add rank warning (#1428)

* add rank warning

* changelog

* use rank_zero_warn

* user trainer_init

* replace warnings

* fix test

* flake8

* docs

* changelog

* bug lol

pytorch_lightning/callbacks/early_stopping.py

@@ -6,12 +6,11 @@ Stop training when a monitored quantity has stopped improving.
 
 """
 
-import warnings
-
 import numpy as np
 
 from pytorch_lightning import _logger as log
 from pytorch_lightning.callbacks.base import Callback
+from pytorch_lightning.utilities import rank_zero_warn
 
 
 class EarlyStopping(Callback):
@@ -80,7 +79,7 @@ class EarlyStopping(Callback):
             if self.strict:
                 raise RuntimeError(error_msg)
             if self.verbose > 0:
-                warnings.warn(error_msg, RuntimeWarning)
+                rank_zero_warn(error_msg, RuntimeWarning)
 
             return False
 
@@ -113,6 +112,6 @@ class EarlyStopping(Callback):
 
     def on_train_end(self, trainer, pl_module):
         if self.stopped_epoch > 0 and self.verbose > 0:
-            warnings.warn('Displayed epoch numbers by `EarlyStopping` start from "1" until v0.6.x,'
-                          ' but will start from "0" in v0.8.0.', DeprecationWarning)
+            rank_zero_warn('Displayed epoch numbers by `EarlyStopping` start from "1" until v0.6.x,'
+                           ' but will start from "0" in v0.8.0.', DeprecationWarning)
             log.info(f'Epoch {self.stopped_epoch + 1:05d}: early stopping')

pytorch_lightning/callbacks/gradient_accumulation_scheduler.py

@@ -6,9 +6,8 @@ Change gradient accumulation factor according to scheduling.
 
 """
 
-import warnings
-
 from pytorch_lightning.callbacks.base import Callback
+from pytorch_lightning.utilities import rank_zero_warn
 
 
 class GradientAccumulationScheduler(Callback):
@@ -46,8 +45,8 @@ class GradientAccumulationScheduler(Callback):
                 raise TypeError("All epoches and accumulation factor must be integers")
 
         minimal_epoch = min(scheduling.keys())
-        warnings.warn('Epochs indexing of `scheduling` starts from "1" until v0.6.x,'
-                      ' but will start from "0" in v0.8.0.', DeprecationWarning)
+        rank_zero_warn('Epochs indexing of `scheduling` starts from "1" until v0.6.x,'
+                       ' but will start from "0" in v0.8.0.', DeprecationWarning)
         if minimal_epoch < 1:
             msg = f"Epochs indexing from 1, epoch {minimal_epoch} cannot be interpreted correct"
             raise IndexError(msg)

pytorch_lightning/callbacks/model_checkpoint.py

@@ -7,14 +7,13 @@ Automatically save model checkpoints during training.
 """
 
 import os
-import shutil
-import warnings
 import re
 
 import numpy as np
 
-from pytorch_lightning.callbacks.base import Callback
 from pytorch_lightning import _logger as log
+from pytorch_lightning.callbacks.base import Callback
+from pytorch_lightning.utilities import rank_zero_warn
 
 
 class ModelCheckpoint(Callback):
@@ -83,7 +82,7 @@ class ModelCheckpoint(Callback):
                  mode: str = 'auto', period: int = 1, prefix: str = ''):
         super().__init__()
         if save_top_k > 0 and os.path.isdir(filepath) and len(os.listdir(filepath)) > 0:
-            warnings.warn(
+            rank_zero_warn(
                 f"Checkpoint directory {filepath} exists and is not empty with save_top_k != 0."
                 "All files in this directory will be deleted when a checkpoint is saved!"
             )
@@ -115,9 +114,7 @@ class ModelCheckpoint(Callback):
         }
 
         if mode not in mode_dict:
-            warnings.warn(
-                f'ModelCheckpoint mode {mode} is unknown, '
-                'fallback to auto mode.', RuntimeWarning)
+            rank_zero_warn(f'ModelCheckpoint mode {mode} is unknown, fallback to auto mode.', RuntimeWarning)
             mode = 'auto'
 
         self.monitor_op, self.kth_value, self.mode = mode_dict[mode]
@@ -206,7 +203,7 @@ class ModelCheckpoint(Callback):
             current = metrics.get(self.monitor)
 
             if current is None:
-                warnings.warn(f'Can save best model only with {self.monitor} available, skipping.', RuntimeWarning)
+                rank_zero_warn(f'Can save best model only with {self.monitor} available, skipping.', RuntimeWarning)
             elif self.check_monitor_top_k(current):
                 self._do_check_save(filepath, current, epoch)
             elif self.verbose > 0: