-
Notifications
You must be signed in to change notification settings - Fork 3.4k
/
trainer.py
2238 lines (1822 loc) · 92.8 KB
/
trainer.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
# 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.
# THIS FILE MUST READ EASILY, FOR UNDERSTANDING AND DEBUGGING PURPOSES.
# DO NOT OBSCURE THE TRAINING LOOP
# THIS IS A HARD REQUIREMENT TO CONTRIBUTING TO LIGHTNING
# WE FAVOR READABILITY OVER ENGINEERING-CONSTRUCTS BY DESIGN
# DO NOT REMOVE THIS NOTICE
# - WILLIAM FALCON
"""Trainer to automate the training."""
import inspect
import logging
import math
import os
import warnings
from argparse import _ArgumentGroup, ArgumentParser, Namespace
from contextlib import contextmanager
from copy import deepcopy
from datetime import timedelta
from pathlib import Path
from typing import Any, Dict, Generator, Iterable, List, Optional, Type, Union
from weakref import proxy
import torch
import torch.distributed as dist
from lightning_utilities.core.apply_func import apply_to_collection
from lightning_utilities.core.imports import module_available
from packaging.version import Version
from torch import Tensor
from torch.optim import Optimizer
from torch.utils.data import DataLoader
from typing_extensions import Literal
import pytorch_lightning as pl
from lightning_lite.utilities.cloud_io import get_filesystem
from lightning_lite.utilities.data import _auto_add_worker_init_fn
from lightning_lite.utilities.types import _PATH
from lightning_lite.utilities.warnings import PossibleUserWarning
from pytorch_lightning.accelerators import Accelerator, HPUAccelerator, 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.loggers import Logger
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 Profiler
from pytorch_lightning.strategies import ParallelStrategy, Strategy
from pytorch_lightning.trainer import call, setup
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 _OUT_DICT, _PBAR_DICT, _ResultCollection
from pytorch_lightning.trainer.connectors.signal_connector import SignalConnector
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 AMPType, GradClipAlgorithmType, parsing
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.data import has_len_all_ranks
from pytorch_lightning.utilities.exceptions import ExitGracefullyException, MisconfigurationException
from pytorch_lightning.utilities.imports import _fault_tolerant_training
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,
_PREDICT_OUTPUT,
EVAL_DATALOADERS,
LRSchedulerConfig,
TRAIN_DATALOADERS,
)
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"
)
class Trainer:
@_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[_PATH] = 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,
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",
inference_mode: bool = True,
) -> 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.
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".
.. deprecated:: v1.8
Setting ``amp_level`` inside the ``Trainer`` is deprecated in v1.8.0 and will be removed
in v1.10.0. Please set it inside the specific precision plugin and pass it to the ``Trainer``.
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
or LightningDataModule depending on your setup.
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, local files (checkpoints, profiler traces, etc.) are saved in the ``log_dir`` of
the first logger.
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``.
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"``.
inference_mode: Whether to use :func:`torch.inference_mode` or :func:`torch.no_grad` during
evaluation (``validate``/``test``/``predict``).
