[BE] [1/3] Rewrite super() calls in caffe2 and benchmarks (pytorch#…

…94587)

Rewrite Python built-in class `super()` calls. Only non-semantic changes should be applied.

- pytorch#94587
- pytorch#94588
- pytorch#94592

Also, methods with only a `super()` call are removed:

```diff
class MyModule(nn.Module):
-   def __init__(self):
-       super().__init__()
-
    def forward(self, ...):
        ...
```

Some cases that change the semantics should be kept unchanged. E.g.:

https://github.com/pytorch/pytorch/blob/f152a79be9612b824e1672b8f8cb88a414ce4c12/caffe2/python/net_printer.py#L184-L190

https://github.com/pytorch/pytorch/blob/f152a79be9612b824e1672b8f8cb88a414ce4c12/test/test_jit_fuser_te.py#L2628-L2635

Pull Request resolved: pytorch#94587
Approved by: https://github.com/ezyang

.circleci/cimodel/data/binary_build_data.py

@@ -57,7 +57,7 @@ def get_processor_arch_name(gpu_version):
 
 class TopLevelNode(ConfigNode):
     def __init__(self, node_name, config_tree_data, smoke):
-        super(TopLevelNode, self).__init__(None, node_name)
+        super().__init__(None, node_name)
 
         self.config_tree_data = config_tree_data
         self.props["smoke"] = smoke
@@ -68,7 +68,7 @@ def get_children(self):
 
 class OSConfigNode(ConfigNode):
     def __init__(self, parent, os_name, gpu_versions, py_tree):
-        super(OSConfigNode, self).__init__(parent, os_name)
+        super().__init__(parent, os_name)
 
         self.py_tree = py_tree
         self.props["os_name"] = os_name
@@ -80,7 +80,7 @@ def get_children(self):
 
 class PackageFormatConfigNode(ConfigNode):
     def __init__(self, parent, package_format, python_versions):
-        super(PackageFormatConfigNode, self).__init__(parent, package_format)
+        super().__init__(parent, package_format)
 
         self.props["python_versions"] = python_versions
         self.props["package_format"] = package_format
@@ -97,7 +97,7 @@ def get_children(self):
 
 class LinuxGccConfigNode(ConfigNode):
     def __init__(self, parent, gcc_config_variant):
-        super(LinuxGccConfigNode, self).__init__(parent, "GCC_CONFIG_VARIANT=" + str(gcc_config_variant))
+        super().__init__(parent, "GCC_CONFIG_VARIANT=" + str(gcc_config_variant))
 
         self.props["gcc_config_variant"] = gcc_config_variant
 
@@ -122,7 +122,7 @@ def get_children(self):
 
 class WindowsLibtorchConfigNode(ConfigNode):
     def __init__(self, parent, libtorch_config_variant):
-        super(WindowsLibtorchConfigNode, self).__init__(parent, "LIBTORCH_CONFIG_VARIANT=" + str(libtorch_config_variant))
+        super().__init__(parent, "LIBTORCH_CONFIG_VARIANT=" + str(libtorch_config_variant))
 
         self.props["libtorch_config_variant"] = libtorch_config_variant
 
@@ -132,7 +132,7 @@ def get_children(self):
 
 class ArchConfigNode(ConfigNode):
     def __init__(self, parent, gpu):
-        super(ArchConfigNode, self).__init__(parent, get_processor_arch_name(gpu))
+        super().__init__(parent, get_processor_arch_name(gpu))
 
         self.props["gpu"] = gpu
 
@@ -142,7 +142,7 @@ def get_children(self):
 
 class PyVersionConfigNode(ConfigNode):
     def __init__(self, parent, pyver):
-        super(PyVersionConfigNode, self).__init__(parent, pyver)
+        super().__init__(parent, pyver)
 
         self.props["pyver"] = pyver
 
@@ -158,14 +158,14 @@ def get_children(self):
 
 class LinkingVariantConfigNode(ConfigNode):
     def __init__(self, parent, linking_variant):
-        super(LinkingVariantConfigNode, self).__init__(parent, linking_variant)
+        super().__init__(parent, linking_variant)
 
     def get_children(self):
         return [DependencyInclusionConfigNode(self, v) for v in DEPS_INCLUSION_DIMENSIONS]
 
 
 class DependencyInclusionConfigNode(ConfigNode):
     def __init__(self, parent, deps_variant):
-        super(DependencyInclusionConfigNode, self).__init__(parent, deps_variant)
+        super().__init__(parent, deps_variant)
 
         self.props["libtorch_variant"] = "-".join([self.parent.get_label(), self.get_label()])

.circleci/cimodel/data/pytorch_build_data.py

@@ -12,7 +12,7 @@ def get_major_pyver(dotted_version):
 
 class TreeConfigNode(ConfigNode):
     def __init__(self, parent, node_name, subtree):
-        super(TreeConfigNode, self).__init__(parent, self.modify_label(node_name))
+        super().__init__(parent, self.modify_label(node_name))
         self.subtree = subtree
         self.init2(node_name)
 
@@ -28,7 +28,7 @@ def get_children(self):
 
 class TopLevelNode(TreeConfigNode):
     def __init__(self, node_name, subtree):
-        super(TopLevelNode, self).__init__(None, node_name, subtree)
+        super().__init__(None, node_name, subtree)
 
     # noinspection PyMethodMayBeStatic
     def child_constructor(self):

android/pytorch_android/generate_test_torchscripts.py

@@ -15,9 +15,6 @@ def scriptAndSave(module, fileName):
     print('=' * 80)
 
 class Test(torch.jit.ScriptModule):
-    def __init__(self):
-        super(Test, self).__init__()
-
     @torch.jit.script_method
     def forward(self, input):
         return None

benchmarks/distributed/ddp/benchmark.py

@@ -173,7 +173,7 @@ def generate_target(self):
 
 class TorchvisionBenchmark(Benchmark):
     def __init__(self, device, distributed_backend, bucket_size, model):
-        super(TorchvisionBenchmark, self).__init__(
+        super().__init__(
             device,
             distributed_backend,
             bucket_size,

benchmarks/distributed/pipeline/pipe.py

@@ -43,7 +43,7 @@ def forward(self, src):
 
 class PositionalEncodingLayer(nn.Module):
     def __init__(self, d_model, dropout=0.1, max_len=5000):
-        super(PositionalEncodingLayer, self).__init__()
+        super().__init__()
         self.dropout = nn.Dropout(p=dropout)
 
         pe = torch.zeros(max_len, d_model)
@@ -99,7 +99,7 @@ def __init__(self, ntokens, ninp, nhead, nhid, dropout, initrange, ndecoder):
             layers.append(TransformerDecoderLayer(ninp, nhead, nhid, dropout))
 
         layers.append(LinearLayer(ninp, ntokens, initrange))
-        super(TransformerLMSequential, self).__init__(*layers)
+        super().__init__(*layers)
 
 
 def make_model(args, device, ntokens):

benchmarks/distributed/rpc/rl/agent.py

@@ -22,7 +22,7 @@ def __init__(self, in_features, nlayers, out_features):
             nlayers (int): Number of layers in the model
             out_features (int): Number of features the model outputs
         """
-        super(Policy, self).__init__()
+        super().__init__()
 
         self.model = nn.Sequential(
             nn.Flatten(1, -1),

benchmarks/dynamo/dist_util.py

@@ -38,7 +38,7 @@ def cleanup():
 
 class CustomLinear(torch.nn.Module):
     def __init__(self, a, b):
-        super(CustomLinear, self).__init__()
+        super().__init__()
         self.weight = nn.Parameter(torch.randn(a, b))
 
     def forward(self, x):
@@ -47,7 +47,7 @@ def forward(self, x):
 
 class MyModule(torch.nn.Module):
     def __init__(self, a, b):
-        super(MyModule, self).__init__()
+        super().__init__()
         self.net = nn.Sequential(
             nn.Linear(a, b),
             nn.ReLU(),
@@ -59,7 +59,7 @@ def forward(self, x):
 
 class ToyModel(nn.Module):
     def __init__(self):
-        super(ToyModel, self).__init__()
+        super().__init__()
         self.net = nn.Sequential(
             *[nn.Linear(10, 10000), nn.ReLU()]
             + [nn.Linear(10000, 10000), nn.ReLU()]

benchmarks/dynamo/huggingface.py

@@ -364,7 +364,7 @@ def rand_int_tensor(device, low, high, shape):
 
 class HuggingfaceRunner(BenchmarkRunner):
     def __init__(self):
-        super(HuggingfaceRunner, self).__init__()
+        super().__init__()
         self.suite_name = "huggingface"
 
     def load_model(

benchmarks/dynamo/timm_models.py

@@ -169,7 +169,7 @@ def populate_family(models):
 
 class TimmRunnner(BenchmarkRunner):
     def __init__(self):
-        super(TimmRunnner, self).__init__()
+        super().__init__()
         self.suite_name = "timm_models"
 
     def load_model(

benchmarks/dynamo/torchbench.py

@@ -195,7 +195,7 @@ def setup_torchbench_cwd():
 
 class TorchBenchmarkRunner(BenchmarkRunner):
     def __init__(self):
-        super(TorchBenchmarkRunner, self).__init__()
+        super().__init__()
         self.suite_name = "torchbench"
         self.optimizer = None
 

benchmarks/fastrnns/custom_lstms.py

@@ -92,7 +92,7 @@ def reverse(lst: List[Tensor]) -> List[Tensor]:
 
 class LSTMCell(jit.ScriptModule):
     def __init__(self, input_size, hidden_size):
-        super(LSTMCell, self).__init__()
+        super().__init__()
         self.input_size = input_size
         self.hidden_size = hidden_size
         self.weight_ih = Parameter(torch.randn(4 * hidden_size, input_size))
@@ -120,7 +120,7 @@ def forward(self, input: Tensor, state: Tuple[Tensor, Tensor]) -> Tuple[Tensor,
 
 class LayerNorm(jit.ScriptModule):
     def __init__(self, normalized_shape):
-        super(LayerNorm, self).__init__()
+        super().__init__()
         if isinstance(normalized_shape, numbers.Integral):
             normalized_shape = (normalized_shape,)
         normalized_shape = torch.Size(normalized_shape)
@@ -146,7 +146,7 @@ def forward(self, input):
 
 class LayerNormLSTMCell(jit.ScriptModule):
     def __init__(self, input_size, hidden_size, decompose_layernorm=False):
-        super(LayerNormLSTMCell, self).__init__()
+        super().__init__()
         self.input_size = input_size
         self.hidden_size = hidden_size
         self.weight_ih = Parameter(torch.randn(4 * hidden_size, input_size))
@@ -183,7 +183,7 @@ def forward(self, input: Tensor, state: Tuple[Tensor, Tensor]) -> Tuple[Tensor,
 
 class LSTMLayer(jit.ScriptModule):
     def __init__(self, cell, *cell_args):
-        super(LSTMLayer, self).__init__()
+        super().__init__()
         self.cell = cell(*cell_args)
 
     @jit.script_method
@@ -198,7 +198,7 @@ def forward(self, input: Tensor, state: Tuple[Tensor, Tensor]) -> Tuple[Tensor,
 
 class ReverseLSTMLayer(jit.ScriptModule):
     def __init__(self, cell, *cell_args):
-        super(ReverseLSTMLayer, self).__init__()
+        super().__init__()
         self.cell = cell(*cell_args)
 
     @jit.script_method
@@ -215,7 +215,7 @@ class BidirLSTMLayer(jit.ScriptModule):
     __constants__ = ['directions']
 
     def __init__(self, cell, *cell_args):
-        super(BidirLSTMLayer, self).__init__()
+        super().__init__()
         self.directions = nn.ModuleList([
             LSTMLayer(cell, *cell_args),
             ReverseLSTMLayer(cell, *cell_args),
@@ -247,7 +247,7 @@ class StackedLSTM(jit.ScriptModule):
     __constants__ = ['layers']  # Necessary for iterating through self.layers
 
     def __init__(self, num_layers, layer, first_layer_args, other_layer_args):
-        super(StackedLSTM, self).__init__()
+        super().__init__()
         self.layers = init_stacked_lstm(num_layers, layer, first_layer_args,
                                         other_layer_args)
 
@@ -274,7 +274,7 @@ class StackedLSTM2(jit.ScriptModule):
     __constants__ = ['layers']  # Necessary for iterating through self.layers
 
     def __init__(self, num_layers, layer, first_layer_args, other_layer_args):
-        super(StackedLSTM2, self).__init__()
+        super().__init__()
         self.layers = init_stacked_lstm(num_layers, layer, first_layer_args,
                                         other_layer_args)
 
@@ -299,7 +299,7 @@ class StackedLSTMWithDropout(jit.ScriptModule):
     __constants__ = ['layers', 'num_layers']
 
     def __init__(self, num_layers, layer, first_layer_args, other_layer_args):
-        super(StackedLSTMWithDropout, self).__init__()
+        super().__init__()
         self.layers = init_stacked_lstm(num_layers, layer, first_layer_args,
                                         other_layer_args)
         # Introduces a Dropout layer on the outputs of each LSTM layer except

benchmarks/framework_overhead_benchmark/SimpleAddModule.py

@@ -9,7 +9,7 @@ def add_tensors_loop(x, y):
 
 class SimpleAddModule(torch.nn.Module):
     def __init__(self, add_op):
-        super(SimpleAddModule, self).__init__()
+        super().__init__()
         self.add_op = add_op
 
     def forward(self, x, y):

benchmarks/functional_autograd_benchmark/torchaudio_models.py

@@ -27,7 +27,7 @@ class Wav2Letter(nn.Module):
     def __init__(self, num_classes: int = 40,
                  input_type: str = "waveform",
                  num_features: int = 1) -> None:
-        super(Wav2Letter, self).__init__()
+        super().__init__()
 
         acoustic_num_features = 250 if input_type == "waveform" else num_features
         acoustic_model = nn.Sequential(
@@ -85,7 +85,7 @@ def __init__(self, module):
         Allows handling of variable sequence lengths and minibatch sizes.
         :param module: Module to apply input to.
         """
-        super(SequenceWise, self).__init__()
+        super().__init__()
         self.module = module
 
     def forward(self, x):
@@ -110,7 +110,7 @@ def __init__(self, seq_module):
         Input needs to be in the shape of (BxCxDxT)
         :param seq_module: The sequential module containing the conv stack.
         """
-        super(MaskConv, self).__init__()
+        super().__init__()
         self.seq_module = seq_module
 
     def forward(self, x, lengths):
@@ -142,7 +142,7 @@ def forward(self, input_):
 
 class BatchRNN(nn.Module):
     def __init__(self, input_size, hidden_size, rnn_type=nn.LSTM, bidirectional=False, batch_norm=True):
-        super(BatchRNN, self).__init__()
+        super().__init__()
         self.input_size = input_size
         self.hidden_size = hidden_size
         self.bidirectional = bidirectional
@@ -170,7 +170,7 @@ class Lookahead(nn.Module):
     # input shape - sequence, batch, feature - TxNxH
     # output shape - same as input
     def __init__(self, n_features, context):
-        super(Lookahead, self).__init__()
+        super().__init__()
         assert context > 0
         self.context = context
         self.n_features = n_features
@@ -193,7 +193,7 @@ def __repr__(self):
 class DeepSpeech(nn.Module):
     def __init__(self, rnn_type, labels, rnn_hidden_size, nb_layers, audio_conf,
                  bidirectional, context=20):
-        super(DeepSpeech, self).__init__()
+        super().__init__()
 
         self.hidden_size = rnn_hidden_size
         self.hidden_layers = nb_layers
@@ -298,7 +298,7 @@ class PositionalEncoding(nn.Module):
     """
 
     def __init__(self, d_model, dropout=0.1, max_len=5000):
-        super(PositionalEncoding, self).__init__()
+        super().__init__()
         self.dropout = nn.Dropout(p=dropout)
 
         pe = torch.zeros(max_len, d_model)
@@ -327,7 +327,7 @@ class TransformerModel(nn.Module):
     """Container module with an encoder, a recurrent or transformer module, and a decoder."""
 
     def __init__(self, ntoken, ninp, nhead, nhid, nlayers, dropout=0.5):
-        super(TransformerModel, self).__init__()
+        super().__init__()
         try:
             from torch.nn import TransformerEncoder, TransformerEncoderLayer
         except Exception as e:
@@ -392,7 +392,7 @@ def __init__(self, nhead, in_proj_container, attention_layer, out_proj):
             >>> print(attn_output.shape)
             >>> torch.Size([21, 64, 10])
         """
-        super(MultiheadAttentionContainer, self).__init__()
+        super().__init__()
         self.nhead = nhead
         self.in_proj_container = in_proj_container
         self.attention_layer = attention_layer
@@ -456,7 +456,7 @@ def __init__(self, dropout=0.0):
             >>> print(attn_output.shape, attn_weights.shape)
             torch.Size([256, 21, 3]) torch.Size([256, 21, 21])
         """
-        super(ScaledDotProduct, self).__init__()
+        super().__init__()
         self.dropout = dropout
 
     def forward(self, query: torch.Tensor, key: torch.Tensor, value: torch.Tensor,
@@ -532,7 +532,7 @@ def __init__(self, query_proj, key_proj, value_proj):
             value_proj: a proj layer for value.
         """
 
-        super(InProjContainer, self).__init__()
+        super().__init__()
         self.query_proj = query_proj
         self.key_proj = key_proj
         self.value_proj = value_proj