replace all pyre-ignore with pyre-fixme

Summary: Consolidate all pyre annotations to pyre-fixme

Differential Revision: D53064137

tests/framework/callbacks/test_base_checkpointer.py

@@ -664,7 +664,7 @@ def test_best_checkpoint_no_top_k(self) -> None:
                     save_every_n_epochs=1,
                     best_checkpoint_config=BestCheckpointConfig(
                         monitored_metric="train_loss",
-                        # pyre-ignore: Incompatible parameter type [6]
+                        # pyre-fixme: Incompatible parameter type [6]
                         mode=mode,
                     ),
                 )

tests/utils/test_swa.py

@@ -191,7 +191,7 @@ def test_input_checks(self) -> None:
         with self.assertRaisesRegex(
             ValueError, "Unknown averaging method: foo. Only ema and swa are supported."
         ):
-            # pyre-ignore On purpose to test run time exception
+            # pyre-fixme On purpose to test run time exception
             AveragedModel(model, averaging_method="foo")
 
     def test_lit_ema(self) -> None:

torchtnt/framework/callbacks/base_csv_writer.py

@@ -59,7 +59,7 @@ def get_step_output_rows(
         self,
         state: State,
         unit: TPredictUnit,
-        # pyre-ignore: Missing parameter annotation [2]
+        # pyre-fixme: Missing parameter annotation [2]
         step_output: Any,
     ) -> Union[List[str], List[List[str]]]:
         ...

torchtnt/framework/callbacks/module_summary.py

@@ -43,7 +43,7 @@ def __init__(
         process_fn: Callable[
             [List[ModuleSummaryObj]], None
         ] = _log_module_summary_tables,
-        # pyre-ignore
+        # pyre-fixme
         module_inputs: Optional[
             MutableMapping[str, Tuple[Tuple[Any, ...], Dict[str, Any]]]
         ] = None,

torchtnt/framework/train.py

@@ -223,7 +223,7 @@ def _train_epoch_impl(
             ):
                 _evaluate_impl(
                     state,
-                    # pyre-ignore: Incompatible parameter type [6]
+                    # pyre-fixme: Incompatible parameter type [6]
                     train_unit,
                     callback_handler,
                 )
@@ -257,7 +257,7 @@ def _train_epoch_impl(
     ):
         _evaluate_impl(
             state,
-            # pyre-ignore: Incompatible parameter type [6]
+            # pyre-fixme: Incompatible parameter type [6]
             train_unit,
             callback_handler,
         )

torchtnt/utils/data/iterators.py

@@ -343,7 +343,7 @@ def __init__(
                 name: torch.IntTensor([idx])
                 for idx, name in enumerate(self._iterator_names)
             }
-            # pyre-ignore[4]: missing attribute annotation
+            # pyre-fixme[4]: missing attribute annotation
             self._process_group = dist.new_group(backend="gloo", ranks=None)
 
         self._iterators_finished: List[str] = []

torchtnt/utils/data/profile_dataloader.py

@@ -52,7 +52,7 @@ def profile_dataloader(
                 with timer.time("copy_data_to_device"), record_function(
                     "copy_data_to_device"
                 ):
-                    # pyre-ignore [6]: device is checked as not None before calling this
+                    # pyre-fixme [6]: device is checked as not None before calling this
                     data = copy_data_to_device(data, device)
 
             steps_completed += 1

torchtnt/utils/device.py

@@ -313,14 +313,14 @@ def collect_system_stats(device: torch.device) -> Dict[str, Any]:
     system_stats: Dict[str, Any] = {}
     cpu_stats = get_psutil_cpu_stats()
 
-    # pyre-ignore
+    # pyre-fixme
     system_stats.update(cpu_stats)
 
     if torch.cuda.is_available():
         try:
             gpu_stats = get_nvidia_smi_gpu_stats(device)
 
-            # pyre-ignore
+            # pyre-fixme
             system_stats.update(gpu_stats)
             system_stats.update(torch.cuda.memory_stats())
         except FileNotFoundError:

torchtnt/utils/flops.py

@@ -21,7 +21,7 @@
 aten: torch._ops._OpNamespace = torch.ops.aten
 
 
-# pyre-ignore [2] we don't care the type in outputs
+# pyre-fixme [2] we don't care the type in outputs
 def _matmul_flop_jit(inputs: Tuple[torch.Tensor], outputs: Tuple[Any]) -> Number:
     """
     Count flops for matmul.
@@ -35,7 +35,7 @@ def _matmul_flop_jit(inputs: Tuple[torch.Tensor], outputs: Tuple[Any]) -> Number
     return flop
 
 
-# pyre-ignore [2] we don't care the type in outputs
+# pyre-fixme [2] we don't care the type in outputs
 def _addmm_flop_jit(inputs: Tuple[torch.Tensor], outputs: Tuple[Any]) -> Number:
     """
     Count flops for fully connected layers.
@@ -53,7 +53,7 @@ def _addmm_flop_jit(inputs: Tuple[torch.Tensor], outputs: Tuple[Any]) -> Number:
     return flops
 
 
-# pyre-ignore [2] we don't care the type in outputs
+# pyre-fixme [2] we don't care the type in outputs
 def _bmm_flop_jit(inputs: Tuple[torch.Tensor], outputs: Tuple[Any]) -> Number:
     """
     Count flops for the bmm operation.
@@ -98,7 +98,7 @@ def _conv_flop_count(
 
 
 def _conv_flop_jit(
-    inputs: Tuple[Any],  # pyre-ignore [2] the inputs can be union of Tensor/bool/Tuple
+    inputs: Tuple[Any],  # pyre-fixme [2] the inputs can be union of Tensor/bool/Tuple
     outputs: Tuple[torch.Tensor],
 ) -> Number:
     """
@@ -118,7 +118,7 @@ def _transpose_shape(shape: torch.Size) -> List[int]:
     return [shape[1], shape[0]] + list(shape[2:])
 
 
-# pyre-ignore [2] the inputs can be union of Tensor/bool/Tuple & we don't care about outputs
+# pyre-fixme [2] the inputs can be union of Tensor/bool/Tuple & we don't care about outputs
 def _conv_backward_flop_jit(inputs: Tuple[Any], outputs: Tuple[Any]) -> Number:
     grad_out_shape, x_shape, w_shape = [i.shape for i in inputs[:3]]
     output_mask = inputs[-1]
@@ -127,7 +127,7 @@ def _conv_backward_flop_jit(inputs: Tuple[Any], outputs: Tuple[Any]) -> Number:
 
     if output_mask[0]:
         grad_input_shape = outputs[0].shape
-        # pyre-ignore [58] this is actually sum of Number and Number
+        # pyre-fixme [58] this is actually sum of Number and Number
         flop_count = flop_count + _conv_flop_count(
             grad_out_shape, w_shape, grad_input_shape, not fwd_transposed
         )
@@ -143,7 +143,7 @@ def _conv_backward_flop_jit(inputs: Tuple[Any], outputs: Tuple[Any]) -> Number:
     return flop_count
 
 
-# pyre-ignore [5]
+# pyre-fixme [5]
 flop_mapping: Dict[Callable[..., Any], Callable[[Tuple[Any], Tuple[Any]], Number]] = {
     aten.mm: _matmul_flop_jit,
     aten.matmul: _matmul_flop_jit,
@@ -163,7 +163,7 @@ def _conv_backward_flop_jit(inputs: Tuple[Any], outputs: Tuple[Any]) -> Number:
 }
 
 
-# pyre-ignore [2, 3] it can be Tuple of anything.
+# pyre-fixme [2, 3] it can be Tuple of anything.
 def _normalize_tuple(x: Any) -> Tuple[Any]:
     if not isinstance(x, tuple):
         return (x,)
@@ -213,33 +213,33 @@ def __init__(self, module: torch.nn.Module) -> None:
         )
         self._parents: List[str] = [""]
 
-    # pyre-ignore
+    # pyre-fixme
     def __exit__(self, exc_type, exc_val, exc_tb):
         for hook_handle in self._all_hooks:
             hook_handle.remove()
         super().__exit__(exc_type, exc_val, exc_tb)
 
     def __torch_dispatch__(
         self,
-        func: Callable[..., Any],  # pyre-ignore [2] func can be any func
-        types: Tuple[Any],  # pyre-ignore [2]
-        args=(),  # pyre-ignore [2]
-        kwargs=None,  # pyre-ignore [2]
+        func: Callable[..., Any],  # pyre-fixme [2] func can be any func
+        types: Tuple[Any],  # pyre-fixme [2]
+        args=(),  # pyre-fixme [2]
+        kwargs=None,  # pyre-fixme [2]
     ) -> PyTree:
         rs = func(*args, **kwargs)
         outs = _normalize_tuple(rs)
 
         if func in flop_mapping:
             flop_count = flop_mapping[func](args, outs)
             for par in self._parents:
-                # pyre-ignore [58]
+                # pyre-fixme [58]
                 self.flop_counts[par][func.__name__] += flop_count
         else:
             logging.debug(f"{func} is not yet supported in FLOPs calculation.")
 
         return rs
 
-    # pyre-ignore [3]
+    # pyre-fixme [3]
     def _create_backwards_push(self, name: str) -> Callable[..., Any]:
         class PushState(torch.autograd.Function):
             @staticmethod
@@ -262,7 +262,7 @@ def backward(ctx, *grad_outs):
         # using a function parameter.
         return PushState.apply
 
-    # pyre-ignore [3]
+    # pyre-fixme [3]
     def _create_backwards_pop(self, name: str) -> Callable[..., Any]:
         class PopState(torch.autograd.Function):
             @staticmethod
@@ -286,9 +286,9 @@ def backward(ctx, *grad_outs):
         # using a function parameter.
         return PopState.apply
 
-    # pyre-ignore [3] Return a callable function
+    # pyre-fixme [3] Return a callable function
     def _enter_module(self, name: str) -> Callable[..., Any]:
-        # pyre-ignore [2, 3]
+        # pyre-fixme [2, 3]
         def f(module: torch.nn.Module, inputs: Tuple[Any]):
             parents = self._parents
             parents.append(name)
@@ -298,9 +298,9 @@ def f(module: torch.nn.Module, inputs: Tuple[Any]):
 
         return f
 
-    # pyre-ignore [3] Return a callable function
+    # pyre-fixme [3] Return a callable function
     def _exit_module(self, name: str) -> Callable[..., Any]:
-        # pyre-ignore [2, 3]
+        # pyre-fixme [2, 3]
         def f(module: torch.nn.Module, inputs: Tuple[Any], outputs: Tuple[Any]):
             parents = self._parents
             assert parents[-1] == name

torchtnt/utils/memory.py

@@ -20,14 +20,14 @@
 
 
 def _is_named_tuple(
-    # pyre-ignore: Missing parameter annotation [2]: Parameter `x` must have a type other than `Any`.
+    # pyre-fixme: Missing parameter annotation [2]: Parameter `x` must have a type other than `Any`.
     x: Any,
 ) -> bool:
     return isinstance(x, tuple) and hasattr(x, "_asdict") and hasattr(x, "_fields")
 
 
 def get_tensor_size_bytes_map(
-    # pyre-ignore: Missing parameter annotation [2]: Parameter `obj` must have a type other than `Any`.
+    # pyre-fixme: Missing parameter annotation [2]: Parameter `obj` must have a type other than `Any`.
     obj: Any,
 ) -> Dict[torch.Tensor, int]:
     tensor_map = {}

torchtnt/utils/module_summary.py

@@ -209,9 +209,9 @@ def _clean_flops(flop: DefaultDict[str, DefaultDict[str, int]], N: int) -> None:
 
 def _get_module_flops_and_activation_sizes(
     module: torch.nn.Module,
-    # pyre-ignore
+    # pyre-fixme
     module_args: Optional[Tuple[Any, ...]] = None,
-    # pyre-ignore
+    # pyre-fixme
     module_kwargs: Optional[MutableMapping[str, Any]] = None,
 ) -> _ModuleSummaryData:
     # a mapping from module name to activation size tuple (in_size, out_size)
@@ -309,9 +309,9 @@ def _has_tensor(item: Optional[PyTree]) -> bool:
 
 def get_module_summary(
     module: torch.nn.Module,
-    # pyre-ignore
+    # pyre-fixme
     module_args: Optional[Tuple[Any, ...]] = None,
-    # pyre-ignore
+    # pyre-fixme
     module_kwargs: Optional[MutableMapping[str, Any]] = None,
 ) -> ModuleSummary:
     """
@@ -669,13 +669,13 @@ def _activation_size_hook(
     activation_sizes: Dict[
         str, Tuple[Union[TUnknown, List[int]], Union[TUnknown, List[int]]]
     ],
-    # pyre-ignore: Invalid type parameters [24]
+    # pyre-fixme: Invalid type parameters [24]
 ) -> Callable[[str], Callable]:
-    # pyre-ignore: Missing parameter annotation [2]
+    # pyre-fixme: Missing parameter annotation [2]
     def intermediate_hook(
         module_name: str,
     ) -> Callable[[torch.nn.Module, Any, Any], None]:
-        # pyre-ignore
+        # pyre-fixme
         def hook(_: torch.nn.Module, inp: Any, out: Any) -> None:
             if len(inp) == 1:
                 inp = inp[0]
@@ -690,9 +690,9 @@ def hook(_: torch.nn.Module, inp: Any, out: Any) -> None:
 
 def _forward_time_pre_hook(
     timer_mapping: Dict[str, float]
-    # pyre-ignore: Invalid type parameters [24]
+    # pyre-fixme: Invalid type parameters [24]
 ) -> Callable[[str], Callable]:
-    # pyre-ignore: Missing parameter annotation [2]
+    # pyre-fixme: Missing parameter annotation [2]
     def intermediate_hook(
         module_name: str,
     ) -> Callable[[torch.nn.Module, Any], None]:
@@ -707,9 +707,9 @@ def hook(_module: torch.nn.Module, _inp: Any) -> None:
 def _forward_time_hook(
     timer_mapping: Dict[str, float],
     elapsed_times: Dict[str, float],
-    # pyre-ignore: Invalid type parameters [24]
+    # pyre-fixme: Invalid type parameters [24]
 ) -> Callable[[str], Callable]:
-    # pyre-ignore: Missing parameter annotation [2]
+    # pyre-fixme: Missing parameter annotation [2]
     def intermediate_hook(
         module_name: str,
     ) -> Callable[[torch.nn.Module, Any, Any], None]:
@@ -725,7 +725,7 @@ def hook(_module: torch.nn.Module, _inp: Any, _out: Any) -> None:
 
 def _register_hooks(
     module: torch.nn.Module,
-    # pyre-ignore: Invalid type parameters [24]
+    # pyre-fixme: Invalid type parameters [24]
     hooks: List[Tuple[Callable, _HookType]],
 ) -> List[RemovableHandle]:
     """

torchtnt/utils/prepare_module.py

@@ -112,7 +112,7 @@ class TorchCompileParams:
 
     fullgraph: bool = False
     dynamic: bool = False
-    # pyre-ignore: Invalid type parameters [24]
+    # pyre-fixme: Invalid type parameters [24]
     backend: Union[str, Callable] = "inductor"
     mode: Union[str, None] = None
     options: Optional[Dict[str, Union[str, int, bool]]] = None

torchtnt/utils/swa.py

@@ -9,7 +9,7 @@
 import torch
 
 # TODO: torch/optim/swa_utils.pyi needs to be updated
-# pyre-ignore Undefined import [21]: Could not find a name `get_ema_multi_avg_fn` defined in module `torch.optim.swa_utils`.
+# pyre-fixme Undefined import [21]: Could not find a name `get_ema_multi_avg_fn` defined in module `torch.optim.swa_utils`.
 from torch.optim.swa_utils import (
     AveragedModel as PyTorchAveragedModel,
     get_ema_multi_avg_fn,
@@ -55,11 +55,11 @@ def __init__(
                 raise ValueError(f"Decay must be between 0 and 1, got {ema_decay}")
 
             # TODO: torch/optim/swa_utils.pyi needs to be updated
-            # pyre-ignore Undefined attribute [16]: Module `torch.optim.swa_utils` has no attribute `get_ema_multi_avg_fn`.
+            # pyre-fixme Undefined attribute [16]: Module `torch.optim.swa_utils` has no attribute `get_ema_multi_avg_fn`.
             multi_avg_fn = get_ema_multi_avg_fn(ema_decay)
         elif averaging_method == "swa":
             # TODO: torch/optim/swa_utils.pyi needs to be updated
-            # pyre-ignore Undefined attribute [16]: Module `torch.optim.swa_utils` has no attribute `get_swa_multi_avg_fn`.
+            # pyre-fixme Undefined attribute [16]: Module `torch.optim.swa_utils` has no attribute `get_swa_multi_avg_fn`.
             multi_avg_fn = get_swa_multi_avg_fn()
 
             if use_lit:
@@ -88,7 +88,7 @@ def __init__(
             # use default init implementation
 
             # TODO: torch/optim/swa_utils.pyi needs to be updated
-            # pyre-ignore Unexpected keyword [28]
+            # pyre-fixme Unexpected keyword [28]
             super().__init__(
                 model,
                 device=device,
@@ -104,6 +104,6 @@ def update_parameters(self, model: torch.nn.Module) -> None:
             )
 
             # TODO: torch/optim/swa_utils.pyi needs to be updated
-            # pyre-ignore Undefined attribute [16]: Module `torch.optim.swa_utils` has no attribute `get_ema_multi_avg_fn`.
+            # pyre-fixme Undefined attribute [16]: Module `torch.optim.swa_utils` has no attribute `get_ema_multi_avg_fn`.
             self.multi_avg_fn = get_ema_multi_avg_fn(decay)
         super().update_parameters(model)

torchtnt/utils/timer.py

@@ -364,7 +364,7 @@ def _sync_durations(
     pg_wrapper.all_gather_object(outputs, recorded_durations)
     ret = defaultdict(list)
     for output in outputs:
-        # pyre-ignore [16]: `Optional` has no attribute `__getitem__`.
+        # pyre-fixme [16]: `Optional` has no attribute `__getitem__`.
         for k, v in output.items():
             if k not in ret:
                 ret[k] = []