Fix Checkpointer cleaner regex on Windows (#5361)

* Fix to work on Windows

* Update CHANGELOG.md

* Update CHANGELOG.md

Co-authored-by: Pete <petew@allenai.org>
Co-authored-by: Pete <epwalsh10@gmail.com>

CHANGELOG.md

@@ -42,6 +42,7 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 - `TransformerTextField` can now take tensors of shape `(1, n)` like the tensors produced from a HuggingFace tokenizer.
 - `tqdm` lock is now set inside `MultiProcessDataLoading` when new workers are spawned to avoid contention when writing output.
 - `ConfigurationError` is now pickleable.
+- Checkpointer cleaning was fixed to work on Windows Paths
 - Multitask models now support `TextFieldTensor` in heads, not just in the backbone.
 - Fixed the signature of `ScaledDotProductAttention` to match the other `Attention` classes
 - Fixed the way names are applied to Tango `Step` instances.

allennlp/training/checkpointer.py

@@ -97,8 +97,8 @@ def _training_state_path(self, epochs_completed: int, batches_in_epoch_completed
         else:
             return path + ".th"
 
-    _model_state_file_re = re.compile(r"(.*/)?model_state_e(\d+)_b(\d+)(_w\d+)?\.th$")
-    _training_state_file_re = re.compile(r"(.*/)?training_state_e(\d+)_b(\d+)(_w\d+)?\.th$")
+    _model_state_file_re = re.compile(r"(.*[/\\])?model_state_e(\d+)_b(\d+)(_w\d+)?\.th$")
+    _training_state_file_re = re.compile(r"(.*[/\\])?training_state_e(\d+)_b(\d+)(_w\d+)?\.th$")
 
     @classmethod
     def _parse_model_state_path(cls, path: Union[str, os.PathLike]) -> Optional[Tuple[int, int]]: