Update black linter in OSS lintrunner (#2229)

Summary:
Pull Request resolved: #2229

Black linter is upgraded to v24.2.0 in fbsource. Let's also upgrade the OSS linter to match this to reduce inconsistency between internal linter and OSS linters.

Currently lintrunner is failing as is: https://hud.pytorch.org/hud/pytorch/executorch/main/1?per_page=50

Context: https://fb.workplace.com/groups/pyfmt/posts/1391116614859184/?fbclid=IwAR1Gag0Bkq2OE_4EeH5XY_iOFgwF6VE-7OAr9kmHeyB3QzkfaGvsWWY3nCo

Reviewed By: kimishpatel

Differential Revision: D54487606

fbshipit-source-id: 82d5eb011e6e96dd2b369299cd4a74724564b94c

Author: mergennachin

backends/apple/mps/operators/node_visitor.py

@@ -235,7 +235,6 @@ def get_serialized_data(
     def get_serialized_id(
         self, node: Union[torch.fx.Node, float, int], mps_graph: MPSGraph
     ) -> int:
-
         """
         Map a tensor to a unique id. If the tensor was already mapped, return
         the existent id.

backends/arm/arm_backend.py

@@ -128,9 +128,11 @@ def preprocess(  # noqa: C901
                 # Add output to TOSA graph
                 tosa_graph.currRegion.currBasicBlock.addTensor(
                     output.name,
-                    inputs[0].shape
-                    if is_permute_node_before_addmm(node)
-                    else output.shape,
+                    (
+                        inputs[0].shape
+                        if is_permute_node_before_addmm(node)
+                        else output.shape
+                    ),
                     ts.DType.INT8 if is_quant_node(node) else output.dtype,
                 )
 

backends/arm/arm_vela.py

@@ -12,6 +12,7 @@
 
 import numpy as np
 
+
 # Pack either input or output tensor block, compose the related arrays into
 # per-io structs to simplify runtime use.
 def vela_bin_pack_io(prefix, data):

backends/qualcomm/passes/decompose_scaled_dot_product_attention.py

@@ -42,9 +42,9 @@ def call(self, graph_module: torch.fx.GraphModule):
                     # In decomposed module, there are only input tensors for placeholder op.
                     for decomposed_node in decomposed_module.graph.nodes:
                         if decomposed_node.op == "placeholder":
-                            decomposed_node_to_subgraph_node[
-                                decomposed_node
-                            ] = name_to_input_tensor_map[decomposed_node.name]
+                            decomposed_node_to_subgraph_node[decomposed_node] = (
+                                name_to_input_tensor_map[decomposed_node.name]
+                            )
 
                         if decomposed_node.op == "output":
                             last_decomposed_node = decomposed_node.args[0]
@@ -76,9 +76,9 @@ def call(self, graph_module: torch.fx.GraphModule):
                         subgraph_node.meta["source_fn_stack"] = [
                             (subgraph_node, subgraph_node.target)
                         ]
-                        decomposed_node_to_subgraph_node[
-                            decomposed_node
-                        ] = subgraph_node
+                        decomposed_node_to_subgraph_node[decomposed_node] = (
+                            subgraph_node
+                        )
 
                     graph.erase_node(node)
 

backends/qualcomm/passes/i64_to_i32.py

@@ -21,9 +21,11 @@ def _update_meta(self, node: torch.fx.node) -> None:
         meta_val = node.meta["val"]
         if isinstance(meta_val, tuple):
             node.meta["val"] = (
-                fake_tensor.to(torch.int32)
-                if fake_tensor.dtype == torch.int64
-                else fake_tensor
+                (
+                    fake_tensor.to(torch.int32)
+                    if fake_tensor.dtype == torch.int64
+                    else fake_tensor
+                )
                 for fake_tensor in meta_val
             )
         else:

examples/arm/aot_arm_compiler.py

@@ -26,6 +26,7 @@
 #       quantization step in our example. This will take the models
 #       from examples/models/ and quantize then export to delegate.
 
+
 # Two simple models
 class AddModule(torch.nn.Module):
     def __init__(self):

examples/models/llama2/quantize.py

@@ -64,7 +64,8 @@ def dynamically_quantize_per_channel(
                         with a final group of a size less than group size.
 
     Assumptions:
-        This function assumes symmetric quantization, axis ==0 and a dense memory format."""
+        This function assumes symmetric quantization, axis ==0 and a dense memory format.
+    """
 
     # assumes symmetric quantization
     # assumes axis == 0

exir/passes/_quant_patterns_and_replacements.py

@@ -337,9 +337,9 @@ def binary_relu_op_replacement(
     ]
 
 
-def _get_binary_ops_patterns_and_replacements() -> List[
-    Tuple[Callable, Callable, List[Callable]]
-]:
+def _get_binary_ops_patterns_and_replacements() -> (
+    List[Tuple[Callable, Callable, List[Callable]]]
+):
 
     # TODO: replace qbinary op with the ops implemented in lean mode
     binary_op_to_qbinary_ops = {
@@ -360,9 +360,9 @@ def _get_binary_ops_patterns_and_replacements() -> List[
     return pattern_and_replacements
 
 
-def _get_reshape_patterns_and_replacements() -> List[
-    Tuple[Callable, Callable, List[Callable]]
-]:
+def _get_reshape_patterns_and_replacements() -> (
+    List[Tuple[Callable, Callable, List[Callable]]]
+):
     def pattern(
         x,
         arg0,
@@ -413,9 +413,9 @@ def replacement(
     ]
 
 
-def _get_slice_patterns_and_replacements() -> List[
-    Tuple[Callable, Callable, List[Callable]]
-]:
+def _get_slice_patterns_and_replacements() -> (
+    List[Tuple[Callable, Callable, List[Callable]]]
+):
     def pattern(x, dim, start, end, x_scale, x_zero_point, x_qmin, x_qmax):
         x = torch.ops.quantized_decomposed.dequantize_per_tensor.default(
             x, x_scale, x_zero_point, x_qmin, x_qmax, torch.uint8
@@ -439,9 +439,9 @@ def replacement(x, dim, start, end, x_scale, x_zero_point, x_qmin, x_qmax):
     ]
 
 
-def _get_embedding_ops_patterns_and_replacements() -> List[
-    Tuple[Callable, Callable, List[Callable]]
-]:
+def _get_embedding_ops_patterns_and_replacements() -> (
+    List[Tuple[Callable, Callable, List[Callable]]]
+):
     def get_pattern_and_replacement():
         @bind_pattern_to_op(quantized_decomposed_lib, "embedding_byte")
         def pattern(
@@ -569,9 +569,9 @@ def replacement(x, x_scale, x_zero_point, x_qmin, x_qmax):
 """
 
 
-def get_quant_patterns_and_replacements() -> List[
-    Tuple[Callable, Callable, List[Callable]]
-]:
+def get_quant_patterns_and_replacements() -> (
+    List[Tuple[Callable, Callable, List[Callable]]]
+):
 
     return copy.copy(
         [

exir/serde/export_serialize.py

@@ -1728,9 +1728,9 @@ def deserialize(
             symbol_name_to_range,
             res.names_to_symbols,
         )
-        model_opset_version: Optional[
-            Dict[str, int]
-        ] = serialized_artifact.exported_program.opset_version  # pyre-ignore
+        model_opset_version: Optional[Dict[str, int]] = (
+            serialized_artifact.exported_program.opset_version  # pyre-ignore
+        )
         self._validate_model_opset_version(model_opset_version)
 
         upgrader = GraphModuleOpUpgrader(

requirements-lintrunner.txt

@@ -13,7 +13,7 @@ pycodestyle==2.10.0
 torchfix==0.1.1
 
 # UFMT
-black==22.12.0
+black==24.2.0
 ufmt==2.0.1
 usort==1.0.5