Shortcuts

Trainer

Once you’ve organized your PyTorch code into a LightningModule, the Trainer automates everything else.


This abstraction achieves the following:

  1. You maintain control over all aspects via PyTorch code without an added abstraction.

  2. The trainer uses best practices embedded by contributors and users from top AI labs such as Facebook AI Research, NYU, MIT, Stanford, etc…

  3. The trainer allows overriding any key part that you don’t want automated.



Basic use

This is the basic use of the trainer:

model = MyLightningModule()

trainer = Trainer()
trainer.fit(model, train_dataloader, val_dataloader)

Under the hood

Under the hood, the Lightning Trainer handles the training loop details for you, some examples include:

  • Automatically enabling/disabling grads

  • Running the training, validation and test dataloaders

  • Calling the Callbacks at the appropriate times

  • Putting batches and computations on the correct devices

Here’s the pseudocode for what the trainer does under the hood (showing the train loop only)

# put model in train mode
model.train()
torch.set_grad_enabled(True)

losses = []
for batch in train_dataloader:
    # calls hooks like this one
    on_train_batch_start()

    # train step
    loss = training_step(batch)

    # clear gradients
    optimizer.zero_grad()

    # backward
    loss.backward()

    # update parameters
    optimizer.step()

    losses.append(loss)

Trainer in Python scripts

In Python scripts, it’s recommended you use a main function to call the Trainer.

from argparse import ArgumentParser


def main(hparams):
    model = LightningModule()
    trainer = Trainer(gpus=hparams.gpus)
    trainer.fit(model)


if __name__ == "__main__":
    parser = ArgumentParser()
    parser.add_argument("--gpus", default=None)
    args = parser.parse_args()

    main(args)

So you can run it like so:

python main.py --gpus 2

Note

Pro-tip: You don’t need to define all flags manually. Lightning can add them automatically

from argparse import ArgumentParser


def main(args):
    model = LightningModule()
    trainer = Trainer.from_argparse_args(args)
    trainer.fit(model)


if __name__ == "__main__":
    parser = ArgumentParser()
    parser = Trainer.add_argparse_args(parser)
    args = parser.parse_args()

    main(args)

So you can run it like so:

python main.py --gpus 2 --max_steps 10 --limit_train_batches 10 --any_trainer_arg x

Note

If you want to stop a training run early, you can press “Ctrl + C” on your keyboard. The trainer will catch the KeyboardInterrupt and attempt a graceful shutdown, including running accelerator callback on_train_end to clean up memory. The trainer object will also set an attribute interrupted to True in such cases. If you have a callback which shuts down compute resources, for example, you can conditionally run the shutdown logic for only uninterrupted runs.


Validation

You can perform an evaluation epoch over the validation set, outside of the training loop, using pytorch_lightning.trainer.trainer.Trainer.validate(). This might be useful if you want to collect new metrics from a model right at its initialization or after it has already been trained.

trainer.validate(dataloaders=val_dataloaders)

Testing

Once you’re done training, feel free to run the test set! (Only right before publishing your paper or pushing to production)

trainer.test(test_dataloaders=test_dataloaders)

Reproducibility

To ensure full reproducibility from run to run you need to set seeds for pseudo-random generators, and set deterministic flag in Trainer.

Example:

from pytorch_lightning import Trainer, seed_everything

seed_everything(42, workers=True)
# sets seeds for numpy, torch, python.random and PYTHONHASHSEED.
model = Model()
trainer = Trainer(deterministic=True)

By setting workers=True in seed_everything(), Lightning derives unique seeds across all dataloader workers and processes for torch, numpy and stdlib random number generators. When turned on, it ensures that e.g. data augmentations are not repeated across workers.


Trainer flags

accelerator

Supports passing different accelerator types ("cpu", "gpu", "tpu", "ipu", "auto") as well as custom accelerator instances.

# CPU accelerator
trainer = Trainer(accelerator="cpu")

# Training with GPU Accelerator using 2 gpus
trainer = Trainer(devices=2, accelerator="gpu")

# Training with TPU Accelerator using 8 tpu cores
trainer = Trainer(devices=8, accelerator="tpu")

# Training with GPU Accelerator using the DistributedDataParallel strategy
trainer = Trainer(devices=4, accelerator="gpu", strategy="ddp")

Note

The "auto" option recognizes the machine you are on, and selects the respective Accelerator.

# If your machine has GPUs, it will use the GPU Accelerator for training
trainer = Trainer(devices=2, accelerator="auto")

You can also modify hardware behavior by subclassing an existing accelerator to adjust for your needs.

Example:

class MyOwnAcc(CPUAccelerator):
    ...

Trainer(accelerator=MyOwnAcc())

Warning

Passing training strategies (e.g., "ddp") to accelerator has been deprecated in v1.5.0 and will be removed in v1.7.0. Please use the strategy argument instead.

accumulate_grad_batches


Accumulates grads every k batches or as set up in the dict. Trainer also calls optimizer.step() for the last indivisible step number.

# default used by the Trainer (no accumulation)
trainer = Trainer(accumulate_grad_batches=1)

Example:

# accumulate every 4 batches (effective batch size is batch*4)
trainer = Trainer(accumulate_grad_batches=4)

# no accumulation for epochs 1-4. accumulate 3 for epochs 5-10. accumulate 20 after that
trainer = Trainer(accumulate_grad_batches={5: 3, 10: 20})

amp_backend


Use PyTorch AMP (‘native’) (available PyTorch 1.6+), or NVIDIA apex (‘apex’).

# using PyTorch built-in AMP, default used by the Trainer
trainer = Trainer(amp_backend="native")

# using NVIDIA Apex
trainer = Trainer(amp_backend="apex")

amp_level


The optimization level to use (O1, O2, etc…) for 16-bit GPU precision (using NVIDIA apex under the hood).

Check NVIDIA apex docs for level

Example:

# default used by the Trainer
trainer = Trainer(amp_level='O2')

auto_scale_batch_size


Automatically tries to find the largest batch size that fits into memory, before any training.

# default used by the Trainer (no scaling of batch size)
trainer = Trainer(auto_scale_batch_size=None)

# run batch size scaling, result overrides hparams.batch_size
trainer = Trainer(auto_scale_batch_size="binsearch")

# call tune to find the batch size
trainer.tune(model)

auto_select_gpus


If enabled and gpus is an integer, pick available gpus automatically. This is especially useful when GPUs are configured to be in “exclusive mode”, such that only one process at a time can access them.

Example:

# no auto selection (picks first 2 gpus on system, may fail if other process is occupying)
trainer = Trainer(gpus=2, auto_select_gpus=False)

# enable auto selection (will find two available gpus on system)
trainer = Trainer(gpus=2, auto_select_gpus=True)

# specifies all GPUs regardless of its availability
Trainer(gpus=-1, auto_select_gpus=False)

# specifies all available GPUs (if only one GPU is not occupied, uses one gpu)
Trainer(gpus=-1, auto_select_gpus=True)

auto_lr_find


Runs a learning rate finder algorithm (see this paper) when calling trainer.tune(), to find optimal initial learning rate.

# default used by the Trainer (no learning rate finder)
trainer = Trainer(auto_lr_find=False)

Example:

# run learning rate finder, results override hparams.learning_rate
trainer = Trainer(auto_lr_find=True)

# call tune to find the lr
trainer.tune(model)

Example:

# run learning rate finder, results override hparams.my_lr_arg
trainer = Trainer(auto_lr_find='my_lr_arg')

# call tune to find the lr
trainer.tune(model)

benchmark


If true enables cudnn.benchmark. This flag is likely to increase the speed of your system if your input sizes don’t change. However, if it does, then it will likely make your system slower.

The speedup comes from allowing the cudnn auto-tuner to find the best algorithm for the hardware [see discussion here].

Example:

# default used by the Trainer
trainer = Trainer(benchmark=False)

deterministic


If true enables cudnn.deterministic. Might make your system slower, but ensures reproducibility. Also sets $HOROVOD_FUSION_THRESHOLD=0.

For more info check [pytorch docs].

Example:

# default used by the Trainer
trainer = Trainer(deterministic=False)

callbacks


Add a list of Callback. Callbacks run sequentially in the order defined here with the exception of ModelCheckpoint callbacks which run after all others to ensure all states are saved to the checkpoints.

# a list of callbacks
callbacks = [PrintCallback()]
trainer = Trainer(callbacks=callbacks)

Example:

from pytorch_lightning.callbacks import Callback

class PrintCallback(Callback):
    def on_train_start(self, trainer, pl_module):
        print("Training is started!")
    def on_train_end(self, trainer, pl_module):
        print("Training is done.")

Model-specific callbacks can also be added inside the LightningModule through configure_callbacks(). Callbacks returned in this hook will extend the list initially given to the Trainer argument, and replace the trainer callbacks should there be two or more of the same type. ModelCheckpoint callbacks always run last.

check_val_every_n_epoch


Check val every n train epochs.

Example:

# default used by the Trainer
trainer = Trainer(check_val_every_n_epoch=1)

# run val loop every 10 training epochs
trainer = Trainer(check_val_every_n_epoch=10)

checkpoint_callback

Warning

checkpoint_callback has been deprecated in v1.5 and will be removed in v1.7. To disable checkpointing, pass enable_checkpointing = False to the Trainer instead.

default_root_dir


Default path for logs and weights when no logger or pytorch_lightning.callbacks.ModelCheckpoint callback passed. On certain clusters you might want to separate where logs and checkpoints are stored. If you don’t then use this argument for convenience. Paths can be local paths or remote paths such as s3://bucket/path or ‘hdfs://path/’. Credentials will need to be set up to use remote filepaths.

# default used by the Trainer
trainer = Trainer(default_root_dir=os.getcwd())

devices

Number of devices to train on (int), which devices to train on (list or str), or "auto". It will be mapped to either gpus, tpu_cores, num_processes or ipus, based on the accelerator type ("cpu", "gpu", "tpu", "ipu", "auto").

# Training with CPU Accelerator using 2 processes
trainer = Trainer(devices=2, accelerator="cpu")

# Training with GPU Accelerator using GPUs 1 and 3
trainer = Trainer(devices=[1, 3], accelerator="gpu")

# Training with TPU Accelerator using 8 tpu cores
trainer = Trainer(devices=8, accelerator="tpu")

Tip

The "auto" option recognizes the devices to train on, depending on the Accelerator being used.

# If your machine has GPUs, it will use all the available GPUs for training
trainer = Trainer(devices="auto", accelerator="auto")

# Training with CPU Accelerator using 1 process
trainer = Trainer(devices="auto", accelerator="cpu")

# Training with TPU Accelerator using 8 tpu cores
trainer = Trainer(devices="auto", accelerator="tpu")

# Training with IPU Accelerator using 4 ipus
trainer = Trainer(devices="auto", accelerator="ipu")

enable_checkpointing


By default Lightning saves a checkpoint for you in your current working directory, with the state of your last training epoch, Checkpoints capture the exact value of all parameters used by a model. To disable automatic checkpointing, set this to False.

# default used by Trainer, saves the most recent model to a single checkpoint after each epoch
trainer = Trainer(enable_checkpointing=True)

# turn off automatic checkpointing
trainer = Trainer(enable_checkpointing=False)

You can override the default behavior by initializing the ModelCheckpoint callback, and adding it to the callbacks list. See Saving and Loading Weights for how to customize checkpointing.

from pytorch_lightning.callbacks import ModelCheckpoint

# Init ModelCheckpoint callback, monitoring 'val_loss'
checkpoint_callback = ModelCheckpoint(monitor="val_loss")

# Add your callback to the callbacks list
trainer = Trainer(callbacks=[checkpoint_callback])

fast_dev_run


Runs n if set to n (int) else 1 if set to True batch(es) of train, val and test to find any bugs (ie: a sort of unit test).

Under the hood the pseudocode looks like this when running fast_dev_run with a single batch:

# loading
__init__()
prepare_data

# test training step
training_batch = next(train_dataloader)
training_step(training_batch)

# test val step
val_batch = next(val_dataloader)
out = validation_step(val_batch)
validation_epoch_end([out])
# default used by the Trainer
trainer = Trainer(fast_dev_run=False)

# runs 1 train, val, test batch and program ends
trainer = Trainer(fast_dev_run=True)

# runs 7 train, val, test batches and program ends
trainer = Trainer(fast_dev_run=7)

Note

This argument is a bit different from limit_train/val/test_batches. Setting this argument will disable tuner, checkpoint callbacks, early stopping callbacks, loggers and logger callbacks like LearningRateLogger and runs for only 1 epoch. This must be used only for debugging purposes. limit_train/val/test_batches only limits the number of batches and won’t disable anything.

flush_logs_every_n_steps

Warning

flush_logs_every_n_steps has been deprecated in v1.5 and will be removed in v1.7. Please configure flushing directly in the logger instead.


Writes logs to disk this often.

# default used by the Trainer
trainer = Trainer(flush_logs_every_n_steps=100)
See Also:

gpus


  • Number of GPUs to train on (int)

  • or which GPUs to train on (list)

  • can handle strings

# default used by the Trainer (ie: train on CPU)
trainer = Trainer(gpus=None)

# equivalent
trainer = Trainer(gpus=0)

Example:

# int: train on 2 gpus
trainer = Trainer(gpus=2)

# list: train on GPUs 1, 4 (by bus ordering)
trainer = Trainer(gpus=[1, 4])
trainer = Trainer(gpus='1, 4') # equivalent

# -1: train on all gpus
trainer = Trainer(gpus=-1)
trainer = Trainer(gpus='-1') # equivalent

# combine with num_nodes to train on multiple GPUs across nodes
# uses 8 gpus in total
trainer = Trainer(gpus=2, num_nodes=4)

# train only on GPUs 1 and 4 across nodes
trainer = Trainer(gpus=[1, 4], num_nodes=4)
See Also:

gradient_clip_val


Gradient clipping value

  • 0 means don’t clip.

# default used by the Trainer
trainer = Trainer(gradient_clip_val=0.0)

limit_train_batches


How much of training dataset to check. Useful when debugging or testing something that happens at the end of an epoch.

# default used by the Trainer
trainer = Trainer(limit_train_batches=1.0)

Example:

# default used by the Trainer
trainer = Trainer(limit_train_batches=1.0)

# run through only 25% of the training set each epoch
trainer = Trainer(limit_train_batches=0.25)

# run through only 10 batches of the training set each epoch
trainer = Trainer(limit_train_batches=10)

limit_test_batches


How much of test dataset to check.

# default used by the Trainer
trainer = Trainer(limit_test_batches=1.0)

# run through only 25% of the test set each epoch
trainer = Trainer(limit_test_batches=0.25)

# run for only 10 batches
trainer = Trainer(limit_test_batches=10)

In the case of multiple test dataloaders, the limit applies to each dataloader individually.

limit_val_batches


How much of validation dataset to check. Useful when debugging or testing something that happens at the end of an epoch.

# default used by the Trainer
trainer = Trainer(limit_val_batches=1.0)

# run through only 25% of the validation set each epoch
trainer = Trainer(limit_val_batches=0.25)

# run for only 10 batches
trainer = Trainer(limit_val_batches=10)

In the case of multiple validation dataloaders, the limit applies to each dataloader individually.

log_every_n_steps


How often to add logging rows (does not write to disk)

# default used by the Trainer
trainer = Trainer(log_every_n_steps=50)
See Also:

logger


Logger (or iterable collection of loggers) for experiment tracking. A True value uses the default TensorBoardLogger shown below. False will disable logging.

from pytorch_lightning.loggers import TensorBoardLogger

# default logger used by trainer
logger = TensorBoardLogger(save_dir=os.getcwd(), version=1, name="lightning_logs")
Trainer(logger=logger)

max_epochs


Stop training once this number of epochs is reached

# default used by the Trainer
trainer = Trainer(max_epochs=1000)

If both max_epochs and max_steps aren’t specified, max_epochs will default to 1000. To enable infinite training, set max_epochs = -1.

min_epochs


Force training for at least these many epochs

# default used by the Trainer
trainer = Trainer(min_epochs=1)

max_steps


Stop training after this number of steps Training will stop if max_steps or max_epochs have reached (earliest).

# Default (disabled)
trainer = Trainer(max_steps=None)

# Stop after 100 steps
trainer = Trainer(max_steps=100)

If max_steps is not specified, max_epochs will be used instead (and max_epochs defaults to 1000 if max_epochs is not specified). To disable this default, set max_steps = -1.

min_steps


Force training for at least these number of steps. Trainer will train model for at least min_steps or min_epochs (latest).

# Default (disabled)
trainer = Trainer(min_steps=None)

# Run at least for 100 steps (disable min_epochs)
trainer = Trainer(min_steps=100, min_epochs=0)

max_time

Set the maximum amount of time for training. Training will get interrupted mid-epoch. For customizable options use the Timer callback.

# Default (disabled)
trainer = Trainer(max_time=None)

# Stop after 12 hours of training or when reaching 10 epochs (string)
trainer = Trainer(max_time="00:12:00:00", max_epochs=10)

# Stop after 1 day and 5 hours (dict)
trainer = Trainer(max_time={"days": 1, "hours": 5})

In case max_time is used together with min_steps or min_epochs, the min_* requirement always has precedence.

num_nodes


Number of GPU nodes for distributed training.

# default used by the Trainer
trainer = Trainer(num_nodes=1)

# to train on 8 nodes
trainer = Trainer(num_nodes=8)

num_processes


Number of processes to train with. Automatically set to the number of GPUs when using strategy="ddp". Set to a number greater than 1 when using accelerator="cpu" and strategy="ddp" to mimic distributed training on a machine without GPUs. This is useful for debugging, but will not provide any speedup, since single-process Torch already makes efficient use of multiple CPUs. While it would typically spawns subprocesses for training, setting num_nodes > 1 and keeping num_processes = 1 runs training in the main process.

# Simulate DDP for debugging on your GPU-less laptop
trainer = Trainer(accelerator="cpu", strategy="ddp", num_processes=2)

num_sanity_val_steps


Sanity check runs n batches of val before starting the training routine. This catches any bugs in your validation without having to wait for the first validation check. The Trainer uses 2 steps by default. Turn it off or modify it here.

# default used by the Trainer
trainer = Trainer(num_sanity_val_steps=2)

# turn it off
trainer = Trainer(num_sanity_val_steps=0)

# check all validation data
trainer = Trainer(num_sanity_val_steps=-1)

This option will reset the validation dataloader unless num_sanity_val_steps=0.

overfit_batches


Uses this much data of the training set. If nonzero, will use the same training set for validation and testing. If the training dataloaders have shuffle=True, Lightning will automatically disable it.

Useful for quickly debugging or trying to overfit on purpose.

# default used by the Trainer
trainer = Trainer(overfit_batches=0.0)

# use only 1% of the train set (and use the train set for val and test)
trainer = Trainer(overfit_batches=0.01)

# overfit on 10 of the same batches
trainer = Trainer(overfit_batches=10)

plugins


Plugins allow you to connect arbitrary backends, precision libraries, clusters etc. For example:

To define your own behavior, subclass the relevant class and pass it in. Here’s an example linking up your own ClusterEnvironment.

from pytorch_lightning.plugins.environments import ClusterEnvironment


class MyCluster(ClusterEnvironment):
    def master_address(self):
        return your_master_address

    def master_port(self):
        return your_master_port

    def world_size(self):
        return the_world_size


trainer = Trainer(plugins=[MyCluster()], ...)

prepare_data_per_node


If True will call prepare_data() on LOCAL_RANK=0 for every node. If False will only call from NODE_RANK=0, LOCAL_RANK=0

# default
Trainer(prepare_data_per_node=True)

# use only NODE_RANK=0, LOCAL_RANK=0
Trainer(prepare_data_per_node=False)

precision


Lightning supports either double precision (64), full precision (32), or half precision (16) training.

Half precision, or mixed precision, is the combined use of 32 and 16 bit floating points to reduce memory footprint during model training. This can result in improved performance, achieving +3X speedups on modern GPUs.

# default used by the Trainer
trainer = Trainer(precision=32, gpus=1)

# 16-bit precision
trainer = Trainer(precision=16, gpus=1)

# 64-bit precision
trainer = Trainer(precision=64, gpus=1)

Note

When running on TPUs, torch.float16 will be used but tensor printing will still show torch.float32.

Note

16-bit precision is not supported on CPUs.

process_position

Warning

process_position has been deprecated in v1.5 and will be removed in v1.7. Please pass TQDMProgressBar with process_position directly to the Trainer’s callbacks argument instead.


Orders the progress bar. Useful when running multiple trainers on the same node.

# default used by the Trainer
trainer = Trainer(process_position=0)

Note

This argument is ignored if a custom callback is passed to callbacks.

profiler


To profile individual steps during training and assist in identifying bottlenecks.

See the profiler documentation. for more details.

from pytorch_lightning.profiler import SimpleProfiler, AdvancedProfiler

# default used by the Trainer
trainer = Trainer(profiler=None)

# to profile standard training events, equivalent to `profiler=SimpleProfiler()`
trainer = Trainer(profiler="simple")

# advanced profiler for function-level stats, equivalent to `profiler=AdvancedProfiler()`
trainer = Trainer(profiler="advanced")

progress_bar_refresh_rate

Warning

progress_bar_refresh_rate has been deprecated in v1.5 and will be removed in v1.7. Please pass TQDMProgressBar with refresh_rate directly to the Trainer’s callbacks argument instead. To disable the progress bar, pass enable_progress_bar = False to the Trainer.


How often to refresh progress bar (in steps).

# default used by the Trainer
trainer = Trainer(progress_bar_refresh_rate=1)

# disable progress bar
trainer = Trainer(progress_bar_refresh_rate=0)
Note:
  • In Google Colab notebooks, faster refresh rates (lower number) is known to crash them because of their screen refresh rates. Lightning will set it to 20 in these environments if the user does not provide a value.

  • This argument is ignored if a custom callback is passed to callbacks.

enable_progress_bar

Whether to enable or disable the progress bar. Defaults to True.

# default used by the Trainer
trainer = Trainer(enable_progress_bar=True)

# disable progress bar
trainer = Trainer(enable_progress_bar=False)

reload_dataloaders_every_n_epochs


Set to a postive integer to reload dataloaders every n epochs.

# if 0 (default)
train_loader = model.train_dataloader()
for epoch in epochs:
    for batch in train_loader:
        ...

# if a positive integer
for epoch in epochs:
    if not epoch % reload_dataloaders_every_n_epochs:
        train_loader = model.train_dataloader()
    for batch in train_loader:
        ...

replace_sampler_ddp


Enables auto adding of DistributedSampler. In PyTorch, you must use it in distributed settings such as TPUs or multi-node. The sampler makes sure each GPU sees the appropriate part of your data. By default it will add shuffle=True for train sampler and shuffle=False for val/test sampler. If you want to customize it, you can set replace_sampler_ddp=False and add your own distributed sampler. If replace_sampler_ddp=True and a distributed sampler was already added, Lightning will not replace the existing one.

# default used by the Trainer
trainer = Trainer(replace_sampler_ddp=True)

By setting to False, you have to add your own distributed sampler:

# in your LightningModule or LightningDataModule
def train_dataloader(self):
    # default used by the Trainer
    sampler = torch.utils.data.distributed.DistributedSampler(dataset, shuffle=True)
    dataloader = DataLoader(dataset, batch_size=32, sampler=sampler)
    return dataloader

Note

For iterable datasets, we don’t do this automatically.

resume_from_checkpoint

Warning

resume_from_checkpoint is deprecated in v1.5 and will be removed in v1.7. Please pass trainer.fit(ckpt_path="some/path/to/my_checkpoint.ckpt") instead.


To resume training from a specific checkpoint pass in the path here. If resuming from a mid-epoch checkpoint, training will start from the beginning of the next epoch.

# default used by the Trainer
trainer = Trainer(resume_from_checkpoint=None)

# resume from a specific checkpoint
trainer = Trainer(resume_from_checkpoint="some/path/to/my_checkpoint.ckpt")

strategy

Supports passing different training strategies with aliases (ddp, ddp_spawn, etc) as well as custom training type plugins.

# Training with the DistributedDataParallel strategy on 4 gpus
trainer = Trainer(strategy="ddp", accelerator="gpu", devices=4)

# Training with the DDP Spawn strategy using 4 cpu processes
trainer = Trainer(strategy="ddp_spawn", accelerator="cpu", devices=4)

Note

Additionally, you can pass your custom training type plugins to the strategy argument.

from pytorch_lightning.plugins import DDPPlugin


class CustomDDPPlugin(DDPPlugin):
    def configure_ddp(self):
        self._model = MyCustomDistributedDataParallel(
            self.model,
            device_ids=...,
        )


trainer = Trainer(strategy=CustomDDPPlugin(), accelerator="gpu", devices=2)
See Also:

sync_batchnorm


Enable synchronization between batchnorm layers across all GPUs.

trainer = Trainer(sync_batchnorm=True)

track_grad_norm


  • no tracking (-1)

  • Otherwise tracks that norm (2 for 2-norm)

# default used by the Trainer
trainer = Trainer(track_grad_norm=-1)

# track the 2-norm
trainer = Trainer(track_grad_norm=2)

tpu_cores


  • How many TPU cores to train on (1 or 8).

  • Which TPU core to train on [1-8]

A single TPU v2 or v3 has 8 cores. A TPU pod has up to 2048 cores. A slice of a POD means you get as many cores as you request.

Your effective batch size is batch_size * total tpu cores.

This parameter can be either 1 or 8.

Example:

# your_trainer_file.py

# default used by the Trainer (ie: train on CPU)
trainer = Trainer(tpu_cores=None)

# int: train on a single core
trainer = Trainer(tpu_cores=1)

# list: train on a single selected core
trainer = Trainer(tpu_cores=[2])

# int: train on all cores few cores
trainer = Trainer(tpu_cores=8)

# for 8+ cores must submit via xla script with
# a max of 8 cores specified. The XLA script
# will duplicate script onto each TPU in the POD
trainer = Trainer(tpu_cores=8)

To train on more than 8 cores (ie: a POD), submit this script using the xla_dist script.

Example:

python -m torch_xla.distributed.xla_dist
--tpu=$TPU_POD_NAME
--conda-env=torch-xla-nightly
--env=XLA_USE_BF16=1
-- python your_trainer_file.py

val_check_interval


How often within one training epoch to check the validation set. Can specify as float or int.

  • use (float) to check within a training epoch

  • use (int) to check every n steps (batches)

# default used by the Trainer
trainer = Trainer(val_check_interval=1.0)

# check validation set 4 times during a training epoch
trainer = Trainer(val_check_interval=0.25)

# check validation set every 1000 training batches
# use this when using iterableDataset and your dataset has no length
# (ie: production cases with streaming data)
trainer = Trainer(val_check_interval=1000)
# Here is the computation to estimate the total number of batches seen within an epoch.

# Find the total number of train batches
total_train_batches = total_train_samples // (train_batch_size * world_size)

# Compute how many times we will call validation during the training loop
val_check_batch = max(1, int(total_train_batches * val_check_interval))
val_checks_per_epoch = total_train_batches / val_check_batch

# Find the total number of validation batches
total_val_batches = total_val_samples // (val_batch_size * world_size)

# Total number of batches run
total_fit_batches = total_train_batches + total_val_batches

weights_save_path


Directory of where to save weights if specified.

# default used by the Trainer
trainer = Trainer(weights_save_path=os.getcwd())

# save to your custom path
trainer = Trainer(weights_save_path="my/path")

Example:

# if checkpoint callback used, then overrides the weights path
# **NOTE: this saves weights to some/path NOT my/path
checkpoint = ModelCheckpoint(dirpath='some/path')
trainer = Trainer(
    callbacks=[checkpoint],
    weights_save_path='my/path'
)

weights_summary

Warning

weights_summary is deprecated in v1.5 and will be removed in v1.7. Please pass ModelSummary directly to the Trainer’s callbacks argument instead. To disable the model summary, pass enable_model_summary = False to the Trainer.


Prints a summary of the weights when training begins. Options: ‘full’, ‘top’, None.

# default used by the Trainer (ie: print summary of top level modules)
trainer = Trainer(weights_summary="top")

# print full summary of all modules and submodules
trainer = Trainer(weights_summary="full")

# don't print a summary
trainer = Trainer(weights_summary=None)

enable_model_summary

Whether to enable or disable the model summarization. Defaults to True.

# default used by the Trainer
trainer = Trainer(enable_model_summary=True)

# disable summarization
trainer = Trainer(enable_model_summary=False)

# enable custom summarization
from pytorch_lightning.callbacks import ModelSummary

trainer = Trainer(enable_model_summary=True, callbacks=[ModelSummary(max_depth=-1)])

Trainer class API

Methods

init

Trainer.__init__(logger=True, checkpoint_callback=None, enable_checkpointing=True, callbacks=None, default_root_dir=None, gradient_clip_val=None, gradient_clip_algorithm=None, process_position=0, num_nodes=1, num_processes=1, devices=None, gpus=None, auto_select_gpus=False, tpu_cores=None, ipus=None, log_gpu_memory=None, progress_bar_refresh_rate=None, enable_progress_bar=True, overfit_batches=0.0, track_grad_norm=- 1, check_val_every_n_epoch=1, fast_dev_run=False, accumulate_grad_batches=None, max_epochs=None, min_epochs=None, max_steps=- 1, min_steps=None, max_time=None, limit_train_batches=1.0, limit_val_batches=1.0, limit_test_batches=1.0, limit_predict_batches=1.0, val_check_interval=1.0, flush_logs_every_n_steps=None, log_every_n_steps=50, accelerator=None, strategy=None, sync_batchnorm=False, precision=32, enable_model_summary=True, weights_summary='top', weights_save_path=None, num_sanity_val_steps=2, resume_from_checkpoint=None, profiler=None, benchmark=False, deterministic=False, reload_dataloaders_every_n_epochs=0, reload_dataloaders_every_epoch=False, auto_lr_find=False, replace_sampler_ddp=True, detect_anomaly=False, auto_scale_batch_size=False, prepare_data_per_node=None, plugins=None, amp_backend='native', amp_level=None, move_metrics_to_cpu=False, multiple_trainloader_mode='max_size_cycle', stochastic_weight_avg=False, terminate_on_nan=None)[source]

Customize every aspect of training via flags.

Parameters
  • accelerator (Union[str, Accelerator, None]) –

    Supports passing different accelerator types (“cpu”, “gpu”, “tpu”, “ipu”, “auto”) as well as custom accelerator instances.

    Deprecated since version v1.5: Passing training strategies (e.g., ‘ddp’) to accelerator has been deprecated in v1.5.0 and will be removed in v1.7.0. Please use the strategy argument instead.

  • accumulate_grad_batches (Union[int, Dict[int, int], None]) – Accumulates grads every k batches or as set up in the dict.

  • amp_backend (str) – The mixed precision backend to use (“native” or “apex”).

  • amp_level (Optional[str]) – The optimization level to use (O1, O2, etc…). By default it will be set to “O2” if amp_backend is set to “apex”.

  • auto_lr_find (Union[bool, str]) – If set to True, will make trainer.tune() run a learning rate finder, trying to optimize initial learning for faster convergence. trainer.tune() method will set the suggested learning rate in self.lr or self.learning_rate in the LightningModule. To use a different key set a string instead of True with the key name.

  • auto_scale_batch_size (Union[str, bool]) – If set to True, will initially run a batch size finder trying to find the largest batch size that fits into memory. The result will be stored in self.batch_size in the LightningModule. Additionally, can be set to either power that estimates the batch size through a power search or binsearch that estimates the batch size through a binary search.

  • auto_select_gpus (bool) – If enabled and gpus is an integer, pick available gpus automatically. This is especially useful when GPUs are configured to be in “exclusive mode”, such that only one process at a time can access them.

  • benchmark (bool) – If true enables cudnn.benchmark.

  • callbacks (Union[List[Callback], Callback, None]) – Add a callback or list of callbacks.

  • checkpoint_callback (Optional[bool]) –

    If True, enable checkpointing.

    Deprecated since version v1.5: checkpoint_callback has been deprecated in v1.5 and will be removed in v1.7. Please consider using enable_checkpointing instead.

  • enable_checkpointing (bool) – If True, enable checkpointing. It will configure a default ModelCheckpoint callback if there is no user-defined ModelCheckpoint in callbacks.

  • check_val_every_n_epoch (int) – Check val every n train epochs.

  • default_root_dir (Optional[str]) – Default path for logs and weights when no logger/ckpt_callback passed. Default: os.getcwd(). Can be remote file paths such as s3://mybucket/path or ‘hdfs://path/’

  • detect_anomaly (bool) – Enable anomaly detection for the autograd engine.

  • deterministic (bool) – If True, sets whether PyTorch operations must use deterministic algorithms. Default: False.

  • devices (Union[int, str, List[int], None]) – Will be mapped to either gpus, tpu_cores, num_processes or ipus, based on the accelerator type.

  • fast_dev_run (Union[int, bool]) – Runs n if set to n (int) else 1 if set to True batch(es) of train, val and test to find any bugs (ie: a sort of unit test).

  • flush_logs_every_n_steps (Optional[int]) –

    How often to flush logs to disk (defaults to every 100 steps).

    Deprecated since version v1.5: flush_logs_every_n_steps has been deprecated in v1.5 and will be removed in v1.7. Please configure flushing directly in the logger instead.

  • gpus (Union[int, str, List[int], None]) – Number of GPUs to train on (int) or which GPUs to train on (list or str) applied per node

  • gradient_clip_val (Union[int, float, None]) – The value at which to clip gradients. Passing gradient_clip_val=None disables gradient clipping. If using Automatic Mixed Precision (AMP), the gradients will be unscaled before.

  • gradient_clip_algorithm (Optional[str]) – The gradient clipping algorithm to use. Pass gradient_clip_algorithm="value" to clip by value, and gradient_clip_algorithm="norm" to clip by norm. By default it will be set to "norm".

  • limit_train_batches (Union[int, float]) – How much of training dataset to check (float = fraction, int = num_batches).

  • limit_val_batches (Union[int, float]) – How much of validation dataset to check (float = fraction, int = num_batches).

  • limit_test_batches (Union[int, float]) – How much of test dataset to check (float = fraction, int = num_batches).

  • limit_predict_batches (Union[int, float]) – How much of prediction dataset to check (float = fraction, int = num_batches).

  • logger (Union[LightningLoggerBase, Iterable[LightningLoggerBase], bool]) – Logger (or iterable collection of loggers) for experiment tracking. A True value uses the default TensorBoardLogger. False will disable logging. If multiple loggers are provided and the save_dir property of that logger is not set, local files (checkpoints, profiler traces, etc.) are saved in default_root_dir rather than in the log_dir of any of the individual loggers.

  • log_gpu_memory (Optional[str]) –

    None, ‘min_max’, ‘all’. Might slow performance.

    Deprecated since version v1.5: Deprecated in v1.5.0 and will be removed in v1.7.0 Please use the DeviceStatsMonitor callback directly instead.

  • log_every_n_steps (int) – How often to log within steps (defaults to every 50 steps).

  • prepare_data_per_node (Optional[bool]) –

    If True, each LOCAL_RANK=0 will call prepare data. Otherwise only NODE_RANK=0, LOCAL_RANK=0 will prepare data

    Deprecated since version v1.5: Deprecated in v1.5.0 and will be removed in v1.7.0 Please set prepare_data_per_node in LightningDataModule or LightningModule directly instead.

  • process_position (int) –

    Orders the progress bar when running multiple models on same machine.

    Deprecated since version v1.5: process_position has been deprecated in v1.5 and will be removed in v1.7. Please pass TQDMProgressBar with process_position directly to the Trainer’s callbacks argument instead.

  • progress_bar_refresh_rate (Optional[int]) –

    How often to refresh progress bar (in steps). Value 0 disables progress bar. Ignored when a custom progress bar is passed to callbacks. Default: None, means a suitable value will be chosen based on the environment (terminal, Google COLAB, etc.).

    Deprecated since version v1.5: progress_bar_refresh_rate has been deprecated in v1.5 and will be removed in v1.7. Please pass TQDMProgressBar with refresh_rate directly to the Trainer’s callbacks argument instead. To disable the progress bar, pass enable_progress_bar = False to the Trainer.

  • enable_progress_bar (bool) – Whether to enable to progress bar by default.

  • profiler (Union[BaseProfiler, str, None]) – To profile individual steps during training and assist in identifying bottlenecks.

  • overfit_batches (Union[int, float]) – Overfit a fraction of training data (float) or a set number of batches (int).

  • plugins (Union[TrainingTypePlugin, PrecisionPlugin, ClusterEnvironment, CheckpointIO, str, List[Union[TrainingTypePlugin, PrecisionPlugin, ClusterEnvironment, CheckpointIO, str]], None]) – Plugins allow modification of core behavior like ddp and amp, and enable custom lightning plugins.

  • precision (Union[int, str]) – Double precision (64), full precision (32), half precision (16) or bfloat16 precision (bf16). Can be used on CPU, GPU or TPUs.

  • max_epochs (Optional[int]) – Stop training once this number of epochs is reached. Disabled by default (None). If both max_epochs and max_steps are not specified, defaults to max_epochs = 1000. To enable infinite training, set max_epochs = -1.

  • min_epochs (Optional[int]) – Force training for at least these many epochs. Disabled by default (None). If both min_epochs and min_steps are not specified, defaults to min_epochs = 1.

  • max_steps (int) – Stop training after this number of steps. Disabled by default (-1). If max_steps = -1 and max_epochs = None, will default to max_epochs = 1000. To enable infinite training, set max_epochs to -1.

  • min_steps (Optional[int]) – Force training for at least these number of steps. Disabled by default (None).

  • max_time (Union[str, timedelta, Dict[str, int], None]) – Stop training after this amount of time has passed. Disabled by default (None). The time duration can be specified in the format DD:HH:MM:SS (days, hours, minutes seconds), as a datetime.timedelta, or a dictionary with keys that will be passed to datetime.timedelta.

  • num_nodes (int) – Number of GPU nodes for distributed training.

  • num_processes (int) – Number of processes for distributed training with accelerator="cpu".

  • num_sanity_val_steps (int) – Sanity check runs n validation batches before starting the training routine. Set it to -1 to run all batches in all validation dataloaders.

  • reload_dataloaders_every_n_epochs (int) – Set to a non-negative integer to reload dataloaders every n epochs.

  • reload_dataloaders_every_epoch (bool) –

    Set to True to reload dataloaders every epoch.

    Deprecated since version v1.4: reload_dataloaders_every_epoch has been deprecated in v1.4 and will be removed in v1.6. Please use reload_dataloaders_every_n_epochs.

  • replace_sampler_ddp (bool) – Explicitly enables or disables sampler replacement. If not specified this will toggled automatically when DDP is used. By default it will add shuffle=True for train sampler and shuffle=False for val/test sampler. If you want to customize it, you can set replace_sampler_ddp=False and add your own distributed sampler.

  • resume_from_checkpoint (Union[str, Path, None]) –

    Path/URL of the checkpoint from which training is resumed. If there is no checkpoint file at the path, an exception is raised. If resuming from mid-epoch checkpoint, training will start from the beginning of the next epoch.

    Deprecated since version v1.5: resume_from_checkpoint is deprecated in v1.5 and will be removed in v1.7. Please pass the path to Trainer.fit(..., ckpt_path=...) instead.

  • strategy (Union[str, TrainingTypePlugin, None]) – Supports different training strategies with aliases as well custom training type plugins.

  • sync_batchnorm (bool) – Synchronize batch norm layers between process groups/whole world.

  • terminate_on_nan (Optional[bool]) –

    If set to True, will terminate training (by raising a ValueError) at the end of each training batch, if any of the parameters or the loss are NaN or +/-inf.

    Deprecated since version v1.5: Trainer argument terminate_on_nan was deprecated in v1.5 and will be removed in 1.7. Please use detect_anomaly instead.

  • detect_anomaly – Enable anomaly detection for the autograd engine.

  • tpu_cores (Union[int, str, List[int], None]) – How many TPU cores to train on (1 or 8) / Single TPU to train on [1]

  • ipus (Optional[int]) – How many IPUs to train on.

  • track_grad_norm (Union[int, float, str]) – -1 no tracking. Otherwise tracks that p-norm. May be set to ‘inf’ infinity-norm. If using Automatic Mixed Precision (AMP), the gradients will be unscaled before logging them.

  • val_check_interval (Union[int, float]) – How often to check the validation set. Use float to check within a training epoch, use int to check every n steps (batches).

  • enable_model_summary (bool) – Whether to enable model summarization by default.

  • weights_summary (Optional[str]) –

    Prints a summary of the weights when training begins.

    Deprecated since version v1.5: weights_summary has been deprecated in v1.5 and will be removed in v1.7. To disable the summary, pass enable_model_summary = False to the Trainer. To customize the summary, pass ModelSummary directly to the Trainer’s callbacks argument.

  • weights_save_path (Optional[str]) – Where to save weights if specified. Will override default_root_dir for checkpoints only. Use this if for whatever reason you need the checkpoints stored in a different place than the logs written in default_root_dir. Can be remote file paths such as s3://mybucket/path or ‘hdfs://path/’ Defaults to default_root_dir.

  • move_metrics_to_cpu (bool) – Whether to force internal logged metrics to be moved to cpu. This can save some gpu memory, but can make training slower. Use with attention.

  • multiple_trainloader_mode (str) – How to loop over the datasets when there are multiple train loaders. In ‘max_size_cycle’ mode, the trainer ends one epoch when the largest dataset is traversed, and smaller datasets reload when running out of their data. In ‘min_size’ mode, all the datasets reload when reaching the minimum length of datasets.

  • stochastic_weight_avg (bool) –

    Whether to use Stochastic Weight Averaging (SWA).

    Deprecated since version v1.5: stochastic_weight_avg has been deprecated in v1.5 and will be removed in v1.7. Please pass StochasticWeightAveraging directly to the Trainer’s callbacks argument instead.

fit

Trainer.fit(model, train_dataloaders=None, val_dataloaders=None, datamodule=None, train_dataloader=None, ckpt_path=None)[source]

Runs the full optimization routine.

Parameters
Return type

None

validate

Trainer.validate(model=None, dataloaders=None, ckpt_path=None, verbose=True, datamodule=None, val_dataloaders=None)[source]

Perform one evaluation epoch over the validation set.

Parameters
Return type

List[Dict[str, float]]

Returns

List of dictionaries with metrics logged during the validation phase, e.g., in model- or callback hooks like validation_step(), validation_epoch_end(), etc. The length of the list corresponds to the number of validation dataloaders used.

test

Trainer.test(model=None, dataloaders=None, ckpt_path=None, verbose=True, datamodule=None, test_dataloaders=None)[source]

Perform one evaluation epoch over the test set. It’s separated from fit to make sure you never run on your test set until you want to.

Parameters
Return type

List[Dict[str, float]]

Returns

List of dictionaries with metrics logged during the test phase, e.g., in model- or callback hooks like test_step(), test_epoch_end(), etc. The length of the list corresponds to the number of test dataloaders used.

predict

Trainer.predict(model=None, dataloaders=None, datamodule=None, return_predictions=None, ckpt_path=None)[source]

Run inference on your data. This will call the model forward function to compute predictions. Useful to perform distributed and batched predictions. Logging is disabled in the predict hooks.

Parameters
Return type

Union[List[Any], List[List[Any]], None]

Returns

Returns a list of dictionaries, one for each provided dataloader containing their respective predictions.

tune

Trainer.tune(model, train_dataloaders=None, val_dataloaders=None, datamodule=None, scale_batch_size_kwargs=None, lr_find_kwargs=None, train_dataloader=None)[source]

Runs routines to tune hyperparameters before training.

Parameters
Return type

Dict[str, Union[int, _LRFinder, None]]

Properties

callback_metrics

The metrics available to callbacks. These are automatically set when you log via self.log

def training_step(self, batch, batch_idx):
    self.log("a_val", 2)


callback_metrics = trainer.callback_metrics
assert callback_metrics["a_val"] == 2

current_epoch

The current epoch

def training_step(self, batch, batch_idx):
    current_epoch = self.trainer.current_epoch
    if current_epoch > 100:
        # do something
        pass

logger (p)

The current logger being used. Here’s an example using tensorboard

def training_step(self, batch, batch_idx):
    logger = self.trainer.logger
    tensorboard = logger.experiment

logged_metrics

The metrics sent to the logger (visualizer).

def training_step(self, batch, batch_idx):
    self.log("a_val", 2, logger=True)


logged_metrics = trainer.logged_metrics
assert logged_metrics["a_val"] == 2

log_dir

The directory for the current experiment. Use this to save images to, etc…

def training_step(self, batch, batch_idx):
    img = ...
    save_img(img, self.trainer.log_dir)

is_global_zero

Whether this process is the global zero in multi-node training

def training_step(self, batch, batch_idx):
    if self.trainer.is_global_zero:
        print("in node 0, accelerator 0")

progress_bar_metrics

The metrics sent to the progress bar.

def training_step(self, batch, batch_idx):
    self.log("a_val", 2, prog_bar=True)


progress_bar_metrics = trainer.progress_bar_metrics
assert progress_bar_metrics["a_val"] == 2