539 Manual Optimization

used optimizer_idx argument in LightningModule.optimizer_zero_grad

use manual optimization

#16539 Manual Optimization

used optimizer_idx argument in LightningModule.lr_scheduler_step

use manual optimization

#16539 Manual Optimization

used declaring optimizer frequencies in the dictionary returned from LightningModule.configure_optimizers

use manual optimization

#16539 Manual Optimization

used optimizer argument in LightningModule.backward

use manual optimization

#16539 Manual Optimization

used optimizer_idx argument in LightningModule.backward

use manual optimization

#16539 Manual Optimization

used optimizer_idx argument in PrecisionPlugin.optimizer_step

use manual optimization

#16539 Manual Optimization

used optimizer_idx argument in PrecisionPlugin.,backward

use manual optimization

#16539 Manual Optimization

used optimizer_idx argument in PrecisionPlugin.optimizer_step

use manual optimization

#16539 Manual Optimization

used optimizer_idx argument in Strategy.backward

use manual optimization

#16539 Manual Optimization

used optimizer_idx argument in Strategy.optimizer_step

use manual optimization

#16539 Manual Optimization

used Trainer’s Trainer.optimizer_frequencies attribute

use manual optimization

Manual Optimization

used PL_INTER_BATCH_PARALLELISM environment flag

#16355

used training integration with Horovod

install standalone package/project

https://github.com/Lightning-AI/lightning-Horovod

used training integration with ColossalAI

install standalone package/project

https://lightning.ai/docs/pytorch/latest/advanced/third_party/colossalai.html

used QuantizationAwareTraining callback

use Torch’s Quantization directly

#16750

had any logic except reducing the DP outputs in LightningModule.training_step_end hook

port it to LightningModule.training_batch_end hook

#16791

had any logic except reducing the DP outputs in LightningModule.validation_step_end hook

port it to LightningModule.validation_batch_end hook

#16791

had any logic except reducing the DP outputs in LightningModule.test_step_end hook

port it to LightningModule.test_batch_end hook

#16791

used pl.strategies.DDPSpawnStrategy

switch to general DDPStrategy(start_method='spawn') with proper starting method

#16809

used the automatic addition of a moving average of the training_step loss in the progress bar

use self.log("loss", ..., prog_bar=True) instead.

#16192

rely on the outputs argument from the on_predict_epoch_end hook

access them via trainer.predict_loop.predictions

#16655

need to pass a dictionary to self.log()

pass them independently.

#16389

Developer

devel 1.7

If

Then

Ref

Removed the legacy Trainer.get_deprecated_arg_names()

#14415

used the generic method Trainer.run_stage

switch to a specific one depending on your purpose Trainer.{fit,validate,test,predict} .

#11000

used rank_zero_only from pl.utilities.distributed

import it from pl.utilities.rank_zero

#11747

used rank_zero_debug from pl.utilities.distributed

import it from pl.utilities.rank_zero

#11747

used rank_zero_info from pl.utilities.distributed

import it from pl.utilities.rank_zero

#11747

used rank_zero_warn from pl.utilities.warnings

import it from pl.utilities.rank_zero

#11747

used rank_zero_deprecation from pl.utilities.warnings

import it from pl.utilities.rank_zero

#11747

used LightningDeprecationWarning from pl.utilities.warnings

import it from pl.utilities.rank_zero

#11747

used LightningDeprecationWarning from pl.utilities.warnings

import it from pl.utilities.rank_zero

#11747

used Trainer.data_parallel_device_ids attribute

switch it to Trainer.device_ids

#12072

derived it from TrainerCallbackHookMixin

use Trainer base class

#14401

used base class pytorch_lightning.profiler.BaseProfilerto

switch to use pytorch_lightning.profiler.Profiler instead

#12150

set distributed backend via the environment variable PL_TORCH_DISTRIBUTED_BACKEND

use process_group_backend in the strategy constructor

#11745

used PrecisionPlugin.on_load_checkpoint hooks

switch to PrecisionPlugin.load_state_dict

#11978

used PrecisionPlugin.on_save_checkpoint hooks

switch to PrecisionPlugin.load_state_dict

#11978

used Trainer.root_gpu attribute

use Trainer.strategy.root_device.index when GPU is used

#12262

used Trainer.use_amp attribute

rely on Torch native AMP

#12312

used LightingModule.use_amp attribute

rely on Torch native AMP

#12315

used Trainer’s attribute Trainer.verbose_evaluate

rely on loop constructor EvaluationLoop(verbose=...)

#10931

used Trainer’s attribute Trainer.should_rank_save_checkpoint

it was removed

#11068

derived from TrainerOptimizersMixin

rely on core/optimizer.py

#11155

derived from TrainerDataLoadingMixin

rely on methods from Trainer and DataConnector

#11282

used Trainer’s attribute Trainer.lightning_optimizers

switch to the Strategy and its attributes.

#11444

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.

#10979

used Profiler’s attribute SimpleProfiler.profile_iterable

it was removed

#12102

used Profiler’s attribute AdvancedProfiler.profile_iterable

it was removed

#12102

used the device_stats_monitor.prefix_metric_keys

#11254

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).

#12182

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).

#12182
devel 1.8

If

Then

Ref

derived from pytorch_lightning.loggers.base.LightningLoggerBase

derive from pytorch_lightning.loggers.logger.Logger

#12014

derived from pytorch_lightning.profiler.base.BaseProfiler

derive from pytorch_lightning.profilers.profiler.Profiler

#12150

derived from pytorch_lightning.profiler.base.AbstractProfiler

derive from pytorch_lightning.profilers.profiler.Profiler

#12106
devel 1.9

If

Then

Ref

passed the pl_module argument to distributed module wrappers

passed the (required) forward_module argument

#16386

used DataParallel and the LightningParallelModule wrapper

use DDP or DeepSpeed instead

#16748 DDP

used pl_module argument from the distributed module wrappers

use DDP or DeepSpeed instead

#16386 DDP

called pl.overrides.base.unwrap_lightning_module function

use DDP or DeepSpeed instead

#16386 DDP

used or derived from pl.overrides.distributed.LightningDistributedModule class

use DDP instead

#16386 DDP

used the pl.plugins.ApexMixedPrecisionPlugin`` plugin

use PyTorch native mixed precision

#16039

used the pl.plugins.NativeMixedPrecisionPlugin plugin

switch to the pl.plugins.MixedPrecisionPlugin plugin

#16039

used the fit_loop.min_steps setters

implement your training loop with Fabric

#16803

used the fit_loop.max_steps setters

implement your training loop with Fabric

#16803

used the data_parallel attribute in Trainer

check the same using isinstance(trainer.strategy, ParallelStrategy)

#16703

used any function from pl.utilities.xla_device

switch to pl.accelerators.TPUAccelerator.is_available()

#14514 #14550

imported functions from pl.utilities.device_parser.*

import them from lightning_fabric.utilities.device_parser.*

#14492 #14753

imported functions from pl.utilities.cloud_io.*

import them from lightning_fabric.utilities.cloud_io.*

#14515

imported functions from pl.utilities.apply_func.*

import them from lightning_utilities.core.apply_func.*

#14516 #14537

used any code from pl.core.mixins

use the base classes

#16424

used any code from pl.utilities.distributed

rely on Pytorch’s native functions

#16390

used any code from pl.utilities.data

it was removed

#16440

used any code from pl.utilities.optimizer

it was removed

#16439

used any code from pl.utilities.seed

it was removed

#16422

were using truncated backpropagation through time (TBPTT) with LightningModule.truncated_bptt_steps

use manual optimization

#16172 Manual Optimization

were using truncated backpropagation through time (TBPTT) with LightningModule.tbptt_split_batch

use manual optimization

#16172 Manual Optimization

were using truncated backpropagation through time (TBPTT) and passing hidden to LightningModule.training_step

use manual optimization

#16172 Manual Optimization

used pl.utilities.finite_checks.print_nan_gradients function

it was removed

used pl.utilities.finite_checks.detect_nan_parameters function

it was removed

used pl.utilities.parsing.flatten_dict function

it was removed

used pl.utilities.metrics.metrics_to_scalars function

it was removed

used pl.utilities.memory.get_model_size_mb function

it was removed

used pl.strategies.utils.on_colab_kaggle function

it was removed

#16437

used LightningDataModule.add_argparse_args() method

switch to using LightningCLI

#16708

used LightningDataModule.parse_argparser() method

switch to using LightningCLI

#16708

used LightningDataModule.from_argparse_args() method

switch to using LightningCLI

#16708

used LightningDataModule.get_init_arguments_and_types() method

switch to using LightningCLI

#16708

used Trainer.default_attributes() method

switch to using LightningCLI

#16708

used Trainer.from_argparse_args() method

switch to using LightningCLI

#16708

used Trainer.parse_argparser() method

switch to using LightningCLI

#16708

used Trainer.match_env_arguments() method

switch to using LightningCLI

#16708

used Trainer.add_argparse_args() method

switch to using LightningCLI

#16708

used pl.utilities.argparse.from_argparse_args() function

switch to using LightningCLI

#16708

used pl.utilities.argparse.parse_argparser() function

switch to using LightningCLI

#16708

used pl.utilities.argparseparse_env_variables() function

switch to using LightningCLI

#16708

used get_init_arguments_and_types() function

switch to using LightningCLI

#16708

used pl.utilities.argparse.add_argparse_args() function

switch to using LightningCLI

#16708

used pl.utilities.parsing.str_to_bool() function

switch to using LightningCLI

#16708

used pl.utilities.parsing.str_to_bool_or_int() function

switch to using LightningCLI

#16708

used pl.utilities.parsing.str_to_bool_or_str() function

switch to using LightningCLI

#16708

derived from pl.utilities.distributed.AllGatherGrad class

switch to PyTorch native equivalent

#15364

used PL_RECONCILE_PROCESS=1 env. variable

customize your logger

#16204

if you derived from mixin’s method pl.core.saving.ModelIO.load_from_checkpoint

rely on pl.core.module.LightningModule

#16999

used Accelerator.setup_environment method

switch to Accelerator.setup_device

#16436

used PL_FAULT_TOLERANT_TRAINING env. variable

implement own logic with Fabric

#16516 #16533

used or derived from public pl.overrides.distributed.IndexBatchSamplerWrapper class

it is set as protected

#16826

used the DataLoaderLoop class

use manual optimization

#16726 Manual Optimization

used the EvaluationEpochLoop class

use manual optimization

#16726 Manual Optimization

used the PredictionEpochLoop class

use manual optimization

#16726 Manual Optimization

used trainer.reset_*_dataloader() methods

use Loop.setup_data() for the top-level loops

#16726

used LightningModule.precision attribute

rely on Trainer precision attribute

#16203

used Trainer.model setter

you shall pass the model in fit/test/predict method

#16462

relied on pl.utilities.supporters.CombinedLoaderIterator class

pass dataloders directly

#16714

relied on pl.utilities.supporters.CombinedLoaderIterator class

pass dataloders directly

#16714

accessed ProgressBarBase.train_batch_idx property

rely on Trainer internal loops’ properties

#16760

accessed ProgressBarBase.val_batch_idx property

rely on Trainer internal loops’ properties

#16760

accessed ProgressBarBase.test_batch_idx property

rely on Trainer internal loops’ properties

#16760

accessed ProgressBarBase.predict_batch_idx property

rely on Trainer internal loops’ properties

#16760

used Trainer.prediction_writer_callbacks property

rely on precision plugin

#16759

used PrecisionPlugin.dispatch

it was removed

#16618

used Strategy.dispatch

it was removed

#16618

© Copyright Copyright (c) 2018-2023, Lightning AI et al...

Built with Sphinx using a theme provided by Read the Docs.

To analyze traffic and optimize your experience, we serve cookies on this site. By clicking or navigating, you agree to allow our usage of cookies. Read PyTorch Lightning's Privacy Policy.