Skip to content

Commit 1b8d8e1

Browse files
sdesrozissdesrozis
committed
fix quotes
1 parent c7c86c0 commit 1b8d8e1

File tree

1 file changed

+53
-53
lines changed

1 file changed

+53
-53
lines changed

‎ignite/handlers/checkpoint.py‎

Lines changed: 53 additions & 53 deletions
Original file line numberDiff line numberDiff line change
@@ -65,92 +65,92 @@ def remove(self, filename: str) -> None:
6565

6666
class Checkpoint(Serializable):
6767
"""Checkpoint handler can be used to periodically save and load objects which have attribute
68-
``state_dict`/`load_state_dict``. This class can use specific save handlers to store on the disk or a cloud
68+
`state_dict`/`load_state_dict`. This class can use specific save handlers to store on the disk or a cloud
6969
storage, etc. The Checkpoint handler (if used with :class:`~ignite.handlers.DiskSaver`) also handles automatically
7070
moving data on TPU to CPU before writing the checkpoint.
7171
7272
Args:
73-
to_save (Mapping): Dictionary with the objects to save. Objects should have implemented ``state_dict`` and
74-
``load_state_dict`` methods. If contains objects of type torch `DistributedDataParallel`_ or
75-
`DataParallel`_, their internal wrapped model is automatically saved (to avoid additional key ``module.`` in
73+
to_save (Mapping): Dictionary with the objects to save. Objects should have implemented `state_dict` and
74+
`load_state_dict` methods. If contains objects of type torch `DistributedDataParallel`_ or
75+
`DataParallel`_, their internal wrapped model is automatically saved (to avoid additional key `module.` in
7676
the state dictionary).
7777
save_handler (callable or :class:`~ignite.handlers.checkpoint.BaseSaveHandler`): Method or callable class to
7878
use to save engine and other provided objects. Function receives two objects: checkpoint as a dictionary
79-
and filename. If ``save_handler`` is callable class, it can
80-
inherit of :class:`~ignite.handlers.checkpoint.BaseSaveHandler` and optionally implement ``remove`` method
79+
and filename. If `save_handler` is callable class, it can
80+
inherit of :class:`~ignite.handlers.checkpoint.BaseSaveHandler` and optionally implement `remove` method
8181
to keep a fixed number of saved checkpoints. In case if user needs to save engine's checkpoint on a disk,
82-
``save_handler`` can be defined with :class:`~ignite.handlers.DiskSaver`.
82+
`save_handler` can be defined with :class:`~ignite.handlers.DiskSaver`.
8383
filename_prefix (str, optional): Prefix for the file name to which objects will be saved. See Note for details.
8484
score_function (callable, optional): If not None, it should be a function taking a single argument,
8585
:class:`~ignite.engine.engine.Engine` object, and returning a score (`float`). Objects with highest scores
8686
will be retained.
87-
score_name (str, optional): If ``score_function`` not None, it is possible to store its value using
88-
``score_name``. See Notes for more details.
87+
score_name (str, optional): If `score_function` not None, it is possible to store its value using
88+
`score_name`. See Notes for more details.
8989
n_saved (int, optional): Number of objects that should be kept on disk. Older files will be removed. If set to
9090
`None`, all objects are kept.
9191
global_step_transform (callable, optional): global step transform function to output a desired global step.
92-
Input of the function is ``(engine, event_name)``. Output of function should be an integer.
92+
Input of the function is `(engine, event_name)`. Output of function should be an integer.
9393
Default is None, global_step based on attached engine. If provided, uses function output as global_step.
9494
To setup global step from another engine, please use :meth:`~ignite.handlers.global_step_from_engine`.
95-
archived (bool, optional): Deprecated argument as models saved by ``torch.save`` are already compressed.
96-
filename_pattern (str, optional): If ``filename_pattern`` is provided, this pattern will be used to render
95+
archived (bool, optional): Deprecated argument as models saved by `torch.save` are already compressed.
96+
filename_pattern (str, optional): If `filename_pattern` is provided, this pattern will be used to render
9797
checkpoint filenames. If the pattern is not defined, the default pattern would be used. See Note for
9898
details.
9999
include_self (bool): Whether to include the `state_dict` of this object in the checkpoint. If `True`, then
100-
there must not be another object in ``to_save`` with key ``checkpointer``.
100+
there must not be another object in `to_save` with key `checkpointer`.
101101
102102
.. _DistributedDataParallel: https://pytorch.org/docs/stable/nn.html#torch.nn.parallel.DistributedDataParallel
103103
.. _DataParallel: https://pytorch.org/docs/stable/nn.html#torch.nn.DataParallel
104104
105105
Note:
106106
This class stores a single file as a dictionary of provided objects to save.
107-
The filename is defined by ``filename_pattern`` and by default has the following
108-
structure: ``{filename_prefix}_{name}_{suffix}.{ext}`` where
107+
The filename is defined by `filename_pattern` and by default has the following
108+
structure: `{filename_prefix}_{name}_{suffix}.{ext}` where
109109
110-
- ``filename_prefix`` is the argument passed to the constructor,
111-
- `name` is the key in ``to_save`` if a single object is to store, otherwise `name` is "checkpoint".
112-
- `suffix` is composed as following ``{global_step}_{score_name}={score}``.
110+
- `filename_prefix` is the argument passed to the constructor,
111+
- `name` is the key in `to_save` if a single object is to store, otherwise `name` is "checkpoint".
112+
- `suffix` is composed as following `{global_step}_{score_name}={score}`.
113113
114114
+----------------+------------+-----------------------+----------------------------------------------+
115115
| score_function | score_name | global_step_transform | suffix |
116116
+================+============+=======================+==============================================+
117-
| None | None | None | ``{engine.state.iteration}`` |
117+
| None | None | None | `{engine.state.iteration}` |
118118
+----------------+------------+-----------------------+----------------------------------------------+
119-
| X | None | None | ``{score}`` |
119+
| X | None | None | `{score}` |
120120
+----------------+------------+-----------------------+----------------------------------------------+
121-
| X | None | X | ``{global_step}_{score}`` |
121+
| X | None | X | `{global_step}_{score}` |
122122
+----------------+------------+-----------------------+----------------------------------------------+
123-
| X | X | X | ``{global_step}_{score_name}={score}`` |
123+
| X | X | X | `{global_step}_{score_name}={score}` |
124124
+----------------+------------+-----------------------+----------------------------------------------+
125-
| None | None | X | ``{global_step}`` |
125+
| None | None | X | `{global_step}` |
126126
+----------------+------------+-----------------------+----------------------------------------------+
127-
| X | X | None | ``{score_name}={score}`` |
127+
| X | X | None | `{score_name}={score}` |
128128
+----------------+------------+-----------------------+----------------------------------------------+
129129
130130
Above `global_step` defined by the output of `global_step_transform` and `score` defined by the output
131131
of `score_function`.
132132
133-
By default, none of ``score_function``, ``score_name``, ``global_step_transform`` is defined, then suffix is
133+
By default, none of `score_function`, `score_name`, `global_step_transform` is defined, then suffix is
134134
setup by attached engine's current iteration. The filename will be
135135
`{filename_prefix}_{name}_{engine.state.iteration}.{ext}`.
136136
137-
For example, ``score_name="neg_val_loss"`` and ``score_function`` that returns `-loss` (as objects with highest
138-
scores will be retained), then saved filename will be ``{filename_prefix}_{name}_neg_val_loss=-0.1234.pt``.
137+
For example, `score_name="neg_val_loss"` and `score_function` that returns `-loss` (as objects with highest
138+
scores will be retained), then saved filename will be `{filename_prefix}_{name}_neg_val_loss=-0.1234.pt`.
139139
140140
Note:
141-
If ``filename_pattern`` is given, it will be used to render the filenames. ``filename_pattern`` is a string
142-
that can contain ``{filename_prefix}``, ``{name}``, ``{score}``, ``{score_name}`` and ``{global_step}`` as
141+
If `filename_pattern` is given, it will be used to render the filenames. `filename_pattern` is a string
142+
that can contain `{filename_prefix}`, `{name}`, `{score}`, `{score_name}` and `{global_step}` as
143143
templates.
144144
145-
For example, let ``filename_pattern="{global_step}-{name}-{score}.pt"`` then the saved filename will be
146-
``30000-checkpoint-94.pt``
145+
For example, let `filename_pattern="{global_step}-{name}-{score}.pt"` then the saved filename will be
146+
`30000-checkpoint-94.pt`
147147
148148
**Warning:** Please, keep in mind that if filename collide with already used one to saved a checkpoint,
149-
new checkpoint will not be stored. This means that filename like ``checkpoint.pt`` will be saved only once
149+
new checkpoint will not be stored. This means that filename like `checkpoint.pt` will be saved only once
150150
and will not be overwritten by newer checkpoints.
151151
152152
Note:
153-
To get the last stored filename, handler exposes attribute ``last_checkpoint``:
153+
To get the last stored filename, handler exposes attribute `last_checkpoint`:
154154
155155
.. code-block:: python
156156
@@ -390,16 +390,16 @@ def setup_filename_pattern(
390390
"""Helper method to get the default filename pattern for a checkpoint.
391391
392392
Args:
393-
with_prefix (bool): If True, the ``filename_prefix`` is added to the filename pattern:
394-
``{filename_prefix}_{name}...``. Default, True.
395-
with_score (bool): If True, ``score`` is added to the filename pattern: ``..._{score}.{ext}``.
396-
Default, True. At least one of ``with_score`` and ``with_global_step`` should be True.
397-
with_score_name (bool): If True, ``score_name`` is added to the filename pattern:
398-
``..._{score_name}={score}.{ext}``. If activated, argument ``with_score`` should be
393+
with_prefix (bool): If True, the `filename_prefix` is added to the filename pattern:
394+
`{filename_prefix}_{name}...`. Default, True.
395+
with_score (bool): If True, `score` is added to the filename pattern: `..._{score}.{ext}`.
396+
Default, True. At least one of `with_score` and `with_global_step` should be True.
397+
with_score_name (bool): If True, `score_name` is added to the filename pattern:
398+
`..._{score_name}={score}.{ext}`. If activated, argument `with_score` should be
399399
also True, otherwise an error is raised. Default, True.
400-
with_global_step (bool): If True, ``{global_step}`` is added to the
401-
filename pattern: ``...{name}_{global_step}...``.
402-
At least one of ``with_score`` and ``with_global_step`` should be True.
400+
with_global_step (bool): If True, `{global_step}` is added to the
401+
filename pattern: `...{name}_{global_step}...`.
402+
At least one of `with_score` and `with_global_step` should be True.
403403
404404
Example:
405405
@@ -441,7 +441,7 @@ def _check_objects(objs: Mapping, attr: str) -> None:
441441

442442
@staticmethod
443443
def load_objects(to_load: Mapping, checkpoint: Mapping, **kwargs) -> None:
444-
"""Helper method to apply ``load_state_dict`` on the objects from ``to_load`` using states from ``checkpoint``.
444+
"""Helper method to apply `load_state_dict` on the objects from `to_load` using states from `checkpoint`.
445445
446446
Exemples:
447447
@@ -464,8 +464,8 @@ def load_objects(to_load: Mapping, checkpoint: Mapping, **kwargs) -> None:
464464
Checkpoint.load_objects(to_load=to_load, checkpoint=checkpoint)
465465
466466
Note:
467-
If ``to_load`` contains objects of type torch `DistributedDataParallel`_ or
468-
`DataParallel`_, method ``load_state_dict`` will applied to their internal wrapped model (``obj.module``).
467+
If `to_load` contains objects of type torch `DistributedDataParallel`_ or
468+
`DataParallel`_, method `load_state_dict` will applied to their internal wrapped model (`obj.module`).
469469
470470
Args:
471471
to_load (Mapping): a dictionary with objects, e.g. `{"model": model, "optimizer": optimizer, ...}`
@@ -523,9 +523,9 @@ class DiskSaver(BaseSaveHandler):
523523
atomic (bool, optional): if True, checkpoint is serialized to a temporary file, and then
524524
moved to final destination, so that files are guaranteed to not be damaged
525525
(for example if exception occurs during saving).
526-
create_dir (bool, optional): if True, will create directory ``dirname`` if it doesnt exist.
526+
create_dir (bool, optional): if True, will create directory `dirname` if it doesnt exist.
527527
require_empty (bool, optional): If True, will raise exception if there are any files in the
528-
directory ``dirname``.
528+
directory `dirname`.
529529
"""
530530

531531
def __init__(
@@ -617,9 +617,9 @@ class ModelCheckpoint(Checkpoint):
617617
618618
Behaviour of this class has been changed since v0.3.0.
619619
620-
Argument ``save_as_state_dict`` is deprecated and should not be used. It is considered as True.
620+
Argument `save_as_state_dict` is deprecated and should not be used. It is considered as True.
621621
622-
Argument ``save_interval`` is deprecated and should not be used. Please, use events filtering instead, e.g.
622+
Argument `save_interval` is deprecated and should not be used. Please, use events filtering instead, e.g.
623623
:attr:`~ignite.engine.events.Events.ITERATION_STARTED(every=1000)`
624624
625625
There is no more internal counter that has been used to indicate the number of save actions. User could
@@ -636,23 +636,23 @@ class ModelCheckpoint(Checkpoint):
636636
score_function (callable, optional): if not None, it should be a function taking a single argument, an
637637
:class:`~ignite.engine.engine.Engine` object, and return a score (`float`). Objects with highest scores
638638
will be retained.
639-
score_name (str, optional): if ``score_function`` not None, it is possible to store its value using
639+
score_name (str, optional): if `score_function` not None, it is possible to store its value using
640640
`score_name`. See Notes for more details.
641641
n_saved (int, optional): Number of objects that should be kept on disk. Older files will be removed. If set to
642642
`None`, all objects are kept.
643643
atomic (bool, optional): If True, objects are serialized to a temporary file, and then moved to final
644644
destination, so that files are guaranteed to not be damaged (for example if exception
645645
occurs during saving).
646646
require_empty (bool, optional): If True, will raise exception if there are any files starting with
647-
``filename_prefix`` in the directory ``dirname``.
648-
create_dir (bool, optional): If True, will create directory ``dirname`` if it does not exist.
647+
`filename_prefix` in the directory `dirname`.
648+
create_dir (bool, optional): If True, will create directory `dirname` if it does not exist.
649649
global_step_transform (callable, optional): global step transform function to output a desired global step.
650650
Input of the function is `(engine, event_name)`. Output of function should be an integer.
651651
Default is None, global_step based on attached engine. If provided, uses function output as global_step.
652652
To setup global step from another engine, please use :meth:`~ignite.handlers.global_step_from_engine`.
653653
archived (bool, optional): Deprecated argument as models saved by `torch.save` are already compressed.
654654
include_self (bool): Whether to include the `state_dict` of this object in the checkpoint. If `True`, then
655-
there must not be another object in ``to_save`` with key ``checkpointer``.
655+
there must not be another object in `to_save` with key `checkpointer`.
656656
657657
Examples:
658658
>>> import os

0 commit comments

Comments
 (0)