Source code for pytorch_lightning.trainer.trainer

# Copyright The PyTorch Lightning team.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#     http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
"""Trainer to automate the training."""
import inspect
import logging
import math
import operator
import os
import traceback
import warnings
from argparse import ArgumentParser, Namespace
from contextlib import contextmanager
from copy import deepcopy
from datetime import timedelta
from functools import partial
from pathlib import Path
from typing import Any, Callable, Dict, Generator, Iterable, List, Optional, Type, Union
from weakref import proxy

import torch
import torch.distributed as dist
from packaging.version import Version
from torch import Tensor
from torch.optim import Optimizer
from torch.utils.data import DataLoader

import pytorch_lightning as pl
from pytorch_lightning.accelerators import (
    Accelerator,
    CUDAAccelerator,
    HPUAccelerator,
    IPUAccelerator,
    MPSAccelerator,
    TPUAccelerator,
)
from pytorch_lightning.callbacks import Callback, Checkpoint, EarlyStopping, ProgressBarBase
from pytorch_lightning.callbacks.prediction_writer import BasePredictionWriter
from pytorch_lightning.core.datamodule import LightningDataModule
from pytorch_lightning.core.optimizer import LightningOptimizer
from pytorch_lightning.loggers import Logger
from pytorch_lightning.loggers.logger import DummyLogger, LoggerCollection
from pytorch_lightning.loggers.tensorboard import TensorBoardLogger
from pytorch_lightning.loops import PredictionLoop, TrainingEpochLoop
from pytorch_lightning.loops.dataloader.evaluation_loop import EvaluationLoop
from pytorch_lightning.loops.fit_loop import FitLoop
from pytorch_lightning.loops.utilities import _parse_loop_limits, _reset_progress
from pytorch_lightning.plugins import (
    ApexMixedPrecisionPlugin,
    NativeMixedPrecisionPlugin,
    PLUGIN_INPUT,
    PrecisionPlugin,
)
from pytorch_lightning.profilers import (
    AdvancedProfiler,
    PassThroughProfiler,
    Profiler,
    PyTorchProfiler,
    SimpleProfiler,
    XLAProfiler,
)
from pytorch_lightning.strategies import ParallelStrategy, Strategy
from pytorch_lightning.strategies.ddp_spawn import DDPSpawnStrategy
from pytorch_lightning.trainer.callback_hook import TrainerCallbackHookMixin
from pytorch_lightning.trainer.configuration_validator import verify_loop_configurations
from pytorch_lightning.trainer.connectors.accelerator_connector import _LITERAL_WARN, AcceleratorConnector
from pytorch_lightning.trainer.connectors.callback_connector import CallbackConnector
from pytorch_lightning.trainer.connectors.checkpoint_connector import CheckpointConnector
from pytorch_lightning.trainer.connectors.data_connector import DataConnector
from pytorch_lightning.trainer.connectors.logger_connector import LoggerConnector
from pytorch_lightning.trainer.connectors.logger_connector.result import _ResultCollection
from pytorch_lightning.trainer.connectors.signal_connector import SignalConnector
from pytorch_lightning.trainer.data_loading import TrainerDataLoadingMixin
from pytorch_lightning.trainer.optimizers import TrainerOptimizersMixin
from pytorch_lightning.trainer.states import RunningStage, TrainerFn, TrainerState, TrainerStatus
from pytorch_lightning.trainer.supporters import CombinedLoader
from pytorch_lightning.tuner.tuning import _TunerResult, Tuner
from pytorch_lightning.utilities import (
    _HPU_AVAILABLE,
    _IPU_AVAILABLE,
    _TPU_AVAILABLE,
    AMPType,
    GradClipAlgorithmType,
    parsing,
)
from pytorch_lightning.utilities.apply_func import apply_to_collection
from pytorch_lightning.utilities.argparse import (
    _defaults_from_env_vars,
    add_argparse_args,
    from_argparse_args,
    parse_argparser,
    parse_env_variables,
)
from pytorch_lightning.utilities.auto_restart import _add_capture_metadata_collate
from pytorch_lightning.utilities.cloud_io import get_filesystem
from pytorch_lightning.utilities.data import _auto_add_worker_init_fn, has_len_all_ranks
from pytorch_lightning.utilities.distributed import distributed_available
from pytorch_lightning.utilities.exceptions import ExitGracefullyException, MisconfigurationException
from pytorch_lightning.utilities.imports import _fault_tolerant_training
from pytorch_lightning.utilities.meta import is_on_meta_device, materialize_module
from pytorch_lightning.utilities.model_helpers import is_overridden
from pytorch_lightning.utilities.rank_zero import rank_zero_deprecation, rank_zero_info, rank_zero_warn
from pytorch_lightning.utilities.seed import isolate_rng
from pytorch_lightning.utilities.types import (
    _EVALUATE_OUTPUT,
    _PATH,
    _PREDICT_OUTPUT,
    EVAL_DATALOADERS,
    LRSchedulerConfig,
    TRAIN_DATALOADERS,
)
from pytorch_lightning.utilities.warnings import PossibleUserWarning

log = logging.getLogger(__name__)
# warnings to ignore in trainer
warnings.filterwarnings(
    "ignore", message="torch.distributed.reduce_op is deprecated, please use torch.distributed.ReduceOp instead"
)


[docs]class Trainer( TrainerCallbackHookMixin, # TODO: Remove in v1.8 TrainerOptimizersMixin, # TODO: Remove in v1.8 TrainerDataLoadingMixin, # TODO: Remove in v1.8 ): @_defaults_from_env_vars def __init__( self, logger: Union[Logger, Iterable[Logger], bool] = True, enable_checkpointing: bool = True, callbacks: Optional[Union[List[Callback], Callback]] = None, default_root_dir: Optional[str] = None, gradient_clip_val: Optional[Union[int, float]] = None, gradient_clip_algorithm: Optional[str] = None, num_nodes: int = 1, num_processes: Optional[int] = None, # TODO: Remove in 2.0 devices: Optional[Union[List[int], str, int]] = None, gpus: Optional[Union[List[int], str, int]] = None, # TODO: Remove in 2.0 auto_select_gpus: bool = False, tpu_cores: Optional[Union[List[int], str, int]] = None, # TODO: Remove in 2.0 ipus: Optional[int] = None, # TODO: Remove in 2.0 enable_progress_bar: bool = True, overfit_batches: Union[int, float] = 0.0, track_grad_norm: Union[int, float, str] = -1, check_val_every_n_epoch: Optional[int] = 1, fast_dev_run: Union[int, bool] = False, accumulate_grad_batches: Optional[Union[int, Dict[int, int]]] = None, max_epochs: Optional[int] = None, min_epochs: Optional[int] = None, max_steps: int = -1, min_steps: Optional[int] = None, max_time: Optional[Union[str, timedelta, Dict[str, int]]] = None, limit_train_batches: Optional[Union[int, float]] = None, limit_val_batches: Optional[Union[int, float]] = None, limit_test_batches: Optional[Union[int, float]] = None, limit_predict_batches: Optional[Union[int, float]] = None, val_check_interval: Optional[Union[int, float]] = None, log_every_n_steps: int = 50, accelerator: Optional[Union[str, Accelerator]] = None, strategy: Optional[Union[str, Strategy]] = None, sync_batchnorm: bool = False, precision: Union[int, str] = 32, enable_model_summary: bool = True, weights_save_path: Optional[str] = None, # TODO: Remove in 1.8 num_sanity_val_steps: int = 2, resume_from_checkpoint: Optional[Union[Path, str]] = None, profiler: Optional[Union[Profiler, str]] = None, benchmark: Optional[bool] = None, deterministic: Optional[Union[bool, _LITERAL_WARN]] = None, reload_dataloaders_every_n_epochs: int = 0, auto_lr_find: Union[bool, str] = False, replace_sampler_ddp: bool = True, detect_anomaly: bool = False, auto_scale_batch_size: Union[str, bool] = False, plugins: Optional[Union[PLUGIN_INPUT, List[PLUGIN_INPUT]]] = None, amp_backend: str = "native", amp_level: Optional[str] = None, move_metrics_to_cpu: bool = False, multiple_trainloader_mode: str = "max_size_cycle", ) -> None: r""" Customize every aspect of training via flags. Args: accelerator: Supports passing different accelerator types ("cpu", "gpu", "tpu", "ipu", "hpu", "mps, "auto") as well as custom accelerator instances. .. deprecated:: 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: Accumulates grads every k batches or as set up in the dict. Default: ``None``. amp_backend: The mixed precision backend to use ("native" or "apex"). Default: ``'native''``. amp_level: 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: 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. Default: ``False``. auto_scale_batch_size: 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. Default: ``False``. auto_select_gpus: If enabled and ``gpus`` or ``devices`` 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. Default: ``False``. benchmark: The value (``True`` or ``False``) to set ``torch.backends.cudnn.benchmark`` to. The value for ``torch.backends.cudnn.benchmark`` set in the current session will be used (``False`` if not manually set). If :paramref:`~pytorch_lightning.trainer.Trainer.deterministic` is set to ``True``, this will default to ``False``. Override to manually set a different value. Default: ``None``. callbacks: Add a callback or list of callbacks. Default: ``None``. enable_checkpointing: If ``True``, enable checkpointing. It will configure a default ModelCheckpoint callback if there is no user-defined ModelCheckpoint in :paramref:`~pytorch_lightning.trainer.trainer.Trainer.callbacks`. Default: ``True``. check_val_every_n_epoch: Perform a validation loop every after every `N` training epochs. If ``None``, validation will be done solely based on the number of training batches, requiring ``val_check_interval`` to be an integer value. Default: ``1``. default_root_dir: 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: Enable anomaly detection for the autograd engine. Default: ``False``. deterministic: If ``True``, sets whether PyTorch operations must use deterministic algorithms. Set to ``"warn"`` to use deterministic algorithms whenever possible, throwing warnings on operations that don't support deterministic mode (requires PyTorch 1.11+). If not set, defaults to ``False``. Default: ``None``. devices: Will be mapped to either `gpus`, `tpu_cores`, `num_processes` or `ipus`, based on the accelerator type. 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). Default: ``False``. gpus: Number of GPUs to train on (int) or which GPUs to train on (list or str) applied per node Default: ``None``. .. deprecated:: v1.7 ``gpus`` has been deprecated in v1.7 and will be removed in v2.0. Please use ``accelerator='gpu'`` and ``devices=x`` instead. gradient_clip_val: 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. Default: ``None``. gradient_clip_algorithm: 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: How much of training dataset to check (float = fraction, int = num_batches). Default: ``1.0``. limit_val_batches: How much of validation dataset to check (float = fraction, int = num_batches). Default: ``1.0``. limit_test_batches: How much of test dataset to check (float = fraction, int = num_batches). Default: ``1.0``. limit_predict_batches: How much of prediction dataset to check (float = fraction, int = num_batches). Default: ``1.0``. logger: 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. Default: ``True``. log_every_n_steps: How often to log within steps. Default: ``50``. enable_progress_bar: Whether to enable to progress bar by default. Default: ``True``. profiler: To profile individual steps during training and assist in identifying bottlenecks. Default: ``None``. overfit_batches: Overfit a fraction of training/validation data (float) or a set number of batches (int). Default: ``0.0``. plugins: Plugins allow modification of core behavior like ddp and amp, and enable custom lightning plugins. Default: ``None``. precision: Double precision (64), full precision (32), half precision (16) or bfloat16 precision (bf16). Can be used on CPU, GPU, TPUs, HPUs or IPUs. Default: ``32``. max_epochs: 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: Force training for at least these many epochs. Disabled by default (None). max_steps: 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: Force training for at least these number of steps. Disabled by default (``None``). max_time: 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 :class:`datetime.timedelta`, or a dictionary with keys that will be passed to :class:`datetime.timedelta`. num_nodes: Number of GPU nodes for distributed training. Default: ``1``. num_processes: Number of processes for distributed training with ``accelerator="cpu"``. Default: ``1``. .. deprecated:: v1.7 ``num_processes`` has been deprecated in v1.7 and will be removed in v2.0. Please use ``accelerator='cpu'`` and ``devices=x`` instead. num_sanity_val_steps: Sanity check runs n validation batches before starting the training routine. Set it to `-1` to run all batches in all validation dataloaders. Default: ``2``. reload_dataloaders_every_n_epochs: Set to a non-negative integer to reload dataloaders every n epochs. Default: ``0``. replace_sampler_ddp: 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: 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:: v1.5 ``resume_from_checkpoint`` is deprecated in v1.5 and will be removed in v2.0. Please pass the path to ``Trainer.fit(..., ckpt_path=...)`` instead. strategy: Supports different training strategies with aliases as well custom strategies. Default: ``None``. sync_batchnorm: Synchronize batch norm layers between process groups/whole world. Default: ``False``. tpu_cores: How many TPU cores to train on (1 or 8) / Single TPU to train on (1) Default: ``None``. .. deprecated:: v1.7 ``tpu_cores`` has been deprecated in v1.7 and will be removed in v2.0. Please use ``accelerator='tpu'`` and ``devices=x`` instead. ipus: How many IPUs to train on. Default: ``None``. .. deprecated:: v1.7 ``ipus`` has been deprecated in v1.7 and will be removed in v2.0. Please use ``accelerator='ipu'`` and ``devices=x`` instead. track_grad_norm: -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. Default: ``-1``. val_check_interval: How often to check the validation set. Pass a ``float`` in the range [0.0, 1.0] to check after a fraction of the training epoch. Pass an ``int`` to check after a fixed number of training batches. An ``int`` value can only be higher than the number of training batches when ``check_val_every_n_epoch=None``, which validates after every ``N`` training batches across epochs or during iteration-based training. Default: ``1.0``. enable_model_summary: Whether to enable model summarization by default. Default: ``True``. weights_save_path: 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`. .. deprecated:: v1.6 ``weights_save_path`` has been deprecated in v1.6 and will be removed in v1.8. Please pass ``dirpath`` directly to the :class:`~pytorch_lightning.callbacks.model_checkpoint.ModelCheckpoint` callback. move_metrics_to_cpu: 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. Default: ``False``. multiple_trainloader_mode: 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. Default: ``"max_size_cycle"``. """ super().__init__() Trainer._log_api_event("init") log.detail(f"{self.__class__.__name__}: Initializing trainer with parameters: {locals()}") self.state = TrainerState() # init connectors self._data_connector = DataConnector(self, multiple_trainloader_mode) self._accelerator_connector = AcceleratorConnector( num_processes=num_processes, devices=devices, tpu_cores=tpu_cores, ipus=ipus, accelerator=accelerator, strategy=strategy, gpus=gpus, num_nodes=num_nodes, sync_batchnorm=sync_batchnorm, benchmark=benchmark, replace_sampler_ddp=replace_sampler_ddp, deterministic=deterministic, auto_select_gpus=auto_select_gpus, precision=precision, amp_type=amp_backend, amp_level=amp_level, plugins=plugins, ) self._logger_connector = LoggerConnector(self) self._callback_connector = CallbackConnector(self) self._checkpoint_connector = CheckpointConnector(self, resume_from_checkpoint) self._signal_connector = SignalConnector(self) self.tuner = Tuner(self) fit_loop = FitLoop(min_epochs=min_epochs, max_epochs=max_epochs) training_epoch_loop = TrainingEpochLoop(min_steps=min_steps, max_steps=max_steps) fit_loop.connect(epoch_loop=training_epoch_loop) # default .fit() loop self.fit_loop = fit_loop # default .validate() loop self.validate_loop = EvaluationLoop() # default .test() loop self.test_loop = EvaluationLoop() # default .predict() loop self.predict_loop = PredictionLoop() # set when a checkpoint is loaded via `Trainer.{fit,validate,test,predict}`. self._ckpt_path: Optional[str] = None # .validate(), predict() and .test() set these when they load a checkpoint. They will be removed in favor of # the unified read-only `Trainer.ckpt_path` attribute in v1.8 self._validated_ckpt_path: Optional[str] = None # TODO: remove in v1.8 self._tested_ckpt_path: Optional[str] = None # TODO: remove in v1.8 self._predicted_ckpt_path: Optional[str] = None # TODO: remove in v1.8 # init callbacks # Declare attributes to be set in _callback_connector on_trainer_init self._callback_connector.on_trainer_init( callbacks, enable_checkpointing, enable_progress_bar, default_root_dir, weights_save_path, enable_model_summary, max_time, accumulate_grad_batches, ) # hook self._call_callback_hooks("on_init_start") # init data flags self.check_val_every_n_epoch: int self._data_connector.on_trainer_init( val_check_interval, reload_dataloaders_every_n_epochs, check_val_every_n_epoch, ) # gradient clipping if gradient_clip_val is not None and not isinstance(gradient_clip_val, (int, float)): raise TypeError(f"`gradient_clip_val` should be an int or a float. Got {gradient_clip_val}.") if gradient_clip_algorithm is not None and not GradClipAlgorithmType.supported_type( gradient_clip_algorithm.lower() ): raise MisconfigurationException( f"`gradient_clip_algorithm` {gradient_clip_algorithm} is invalid. " f"Allowed algorithms: {GradClipAlgorithmType.supported_types()}." ) # gradient norm tracking if track_grad_norm != -1 and not ( (isinstance(track_grad_norm, (int, float)) or track_grad_norm == "inf") and float(track_grad_norm) > 0 ): raise MisconfigurationException( f"`track_grad_norm` must be a positive number or 'inf' (infinity norm). Got {track_grad_norm}." ) self.gradient_clip_val: Union[int, float] = gradient_clip_val self.gradient_clip_algorithm: Optional[GradClipAlgorithmType] = ( GradClipAlgorithmType(gradient_clip_algorithm.lower()) if gradient_clip_algorithm is not None else None ) self.track_grad_norm: float = float(track_grad_norm) self._detect_anomaly: bool = detect_anomaly self._setup_on_init() # configure tuner self.tuner.on_trainer_init(auto_lr_find, auto_scale_batch_size) # configure profiler self.__init_profiler(profiler) # init logger flags self._loggers: List[Logger] self._logger_connector.on_trainer_init(logger, log_every_n_steps, move_metrics_to_cpu) # init debugging flags self.val_check_interval: Union[int, float] self._init_debugging_flags( limit_train_batches, limit_val_batches, limit_test_batches, limit_predict_batches, fast_dev_run, overfit_batches, val_check_interval, num_sanity_val_steps, ) # Callback system self._call_callback_hooks("on_init_end") def _init_debugging_flags( self, limit_train_batches: Optional[Union[int, float]], limit_val_batches: Optional[Union[int, float]], limit_test_batches: Optional[Union[int, float]], limit_predict_batches: Optional[Union[int, float]], fast_dev_run: Union[int, bool], overfit_batches: Union[int, float], val_check_interval: Optional[Union[int, float]], num_sanity_val_steps: int, ): # init debugging flags if isinstance(fast_dev_run, int) and (fast_dev_run < 0): raise MisconfigurationException( f"fast_dev_run={fast_dev_run!r} is not a valid configuration. It should be >= 0." ) self.fast_dev_run = fast_dev_run # set fast_dev_run=True when it is 1, used while logging if fast_dev_run == 1: self.fast_dev_run = True self.overfit_batches = _determine_batch_limits(overfit_batches, "overfit_batches") overfit_batches_enabled = overfit_batches > 0 if fast_dev_run: num_batches = int(fast_dev_run) if not overfit_batches_enabled: self.limit_train_batches = num_batches self.limit_val_batches = num_batches self.limit_test_batches = num_batches self.limit_predict_batches = num_batches self.fit_loop.max_steps = num_batches self.num_sanity_val_steps = 0 self.fit_loop.max_epochs = 1 self.val_check_interval = 1.0 self.check_val_every_n_epoch = 1 self.loggers = [DummyLogger()] if self.loggers else [] rank_zero_info( f"Running in `fast_dev_run` mode: will run the requested loop using {num_batches} batch(es). " "Logging and checkpointing is suppressed." ) else: if not overfit_batches_enabled: self.limit_train_batches = _determine_batch_limits(limit_train_batches, "limit_train_batches") self.limit_val_batches = _determine_batch_limits(limit_val_batches, "limit_val_batches") self.limit_test_batches = _determine_batch_limits(limit_test_batches, "limit_test_batches") self.limit_predict_batches = _determine_batch_limits(limit_predict_batches, "limit_predict_batches") self.num_sanity_val_steps = float("inf") if num_sanity_val_steps == -1 else num_sanity_val_steps self.val_check_interval = _determine_batch_limits(val_check_interval, "val_check_interval") if overfit_batches_enabled: self.limit_train_batches = overfit_batches self.limit_val_batches = overfit_batches def _setup_on_init(self) -> None: self._log_device_info() self.should_stop = False self.state = TrainerState() self.num_training_batches = float("inf") self.train_dataloader = None self.num_sanity_val_batches = [] self.num_test_batches = [] self.num_val_batches = [] self.test_dataloaders = None self.val_dataloaders = None self._last_train_dl_reload_epoch = float("-inf") self._last_val_dl_reload_epoch = float("-inf") self.num_predict_batches = [] def _call_and_handle_interrupt(self, trainer_fn: Callable, *args: Any, **kwargs: Any) -> Any: r""" Error handling, intended to be used only for main trainer function entry points (fit, validate, test, predict) as all errors should funnel through them Args: trainer_fn: one of (fit, validate, test, predict) *args: positional arguments to be passed to the `trainer_fn` **kwargs: keyword arguments to be passed to `trainer_fn` """ try: if self.strategy.launcher is not None: return self.strategy.launcher.launch(trainer_fn, *args, trainer=self, **kwargs) else: return trainer_fn(*args, **kwargs) # TODO(awaelchli): Unify both exceptions below, where `KeyboardError` doesn't re-raise except KeyboardInterrupt as exception: rank_zero_warn("Detected KeyboardInterrupt, attempting graceful shutdown...") # user could press Ctrl+c many times... only shutdown once if not self.interrupted: self.state.status = TrainerStatus.INTERRUPTED self._call_callback_hooks("on_exception", exception) except BaseException as exception: self.state.status = TrainerStatus.INTERRUPTED if distributed_available() and self.world_size > 1: # try syncing remaining processes, kill otherwise self.strategy.reconciliate_processes(traceback.format_exc()) self._call_callback_hooks("on_exception", exception) self._teardown() # teardown might access the stage so we reset it after self.state.stage = None raise def fit( self, model: "pl.LightningModule", train_dataloaders: Optional[Union[TRAIN_DATALOADERS, LightningDataModule]] = None, val_dataloaders: Optional[EVAL_DATALOADERS] = None, datamodule: Optional[LightningDataModule] = None, ckpt_path: Optional[str] = None, ) -> None: r""" Runs the full optimization routine. Args: model: Model to fit. train_dataloaders: A collection of :class:`torch.utils.data.DataLoader` or a :class:`~pytorch_lightning.core.datamodule.LightningDataModule` specifying training samples. In the case of multiple dataloaders, please see this :ref:`section <multiple-dataloaders>`. val_dataloaders: A :class:`torch.utils.data.DataLoader` or a sequence of them specifying validation samples. ckpt_path: 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. datamodule: An instance of :class:`~pytorch_lightning.core.datamodule.LightningDataModule`. """ self.strategy.model = model self._call_and_handle_interrupt( self._fit_impl, model, train_dataloaders, val_dataloaders, datamodule, ckpt_path ) def _fit_impl( self, model: "pl.LightningModule", train_dataloaders: Optional[Union[TRAIN_DATALOADERS, LightningDataModule]] = None, val_dataloaders: Optional[EVAL_DATALOADERS] = None, datamodule: Optional[LightningDataModule] = None, ckpt_path: Optional[str] = None, ) -> None: Trainer._log_api_event("fit") log.detail(f"{self.__class__.__name__}: trainer fit stage") self.state.fn = TrainerFn.FITTING self.state.status = TrainerStatus.RUNNING self.training = True self._last_train_dl_reload_epoch = float("-inf") self._last_val_dl_reload_epoch = float("-inf") # if a datamodule comes in as the second arg, then fix it for the user if isinstance(train_dataloaders, LightningDataModule): datamodule = train_dataloaders train_dataloaders = None # If you supply a datamodule you can't supply train_dataloader or val_dataloaders if (train_dataloaders is not None or val_dataloaders is not None) and datamodule is not None: raise MisconfigurationException( "You cannot pass `train_dataloader` or `val_dataloaders` to `trainer.fit(datamodule=...)`" ) # links data to the trainer self._data_connector.attach_data( model, train_dataloaders=train_dataloaders, val_dataloaders=val_dataloaders, datamodule=datamodule ) # TODO: ckpt_path only in v2.0 ckpt_path = ckpt_path or self.resume_from_checkpoint self._ckpt_path = self.__set_ckpt_path( ckpt_path, model_provided=True, model_connected=self.lightning_module is not None ) results = self._run(model, ckpt_path=self.ckpt_path) assert self.state.stopped self.training = False return results def validate( self, model: Optional["pl.LightningModule"] = None, dataloaders: Optional[Union[EVAL_DATALOADERS, LightningDataModule]] = None, ckpt_path: Optional[str] = None, verbose: bool = True, datamodule: Optional[LightningDataModule] = None, ) -> _EVALUATE_OUTPUT: r""" Perform one evaluation epoch over the validation set. Args: model: The model to validate. dataloaders: A :class:`torch.utils.data.DataLoader` or a sequence of them, or a :class:`~pytorch_lightning.core.datamodule.LightningDataModule` specifying validation samples. ckpt_path: Either ``best`` or path to the checkpoint you wish to validate. If ``None`` and the model instance was passed, use the current weights. Otherwise, the best model checkpoint from the previous ``trainer.fit`` call will be loaded if a checkpoint callback is configured. verbose: If True, prints the validation results. datamodule: An instance of :class:`~pytorch_lightning.core.datamodule.LightningDataModule`. Returns: List of dictionaries with metrics logged during the validation phase, e.g., in model- or callback hooks like :meth:`~pytorch_lightning.core.module.LightningModule.validation_step`, :meth:`~pytorch_lightning.core.module.LightningModule.validation_epoch_end`, etc. The length of the list corresponds to the number of validation dataloaders used. """ self.strategy.model = model or self.lightning_module return self._call_and_handle_interrupt(self._validate_impl, model, dataloaders, ckpt_path, verbose, datamodule) def _validate_impl( self, model: Optional["pl.LightningModule"] = None, dataloaders: Optional[Union[EVAL_DATALOADERS, LightningDataModule]] = None, ckpt_path: Optional[str] = None, verbose: bool = True, datamodule: Optional[LightningDataModule] = None, ) -> _EVALUATE_OUTPUT: # -------------------- # SETUP HOOK # -------------------- Trainer._log_api_event("validate") log.detail(f"{self.__class__.__name__}: trainer validate stage") self.state.fn = TrainerFn.VALIDATING self.state.status = TrainerStatus.RUNNING self.validating = True # if a datamodule comes in as the second arg, then fix it for the user if isinstance(dataloaders, LightningDataModule): datamodule = dataloaders dataloaders = None # If you supply a datamodule you can't supply val_dataloaders if dataloaders is not None and datamodule: raise MisconfigurationException("You cannot pass both `trainer.validate(dataloaders=..., datamodule=...)`") model_provided = model is not None model = model or self.lightning_module if model is None: raise MisconfigurationException( "`model` must be provided to `trainer.validate()` when it hasn't been passed in a previous run" ) self.validate_loop.verbose = verbose # links data to the trainer self._data_connector.attach_data(model, val_dataloaders=dataloaders, datamodule=datamodule) self._ckpt_path = self.__set_ckpt_path( ckpt_path, model_provided=model_provided, model_connected=self.lightning_module is not None ) self._validated_ckpt_path = self.ckpt_path # TODO: remove in v1.8 # run validate results = self._run(model, ckpt_path=self.ckpt_path) assert self.state.stopped self.validating = False return results def test( self, model: Optional["pl.LightningModule"] = None, dataloaders: Optional[Union[EVAL_DATALOADERS, LightningDataModule]] = None, ckpt_path: Optional[str] = None, verbose: bool = True, datamodule: Optional[LightningDataModule] = None, ) -> _EVALUATE_OUTPUT: r""" 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. Args: model: The model to test. dataloaders: A :class:`torch.utils.data.DataLoader` or a sequence of them, or a :class:`~pytorch_lightning.core.datamodule.LightningDataModule` specifying test samples. ckpt_path: Either ``best`` or path to the checkpoint you wish to test. If ``None`` and the model instance was passed, use the current weights. Otherwise, the best model checkpoint from the previous ``trainer.fit`` call will be loaded if a checkpoint callback is configured. verbose: If True, prints the test results. datamodule: An instance of :class:`~pytorch_lightning.core.datamodule.LightningDataModule`. Returns: List of dictionaries with metrics logged during the test phase, e.g., in model- or callback hooks like :meth:`~pytorch_lightning.core.module.LightningModule.test_step`, :meth:`~pytorch_lightning.core.module.LightningModule.test_epoch_end`, etc. The length of the list corresponds to the number of test dataloaders used. """ self.strategy.model = model or self.lightning_module return self._call_and_handle_interrupt(self._test_impl, model, dataloaders, ckpt_path, verbose, datamodule) def _test_impl( self, model: Optional["pl.LightningModule"] = None, dataloaders: Optional[Union[EVAL_DATALOADERS, LightningDataModule]] = None, ckpt_path: Optional[str] = None, verbose: bool = True, datamodule: Optional[LightningDataModule] = None, ) -> _EVALUATE_OUTPUT: # -------------------- # SETUP HOOK # -------------------- Trainer._log_api_event("test") log.detail(f"{self.__class__.__name__}: trainer test stage") self.state.fn = TrainerFn.TESTING self.state.status = TrainerStatus.RUNNING self.testing = True # if a datamodule comes in as the second arg, then fix it for the user if isinstance(dataloaders, LightningDataModule): datamodule = dataloaders dataloaders = None # If you supply a datamodule you can't supply test_dataloaders if dataloaders is not None and datamodule: raise MisconfigurationException("You cannot pass both `trainer.test(dataloaders=..., datamodule=...)`") model_provided = model is not None model = model or self.lightning_module if model is None: raise MisconfigurationException( "`model` must be provided to `trainer.test()` when it hasn't been passed in a previous run" ) self.test_loop.verbose = verbose # links data to the trainer self._data_connector.attach_data(model, test_dataloaders=dataloaders, datamodule=datamodule) self._ckpt_path = self.__set_ckpt_path( ckpt_path, model_provided=model_provided, model_connected=self.lightning_module is not None ) self._tested_ckpt_path = self.ckpt_path # TODO: remove in v1.8 # run test results = self._run(model, ckpt_path=self.ckpt_path) assert self.state.stopped self.testing = False return results def predict( self, model: Optional["pl.LightningModule"] = None, dataloaders: Optional[Union[EVAL_DATALOADERS, LightningDataModule]] = None, datamodule: Optional[LightningDataModule] = None, return_predictions: Optional[bool] = None, ckpt_path: Optional[str] = None, ) -> Optional[_PREDICT_OUTPUT]: r""" 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. Args: model: The model to predict with. dataloaders: A :class:`torch.utils.data.DataLoader` or a sequence of them, or a :class:`~pytorch_lightning.core.datamodule.LightningDataModule` specifying prediction samples. datamodule: The datamodule with a predict_dataloader method that returns one or more dataloaders. return_predictions: Whether to return predictions. ``True`` by default except when an accelerator that spawns processes is used (not supported). ckpt_path: Either ``best`` or path to the checkpoint you wish to predict. If ``None`` and the model instance was passed, use the current weights. Otherwise, the best model checkpoint from the previous ``trainer.fit`` call will be loaded if a checkpoint callback is configured. Returns: Returns a list of dictionaries, one for each provided dataloader containing their respective predictions. """ self.strategy.model = model or self.lightning_module return self._call_and_handle_interrupt( self._predict_impl, model, dataloaders, datamodule, return_predictions, ckpt_path ) def _predict_impl( self, model: Optional["pl.LightningModule"] = None, dataloaders: Optional[Union[EVAL_DATALOADERS, LightningDataModule]] = None, datamodule: Optional[LightningDataModule] = None, return_predictions: Optional[bool] = None, ckpt_path: Optional[str] = None, ) -> Optional[_PREDICT_OUTPUT]: # -------------------- # SETUP HOOK # -------------------- Trainer._log_api_event("predict") log.detail(f"{self.__class__.__name__}: trainer predict stage") self.state.fn = TrainerFn.PREDICTING self.state.status = TrainerStatus.RUNNING self.predicting = True self.predict_loop.return_predictions = return_predictions # if a datamodule comes in as the second arg, then fix it for the user if isinstance(dataloaders, LightningDataModule): datamodule = dataloaders dataloaders = None if dataloaders is not None and datamodule: raise MisconfigurationException("You cannot pass both `trainer.predict(dataloaders=..., datamodule=...)`") model_provided = model is not None model = model or self.lightning_module if model is None: raise MisconfigurationException( "`model` must be provided to `trainer.predict()` when it hasn't been passed in a previous run" ) # links data to the trainer self._data_connector.attach_data(model, predict_dataloaders=dataloaders, datamodule=datamodule) self._ckpt_path = self.__set_ckpt_path( ckpt_path, model_provided=model_provided, model_connected=self.lightning_module is not None ) self._predicted_ckpt_path = self.ckpt_path # TODO: remove in v1.8 results = self._run(model, ckpt_path=self.ckpt_path) assert self.state.stopped self.predicting = False return results def tune( self, model: "pl.LightningModule", train_dataloaders: Optional[Union[TRAIN_DATALOADERS, LightningDataModule]] = None, val_dataloaders: Optional[EVAL_DATALOADERS] = None, datamodule: Optional[LightningDataModule] = None, scale_batch_size_kwargs: Optional[Dict[str, Any]] = None, lr_find_kwargs: Optional[Dict[str, Any]] = None, ) -> _TunerResult: r""" Runs routines to tune hyperparameters before training. Args: model: Model to tune. train_dataloaders: A collection of :class:`torch.utils.data.DataLoader` or a :class:`~pytorch_lightning.core.datamodule.LightningDataModule` specifying training samples. In the case of multiple dataloaders, please see this :ref:`section <multiple-dataloaders>`. val_dataloaders: A :class:`torch.utils.data.DataLoader` or a sequence of them specifying validation samples. datamodule: An instance of :class:`~pytorch_lightning.core.datamodule.LightningDataModule`. scale_batch_size_kwargs: Arguments for :func:`~pytorch_lightning.tuner.batch_size_scaling.scale_batch_size` lr_find_kwargs: Arguments for :func:`~pytorch_lightning.tuner.lr_finder.lr_find` """ Trainer._log_api_event("tune") self.state.fn = TrainerFn.TUNING self.state.status = TrainerStatus.RUNNING self.tuning = True # if a datamodule comes in as the second arg, then fix it for the user if isinstance(train_dataloaders, LightningDataModule): datamodule = train_dataloaders train_dataloaders = None # If you supply a datamodule you can't supply train_dataloader or val_dataloaders if (train_dataloaders is not None or val_dataloaders is not None) and datamodule is not None: raise MisconfigurationException( "You cannot pass `train_dataloader` or `val_dataloaders` to `trainer.tune(datamodule=...)`" ) # links data to the trainer self._data_connector.attach_data( model, train_dataloaders=train_dataloaders, val_dataloaders=val_dataloaders, datamodule=datamodule ) with isolate_rng(): result = self.tuner._tune( model, scale_batch_size_kwargs=scale_batch_size_kwargs, lr_find_kwargs=lr_find_kwargs ) assert self.state.stopped self.tuning = False return result def _restore_modules_and_callbacks(self, checkpoint_path: Optional[_PATH] = None) -> None: # restore modules after setup self._checkpoint_connector.resume_start(checkpoint_path) self._checkpoint_connector._restore_quantization_callbacks() self._checkpoint_connector.restore_model() self._checkpoint_connector.restore_datamodule() if self.state.fn == TrainerFn.FITTING: # restore callback states self._checkpoint_connector.restore_callbacks() def _run( self, model: "pl.LightningModule", ckpt_path: Optional[str] = None ) -> Optional[Union[_EVALUATE_OUTPUT, _PREDICT_OUTPUT]]: if self.state.fn in (TrainerFn.FITTING, TrainerFn.TUNING): min_epochs, max_epochs = _parse_loop_limits( self.min_steps, self.max_steps, self.min_epochs, self.max_epochs, self ) self.fit_loop.min_epochs = min_epochs self.fit_loop.max_epochs = max_epochs # clean hparams if hasattr(model, "hparams"): parsing.clean_namespace(model.hparams) # attach model to the strategy self.strategy.connect(model) self._callback_connector._attach_model_callbacks() self._callback_connector._attach_model_logging_functions() verify_loop_configurations(self) # hook log.detail(f"{self.__class__.__name__}: preparing data") self._data_connector.prepare_data() # ---------------------------- # SET UP TRAINING # ---------------------------- self._call_callback_hooks("on_before_accelerator_backend_setup") log.detail(f"{self.__class__.__name__}: setting up strategy environment") self.strategy.setup_environment() self.__setup_profiler() self._call_setup_hook() # allow user to setup lightning_module in accelerator environment # check if we should delay restoring checkpoint till later if not self.strategy.restore_checkpoint_after_setup: log.detail(f"{self.__class__.__name__}: restoring module and callbacks from checkpoint path: {ckpt_path}") self._restore_modules_and_callbacks(ckpt_path) log.detail(f"{self.__class__.__name__}: configuring sharded model") self._call_configure_sharded_model() # allow user to setup in model sharded environment # ---------------------------- # INSPECT THE CORE LOOPS # ---------------------------- rf""" Lightning internal flow looks like this: {Trainer.fit} or {Trainer.test} or {Trainer.predict} || | || spawn processes || {self.strategy.setup_environment} || | || setup accelerator || and strategy || LIGHTNING | || {self._run_stage} || FLOW | || {self._run_train} || DIRECTION or {self._run_evaluate} || or {self._run_predict} || | || results \/ This is used to guide readers to the core loops: train, test, predict. {self._run_predict} is the simplest to understand, use `Go to Definition` to read it :) """ # ---------------------------- # TRAIN # ---------------------------- # reset logger connector self._logger_connector.reset_results() self._logger_connector.reset_metrics() # strategy will configure model and move it to the device self.strategy.setup(self) # hook if self.state.fn == TrainerFn.FITTING: self._call_callback_hooks("on_fit_start") self._call_lightning_module_hook("on_fit_start") self._log_hyperparams() if self.strategy.restore_checkpoint_after_setup: log.detail(f"{self.__class__.__name__}: restoring module and callbacks from checkpoint path: {ckpt_path}") self._restore_modules_and_callbacks(ckpt_path) # restore optimizers, etc. log.detail(f"{self.__class__.__name__}: restoring training state") self._checkpoint_connector.restore_training_state() self._checkpoint_connector.resume_end() results = self._run_stage() log.detail(f"{self.__class__.__name__}: trainer tearing down") self._teardown() # ---------------------------- # POST-Training CLEAN UP # ---------------------------- # hook if self.state.fn == TrainerFn.FITTING: self._call_callback_hooks("on_fit_end") self._call_lightning_module_hook("on_fit_end") log.detail(f"{self.__class__.__name__}: calling teardown hooks") self._call_teardown_hook() self.state.status = TrainerStatus.FINISHED self.state.stage = None return results def _log_hyperparams(self) -> None: if not self.loggers: return # log hyper-parameters hparams_initial = None # save exp to get started (this is where the first experiment logs are written) datamodule_log_hyperparams = self.datamodule._log_hyperparams if self.datamodule is not None else False if self.lightning_module._log_hyperparams and datamodule_log_hyperparams: datamodule_hparams = self.datamodule.hparams_initial lightning_hparams = self.lightning_module.hparams_initial inconsistent_keys = [] for key in lightning_hparams.keys() & datamodule_hparams.keys(): lm_val, dm_val = lightning_hparams[key], datamodule_hparams[key] if type(lm_val) != type(dm_val): inconsistent_keys.append(key) elif isinstance(lm_val, Tensor) and id(lm_val) != id(dm_val): inconsistent_keys.append(key) elif lm_val != dm_val: inconsistent_keys.append(key) if inconsistent_keys: raise MisconfigurationException( f"Error while merging hparams: the keys {inconsistent_keys} are present " "in both the LightningModule's and LightningDataModule's hparams " "but have different values." ) hparams_initial = {**lightning_hparams, **datamodule_hparams} elif self.lightning_module._log_hyperparams: hparams_initial = self.lightning_module.hparams_initial elif datamodule_log_hyperparams: hparams_initial = self.datamodule.hparams_initial for logger in self.loggers: if hparams_initial is not None: logger.log_hyperparams(hparams_initial) logger.log_graph(self.lightning_module) logger.save() def _teardown(self): """This is the Trainer's internal teardown, unrelated to the `teardown` hooks in LightningModule and Callback; those are handled by :meth:`_call_teardown_hook`.""" self.strategy.teardown() loop = self._active_loop # loop should never be `None` here but it can because we don't know the trainer stage with `ddp_spawn` if loop is not None: loop.teardown() self._logger_connector.teardown() self._signal_connector.teardown() def run_stage(self) -> None: rank_zero_deprecation( "`Trainer.run_stage` is deprecated in v1.6 and will be removed in v1.8. Use" " `Trainer.{fit,validate,test,predict}` instead." ) return self._run_stage() def _run_stage(self): self.strategy.barrier("run-stage") self.strategy.dispatch(self) if self.evaluating: return self._run_evaluate() if self.predicting: return self._run_predict() return self._run_train() def _pre_training_routine(self): # wait for all to join if on distributed self.strategy.barrier("setup_training") # register signals self._signal_connector.register_signal_handlers() # -------------------------- # Pre-train # -------------------------- self._call_callback_hooks("on_pretrain_routine_start") self._call_lightning_module_hook("on_pretrain_routine_start") self._call_callback_hooks("on_pretrain_routine_end") self._call_lightning_module_hook("on_pretrain_routine_end") def _run_train(self) -> None: self._pre_training_routine() with isolate_rng(): self._run_sanity_check() # enable train mode self.model.train() torch.set_grad_enabled(True) self.fit_loop.trainer = self with torch.autograd.set_detect_anomaly(self._detect_anomaly): self.fit_loop.run() def _run_evaluate(self) -> _EVALUATE_OUTPUT: assert self.evaluating # reload dataloaders self._evaluation_loop._reload_evaluation_dataloaders() # reset trainer on this loop and all child loops in case user connected a custom loop self._evaluation_loop.trainer = self with self.profiler.profile(f"run_{self.state.stage}_evaluation"), _evaluation_context(self.accelerator): eval_loop_results = self._evaluation_loop.run() # remove the tensors from the eval results for result in eval_loop_results: if isinstance(result, dict): for k, v in result.items(): if isinstance(v, Tensor): result[k] = v.cpu().item() return eval_loop_results def _run_predict(self) -> Optional[_PREDICT_OUTPUT]: self.reset_predict_dataloader(self.lightning_module) # reset trainer on this loop and all child loops in case user connected a custom loop self.predict_loop.trainer = self with _evaluation_context(self.accelerator): return self.predict_loop.run() def _run_sanity_check(self) -> None: val_loop = self.fit_loop.epoch_loop.val_loop should_sanity_check = ( self.enable_validation and self.num_sanity_val_steps > 0 # do not sanity check if restarting because it would mess up the loaded state and not val_loop.restarting ) # run tiny validation (if validation defined) # to make sure program won't crash during val if should_sanity_check: stage = self.state.stage self.sanity_checking = True # reset logger connector self._logger_connector.reset_results() self._logger_connector.reset_metrics() self._call_callback_hooks("on_sanity_check_start") # reload dataloaders val_loop._reload_evaluation_dataloaders() self.num_sanity_val_batches = [ min(self.num_sanity_val_steps, val_batches) for val_batches in self.num_val_batches ] # run eval step with torch.no_grad(): val_loop.run() self._call_callback_hooks("on_sanity_check_end") # reset logger connector self._logger_connector.reset_results() self._logger_connector.reset_metrics() # reset the progress tracking state after sanity checking. we don't need to set the state before # because sanity check only runs when we are not restarting _reset_progress(val_loop) # restore the previous stage when the sanity check if finished self.state.stage = stage def __set_ckpt_path(self, ckpt_path: Optional[str], model_provided: bool, model_connected: bool) -> Optional[str]: # fault-tolerance takes precedence from pytorch_lightning.callbacks.fault_tolerance import _FaultToleranceCheckpoint ft_checkpoints = [cb for cb in self.callbacks if isinstance(cb, _FaultToleranceCheckpoint)] fn = self.state.fn.value if ckpt_path is None and ft_checkpoints and self.state.fn == TrainerFn.FITTING: ckpt_path = "last" rank_zero_warn( f"`.{fn}(ckpt_path=None)` was called without a model." " Because fault tolerance is enabled, the last model of the previous `fit` call will be used." f" You can pass `{fn}(ckpt_path='best')` to use the best model or" f" `{fn}(ckpt_path='last')` to use the last model." " If you pass a value, this warning will be silenced." ) if model_provided and ckpt_path is None: # use passed model to function without loading weights return if model_connected and ckpt_path is None: ckpt_path = "best" ft_tip = ( " There is also a fault-tolerant checkpoint available, however it is used by default only when fitting." if ft_checkpoints else "" ) rank_zero_warn( f"`.{fn}(ckpt_path=None)` was called without a model." " The best model of the previous `fit` call will be used." + ft_tip + f" You can pass `.{fn}(ckpt_path='best')` to use the best model or" f" `.{fn}(ckpt_path='last')` to use the last model." " If you pass a value, this warning will be silenced." ) if ckpt_path == "best": if len(self.checkpoint_callbacks) > 1: rank_zero_warn( f'`.{fn}(ckpt_path="best")` is called with Trainer configured with multiple `ModelCheckpoint`' " callbacks. It will use the best checkpoint path from first checkpoint callback." ) if not self.checkpoint_callback: raise MisconfigurationException( f'`.{fn}(ckpt_path="best")` is set but `ModelCheckpoint` is not configured.' ) if hasattr(self.checkpoint_callback, "best_model_path") and not self.checkpoint_callback.best_model_path: if self.fast_dev_run: raise MisconfigurationException( f'You cannot execute `.{fn}(ckpt_path="best")` with `fast_dev_run=True`.' f" Please pass an exact checkpoint path to `.{fn}(ckpt_path=...)`" ) raise MisconfigurationException( f'`.{fn}(ckpt_path="best")` is set but `ModelCheckpoint` is not configured to save the best model.' ) # load best weights ckpt_path = getattr(self.checkpoint_callback, "best_model_path", None) if ckpt_path == "last": candidates = [getattr(ft, "ckpt_path", None) for ft in ft_checkpoints] + [ getattr(cb, "last_model_path", None) for cb in self.checkpoint_callbacks ] candidates_fs = {path: get_filesystem(path) for path in candidates if path} candidates_ts = {path: fs.modified(path) for path, fs in candidates_fs.items() if fs.exists(path)} if not candidates_ts: # not an error so it can be set and forget before the first `fit` run rank_zero_warn( f'.{fn}(ckpt_path="last") is set, but there is no fault tolerant' " or last checkpoint available. No checkpoint will be loaded." ) return ckpt_path = max(candidates_ts.keys(), key=partial(operator.getitem, candidates_ts)) if not ckpt_path: raise MisconfigurationException( f"`.{fn}()` found no path for the best weights: {ckpt_path!r}. Please" f" specify a path for a checkpoint `.{fn}(ckpt_path=PATH)`" ) return ckpt_path def _call_setup_hook(self) -> None: fn = self.state.fn._setup_fn self.strategy.barrier("pre_setup") if self.datamodule is not None: self._call_lightning_datamodule_hook("setup", stage=fn) self._call_callback_hooks("setup", stage=fn) self._call_lightning_module_hook("setup", stage=fn) self.strategy.barrier("post_setup") def _call_configure_sharded_model(self) -> None: with self.strategy.model_sharded_context(): self._handle_meta_model() self._call_lightning_module_hook("configure_sharded_model") self._call_callback_hooks("on_configure_sharded_model") def _handle_meta_model(self) -> None: if not is_on_meta_device(self.lightning_module): return if isinstance(self.strategy, DDPSpawnStrategy): raise MisconfigurationException("LightningModule on meta device isn't supported with spawn.") materialize_module(self.lightning_module) # the trainer reference is lost during materialization self.lightning_module.trainer = proxy(self) def _call_teardown_hook(self) -> None: fn = self.state.fn._setup_fn if self.datamodule is not None: self._call_lightning_datamodule_hook("teardown", stage=fn) self._call_callback_hooks("teardown", stage=fn) self._call_lightning_module_hook("teardown", stage=fn) self.lightning_module._current_fx_name = None # these could have become stale if metrics are defined in `setup` self.lightning_module._metric_attributes = None # todo: TPU 8 cores hangs in flush with TensorBoard. Might do for all loggers. # It might be related to xla tensors blocked when moving the cpu kill loggers. for logger in self.loggers: logger.finalize("success") # summarize profile results self.profiler.describe() def call_hook( self, hook_name: str, *args: Any, pl_module: Optional["pl.LightningModule"] = None, **kwargs: Any ) -> Any: r""" .. deprecated:: v1.6 The Trainer's `call_hook` method was deprecated in v1.6 and will be removed in v1.8. """ rank_zero_deprecation("The Trainer's `call_hook` method was deprecated in v1.6 and will be removed in v1.8.") pl_module = self.lightning_module or pl_module if pl_module: prev_fx_name = pl_module._current_fx_name pl_module._current_fx_name = hook_name # always profile hooks with self.profiler.profile(hook_name): # first call trainer hook callback_fx = getattr(self, hook_name, None) if callable(callback_fx): callback_fx(*args, **kwargs) # next call hook in lightningModule output = None model_fx = getattr(pl_module, hook_name, None) if callable(model_fx): output = model_fx(*args, **kwargs) # call the strategy hook if hook_name not in ("setup", "teardown", "on_train_start") and hasattr(self.strategy, hook_name): strategy_hook = getattr(self.strategy, hook_name) strategy_output = strategy_hook(*args, **kwargs) output = strategy_output if output is None else output if pl_module: # restore current_fx when nested context pl_module._current_fx_name = prev_fx_name return output def _call_lightning_module_hook( self, hook_name: str, *args: Any, pl_module: Optional["pl.LightningModule"] = None, **kwargs: Any, ) -> Any: pl_module = pl_module or self.lightning_module if pl_module is None: raise TypeError("No `LightningModule` is available to call hooks on.") fn = getattr(pl_module, hook_name) if not callable(fn): return prev_fx_name = pl_module._current_fx_name pl_module._current_fx_name = hook_name with self.profiler.profile(f"[LightningModule]{pl_module.__class__.__name__}.{hook_name}"): output = fn(*args, **kwargs) # restore current_fx when nested context pl_module._current_fx_name = prev_fx_name return output def _call_lightning_datamodule_hook( self, hook_name: str, *args: Any, **kwargs: Any, ) -> Any: if self.datamodule is None: raise TypeError("No `LightningDataModule` is available to call hooks on.") fn = getattr(self.datamodule, hook_name) if callable(fn): with self.profiler.profile(f"[LightningDataModule]{self.datamodule.__class__.__name__}.{hook_name}"): return fn(*args, **kwargs) def _call_callback_hooks( self, hook_name: str, *args: Any, **kwargs: Any, ) -> None: log.debug(f"{self.__class__.__name__}: calling callback hook: {hook_name}") # TODO: remove if block in v1.8 if hook_name in ("on_init_start", "on_init_end"): # these `Callback` hooks are the only ones that do not take a lightning module. # we also don't profile bc profiler hasn't been set yet for callback in self.callbacks: fn = getattr(callback, hook_name) if callable(fn): fn(self, *args, **kwargs) return pl_module = self.lightning_module if pl_module: prev_fx_name = pl_module._current_fx_name pl_module._current_fx_name = hook_name for callback in self.callbacks: fn = getattr(callback, hook_name) if callable(fn): with self.profiler.profile(f"[Callback]{callback.state_key}.{hook_name}"): fn(self, self.lightning_module, *args, **kwargs) if pl_module: # restore current_fx when nested context pl_module._current_fx_name = prev_fx_name def _call_callbacks_state_dict(self) -> Dict[str, dict]: """Called when saving a model checkpoint, calls and returns every callback's `state_dict`, keyed by `Callback.state_key`.""" callback_state_dicts = {} for callback in self.callbacks: state_dict = callback.state_dict() if state_dict: callback_state_dicts[callback.state_key] = state_dict return callback_state_dicts def _call_callbacks_on_save_checkpoint(self, checkpoint: Dict[str, Any]) -> None: """Called when saving a model checkpoint, calls every callback's `on_save_checkpoint` hook. Will be removed in v1.8: If state is returned, we insert the callback state into ``checkpoint["callbacks"][Callback.state_key]``. It overrides ``state_dict`` if already present. """ pl_module = self.lightning_module if pl_module: prev_fx_name = pl_module._current_fx_name pl_module._current_fx_name = "on_save_checkpoint" for callback in self.callbacks: with self.profiler.profile(f"[Callback]{callback.state_key}.on_save_checkpoint"): state = callback.on_save_checkpoint(self, self.lightning_module, checkpoint) if state: rank_zero_deprecation( f"Returning a value from `{callback.__class__.__name__}.on_save_checkpoint` is deprecated in v1.6" " and will be removed in v1.8. Please override `Callback.state_dict`" " to return state to be saved." ) checkpoint["callbacks"][callback.state_key] = state if pl_module: # restore current_fx when nested context pl_module._current_fx_name = prev_fx_name def _call_callbacks_on_load_checkpoint(self, checkpoint: Dict[str, Any]) -> None: """Called when loading a model checkpoint. Calls every callback's `on_load_checkpoint` hook. We have a dedicated function for this rather than using `_call_callback_hooks` because we have special logic for getting callback_states. """ pl_module = self.lightning_module if pl_module: prev_fx_name = pl_module._current_fx_name pl_module._current_fx_name = "on_load_checkpoint" callback_states: Dict[Union[Type, str], Dict] = checkpoint.get("callbacks") if callback_states is None: return is_legacy_ckpt = Version(checkpoint["pytorch-lightning_version"]) < Version("1.5.0dev") current_callbacks_keys = {cb._legacy_state_key if is_legacy_ckpt else cb.state_key for cb in self.callbacks} difference = callback_states.keys() - current_callbacks_keys if difference: rank_zero_warn( "Be aware that when using `ckpt_path`," " callbacks used to create the checkpoint need to be provided during `Trainer` instantiation." f" Please add the following callbacks: {list(difference)}.", ) for callback in self.callbacks: state = callback_states.get(callback.state_key, callback_states.get(callback._legacy_state_key)) if state: state = deepcopy(state) with self.profiler.profile(f"[Callback]{callback.state_key}.on_load_checkpoint"): callback.on_load_checkpoint(self, self.lightning_module, state) if pl_module: # restore current_fx when nested context pl_module._current_fx_name = prev_fx_name def _call_callbacks_load_state_dict(self, checkpoint: Dict[str, Any]) -> None: """Called when loading a model checkpoint, calls every callback's `load_state_dict`.""" callback_states: Dict[Union[Type, str], Dict] = checkpoint.get("callbacks") if callback_states is None: return for callback in self.callbacks: state = callback_states.get(callback.state_key, callback_states.get(callback._legacy_state_key)) if state: state = deepcopy(state) callback.load_state_dict(state) def _call_strategy_hook( self, hook_name: str, *args: Any, **kwargs: Any, ) -> Any: pl_module = self.lightning_module prev_fx_name = pl_module._current_fx_name pl_module._current_fx_name = hook_name fn = getattr(self.strategy, hook_name) if not callable(fn): return with self.profiler.profile(f"[Strategy]{self.strategy.__class__.__name__}.{hook_name}"): output = fn(*args, **kwargs) # restore current_fx when nested context pl_module._current_fx_name = prev_fx_name return output @staticmethod def _log_api_event(event: str) -> None: torch._C._log_api_usage_once("lightning.trainer." + event) def __init_profiler(self, profiler: Optional[Union[Profiler, str]]) -> None: if isinstance(profiler, str): PROFILERS = { "simple": SimpleProfiler, "advanced": AdvancedProfiler, "pytorch": PyTorchProfiler, "xla": XLAProfiler, } profiler = profiler.lower() if profiler not in PROFILERS: raise MisconfigurationException( "When passing string value for the `profiler` parameter of `Trainer`," f" it can only be one of {list(PROFILERS.keys())}" ) profiler_class = PROFILERS[profiler] profiler = profiler_class() self.profiler: Profiler = profiler or PassThroughProfiler() def __setup_profiler(self) -> None: local_rank = self.local_rank if self.world_size > 1 else None self.profiler._lightning_module = proxy(self.lightning_module) self.profiler.setup(stage=self.state.fn._setup_fn, local_rank=local_rank, log_dir=self.log_dir) def _log_device_info(self) -> None: if CUDAAccelerator.is_available(): gpu_available = True gpu_type = " (cuda)" elif MPSAccelerator.is_available(): gpu_available = True gpu_type = " (mps)" else: gpu_available = False gpu_type = "" gpu_used = isinstance(self.accelerator, (CUDAAccelerator, MPSAccelerator)) rank_zero_info(f"GPU available: {gpu_available}{gpu_type}, used: {gpu_used}") num_tpu_cores = self.num_devices if isinstance(self.accelerator, TPUAccelerator) else 0 rank_zero_info(f"TPU available: {_TPU_AVAILABLE}, using: {num_tpu_cores} TPU cores") num_ipus = self.num_devices if isinstance(self.accelerator, IPUAccelerator) else 0 rank_zero_info(f"IPU available: {_IPU_AVAILABLE}, using: {num_ipus} IPUs") num_hpus = self.num_devices if isinstance(self.accelerator, HPUAccelerator) else 0 rank_zero_info(f"HPU available: {_HPU_AVAILABLE}, using: {num_hpus} HPUs") # TODO: Integrate MPS Accelerator here, once gpu maps to both if CUDAAccelerator.is_available() and not isinstance(self.accelerator, CUDAAccelerator): rank_zero_warn( "GPU available but not used. Set `accelerator` and `devices` using" f" `Trainer(accelerator='gpu', devices={CUDAAccelerator.auto_device_count()})`.", category=PossibleUserWarning, ) if _TPU_AVAILABLE and not isinstance(self.accelerator, TPUAccelerator): rank_zero_warn( "TPU available but not used. Set `accelerator` and `devices` using" f" `Trainer(accelerator='tpu', devices={TPUAccelerator.auto_device_count()})`." ) if _IPU_AVAILABLE and not isinstance(self.accelerator, IPUAccelerator): rank_zero_warn( "IPU available but not used. Set `accelerator` and `devices` using" f" `Trainer(accelerator='ipu', devices={IPUAccelerator.auto_device_count()})`." ) if _HPU_AVAILABLE and not isinstance(self.accelerator, HPUAccelerator): rank_zero_warn( "HPU available but not used. Set `accelerator` and `devices` using" f" `Trainer(accelerator='hpu', devices={HPUAccelerator.auto_device_count()})`." ) if MPSAccelerator.is_available() and not isinstance(self.accelerator, MPSAccelerator): rank_zero_warn( "MPS available but not used. Set `accelerator` and `devices` using" f" `Trainer(accelerator='mps', devices={MPSAccelerator.auto_device_count()})`." ) """ Data loading methods """ def reset_train_dataloader(self, model: Optional["pl.LightningModule"] = None) -> None: """Resets the train dataloader and initialises required variables (number of batches, when to validate, etc.). Args: model: The ``LightningModule`` if calling this outside of the trainer scope. """ source = self._data_connector._train_dataloader_source pl_module = model or self.lightning_module has_step = is_overridden("training_step", pl_module) enable_training = self.limit_train_batches > 0 if not (source.is_defined() and has_step and enable_training): return self.train_dataloader = self._data_connector._request_dataloader(RunningStage.TRAINING) if self.overfit_batches > 0: self.train_dataloader = self._data_connector._resolve_overfit_batches( self.train_dataloader, mode=RunningStage.TRAINING ) # automatically add samplers self.train_dataloader = apply_to_collection( self.train_dataloader, (DataLoader, CombinedLoader), self._data_connector._prepare_dataloader, mode=RunningStage.TRAINING, ) loaders = ( self.train_dataloader.loaders if isinstance(self.train_dataloader, CombinedLoader) else self.train_dataloader ) # check the workers recursively apply_to_collection(loaders, DataLoader, self._data_connector._worker_check, "train_dataloader") # add worker_init_fn for correct seeding in worker processes apply_to_collection(loaders, DataLoader, _auto_add_worker_init_fn, rank=self.global_rank) # add collate_fn to collect metadata for fault tolerant training if _fault_tolerant_training(): apply_to_collection(loaders, DataLoader, _add_capture_metadata_collate) # wrap the sequence of train loaders to a CombinedLoader object for computing the num_training_batches if not isinstance(self.train_dataloader, CombinedLoader): self.train_dataloader = CombinedLoader(loaders, self._data_connector.multiple_trainloader_mode) module = model or self.lightning_module or self.datamodule orig_train_batches = self.num_training_batches = ( len(self.train_dataloader) if has_len_all_ranks(self.train_dataloader, self.strategy, module) else float("inf") ) if orig_train_batches == 0: return # store epoch of dataloader reset for reload_dataloaders_every_n_epochs self._last_train_dl_reload_epoch = self.current_epoch if isinstance(self.limit_train_batches, int): self.num_training_batches = min(orig_train_batches, self.limit_train_batches) elif self.num_training_batches != float("inf"): self.num_training_batches = int(orig_train_batches * self.limit_train_batches) elif self.limit_train_batches != 1.0: raise MisconfigurationException( "When using an `IterableDataset`, `Trainer(limit_train_batches)` must be `1.0` or an int." "An int specifies `num_training_batches` to use." ) if isinstance(self.val_check_interval, int): self.val_check_batch = self.val_check_interval if self.val_check_batch > self.num_training_batches and self.check_val_every_n_epoch is not None: raise ValueError( f"`val_check_interval` ({self.val_check_interval}) must be less than or equal " f"to the number of the training batches ({self.num_training_batches}). " "If you want to disable validation set `limit_val_batches` to 0.0 instead." "If you want to validate based on the total training batches, set `check_val_every_n_epoch=None`." ) else: if not has_len_all_ranks(self.train_dataloader, self.strategy, module): if self.val_check_interval == 1.0: self.val_check_batch = float("inf") else: raise MisconfigurationException( "When using an IterableDataset for `train_dataloader`," " `Trainer(val_check_interval)` must be `1.0` or an int. An int k specifies" " checking validation every k training batches." ) else: self.val_check_batch = int(self.num_training_batches * self.val_check_interval) self.val_check_batch = max(1, self.val_check_batch) if self.loggers and self.num_training_batches < self.log_every_n_steps: rank_zero_warn( f"The number of training batches ({self.num_training_batches}) is smaller than the logging interval" f" Trainer(log_every_n_steps={self.log_every_n_steps}). Set a lower value for log_every_n_steps if" " you want to see logs for the training epoch.", category=PossibleUserWarning, ) if ( self.num_training_batches == 0 and self.limit_train_batches > 0.0 and isinstance(self.limit_train_batches, float) and orig_train_batches != float("inf") ): min_percentage = 1.0 / orig_train_batches raise MisconfigurationException( f"You requested to check {self.limit_train_batches} of the `train_dataloader` but" f" {self.limit_train_batches} * {orig_train_batches} < 1. Please increase the" f" `limit_train_batches` argument. Try at least" f" `limit_train_batches={min_percentage}`" ) def reset_val_dataloader(self, model: Optional["pl.LightningModule"] = None) -> None: """Resets the validation dataloader and determines the number of batches. Args: model: The ``LightningModule`` if called outside of the trainer scope. """ source = self._data_connector._val_dataloader_source pl_module = self.lightning_module or model has_step = is_overridden("validation_step", pl_module) enable_validation = self.limit_val_batches > 0 if source.is_defined() and has_step and enable_validation: self.num_val_batches, self.val_dataloaders = self._data_connector._reset_eval_dataloader( RunningStage.VALIDATING, model=pl_module ) # store epoch of dataloader reset for reload_dataloaders_every_n_epochs self._last_val_dl_reload_epoch = self.current_epoch def reset_test_dataloader(self, model: Optional["pl.LightningModule"] = None) -> None: """Resets the test dataloader and determines the number of batches. Args: model: The ``LightningModule`` if called outside of the trainer scope. """ source = self._data_connector._test_dataloader_source pl_module = self.lightning_module or model has_step = is_overridden("test_step", pl_module) enable_testing = self.limit_test_batches > 0 if source.is_defined() and has_step and enable_testing: self.num_test_batches, self.test_dataloaders = self._data_connector._reset_eval_dataloader( RunningStage.TESTING, model=pl_module ) def reset_predict_dataloader(self, model: Optional["pl.LightningModule"] = None) -> None: """Resets the predict dataloader and determines the number of batches. Args: model: The ``LightningModule`` if called outside of the trainer scope. """ source = self._data_connector._predict_dataloader_source pl_module = self.lightning_module or model enable_prediction = self.limit_predict_batches > 0 if source.is_defined() and enable_prediction: self.num_predict_batches, self.predict_dataloaders = self._data_connector._reset_eval_dataloader( RunningStage.PREDICTING, model=pl_module ) def reset_train_val_dataloaders(self, model: Optional["pl.LightningModule"] = None) -> None: """Resets train and val dataloaders if none are attached to the trainer. The val dataloader must be initialized before training loop starts, as the training loop inspects the val dataloader to determine whether to run the evaluation loop. Args: model: The ``LightningModule`` if called outside of the trainer scope. .. deprecated:: v1.7 This method is deprecated in v1.7 and will be removed in v1.9. Please use ``Trainer.reset_{train,val}_dataloader`` instead. """ rank_zero_deprecation( "`Trainer.reset_train_val_dataloaders` has been deprecated in v1.7 and will be removed in v1.9." " Use `Trainer.reset_{train,val}_dataloader` instead" ) if self.train_dataloader is None: self.reset_train_dataloader(model=model) if self.val_dataloaders is None: self.reset_val_dataloader(model=model) """ Accelerator properties """ @property def accelerator(self) -> Accelerator: return self.strategy.accelerator @property def strategy(self) -> Strategy: return self._accelerator_connector.strategy @property def training_type_plugin(self) -> Strategy: rank_zero_deprecation( "`Trainer.training_type_plugin` is deprecated in v1.6 and will be removed in v1.8. Use" " `Trainer.strategy` instead." ) return self.strategy @property def precision_plugin(self) -> PrecisionPlugin: return self.strategy.precision_plugin @property def global_rank(self) -> int: return self.strategy.global_rank @property def local_rank(self) -> int: # some strategies define a local rank return getattr(self.strategy, "local_rank", 0) @property def node_rank(self) -> int: # some strategies define a node rank return getattr(self.strategy, "node_rank", 0) @property def world_size(self) -> int: # some strategies define a world size return getattr(self.strategy, "world_size", 1) @property def should_rank_save_checkpoint(self) -> bool: rank_zero_deprecation( "`Trainer.should_rank_save_checkpoint` is deprecated in v1.6 and will be removed in v1.8.", stacklevel=5 ) strategy = self.strategy return ( isinstance(strategy, pl.strategies.TPUSpawnStrategy) and strategy.local_rank == 0 or strategy.is_global_zero ) @property def num_nodes(self) -> int: return getattr(self.strategy, "num_nodes", 1) @property def device_ids(self) -> List[int]: """List of device indexes per node.""" devices = ( self.strategy.parallel_devices if isinstance(self.strategy, ParallelStrategy) else [self.strategy.root_device] ) device_ids = [] for idx, device in enumerate(devices): if isinstance(device, torch.device): device_ids.append(device.index or idx) elif isinstance(device, int): device_ids.append(device) return device_ids @property def num_devices(self) -> int: """Number of devices the trainer uses per node.""" return len(self.device_ids) @property def num_processes(self) -> int: rank_zero_deprecation( "`Trainer.num_processes` is deprecated in v1.6 and will be removed in v1.8. " "Please use `Trainer.num_devices` instead." ) return self.num_devices @property def root_gpu(self) -> Optional[int]: rank_zero_deprecation( "`Trainer.root_gpu` is deprecated in v1.6 and will be removed in v1.8. " "Please use `Trainer.strategy.root_device.index` instead." ) return self.strategy.root_device.index if isinstance(self.accelerator, CUDAAccelerator) else None @property def tpu_cores(self) -> int: rank_zero_deprecation( "`Trainer.tpu_cores` is deprecated in v1.6 and will be removed in v1.8. " "Please use `Trainer.num_devices` instead." ) return self.num_devices if isinstance(self.accelerator, TPUAccelerator) else 0 @property def ipus(self) -> int: rank_zero_deprecation( "`Trainer.ipus` was deprecated in v1.6 and will be removed in v1.8." " Please use `Trainer.num_devices` instead." ) return self.num_devices if isinstance(self.accelerator, IPUAccelerator) else 0 @property def num_gpus(self) -> int: rank_zero_deprecation( "`Trainer.num_gpus` was deprecated in v1.6 and will be removed in v1.8." " Please use `Trainer.num_devices` instead." ) return self.num_devices if isinstance(self.accelerator, CUDAAccelerator) else 0 @property def devices(self) -> int: rank_zero_deprecation( "`Trainer.devices` was deprecated in v1.6 and will be removed in v1.8." " Please use `Trainer.num_devices` or `Trainer.device_ids` to get device information instead." ) return self.num_devices @property def data_parallel_device_ids(self) -> Optional[List[int]]: rank_zero_deprecation( "`Trainer.data_parallel_device_ids` was deprecated in v1.6 and will be removed in v1.8." " Please use `Trainer.device_ids` instead." ) return self.device_ids if isinstance(self.accelerator, CUDAAccelerator) else None @property def lightning_module(self) -> "pl.LightningModule": # TODO: this is actually an optional return return self.strategy.lightning_module @property def optimizers(self) -> List[Optimizer]: return self.strategy.optimizers @optimizers.setter def optimizers(self, new_optims: Optional[List[Optimizer]]) -> None: self.strategy.optimizers = new_optims @property def lightning_optimizers(self) -> Dict[int, LightningOptimizer]: rank_zero_deprecation( "`Trainer.lightning_optimizers` is deprecated in v1.6 and will be removed in v1.8", stacklevel=5 ) return self.strategy._lightning_optimizers @property def lr_scheduler_configs(self) -> List[LRSchedulerConfig]: return self.strategy.lr_scheduler_configs @property def lr_schedulers(self) -> List[Dict[str, Any]]: rank_zero_deprecation( "`Trainer.lr_schedulers` is deprecated in v1.6 and will be removed in v1.8." " You can use `trainer.lr_scheduler_configs` instead which contains dataclasses instead of dictionaries.", stacklevel=5, ) from dataclasses import asdict return [asdict(config) for config in self.strategy.lr_scheduler_configs] @property def optimizer_frequencies(self) -> List[int]: return self.strategy.optimizer_frequencies @optimizer_frequencies.setter def optimizer_frequencies(self, new_freqs: List[int]) -> None: self.strategy.optimizer_frequencies = new_freqs @property def amp_backend(self) -> Optional[AMPType]: if isinstance(self.precision_plugin, ApexMixedPrecisionPlugin): return AMPType.APEX if isinstance(self.precision_plugin, NativeMixedPrecisionPlugin): return AMPType.NATIVE return None @property def precision(self) -> Union[str, int]: return self.strategy.precision_plugin.precision @property def scaler(self) -> Optional[Any]: return getattr(self.precision_plugin, "scaler", None) @property def gpus(self) -> Optional[Union[List[int], str, int]]: rank_zero_deprecation( "`Trainer.gpus` was deprecated in v1.6 and will be removed in v1.8." " Please use `Trainer.num_devices` or `Trainer.device_ids` to get device information instead." ) return self._accelerator_connector._gpus @property def model(self) -> torch.nn.Module: """The LightningModule, but possibly wrapped into DataParallel or DistributedDataParallel. To access the pure LightningModule, use :meth:`~pytorch_lightning.trainer.trainer.Trainer.lightning_module` instead. """ return self.strategy.model @model.setter def model(self, model: torch.nn.Module) -> None: """Setter for the model, pass-through to accelerator and plugin where the model reference is stored. Used by the Tuner to reset the state of Trainer and Accelerator. Args: model: The LightningModule, possibly wrapped into DataParallel or DistributedDataParallel, depending on the backend. """ self.strategy.model = model """ General properties """ @property def log_dir(self) -> Optional[str]: if len(self.loggers) == 1: if isinstance(self.logger, TensorBoardLogger): dirpath = self.logger.log_dir else: dirpath = self.logger.save_dir else: dirpath = self.default_root_dir dirpath = self.strategy.broadcast(dirpath) return dirpath @property def use_amp(self) -> bool: rank_zero_deprecation( "`Trainer.use_amp` is deprecated in v1.6.0 and will be removed in v1.8.0." " Please use `Trainer.amp_backend` instead." ) return self.precision == 16 @property def is_global_zero(self) -> bool: return self.strategy.is_global_zero @property def distributed_sampler_kwargs(self) -> Optional[dict]: if isinstance(self.strategy, ParallelStrategy): return self.strategy.distributed_sampler_kwargs @property def data_parallel(self) -> bool: return isinstance(self.strategy, ParallelStrategy) @property def enable_validation(self) -> bool: """Check if we should run validation during training.""" return ( self._data_connector._val_dataloader_source.is_defined() and is_overridden("validation_step", self.lightning_module) and self.limit_val_batches > 0 ) @property def default_root_dir(self) -> str: """The default location to save artifacts of loggers, checkpoints etc. It is used as a fallback if logger or checkpoint callback do not define specific save paths. """ if get_filesystem(self._default_root_dir).protocol == "file": return os.path.normpath(self._default_root_dir) return self._default_root_dir @property def weights_save_path(self) -> str: """ The default root location to save weights (checkpoints), e.g., when the :class:`~pytorch_lightning.callbacks.model_checkpoint.ModelCheckpoint` does not define a file path. .. deprecated:: v1.6 `Trainer.weights_save_path` has been deprecated in v1.6 and will be removed in v1.8. """ rank_zero_deprecation("`Trainer.weights_save_path` has been deprecated in v1.6 and will be removed in v1.8.") return self._weights_save_path_internal # TODO: Remove _weights_save_path_internal in v1.8 @property def _weights_save_path_internal(self) -> str: """This is an internal implementation of weights_save_path which allows weights_save_path to be used internally by the framework without emitting a deprecation warning. To be removed in v1.8. """ if get_filesystem(self._weights_save_path).protocol == "file": return os.path.normpath(self._weights_save_path) return self._weights_save_path @property def early_stopping_callback(self) -> Optional[EarlyStopping]: """The first :class:`~pytorch_lightning.callbacks.early_stopping.EarlyStopping` callback in the Trainer.callbacks list, or ``None`` if it doesn't exist.""" callbacks = self.early_stopping_callbacks return callbacks[0] if len(callbacks) > 0 else None @property def early_stopping_callbacks(self) -> List[EarlyStopping]: """A list of all instances of :class:`~pytorch_lightning.callbacks.early_stopping.EarlyStopping` found in the Trainer.callbacks list.""" return [c for c in self.callbacks if isinstance(c, EarlyStopping)] @property def prediction_writer_callbacks(self) -> List[BasePredictionWriter]: """A list of all instances of :class:`~pytorch_lightning.callbacks.prediction_writer.BasePredictionWriter` found in the Trainer.callbacks list.""" return [cb for cb in self.callbacks if isinstance(cb, BasePredictionWriter)] @property def checkpoint_callback(self) -> Optional[Checkpoint]: """The first :class:`~pytorch_lightning.callbacks.model_checkpoint.ModelCheckpoint` callback in the Trainer.callbacks list, or ``None`` if it doesn't exist.""" callbacks = self.checkpoint_callbacks return callbacks[0] if len(callbacks) > 0 else None @property def checkpoint_callbacks(self) -> List[Checkpoint]: """A list of all instances of :class:`~pytorch_lightning.callbacks.model_checkpoint.ModelCheckpoint` found in the Trainer.callbacks list.""" return [c for c in self.callbacks if isinstance(c, Checkpoint)] @property def progress_bar_callback(self) -> Optional[ProgressBarBase]: """An instance of :class:`~pytorch_lightning.callbacks.progress.base.ProgressBarBase` found in the Trainer.callbacks list, or ``None`` if one doesn't exist.""" for c in self.callbacks: if isinstance(c, ProgressBarBase): return c return None @property def resume_from_checkpoint(self) -> Optional[Union[str, Path]]: resume_from_checkpoint = self._checkpoint_connector.resume_from_checkpoint_fit_path if resume_from_checkpoint is not None: rank_zero_deprecation( "`trainer.resume_from_checkpoint` is deprecated in v1.5 and will be removed in v2.0." " Specify the fit checkpoint path with `trainer.fit(ckpt_path=)` instead.", stacklevel=5, ) return resume_from_checkpoint @property def ckpt_path(self) -> Optional[str]: """Set to the path/URL of a checkpoint loaded via :meth:`~pytorch_lightning.trainer.trainer.Trainer.fit`, :meth:`~pytorch_lightning.trainer.trainer.Trainer.validate`, :meth:`~pytorch_lightning.trainer.trainer.Trainer.test`, or :meth:`~pytorch_lightning.trainer.trainer.Trainer.predict`. ``None`` otherwise.""" return self._ckpt_path @property def validated_ckpt_path(self) -> Optional[str]: rank_zero_deprecation( "The `Trainer.validated_ckpt_path` attribute was deprecated in v1.6 and will be removed in v1.8. The" " path of a checkpoint loaded via `Trainer.{fit,validate,test,predict}` should be accessed via" " `Trainer.ckpt_path` instead.", stacklevel=5, ) return self._validated_ckpt_path @validated_ckpt_path.setter def validated_ckpt_path(self, ckpt_path: Optional[str]) -> None: rank_zero_deprecation( "The `Trainer.validated_ckpt_path` attribute was deprecated in v1.6 and will be removed in v1.8. The" " path of a checkpoint loaded via `Trainer.{fit,validate,test,predict}` should be accessed via the" " read-only `Trainer.ckpt_path`.", stacklevel=5, ) self._validated_ckpt_path = ckpt_path @property def tested_ckpt_path(self) -> Optional[str]: rank_zero_deprecation( "The `Trainer.tested_ckpt_path` attribute was deprecated in v1.6 and will be removed in v1.8. The" " path of a checkpoint loaded via `Trainer.{fit,validate,test,predict}` should be accessed via" " `Trainer.ckpt_path` instead.", stacklevel=5, ) return self._tested_ckpt_path @tested_ckpt_path.setter def tested_ckpt_path(self, ckpt_path: Optional[str]) -> None: rank_zero_deprecation( "The `Trainer.tested_ckpt_path` attribute was deprecated in v1.6 and will be removed in v1.8. The" " path of a checkpoint loaded via `Trainer.{fit,validate,test,predict}` should be accessed via the" " read-only `Trainer.ckpt_path` instead.", stacklevel=5, ) self._tested_ckpt_path = ckpt_path @property def predicted_ckpt_path(self) -> Optional[str]: rank_zero_deprecation( "The `Trainer.predicted_ckpt_path` attribute was deprecated in v1.6 and will be removed in v1.8. The" " path of a checkpoint loaded via `Trainer.{fit,validate,test,predict}` should be accessed via" " `Trainer.ckpt_path` instead.", stacklevel=5, ) return self._predicted_ckpt_path @predicted_ckpt_path.setter def predicted_ckpt_path(self, ckpt_path: Optional[str]) -> None: rank_zero_deprecation( "The `Trainer.predicted_ckpt_path` attribute was deprecated in v1.6 and will be removed in v1.8. The" " path of a checkpoint loaded via `Trainer.{fit,validate,test,predict}` should be accessed via the" " read-only `Trainer.ckpt_path` instead.", stacklevel=5, ) self._predicted_ckpt_path = ckpt_path def save_checkpoint( self, filepath: _PATH, weights_only: bool = False, storage_options: Optional[Any] = None ) -> None: r""" Runs routine to create a checkpoint. Args: filepath: Path where checkpoint is saved. weights_only: If ``True``, will only save the model weights. storage_options: parameter for how to save to storage, passed to ``CheckpointIO`` plugin """ if self.model is None: raise AttributeError( "Saving a checkpoint is only possible if a model is attached to the Trainer. Did you call" " `Trainer.save_checkpoint()` before calling `Trainer.{fit,validate,test,predict}`?" ) self._checkpoint_connector.save_checkpoint(filepath, weights_only=weights_only, storage_options=storage_options) """ Parsing properties """ @classmethod def default_attributes(cls) -> dict: init_signature = inspect.signature(cls) return {k: v.default for k, v in init_signature.parameters.items()} @classmethod def get_deprecated_arg_names(cls) -> List: """Returns a list with deprecated Trainer arguments.""" depr_arg_names = [] for name, val in cls.__dict__.items(): if name.startswith("DEPRECATED") and isinstance(val, (tuple, list)): depr_arg_names.extend(val) return depr_arg_names @classmethod def from_argparse_args(cls: Any, args: Union[Namespace, ArgumentParser], **kwargs) -> Any: return from_argparse_args(cls, args, **kwargs) @classmethod def parse_argparser(cls, arg_parser: Union[ArgumentParser, Namespace]) -> Namespace: return parse_argparser(cls, arg_parser) @classmethod def match_env_arguments(cls) -> Namespace: return parse_env_variables(cls) @classmethod def add_argparse_args(cls, parent_parser: ArgumentParser, **kwargs) -> ArgumentParser: return add_argparse_args(cls, parent_parser, **kwargs) """ State properties """ @property def interrupted(self) -> bool: return self.state.status == TrainerStatus.INTERRUPTED @property def training(self) -> bool: return self.state.stage == RunningStage.TRAINING @training.setter def training(self, val: bool) -> None: if val: self.state.stage = RunningStage.TRAINING elif self.training: self.state.stage = None @property def testing(self) -> bool: return self.state.stage == RunningStage.TESTING @testing.setter def testing(self, val: bool) -> None: if val: self.state.stage = RunningStage.TESTING elif self.testing: self.state.stage = None @property def predicting(self) -> bool: return self.state.stage == RunningStage.PREDICTING @predicting.setter def predicting(self, val: bool) -> None: if val: self.state.stage = RunningStage.PREDICTING elif self.predicting: self.state.stage = None @property def tuning(self) -> bool: return self.state.stage == RunningStage.TUNING @tuning.setter def tuning(self, val: bool) -> None: if val: self.state.stage = RunningStage.TUNING elif self.tuning: self.state.stage = None @property def validating(self) -> bool: return self.state.stage == RunningStage.VALIDATING @validating.setter def validating(self, val: bool) -> None: if val: self.state.stage = RunningStage.VALIDATING elif self.validating: self.state.stage = None @property def evaluating(self) -> bool: return self.state.stage and self.state.stage.evaluating @property def sanity_checking(self) -> bool: return self.state.stage == RunningStage.SANITY_CHECKING @sanity_checking.setter def sanity_checking(self, val: bool) -> None: if val: self.state.stage = RunningStage.SANITY_CHECKING elif self.sanity_checking: self.state.stage = None """ Loop properties """ @property def global_step(self) -> int: """The number of optimizer steps taken (does not reset each epoch). This includes multiple optimizers and TBPTT steps (if enabled). """ return self.fit_loop.epoch_loop.global_step @property def current_epoch(self) -> int: """The current epoch, updated after the epoch end hooks are run.""" return self.fit_loop.epoch_progress.current.completed @property def max_epochs(self) -> int: return self.fit_loop.max_epochs @property def min_epochs(self) -> int: return self.fit_loop.min_epochs @property def max_steps(self) -> int: return self.fit_loop.max_steps @property def min_steps(self) -> Optional[int]: return self.fit_loop.min_steps @property def is_last_batch(self) -> bool: """Whether trainer is executing the last batch.""" return self.fit_loop.epoch_loop.batch_progress.is_last_batch @property def fit_loop(self) -> FitLoop: return self._fit_loop @fit_loop.setter def fit_loop(self, loop: FitLoop): """Attach a custom fit loop to this Trainer. It will run with :meth:`~pytorch_lightning.trainer.trainer.Trainer.fit`. """ loop.trainer = self self._fit_loop = loop @property def validate_loop(self) -> EvaluationLoop: return self._validate_loop @validate_loop.setter def validate_loop(self, loop: EvaluationLoop): """Attach a custom validation loop to this Trainer. It will run with :meth:`~pytorch_lightning.trainer.trainer.Trainer.validate`. Note that this loop is different from the one running during training inside the :meth:`pytorch_lightning.trainer.trainer.Trainer.fit` call. """ loop.trainer = self self._validate_loop = loop @property def test_loop(self) -> EvaluationLoop: return self._test_loop @test_loop.setter def test_loop(self, loop: EvaluationLoop): """Attach a custom test loop to this Trainer. It will run with :meth:`~pytorch_lightning.trainer.trainer.Trainer.test`. """ loop.trainer = self self._test_loop = loop @property def predict_loop(self) -> PredictionLoop: return self._predict_loop @predict_loop.setter def predict_loop(self, loop: PredictionLoop): """Attach a custom prediction loop to this Trainer. It will run with :meth:`~pytorch_lightning.trainer.trainer.Trainer.predict`. """ loop.trainer = self self._predict_loop = loop @property def verbose_evaluate(self) -> bool: rank_zero_deprecation( "The `Trainer.verbose_evaluate` property has been deprecated and will be removed in v1.8. The current value" " returned is the union of the validate and test loop values. You can choose which one to access with" " `trainer.{validate,test}_loop.verbose`.", stacklevel=5, ) return self.validate_loop.verbose or self.test_loop.verbose @verbose_evaluate.setter def verbose_evaluate(self, verbose: bool) -> None: rank_zero_deprecation( "The `Trainer.verbose_evaluate` property has been deprecated and will be removed in v1.8. This will set" " the value for both trainer.{validate,test}_loop.verbose`.", stacklevel=5, ) self.validate_loop.verbose = verbose self.test_loop.verbose = verbose @property def _evaluation_loop(self) -> EvaluationLoop: if self.state.fn in (TrainerFn.FITTING, TrainerFn.TUNING): return self.fit_loop.epoch_loop.val_loop if self.state.fn == TrainerFn.VALIDATING: return self.validate_loop if self.state.fn == TrainerFn.TESTING: return self.test_loop raise RuntimeError("The `Trainer._evaluation_loop` property isn't defined. Accessed outside of scope") @property def _active_loop(self) -> Optional[Union[FitLoop, EvaluationLoop, PredictionLoop]]: if self.training: return self.fit_loop if self.sanity_checking or self.evaluating: return self._evaluation_loop if self.predicting: return self.predict_loop """ Logging properties """ @property def logger(self) -> Optional[Logger]: loggers = self.loggers if len(loggers) == 0: return None if len(loggers) == 1: return loggers[0] else: rank_zero_deprecation( "Using `trainer.logger` when multiple loggers are configured." " This behavior will change in v1.8 when `LoggerCollection` is removed, and" " `trainer.logger` will return the first logger available.", stacklevel=5, ) with warnings.catch_warnings(): warnings.simplefilter("ignore") return LoggerCollection(loggers) @logger.setter def logger(self, logger: Optional[Logger]) -> None: if not logger: self.loggers = [] elif isinstance(logger, LoggerCollection): self.loggers = list(logger) else: self.loggers = [logger] @property def loggers(self) -> List[Logger]: return self._loggers @loggers.setter def loggers(self, loggers: Optional[List[Logger]]) -> None: self._loggers = loggers if loggers else [] @property def callback_metrics(self) -> Dict[str, Tensor]: # TODO: the true typing return can include dictionaries as defined in # `pytorch_lightning.trainer.connectors.logger_connector.result._OUT_DICT` return self._logger_connector.callback_metrics @property def logged_metrics(self) -> dict: return self._logger_connector.logged_metrics @property def progress_bar_metrics(self) -> dict: return self._logger_connector.progress_bar_metrics @property def _results(self) -> Optional[_ResultCollection]: active_loop = self._active_loop if active_loop is not None: return active_loop._results def _exit_gracefully_on_signal(self) -> None: if not _fault_tolerant_training() or not self._should_terminate_gracefully(): return raise ExitGracefullyException(0) def _should_terminate_gracefully(self) -> bool: value = torch.tensor(int(self._terminate_gracefully), device=self.strategy.root_device) return self.strategy.reduce(value, reduce_op="sum") > 0 """ Other """ @property def estimated_stepping_batches(self) -> Union[int, float]: r""" Estimated stepping batches for the complete training inferred from DataLoaders, gradient accumulation factor and distributed setup. Examples:: def configure_optimizers(self): optimizer = ... scheduler = torch.optim.lr_scheduler.OneCycleLR( optimizer, max_lr=1e-3, total_steps=self.trainer.estimated_stepping_batches ) return [optimizer], [scheduler] """ accumulation_scheduler = self.accumulation_scheduler if accumulation_scheduler.epochs != [0]: raise MisconfigurationException( "Estimated stepping batches cannot be computed with different" " `accumulate_grad_batches` at different epochs." ) # infinite training if self.max_epochs == -1 and self.max_steps == -1: return float("inf") if self.train_dataloader is None: rank_zero_info("Loading `train_dataloader` to estimate number of stepping batches.") self.reset_train_dataloader() total_batches = self.num_training_batches # iterable dataset if total_batches == float("inf"): return self.max_steps self.accumulate_grad_batches = accumulation_scheduler.get_accumulate_grad_batches(self.current_epoch) effective_batch_size = self.accumulate_grad_batches max_estimated_steps = math.ceil(total_batches / effective_batch_size) * max(self.max_epochs, 1) max_estimated_steps = min(max_estimated_steps, self.max_steps) if self.max_steps != -1 else max_estimated_steps return max_estimated_steps
@contextmanager def _evaluation_context(accelerator: Accelerator) -> Generator: # inference mode is not supported with gloo backend (#9431), # and HPU & TPU accelerators. context_manager_class = ( torch.inference_mode if not (dist.is_initialized() and dist.get_backend() == "gloo") and not isinstance(accelerator, HPUAccelerator) and not isinstance(accelerator, TPUAccelerator) else torch.no_grad ) with context_manager_class(): yield def _determine_batch_limits(batches: Optional[Union[int, float]], name: str) -> Union[int, float]: if batches is None: # batches is optional to know if the user passed a value so that we can show the above info messages only to the # users that set a value explicitly return 1.0 # differentiating based on the type can be error-prone for users. show a message describing the chosen behaviour if isinstance(batches, int) and batches == 1: if name == "limit_train_batches": message = "1 batch per epoch will be used." elif name == "val_check_interval": message = "validation will run after every batch." else: message = "1 batch will be used." rank_zero_info(f"`Trainer({name}=1)` was configured so {message}") elif isinstance(batches, float) and batches == 1.0: if name == "limit_train_batches": message = "100% of the batches per epoch will be used." elif name == "val_check_interval": message = "validation will run at the end of the training epoch." else: message = "100% of the batches will be used." rank_zero_info(f"`Trainer({name}=1.0)` was configured so {message}.") if 0 <= batches <= 1: return batches if batches > 1 and batches % 1.0 == 0: return int(batches) raise MisconfigurationException( f"You have passed invalid value {batches} for {name}, it has to be in [0.0, 1.0] or an int." )