Update versionadded/versionchanged in master according to the latest releases (#2164)

* Update versionadded/versionchanged in master according to the latest releases

* Revert master version

* Update missing file

Co-authored-by: Taras Savchyn <[email protected]>

Author: Chandan-h-509

ignite/contrib/metrics/regression/fractional_absolute_error.py

@@ -34,7 +34,7 @@ class FractionalAbsoluteError(_BaseRegression):
             metric's device to be the same as your ``update`` arguments ensures the ``update`` method is
             non-blocking. By default, CPU.
 
-    .. versionchanged:: 0.5.0
+    .. versionchanged:: 0.4.5
         - Works with DDP.
     """
 

ignite/contrib/metrics/regression/fractional_bias.py

@@ -34,7 +34,7 @@ class FractionalBias(_BaseRegression):
             metric's device to be the same as your ``update`` arguments ensures the ``update`` method is
             non-blocking. By default, CPU.
 
-    .. versionchanged:: 0.5.0
+    .. versionchanged:: 0.4.5
         - Works with DDP.
     """
 

ignite/contrib/metrics/regression/geometric_mean_absolute_error.py

@@ -34,7 +34,7 @@ class GeometricMeanAbsoluteError(_BaseRegression):
             metric's device to be the same as your ``update`` arguments ensures the ``update`` method is
             non-blocking. By default, CPU.
 
-    .. versionchanged:: 0.5.0
+    .. versionchanged:: 0.4.5
         - Works with DDP.
     """
 

ignite/contrib/metrics/regression/maximum_absolute_error.py

@@ -34,7 +34,7 @@ class MaximumAbsoluteError(_BaseRegression):
             metric's device to be the same as your ``update`` arguments ensures the ``update`` method is
             non-blocking. By default, CPU.
 
-    .. versionchanged:: 0.5.0
+    .. versionchanged:: 0.4.5
         - Works with DDP.
     """
 

ignite/contrib/metrics/regression/mean_absolute_relative_error.py

@@ -34,7 +34,7 @@ class MeanAbsoluteRelativeError(_BaseRegression):
             metric's device to be the same as your ``update`` arguments ensures the ``update`` method is
             non-blocking. By default, CPU.
 
-    .. versionchanged:: 0.5.0
+    .. versionchanged:: 0.4.5
         - Works with DDP.
     """
 

ignite/contrib/metrics/regression/mean_normalized_bias.py

@@ -34,7 +34,7 @@ class MeanNormalizedBias(_BaseRegression):
             metric's device to be the same as your ``update`` arguments ensures the ``update`` method is
             non-blocking. By default, CPU.
 
-    .. versionchanged:: 0.5.0
+    .. versionchanged:: 0.4.5
         - Works with DDP.
     """
 

ignite/contrib/metrics/regression/wave_hedges_distance.py

@@ -33,7 +33,7 @@ class WaveHedgesDistance(_BaseRegression):
             metric's device to be the same as your ``update`` arguments ensures the ``update`` method is
             non-blocking. By default, CPU.
 
-    .. versionchanged:: 0.5.0
+    .. versionchanged:: 0.4.5
         - Works with DDP.
     """
 

ignite/distributed/comp_models/native.py

@@ -437,7 +437,7 @@ def _expand_hostlist(nodelist: str) -> List[str]:
         Args:
             nodelist: Compressed hostlist string
 
-        .. versionadded:: 0.5.1
+        .. versionadded:: 0.4.6
         """
         node_list = nodelist.split(", ")
 

ignite/distributed/launcher.py

@@ -208,7 +208,7 @@ def training(local_rank, config, **kwargs):
     .. versionchanged:: 0.4.2
         ``backend`` now accepts `horovod` distributed framework.
 
-    .. versionchanged:: 0.5.0
+    .. versionchanged:: 0.4.5
         ``init_method`` added.
 
     """

ignite/distributed/utils.py

@@ -414,7 +414,7 @@ def broadcast(
 
     .. versionadded:: 0.4.2
 
-    .. versionchanged:: 0.5.0
+    .. versionchanged:: 0.4.5
         added ``safe_mode``
     """
     if _need_to_sync and isinstance(_model, _SerialModel):
@@ -523,7 +523,7 @@ def train_fn(local_rank, a, b, c):
     .. versionchanged:: 0.4.2
         ``backend`` now accepts `horovod` distributed framework.
 
-    .. versionchanged:: 0.5.0
+    .. versionchanged:: 0.4.5
         ``kwargs`` now accepts ``init_method``, ``rank``, ``world_size`` for PyTorch native distributed backend.
     """
     if not (has_xla_support or has_native_dist_support or has_hvd_support):

ignite/engine/__init__.py

@@ -82,7 +82,7 @@ def supervised_training_step(
         update_fn = supervised_training_step(model, optimizer, loss_fn, 'cuda')
         trainer = Engine(update_fn)
 
-    .. versionadded:: 0.5.0
+    .. versionadded:: 0.4.5
     """
 
     def update(engine: Engine, batch: Sequence[torch.Tensor]) -> Union[Any, Tuple[torch.Tensor]]:
@@ -140,7 +140,7 @@ def supervised_training_step_amp(
         update_fn = supervised_training_step_amp(model, optimizer, loss_fn, 'cuda', scaler=scaler)
         trainer = Engine(update_fn)
 
-    .. versionadded:: 0.5.0
+    .. versionadded:: 0.4.5
     """
 
     try:
@@ -206,7 +206,7 @@ def supervised_training_step_apex(
         update_fn = supervised_training_step_apex(model, optimizer, loss_fn, 'cuda')
         trainer = Engine(update_fn)
 
-    .. versionadded:: 0.5.0
+    .. versionadded:: 0.4.5
     """
 
     try:
@@ -267,7 +267,7 @@ def supervised_training_step_tpu(
         update_fn = supervised_training_step_tpu(model, optimizer, loss_fn, 'xla')
         trainer = Engine(update_fn)
 
-    .. versionadded:: 0.5.0
+    .. versionadded:: 0.4.5
     """
     try:
         import torch_xla.core.xla_model as xm
@@ -380,7 +380,7 @@ def create_supervised_trainer(
 
         See more: https://nvidia.github.io/apex/amp.html#module-apex.amp
 
-    .. versionchanged:: 0.5.0
+    .. versionchanged:: 0.4.5
 
         - Added ``amp_mode`` argument for automatic mixed precision.
         - Added ``scaler`` argument for gradient scaling.
@@ -449,7 +449,7 @@ def supervised_evaluation_step(
         `device` will now *only* be used to move the input data to the correct device.
         The `model` should be moved by the user before creating an optimizer.
 
-    .. versionadded:: 0.5.0
+    .. versionadded:: 0.4.5
     """
 
     def evaluate_step(engine: Engine, batch: Sequence[torch.Tensor]) -> Union[Any, Tuple[torch.Tensor]]:
@@ -497,7 +497,7 @@ def supervised_evaluation_step_amp(
         `device` will now *only* be used to move the input data to the correct device.
         The `model` should be moved by the user before creating an optimizer.
 
-    .. versionadded:: 0.5.0
+    .. versionadded:: 0.4.5
     """
     try:
         from torch.cuda.amp import autocast
@@ -561,7 +561,7 @@ def create_supervised_evaluator(
 
         - `PyTorch's Explanation <https://github.com/pytorch/pytorch/issues/7844#issuecomment-503713840>`_
 
-    .. versionchanged:: 0.5.0
+    .. versionchanged:: 0.4.5
         - Added ``amp_mode`` argument for automatic mixed precision.
     """
     device_type = device.type if isinstance(device, torch.device) else device

ignite/handlers/ema_handler.py

@@ -114,7 +114,7 @@ def print_ema_momentum(engine):
 
               engine.run(...)
 
-    .. versionadded:: 0.5.0
+    .. versionadded:: 0.4.6
 
     """
 

ignite/handlers/lr_finder.py

@@ -70,6 +70,8 @@ class FastaiLRFinder:
         https://arxiv.org/abs/1506.01186
 
         fastai/lr_find: https://github.com/fastai/fastai
+
+    .. versionadded:: 0.4.6
     """
 
     def __init__(self) -> None:

ignite/handlers/param_scheduler.py

@@ -32,7 +32,7 @@ class ParamScheduler(metaclass=ABCMeta):
         More precisely, whatever the state of the optimizer (newly created or used by another scheduler) the scheduler
         sets defined absolute values.
 
-    .. versionadded:: 0.5.1
+    .. versionadded:: 0.4.5
     """
 
     def __init__(
@@ -251,7 +251,7 @@ class CyclicalScheduler(ParamScheduler):
         If the scheduler is bound to an 'ITERATION_*' event, 'cycle_size' should
         usually be the number of batches in an epoch.
 
-    .. versionadded:: 0.5.1
+    .. versionadded:: 0.4.5
     """
 
     def __init__(
@@ -340,7 +340,7 @@ class LinearCyclicalScheduler(CyclicalScheduler):
         # over the course of 1 epoch
         #
 
-    .. versionadded:: 0.5.1
+    .. versionadded:: 0.4.5
     """
 
     def get_param(self) -> float:
@@ -408,7 +408,7 @@ class CosineAnnealingScheduler(CyclicalScheduler):
     .. [Smith17] Smith, Leslie N. "Cyclical learning rates for training neural networks."
                  Applications of Computer Vision (WACV), 2017 IEEE Winter Conference on. IEEE, 2017
 
-    .. versionadded:: 0.5.1
+    .. versionadded:: 0.4.5
     """
 
     def get_param(self) -> float:
@@ -449,7 +449,7 @@ class ConcatScheduler(ParamScheduler):
         # The annealing cycles are repeated indefinitely.
         #
 
-    .. versionadded:: 0.5.1
+    .. versionadded:: 0.4.5
     """
 
     def __init__(self, schedulers: List[ParamScheduler], durations: List[int], save_history: bool = False):
@@ -675,7 +675,7 @@ class LRScheduler(ParamScheduler):
         # the first lr value from the optimizer, otherwise it is will be skipped:
         trainer.add_event_handler(Events.ITERATION_COMPLETED, scheduler)
 
-    .. versionadded:: 0.5.1
+    .. versionadded:: 0.4.5
     """
 
     def __init__(self, lr_scheduler: _LRScheduler, save_history: bool = False):
@@ -806,7 +806,7 @@ def create_lr_scheduler_with_warmup(
             # Attach to the trainer
             trainer.add_event_handler(Events.ITERATION_STARTED, scheduler)
 
-    .. versionadded:: 0.5.1
+    .. versionadded:: 0.4.5
     """
     if not isinstance(lr_scheduler, (ParamScheduler, _LRScheduler)):
         raise TypeError(
@@ -905,7 +905,7 @@ class PiecewiseLinear(ParamScheduler):
         # from 0.3 to 0.1 between 21st and 30th iterations and remains 0.1 until the end of the iterations.
         #
 
-    .. versionadded:: 0.5.1
+    .. versionadded:: 0.4.5
     """
 
     def __init__(
@@ -995,7 +995,7 @@ class ParamGroupScheduler:
         # Attach single scheduler to the trainer
         trainer.add_event_handler(Events.ITERATION_STARTED, scheduler)
 
-    .. versionadded:: 0.5.1
+    .. versionadded:: 0.4.5
     """
 
     def __init__(self, schedulers: List[ParamScheduler], names: Optional[List[str]] = None, save_history: bool = False):

ignite/handlers/stores.py

@@ -34,8 +34,8 @@ def log_training_results(engine):
             # output = [(y_pred0, y0), (y_pred1, y1), ...]
             # do something with output, e.g., plotting
 
-    .. versionadded:: 0.5.0
-    .. versionchanged:: 0.5.0
+    .. versionadded:: 0.4.5
+    .. versionchanged:: 0.4.5
         `attach` now accepts an optional argument `name`
     """
 

ignite/metrics/gan/fid.py

@@ -106,7 +106,7 @@ class FID(_BaseInceptionMetric):
             m.update((y_pred, y))
             print(m.compute())
 
-    .. versionadded:: 0.5.0
+    .. versionadded:: 0.4.6
     """
 
     def __init__(

ignite/metrics/gan/inception_score.py

@@ -60,7 +60,7 @@ class InceptionScore(_BaseInceptionMetric):
             m.update(images)
             print(m.compute())
 
-    .. versionadded:: 0.5.0
+    .. versionadded:: 0.4.6
     """
 
     def __init__(

ignite/metrics/metric.py

@@ -556,7 +556,7 @@ def sync_all_reduce(*attrs: Any) -> Callable:
     Args:
         attrs: attribute names of decorated class
 
-    .. versionchanged:: 0.5.0
+    .. versionchanged:: 0.4.5
         - Ability to handle different reduction operations (SUM, MAX, MIN, PRODUCT).
     """
 

ignite/metrics/multilabel_confusion_matrix.py

@@ -38,7 +38,7 @@ class MultiLabelConfusionMatrix(Metric):
             default, CPU.
         normalized: whether to normalize confusion matrix by its sum or not.
 
-    .. versionadded:: 0.5.0
+    .. versionadded:: 0.4.5
 
     """
 

ignite/metrics/nlp/bleu.py

@@ -109,7 +109,7 @@ class Bleu(Metric):
 
         print(m.compute())
 
-    .. versionadded:: 0.5.0
+    .. versionadded:: 0.4.5
     """
 
     def __init__(

ignite/metrics/nlp/rouge.py

@@ -44,7 +44,7 @@ def compute_ngram_scores(candidate: Sequence[Any], reference: Sequence[Any], n:
     Returns:
         The score containing the number of ngram co-occurences
 
-    .. versionadded:: 0.5.0
+    .. versionadded:: 0.4.5
     """
 
     # ngrams of the candidate
@@ -73,7 +73,7 @@ def compute_lcs_scores(candidate: Sequence[Any], reference: Sequence[Any]) -> Sc
     Returns:
         The score containing the length of longest common subsequence
 
-    .. versionadded:: 0.5.0
+    .. versionadded:: 0.4.5
     """
 
     # lcs of candidate and reference
@@ -228,7 +228,7 @@ class RougeN(_BaseRouge):
         m.compute()
         >>> {'Rouge-2-P': 0.5, 'Rouge-2-R': 0.4, 'Rouge-2-F': 0.4}
 
-    .. versionadded:: 0.5.0
+    .. versionadded:: 0.4.5
     """
 
     def __init__(
@@ -295,7 +295,7 @@ class RougeL(_BaseRouge):
        m.compute()
        >>> {'Rouge-L-P': 0.6, 'Rouge-L-R': 0.5, 'Rouge-L-F': 0.5}
 
-    .. versionadded:: 0.5.0
+    .. versionadded:: 0.4.5
     """
 
     def __init__(
@@ -357,7 +357,7 @@ class Rouge(Metric):
         m.compute()
         >>> {'Rouge-L-P': 0.6, 'Rouge-L-R': 0.5, 'Rouge-L-F': 0.5, 'Rouge-2-P': 0.5, 'Rouge-2-R': 0.4, 'Rouge-2-F': 0.4}
 
-    .. versionadded:: 0.5.0
+    .. versionadded:: 0.4.5
     """
 
     def __init__(

ignite/metrics/nlp/utils.py

@@ -15,7 +15,7 @@ def ngrams(sequence: Sequence[Any], n: int) -> Counter:
     Returns:
         A counter of ngram objects
 
-    .. versionadded:: 0.5.0
+    .. versionadded:: 0.4.5
     """
     return Counter([tuple(sequence[i : i + n]) for i in range(len(sequence) - n + 1)])
 
@@ -32,7 +32,7 @@ def lcs(seq_a: Sequence[Any], seq_b: Sequence[Any]) -> int:
     Returns:
         The length of the longest common subsequence
 
-    .. versionadded:: 0.5.0
+    .. versionadded:: 0.4.5
     """
     m = len(seq_a)
     n = len(seq_b)
@@ -73,7 +73,7 @@ def modified_precision(references: Sequence[Sequence[Any]], candidate: Any, n: i
     Returns:
         The length of the longest common subsequence
 
-    .. versionadded:: 0.5.0
+    .. versionadded:: 0.4.5
     """
     # ngrams of the candidate
     counts = ngrams(candidate, n)