[codemod] Add type: ignore comments everywhere

ghstack-source-id: 7a55ad1e8187a1fc55954fdab088b5ce5cdb4c1d
Pull Request resolved: pytorch#57995

caffe2/contrib/aten/aten_test.py

@@ -25,7 +25,7 @@ def ref(X, Y):
             return [X + Y]
         self.assertReferenceChecks(gc, op, inputs, ref)
 
-    @given(inputs=hu.tensors(n=2, dtype=np.float16), **hu.gcs_gpu_only)
+    @given(inputs=hu.tensors(n=2, dtype=np.float16), **hu.gcs_gpu_only)  # type: ignore[arg-type]
     def test_add_half(self, inputs, gc, dc):
         op = core.CreateOperator(
             "ATen",

caffe2/contrib/gloo/gloo_test.py

@@ -17,7 +17,7 @@
 
 from caffe2.python import core, workspace, dyndep
 import caffe2.python.hypothesis_test_util as hu
-from gloo.python import IoError
+from gloo.python import IoError  # type: ignore[import]
 
 dyndep.InitOpsLibrary("@/caffe2/caffe2/distributed:file_store_handler_ops")
 dyndep.InitOpsLibrary("@/caffe2/caffe2/distributed:redis_store_handler_ops")
@@ -42,7 +42,7 @@ class TestCase(hu.HypothesisTestCase):
 
     def run_test_locally(self, fn, device_option=None, **kwargs):
         # Queue for assertion errors on subprocesses
-        queue = Queue()
+        queue = Queue()  # type: ignore[var-annotated]
 
         # Capture any exception thrown by the subprocess
         def run_fn(*args, **kwargs):
@@ -87,8 +87,8 @@ def run_test_distributed(self, fn, device_option=None, **kwargs):
         self.assertIsNotNone(comm_rank)
         comm_size = os.getenv('COMM_SIZE')
         self.assertIsNotNone(comm_size)
-        kwargs['comm_rank'] = int(comm_rank)
-        kwargs['comm_size'] = int(comm_size)
+        kwargs['comm_rank'] = int(comm_rank)  # type: ignore[arg-type]
+        kwargs['comm_size'] = int(comm_size)  # type: ignore[arg-type]
         with core.DeviceScope(device_option):
             fn(**kwargs)
             workspace.ResetWorkspace()

caffe2/contrib/nccl/nccl_ops_test.py

@@ -61,7 +61,7 @@ def allreduce(*args):
             allreduce, input_device_options)
         for output in outputs:
             np.testing.assert_array_equal(outputs[0], output)
-            self.assertEqual(outputs[0].tobytes(), output.tobytes())
+            self.assertEqual(outputs[0].tobytes(), output.tobytes())  # type: ignore[attr-defined]
 
     @given(n=st.integers(min_value=2, max_value=workspace.NumGpuDevices()),
            m=st.integers(min_value=1, max_value=1000),
@@ -122,7 +122,7 @@ def allgather(*args):
             allgather, input_device_options)
         for output in outputs:
             np.testing.assert_array_equal(outputs[0], output)
-            self.assertEqual(outputs[0].tobytes(), output.tobytes())
+            self.assertEqual(outputs[0].tobytes(), output.tobytes())  # type: ignore[attr-defined]
 
     @given(n=st.integers(min_value=2, max_value=workspace.NumGpuDevices()),
            m=st.integers(min_value=1, max_value=1000))
@@ -136,8 +136,8 @@ def test_nccl_reduce_scatter(self, n, m):
         def reduce_scatter(*args):
             assert len(args) == n
             reduced = sum(args)
-            assert len(reduced.shape) > 1
-            ref = [reduced[i, :] for i in range(n)]
+            assert len(reduced.shape) > 1  # type: ignore[union-attr]
+            ref = [reduced[i, :] for i in range(n)]  # type: ignore[index]
             return ref
 
         self.assertReferenceChecks(

caffe2/contrib/playground/AnyExp.py

@@ -209,7 +209,7 @@ def fun_per_epoch_aftRunNet(self, epoch):
 
     def checkpoint(self, epoch):
         self.model_path = checkpoint.save_model_params(
-            True, self.train_model, self.gen_checkpoint_path(True, epoch + 1),
+            True, self.train_model, self.gen_checkpoint_path(True, epoch + 1),  # type: ignore[attr-defined]
             epoch + 1, self.opts, float('-inf'))
 
     def gen_checkpoint_path(self, is_checkpoint, epoch):
@@ -273,11 +273,11 @@ def run_training_net(self):
 
