Fix typos under torch/utils directory (pytorch#97516)

This PR fixes typos in comments and messages of `.py` files under `torch/utils` directory

Pull Request resolved: pytorch#97516
Approved by: https://github.com/ezyang

torch/utils/benchmark/utils/fuzzer.py

@@ -216,7 +216,7 @@ def __init__(
             min_elements:
                 The minimum number of parameters that this Tensor must have for a
                 set of parameters to be valid. (Otherwise they are resampled.)
-            max_elemnts:
+            max_elements:
                 Like `min_elements`, but setting an upper bound.
             max_allocation_bytes:
                 Like `max_elements`, but for the size of Tensor that must be

torch/utils/checkpoint.py

@@ -413,7 +413,7 @@ def forward(input):
 # with retain_graph=True, we would store recomputed variables as the values of a
 # WeakKeyDictionary and pack strong references to the keys, so that as we
 # backward, those packed keys would be cleared as long as retain_graph=False.
-# Clearing the packed key clears the corresonding entry in the WKD.
+# Clearing the packed key clears the corresponding entry in the WKD.
 #
 # If we wish recomputed variables to be immediately cleared as we unpack them in
 # the retain_graph=True case, we cannot rely on the packed keys to be cleared by

torch/utils/cpp_extension.py

@@ -299,7 +299,7 @@ def check_compiler_ok_for_platform(compiler: str) -> bool:
     env['LC_ALL'] = 'C'  # Don't localize output
     version_string = subprocess.check_output([compiler, '-v'], stderr=subprocess.STDOUT, env=env).decode(*SUBPROCESS_DECODE_ARGS)
     if IS_LINUX:
-        # Check for 'gcc' or 'g++' for sccache warpper
+        # Check for 'gcc' or 'g++' for sccache wrapper
         pattern = re.compile("^COLLECT_GCC=(.*)$", re.MULTILINE)
         results = re.findall(pattern, version_string)
         if len(results) != 1:
@@ -1906,7 +1906,7 @@ def _run_ninja_build(build_directory: str, verbose: bool, error_prefix: str) ->
         _, error, _ = sys.exc_info()
         # error.output contains the stdout and stderr of the build attempt.
         message = error_prefix
-        # `error` is a CalledProcessError (which has an `ouput`) attribute, but
+        # `error` is a CalledProcessError (which has an `output`) attribute, but
         # mypy thinks it's Optional[BaseException] and doesn't narrow
         if hasattr(error, 'output') and error.output:  # type: ignore[union-attr]
             message += f": {error.output.decode(*SUBPROCESS_DECODE_ARGS)}"  # type: ignore[union-attr]

torch/utils/data/dataloader.py

@@ -428,7 +428,7 @@ def __setattr__(self, attr, val):
     # since '_BaseDataLoaderIter' references 'DataLoader'.
     def __iter__(self) -> '_BaseDataLoaderIter':
         # When using a single worker the returned iterator should be
-        # created everytime to avoid reseting its state
+        # created everytime to avoid resetting its state
         # However, in the case of a multiple workers iterator
         # the iterator is only created once in the lifetime of the
         # DataLoader object so that workers can be reused
@@ -545,7 +545,7 @@ def _create_warning_msg(num_worker_suggest, num_worker_created, cpuset_checked):
                 pass
         if max_num_worker_suggest is None:
             # os.cpu_count() could return Optional[int]
-            # get cpu count first and check None in order to satify mypy check
+            # get cpu count first and check None in order to satisfy mypy check
             cpu_count = os.cpu_count()
             if cpu_count is not None:
                 max_num_worker_suggest = cpu_count
@@ -733,7 +733,7 @@ class _MultiProcessingDataLoaderIter(_BaseDataLoaderIter):
     #
     #      First of all, `__del__` is **not** guaranteed to be called when
     #      interpreter exits. Even if it is called, by the time it executes,
-    #      many Python core library resources may alreay be freed, and even
+    #      many Python core library resources may already be freed, and even
     #      simple things like acquiring an internal lock of a queue may hang.
     #      Therefore, in this case, we actually need to prevent `__del__` from
     #      being executed, and rely on the automatic termination of daemonic
@@ -978,7 +978,7 @@ class _MultiProcessingDataLoaderIter(_BaseDataLoaderIter):
     #        NOTE: (c) is better placed after (b) because it may leave corrupted
     #              data in `worker_result_queue`, which `pin_memory_thread`
     #              reads from, in which case the `pin_memory_thread` can only
-    #              happen at timeing out, which is slow. Nonetheless, same thing
+    #              happen at timing out, which is slow. Nonetheless, same thing
     #              happens if a worker is killed by signal at unfortunate times,
     #              but in other cases, we are better off having a non-corrupted
     #              `worker_result_queue` for `pin_memory_thread`.

torch/utils/data/datapipes/dataframe/dataframes.py

@@ -217,7 +217,7 @@ def __str__(self):
 
     def execute(self):
 
-        # TODO: VitalyFedyunin execute kwargs and maybe nestted structures
+        # TODO: VitalyFedyunin execute kwargs and maybe nested structures
         executed_args = []
         for arg in self.kwargs['args']:
             if isinstance(arg, Capture):

torch/utils/data/datapipes/iter/combining.py

@@ -98,7 +98,7 @@ def __new__(
         copy: Optional[Literal["shallow", "deep"]] = None
     ):
         if num_instances < 1:
-            raise ValueError(f"Expected `num_instaces` larger than 0, but {num_instances} is found")
+            raise ValueError(f"Expected `num_instances` larger than 0, but {num_instances} is found")
         if num_instances == 1:
             return datapipe
         container = _ForkerIterDataPipe(datapipe, num_instances, buffer_size, copy)
@@ -191,7 +191,7 @@ def get_next_element_by_instance(self, instance_id: int):
                 if self.buffer and self.child_pointers[instance_id] <= self.leading_ptr:
                     idx = self.child_pointers[instance_id] - self.slowest_ptr - 1
                     return_val = self.buffer[idx]
-                else:  # Retreive one element from main datapipe
+                else:  # Retrieve one element from main datapipe
                     self.leading_ptr = self.child_pointers[instance_id]
                     try:
                         return_val = next(self._datapipe_iterator)  # type: ignore[arg-type]
@@ -385,7 +385,7 @@ class DemultiplexerIterDataPipe(IterDataPipe):
     def __new__(cls, datapipe: IterDataPipe, num_instances: int,
                 classifier_fn: Callable[[T_co], Optional[int]], drop_none: bool = False, buffer_size: int = 1000):
         if num_instances < 1:
-            raise ValueError(f"Expected `num_instaces` larger than 0, but {num_instances} is found")
+            raise ValueError(f"Expected `num_instances` larger than 0, but {num_instances} is found")
 
         _check_unpickable_fn(classifier_fn)
 

torch/utils/file_baton.py

@@ -11,7 +11,7 @@ def __init__(self, lock_file_path, wait_seconds=0.1):
 
         Args:
             lock_file_path: The path to the file used for locking.
-            wait_seconds: The seconds to periorically sleep (spin) when
+            wait_seconds: The seconds to periodically sleep (spin) when
                 calling ``wait()``.
         '''
         self.lock_file_path = lock_file_path