"""
super().__init__()
Trainer._log_api_event("init")
log.detail(f"{self.__class__.__name__}: Initializing trainer with parameters: {locals()}")
self.state = TrainerState()
if default_root_dir is not None:
default_root_dir = os.fspath(default_root_dir)
# 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
# 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,
enable_model_summary,
max_time,
accumulate_grad_batches,
)
# init data flags
self.check_val_every_n_epoch: Optional[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: Optional[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._inference_mode: bool = inference_mode
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
setup._init_profiler(self, 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_batch: Union[int, float]
self.val_check_interval: Union[int, float]
self.num_sanity_val_steps: Union[int, float]
self.limit_train_batches: Union[int, float]
self.limit_val_batches: Union[int, float]
self.limit_test_batches: Union[int, float]
self.limit_predict_batches: Union[int, float]
setup._init_debugging_flags(
self,
limit_train_batches,
limit_val_batches,
limit_test_batches,
limit_predict_batches,
fast_dev_run,
overfit_batches,
val_check_interval,
num_sanity_val_steps,
)
def _setup_on_init(self) -> None:
setup._log_device_info(self)
self.should_stop = False
self.state = TrainerState()
self.num_training_batches = float("inf")
self.train_dataloader: Optional[Union[CombinedLoader, TRAIN_DATALOADERS]] = None
self.num_sanity_val_batches: List[Union[int, float]] = []
self.num_test_batches: List[Union[int, float]] = []
self.num_val_batches: List[Union[int, float]] = []
self.num_predict_batches: List[Union[int, float]] = []
self.test_dataloaders: Optional[List[DataLoader]] = None
self.val_dataloaders: Optional[List[DataLoader]] = None
self.predict_dataloaders: Optional[List[DataLoader]] = None
self._last_train_dl_reload_epoch = float("-inf")
self._last_val_dl_reload_epoch = float("-inf")
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. Could also be one of two special
keywords ``"last"`` and ``"hpc"``. 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`.
"""
if not isinstance(model, pl.LightningModule):
raise TypeError(f"`Trainer.fit()` requires a `LightningModule`, got: {model.__class__.__qualname__}")
self.strategy._lightning_module = model
call._call_and_handle_interrupt(
self, 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
# 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._checkpoint_connector._set_ckpt_path(
self.state.fn,
ckpt_path, # type: ignore[arg-type]
model_provided=True,
model_connected=self.lightning_module is not None,
)
self._run(model, ckpt_path=self.ckpt_path)
assert self.state.stopped
self.training = False
return
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"``, ``"last"``, ``"hpc"`` 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.
"""
if model is not None and not isinstance(model, pl.LightningModule):
raise TypeError(f"`Trainer.validate()` requires a `LightningModule`, got: {model.__class__.__qualname__}")
self.strategy._lightning_module = model or self.lightning_module
return call._call_and_handle_interrupt(
self, 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,
) -> Optional[Union[_PREDICT_OUTPUT, _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._checkpoint_connector._set_ckpt_path(
self.state.fn, 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"``, ``"last"``, ``"hpc"`` 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.
"""
if model is not None and not isinstance(model, pl.LightningModule):
raise TypeError(f"`Trainer.test()` requires a `LightningModule`, got: {model.__class__.__qualname__}")
self.strategy._lightning_module = model or self.lightning_module
return call._call_and_handle_interrupt(
self, 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,
) -> Optional[Union[_PREDICT_OUTPUT, _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._checkpoint_connector._set_ckpt_path(
self.state.fn, 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"``, ``"last"``, ``"hpc"`` 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.
See :ref:`Lightning inference section<deploy/production_basic:Predict step with your LightningModule>` for more.
"""
if model is not None and not isinstance(model, pl.LightningModule):
raise TypeError(f"`Trainer.predict()` requires a `LightningModule`, got: {model.__class__.__qualname__}")
self.strategy._lightning_module = model or self.lightning_module
return call._call_and_handle_interrupt(
self, 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 # type: ignore[assignment]
# 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._checkpoint_connector._set_ckpt_path(
self.state.fn, 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,
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,
method: Literal["fit", "validate", "test", "predict"] = "fit",
) -> _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.
dataloaders: A :class:`torch.utils.data.DataLoader` or a sequence of them specifying val/test/predict
samples used for running tuner on validation/testing/prediction.
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`
method: Method to run tuner on. It can be any of ``("fit", "validate", "test", "predict")``.
"""
if not isinstance(model, pl.LightningModule):
raise TypeError(f"`Trainer.tune()` requires a `LightningModule`, got: {model.__class__.__qualname__}")
Trainer._log_api_event("tune")
with isolate_rng():
result = self.tuner._tune(
model,
train_dataloaders,
val_dataloaders,
dataloaders,
datamodule,
scale_batch_size_kwargs=scale_batch_size_kwargs,
lr_find_kwargs=lr_find_kwargs,
method=method,
)
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 == TrainerFn.FITTING:
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
# ----------------------------
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}")