     @abstractmethod
     def run_testing_net(self):
-        if self.test_model is None:
+        if self.test_model is None:  # type: ignore[attr-defined]
             return
         timeout = 2000.0
         with timeout_guard.CompleteInTimeOrDie(timeout):
-            workspace.RunNet(self.test_model.net.Proto().name)
+            workspace.RunNet(self.test_model.net.Proto().name)  # type: ignore[attr-defined]
 
     # @abstractmethod
     def planning_output(self):
@@ -286,9 +286,9 @@ def planning_output(self):
         self.init_logs()
 
     def prep_data_parallel_models(self):
-        self.prep_a_data_parallel_model(self.train_model,
+        self.prep_a_data_parallel_model(self.train_model,  # type: ignore[attr-defined]
                                         self.train_dataset, True)
-        self.prep_a_data_parallel_model(self.test_model,
+        self.prep_a_data_parallel_model(self.test_model,  # type: ignore[attr-defined]
                                         self.test_dataset, False)
 
     def prep_a_data_parallel_model(self, model, dataset, is_train):
@@ -359,7 +359,7 @@ def loadCheckpoint(self):
                 ))
                 start_epoch, prev_checkpointed_lr, _best_metric = \
                     checkpoint.initialize_params_from_file(
-                        model=self.train_model,
+                        model=self.train_model,  # type: ignore[attr-defined]
                         weights_file=previous_checkpoint,
                         num_xpus=num_xpus,
                         opts=opts,
@@ -370,15 +370,15 @@ def loadCheckpoint(self):
             log.info("Load pretrained model: {}".format(pretrained_model))
             start_epoch, prev_checkpointed_lr, best_metric = \
                 checkpoint.initialize_params_from_file(
-                    model=self.train_model,
+                    model=self.train_model,  # type: ignore[attr-defined]
                     weights_file=pretrained_model,
                     num_xpus=num_xpus,
                     opts=opts,
                     broadcast_computed_param=True,
                     reset_epoch=opts['model_param']['reset_epoch'],
                 )
 
-        data_parallel_model.FinalizeAfterCheckpoint(self.train_model)
+        data_parallel_model.FinalizeAfterCheckpoint(self.train_model)  # type: ignore[attr-defined]
 
     def buildModelAndTrain(self, opts):
         log.info('in buildModelAndTrain, trainer_input: {}'.format(str(opts)))
@@ -424,7 +424,7 @@ def buildModelAndTrain(self, opts):
 
                 self.fun_per_iter_b4RunNet(epoch, epoch_iter)
 
-                if self.train_model is not None:
+                if self.train_model is not None:  # type: ignore[attr-defined]
                     self.run_training_net()
 
                 self.fun_per_iter_aftRunNetB4Test(epoch, epoch_iter)
@@ -484,7 +484,7 @@ def buildModelAndTrain(self, opts):
 
             self.fun_per_epoch_aftRunNet(epoch)
 
-        self.fun_conclude_operator()
+        self.fun_conclude_operator()  # type: ignore[call-arg]
 
         self.createMetricsPlotsModelsOutputs()
 

caffe2/contrib/playground/AnyExpOnTerm.py

@@ -35,14 +35,14 @@ def runShardedTrainLoop(opts, myTrainFun):
     ret = None
 
     pretrained_model = ""
-    shard_results = []
+    shard_results = []  # type: ignore[var-annotated]
 
     for epoch in range(start_epoch,
                        opts['epoch_iter']['num_epochs'],
                        opts['epoch_iter']['num_epochs_per_flow_schedule']):
         # must support checkpoint or the multiple schedule will always
         # start from initial state
-        checkpoint_model = None if epoch == start_epoch else ret['model']
+        checkpoint_model = None if epoch == start_epoch else ret['model']  # type: ignore[index]
         pretrained_model = None if epoch > start_epoch else pretrained_model
         shard_results = []
         # with LexicalContext('epoch{}_gang'.format(epoch),gang_schedule=False):

caffe2/contrib/playground/checkpoint.py

@@ -29,7 +29,7 @@ def initialize_params_from_file(
 def initialize_master_xpu_model_params(model, weights_file, opts, reset_epoch):
     log.info("Initializing model params from file: {}".format(weights_file))
     with open(weights_file, 'r') as fopen:
-        blobs = pickle.load(fopen)
+        blobs = pickle.load(fopen)  # type: ignore[arg-type]
     if 'blobs' in blobs:
         blobs = blobs['blobs']
 
@@ -161,7 +161,7 @@ def save_model_params_blob(model, params_file, epoch, opts, best_metric):
     log.info('to weights file {}'.format(params_file))
     try:
         with open(params_file, 'w') as fwrite:
-            pickle.dump(dict(blobs=save_blobs), fwrite, pickle.HIGHEST_PROTOCOL)
+            pickle.dump(dict(blobs=save_blobs), fwrite, pickle.HIGHEST_PROTOCOL)  # type: ignore[arg-type]
     except IOError as e:
         log.error('I/O error({0}): {1}'.format(e.errno, e.strerror))
 

caffe2/contrib/tensorboard/tensorboard_exporter.py

@@ -44,8 +44,8 @@ def _convert_to_ssa(shapes, track_blob_names, ops):
     I.e. blobs will be renamed so that each blob is produced only once.
     """
     ir = core.IR(ops)
-    seen = set()
-    versioned = {}
+    seen = set()  # type: ignore[var-annotated]
+    versioned = {}  # type: ignore[var-annotated]
     shapes2 = {}
     track_blob_names2 = {}
 
@@ -98,8 +98,8 @@ def _remap_keys(m, f):
 
 
 def _rename_all(shapes, track_blob_names, ops, f):
-    seen = set()
-    renamed = {}
+    seen = set()  # type: ignore[var-annotated]
+    renamed = {}  # type: ignore[var-annotated]
 
     def g(name):
         """ Collision-free version of f.
@@ -288,7 +288,7 @@ def _operators_to_graph_def(
         _add_gradient_scope(shapes, track_blob_names, ops)
     _fill_missing_operator_names(ops)
     g = GraphDef()
-    producing_ops = {}
+    producing_ops = {}  # type: ignore[var-annotated]
     blobs = set()
     for op in ops:
         g.node.extend([_operator_to_node(shapes, op)])

caffe2/contrib/tensorboard/tensorboard_exporter_test.py

@@ -674,7 +674,7 @@ def test_simple_cnnmodel(self):
         model.net.RunAllOnGPU()
         model.param_init_net.RunAllOnGPU()
         model.AddGradientOperators([loss], skip=1)
-        track_blob_names = {}
+        track_blob_names = {}  # type: ignore[var-annotated]
         graph = tb.cnn_to_graph_def(
             model,
             track_blob_names=track_blob_names,

caffe2/core/nomnigraph/op_gen.py

@@ -19,7 +19,7 @@ def parse_lines(lines):
 
     # Preprocess the macros
     curr_macro = ""
-    macros = {}
+    macros = {}  # type: ignore[var-annotated]
 
     index = 0
     while index < len(lines):
@@ -60,7 +60,7 @@ def parse_lines(lines):
     curr_op = ""
     # dict of the form
     #  opName : { attributes: [], ... }
-    ops = {}
+    ops = {}  # type: ignore[var-annotated]
     # To preserve parsing order for dependencies (for things like init_from)
     op_list = []
 

caffe2/distributed/store_ops_test_util.py

@@ -48,7 +48,7 @@ def _test_set_get(cls, queue, create_store_handler_fn, index, num_procs):
     @classmethod
     def test_set_get(cls, create_store_handler_fn):
         # Queue for assertion errors on subprocesses
-        queue = Queue()
+        queue = Queue()  # type: ignore[var-annotated]
 
         # Start N processes in the background
         num_procs = 4

caffe2/experiments/python/SparseTransformer.py

@@ -175,7 +175,7 @@ def net2list(net_root):
 def netbuilder(model):
     print("Welcome to model checker")
     proto = model.net.Proto()
-    net_name2id = {}
+    net_name2id = {}  # type: ignore[var-annotated]
     net_id2node = {}
     net_root = NetDefNode("net_root", "root", None)
 

caffe2/experiments/python/device_reduce_sum_bench.py

@@ -68,7 +68,7 @@ def run(self):
             ["y"]
         )
 
-        for n in itertools.imap(pow, itertools.cycle([10]), range(10)):
+        for n in itertools.imap(pow, itertools.cycle([10]), range(10)):  # type: ignore[attr-defined]
             X = np.random.rand(n).astype(np.float32)
             logger.info('Running benchmark for n = {}'.format(n))
             ret = runOpBenchmark(gpu_do, op, inputs=[X])
@@ -83,7 +83,7 @@ def run(self):
             ["y"]
         )
 
-        for n in itertools.imap(pow, itertools.cycle([10]), range(10)):
+        for n in itertools.imap(pow, itertools.cycle([10]), range(10)):  # type: ignore[attr-defined]
             X = np.random.rand(n).astype(np.float32)
             logger.info('Running benchmark for n = {}'.format(n))
             ret = runOpBenchmark(gpu_do, op, inputs=[X])
@@ -98,8 +98,8 @@ def run(self):
             ["probs", "avgloss"],
         )
 
-        for n in itertools.imap(pow, itertools.cycle([10]), range(8)):
-            for D in itertools.imap(pow, itertools.cycle([10]), range(3)):
+        for n in itertools.imap(pow, itertools.cycle([10]), range(8)):  # type: ignore[attr-defined]
+            for D in itertools.imap(pow, itertools.cycle([10]), range(3)):  # type: ignore[attr-defined]
                 X = np.random.rand(n, D).astype(np.float32)
                 label = (np.random.rand(n) * D).astype(np.int32)
                 logger.info('Running benchmark for n = {}, D= {}'.format(n, D))

caffe2/proto/__init__.py

@@ -10,6 +10,6 @@
 # This has to be done for all python targets, so listing them here
 from caffe2.proto import caffe2_pb2, metanet_pb2, torch_pb2
 try:
-    from caffe2.caffe2.fb.session.proto import session_pb2
+    from caffe2.caffe2.fb.session.proto import session_pb2  # type: ignore[import]
 except ImportError:
     pass

caffe2/proto/caffe2_legacy_pb2.pyi

@@ -13,7 +13,7 @@ import typing_extensions
 DESCRIPTOR: google.protobuf.descriptor.FileDescriptor = ...
 
 global___LegacyPadding = LegacyPadding
-class _LegacyPadding(google.protobuf.internal.enum_type_wrapper._EnumTypeWrapper[LegacyPadding], type):
+class _LegacyPadding(google.protobuf.internal.enum_type_wrapper._EnumTypeWrapper[LegacyPadding], type):  # type: ignore[type-var]
     DESCRIPTOR: google.protobuf.descriptor.EnumDescriptor = ...
     NOTSET = LegacyPadding.V(0)
     VALID = LegacyPadding.V(1)

caffe2/proto/caffe2_pb2.pyi

@@ -13,7 +13,7 @@ import typing_extensions
 DESCRIPTOR: google.protobuf.descriptor.FileDescriptor = ...
 
 global___DeviceTypeProto = DeviceTypeProto
-class _DeviceTypeProto(google.protobuf.internal.enum_type_wrapper._EnumTypeWrapper[DeviceTypeProto], type):
+class _DeviceTypeProto(google.protobuf.internal.enum_type_wrapper._EnumTypeWrapper[DeviceTypeProto], type):  # type: ignore[type-var]
     DESCRIPTOR: google.protobuf.descriptor.EnumDescriptor = ...
     PROTO_CPU = DeviceTypeProto.V(0)
     PROTO_CUDA = DeviceTypeProto.V(1)
@@ -44,7 +44,7 @@ PROTO_COMPILE_TIME_MAX_DEVICE_TYPES = DeviceTypeProto.V(11)
 
 class TensorProto(google.protobuf.message.Message):
     DESCRIPTOR: google.protobuf.descriptor.Descriptor = ...
-    class _DataType(google.protobuf.internal.enum_type_wrapper._EnumTypeWrapper[DataType], type):
+    class _DataType(google.protobuf.internal.enum_type_wrapper._EnumTypeWrapper[DataType], type):  # type: ignore[type-var]
         DESCRIPTOR: google.protobuf.descriptor.EnumDescriptor = ...
         UNDEFINED = TensorProto.DataType.V(0)
         FLOAT = TensorProto.DataType.V(1)
@@ -79,7 +79,7 @@ class TensorProto(google.protobuf.message.Message):
     ZERO_COLLISION_HASH = TensorProto.DataType.V(14)
     REBATCHING_BUFFER = TensorProto.DataType.V(15)
 
-    class _SerializationFormat(google.protobuf.internal.enum_type_wrapper._EnumTypeWrapper[SerializationFormat], type):
+    class _SerializationFormat(google.protobuf.internal.enum_type_wrapper._EnumTypeWrapper[SerializationFormat], type):  # type: ignore[type-var]
         DESCRIPTOR: google.protobuf.descriptor.EnumDescriptor = ...
         FMT_PROTOBUF = TensorProto.SerializationFormat.V(0)
         FMT_BFLOAT16 = TensorProto.SerializationFormat.V(1)
@@ -255,7 +255,7 @@ global___TensorShapes = TensorShapes
 
 class TensorBoundShape(google.protobuf.message.Message):
     DESCRIPTOR: google.protobuf.descriptor.Descriptor = ...
-    class _DimType(google.protobuf.internal.enum_type_wrapper._EnumTypeWrapper[DimType], type):
+    class _DimType(google.protobuf.internal.enum_type_wrapper._EnumTypeWrapper[DimType], type):  # type: ignore[type-var]
         DESCRIPTOR: google.protobuf.descriptor.EnumDescriptor = ...
         UNKNOWN = TensorBoundShape.DimType.V(0)
         CONSTANT = TensorBoundShape.DimType.V(1)
@@ -712,7 +712,7 @@ global___DBReaderProto = DBReaderProto
 
 class BlobSerializationOptions(google.protobuf.message.Message):
     DESCRIPTOR: google.protobuf.descriptor.Descriptor = ...
-    class _FloatFormat(google.protobuf.internal.enum_type_wrapper._EnumTypeWrapper[FloatFormat], type):
+    class _FloatFormat(google.protobuf.internal.enum_type_wrapper._EnumTypeWrapper[FloatFormat], type):  # type: ignore[type-var]
         DESCRIPTOR: google.protobuf.descriptor.EnumDescriptor = ...
         FLOAT_DEFAULT = BlobSerializationOptions.FloatFormat.V(0)
         FLOAT_PROTOBUF = BlobSerializationOptions.FloatFormat.V(1)
@@ -757,11 +757,11 @@ global___SerializationOptions = SerializationOptions
 DeviceType = int
 
 # These are freedom-patched into caffe2_pb2 in caffe2/proto/__init__.py
-CPU: int = DeviceType.PROTO_CPU
-CUDA: int = DeviceType.PROTO_CUDA
-MKLDNN: int = DeviceType.PROTO_MKLDNN
-OPENGL: int = DeviceType.PROTO_OPENGL
-OPENCL: int = DeviceType.PROTO_OPENCL
-IDEEP: int = DeviceType.PROTO_IDEEP
-HIP: int = DeviceType.PROTO_HIP
-COMPILE_TIME_MAX_DEVICE_TYPES: int = DeviceType.PROTO_COMPILE_TIME_MAX_DEVICE_TYPES
+CPU: int = DeviceType.PROTO_CPU  # type: ignore[attr-defined]
+CUDA: int = DeviceType.PROTO_CUDA  # type: ignore[attr-defined]
+MKLDNN: int = DeviceType.PROTO_MKLDNN  # type: ignore[attr-defined]
+OPENGL: int = DeviceType.PROTO_OPENGL  # type: ignore[attr-defined]
+OPENCL: int = DeviceType.PROTO_OPENCL  # type: ignore[attr-defined]
+IDEEP: int = DeviceType.PROTO_IDEEP  # type: ignore[attr-defined]
+HIP: int = DeviceType.PROTO_HIP  # type: ignore[attr-defined]
+COMPILE_TIME_MAX_DEVICE_TYPES: int = DeviceType.PROTO_COMPILE_TIME_MAX_DEVICE_TYPES  # type: ignore[attr-defined]

caffe2/proto/torch_pb2.pyi

@@ -14,7 +14,7 @@ import typing_extensions
 DESCRIPTOR: google.protobuf.descriptor.FileDescriptor = ...
 
 global___ProtoVersion = ProtoVersion
-class _ProtoVersion(google.protobuf.internal.enum_type_wrapper._EnumTypeWrapper[ProtoVersion], type):
+class _ProtoVersion(google.protobuf.internal.enum_type_wrapper._EnumTypeWrapper[ProtoVersion], type):  # type: ignore[type-var]
     DESCRIPTOR: google.protobuf.descriptor.EnumDescriptor = ...
     PROTO_VERSION_NEWEST = ProtoVersion.V(6)
 class ProtoVersion(metaclass=_ProtoVersion):

caffe2/python/allcompare_test.py

@@ -47,7 +47,7 @@ def allcompare_process(filestore_dir, process_id, data, num_procs):
     )
 
     model = ModelHelper()
-    model._rendezvous = rendezvous
+    model._rendezvous = rendezvous  # type: ignore[attr-defined]
 
     workspace.FeedBlob("test_data", data)
 

caffe2/python/benchmark_generator.py

@@ -16,7 +16,7 @@
 
 
 def parse_kwarg(kwarg_str):
-    key, value = map(string.strip, kwarg_str.split("=", 1))
+    key, value = map(string.strip, kwarg_str.split("=", 1))  # type: ignore[attr-defined]
     try:
         value = int(value)
     except ValueError: