[Docs] update docs for resume_from_checkpoint (#5164)

* update docs and add pathlib support

* fix

(cherry picked from commit dd442b6)

docs/source/trainer.rst

@@ -1328,7 +1328,8 @@ resume_from_checkpoint
 
 |
 
-To resume training from a specific checkpoint pass in the path here.
+To resume training from a specific checkpoint pass in the path here. If resuming from a mid-epoch
+checkpoint, training will start from the beginning of the next epoch.
 
 .. testcode::
 

pytorch_lightning/trainer/connectors/checkpoint_connector.py

@@ -15,7 +15,7 @@
 import os
 import re
 from pathlib import Path
-from typing import Union, Optional
+from typing import Optional, Union
 
 import torch
 
@@ -24,8 +24,8 @@
 from pytorch_lightning.utilities import _APEX_AVAILABLE, AMPType, _OMEGACONF_AVAILABLE, rank_zero_info, rank_zero_warn
 from pytorch_lightning.utilities.cloud_io import atomic_save, get_filesystem
 from pytorch_lightning.utilities.cloud_io import load as pl_load
-from pytorch_lightning.utilities.upgrade_checkpoint import KEYS_MAPPING as DEPRECATED_CHECKPOINT_KEYS
 from pytorch_lightning.utilities.exceptions import MisconfigurationException
+from pytorch_lightning.utilities.upgrade_checkpoint import KEYS_MAPPING as DEPRECATED_CHECKPOINT_KEYS
 
 if _APEX_AVAILABLE:
     from apex import amp
@@ -156,9 +156,10 @@ def restore_training_state(self, checkpoint):
         expected_steps = self.trainer.num_training_batches / n_accum
         if self.trainer.num_training_batches != 0 and self.trainer.global_step % expected_steps > 1:
             rank_zero_warn(
-                "You're resuming from a checkpoint that ended mid-epoch. "
-                "This can cause unreliable results if further training is done, "
-                "consider using an end of epoch checkpoint. "
+                "You're resuming from a checkpoint that ended mid-epoch."
+                " Training will start from the beginning of the next epoch."
+                " This can cause unreliable results if further training is done,"
+                " consider using an end of epoch checkpoint."
             )
 
         # restore the optimizers

pytorch_lightning/trainer/trainer.py

@@ -16,6 +16,7 @@
 
 import os
 import warnings
+from pathlib import Path
 from typing import Dict, Iterable, List, Optional, Union
 
 import torch
@@ -117,7 +118,7 @@ def __init__(
         weights_save_path: Optional[str] = None,
         num_sanity_val_steps: int = 2,
         truncated_bptt_steps: Optional[int] = None,
-        resume_from_checkpoint: Optional[str] = None,
+        resume_from_checkpoint: Optional[Union[Path, str]] = None,
         profiler: Optional[Union[BaseProfiler, bool, str]] = None,
         benchmark: bool = False,
         deterministic: bool = False,
@@ -252,7 +253,8 @@ def __init__(
                 you can set ``replace_sampler_ddp=False`` and add your own distributed sampler.
 
             resume_from_checkpoint: To resume training from a specific checkpoint pass in the path here.
-                This can be a URL.
+                This can be a URL. If resuming from mid-epoch checkpoint, training will start from
+                the beginning of the next epoch.
 
             sync_batchnorm: Synchronize batch norm layers between process groups/whole world.