[BE]: Enable ruff rule SIM113 (#147290)

Lint rules that tells the user to avoid keeping track of their own counter and use the builtin enumerate when possible.

Pull Request resolved: https://github.com/pytorch/pytorch/pull/147290
Approved by: https://github.com/jansel

benchmarks/dynamo/parse_logs.py

@@ -192,7 +192,7 @@ for name, name2, log in chunker(entries, 3):
             "unique_graph_breaks": unique_graph_breaks,
         }
     )
-    i += 1
+    i += 1  # noqa: SIM113
 
 if c:
     print(f"failed to classify {c} entries", file=sys.stderr)

benchmarks/fastrnns/custom_lstms.py

@@ -277,7 +277,7 @@ class BidirLSTMLayer(jit.ScriptModule):
             out, out_state = direction(input, state)
             outputs += [out]
             output_states += [out_state]
-            i += 1
+            i += 1  # noqa: SIM113
         return torch.cat(outputs, -1), output_states
 
 
@@ -310,7 +310,7 @@ class StackedLSTM(jit.ScriptModule):
             state = states[i]
             output, out_state = rnn_layer(output, state)
             output_states += [out_state]
-            i += 1
+            i += 1  # noqa: SIM113
         return output, output_states
 
 
@@ -341,7 +341,7 @@ class StackedLSTM2(jit.ScriptModule):
             state = states[i]
             output, out_state = rnn_layer(output, state)
             output_states += [out_state]
-            i += 1
+            i += 1  # noqa: SIM113
         return output, output_states
 
 
@@ -383,7 +383,7 @@ class StackedLSTMWithDropout(jit.ScriptModule):
             if i < self.num_layers - 1:
                 output = self.dropout_layer(output)
             output_states += [out_state]
-            i += 1
+            i += 1  # noqa: SIM113
         return output, output_states
 
 

benchmarks/sparse/dlmc/utils.py

@@ -94,7 +94,7 @@ def load_spmv_dataset(dataset_path, hidden_size, sparsity, device, n_limit=math.
             x_files.append(f.as_posix())
         if size[0] == hidden_size:
             y_files.append(f.as_posix())
-        index += 1
+        index += 1  # noqa: SIM113
     print()
 
     for fx, fy in zip(x_files, y_files):
@@ -136,7 +136,7 @@ def load_spmm_dataset(
             x_files.append(f.as_posix())
         if size[0] == hidden_size:
             y_files.append(f.as_posix())
-        index += 1
+        index += 1  # noqa: SIM113
     print()
 
     for fx, fy in zip(x_files, y_files):

pyproject.toml

@@ -77,7 +77,6 @@ ignore = [
     "PYI041",
     "PYI056",
     "SIM102", "SIM103", "SIM112", # flake8-simplify code styles
-    "SIM113", # please fix
     "SIM105", # these ignores are from flake8-simplify. please fix or ignore with commented reason
     "SIM108", # SIM108 ignored because we prefer if-else-block instead of ternary expression
     "SIM110",

test/dynamo/test_aot_autograd_cache.py

@@ -597,7 +597,7 @@ class AOTAutogradCacheTests(InductorTestCase):
             # see a recompilation (along with a cache miss).
             res1 = compiled_fn(a, b)
             # A first call should miss in the cache.
-            expected_misses += 1
+            expected_misses += 1  # noqa: SIM113
             self.assertEqual(
                 counters["aot_autograd"]["autograd_cache_miss"], expected_misses
             )
@@ -611,7 +611,7 @@ class AOTAutogradCacheTests(InductorTestCase):
             )
             # Because dynamic shapes are enabled, we expect backwards to be compiled ahead of time
             # So we should see a cache save here
-            expected_saves += 1
+            expected_saves += 1  # noqa: SIM113
             self.assertEqual(
                 counters["aot_autograd"]["autograd_cache_saved"], expected_saves
             )
@@ -632,7 +632,7 @@ class AOTAutogradCacheTests(InductorTestCase):
             # shape will still trigger a second call to autograd_cache.
             self._clear_dynamo_and_codecache()
             res2 = compiled_fn(a2, b2)
-            expected_hits += 1
+            expected_hits += 1  # noqa: SIM113
             self.assertEqual(
                 counters["aot_autograd"]["autograd_cache_miss"], expected_misses
             )
@@ -641,7 +641,7 @@ class AOTAutogradCacheTests(InductorTestCase):
                 expected_guard_misses,
             )
             # First compile is a regular cache miss, subsequent are guard misses
-            expected_guard_misses += 1
+            expected_guard_misses += 1  # noqa: SIM113
             self.assertEqual(
                 counters["aot_autograd"]["autograd_cache_hit"], expected_hits
             )

test/functorch/test_eager_transforms.py

@@ -969,7 +969,7 @@ class TestGradTransform(TestCase):
                 expected = f"{repr(x)}"
                 level = 0
                 for op in op_list:
-                    level += 1
+                    level += 1  # noqa: SIM113
                     if op == grad:
                         expected = f"GradTrackingTensor(lvl={level}, value={expected})"
                     elif op == vmap:

test/onnx/test_pytorch_onnx_onnxruntime.py

@@ -6914,7 +6914,7 @@ class TestONNXRuntime(onnx_test_common._TestONNXRuntime):
                     beam_idx = 0
                     for _, token in enumerate(x[i]):
                         beam_hyps.append(token)
-                        beam_idx += 1
+                        beam_idx += 1  # noqa: SIM113
 
                         if beam_idx == 6:
                             break

torch/_strobelight/examples/cli_function_profiler_example.py

@@ -30,6 +30,6 @@ if __name__ == "__main__":
     def work2():
         sum = 0
         for _ in range(100000000):
-            sum += 1
+            sum += 1  # noqa: SIM113
 
     work2()

torch/ao/ns/fx/mappings.py

@@ -445,10 +445,8 @@ def get_base_name_to_sets_of_related_ops() -> dict[str, set[NSNodeTargetType]]:
 
     base_name_to_sets_of_related_ops: dict[str, set[NSNodeTargetType]] = {}
 
-    counter = 0
-    for set_of_related_ops in sets_of_related_ops:
+    for counter, set_of_related_ops in enumerate(sets_of_related_ops):
         base_name = str(counter)
-        counter += 1
         base_name_to_sets_of_related_ops[base_name] = set_of_related_ops
 
     return base_name_to_sets_of_related_ops

torch/ao/quantization/_correct_bias.py

@@ -128,10 +128,8 @@ def bias_correction(
         quantized_submodule = get_module(quantized_model, uncorrected_module)
         bias = get_param(quantized_submodule, "bias")
         if bias is not None:
-            count = 0
-            for data in img_data:
+            for count, data in enumerate(img_data, start=1):
                 quantized_model(data[0])
-                count += 1
                 if count == neval_batches:
                     break
             ob_dict = ns.get_logger_dict(quantized_model)

torch/export/unflatten.py

@@ -1227,10 +1227,8 @@ class _ModuleFrame:
         self.seen_nodes[node.name] = node
 
     def run_outer(self):
-        i = 0
-        for node in self.flat_graph.nodes:
+        for i, node in enumerate(self.flat_graph.nodes):
             self.print(i, node.meta.get("nn_module_stack"), node.format_node())
-            i += 1
 
         # Copy all graph inputs
         node_idx: int = 0

torch/testing/_internal/common_quantization.py

@@ -195,10 +195,8 @@ def accuracy(output, target, topk=(1,)):
 
 def train_one_epoch(model, criterion, optimizer, data_loader, device, ntrain_batches):
     model.train()
-    cnt = 0
-    for image, target in data_loader:
+    for cnt, (image, target) in enumerate(data_loader, start=1):
         print('.', end='')
-        cnt += 1
         image, target = image.to(device), target.to(device)
         output = model(image)
         loss = criterion(output, target)

torch/testing/_internal/distributed/rpc/rpc_test.py

@@ -2573,12 +2573,10 @@ class RpcTest(RpcAgentTestFixture, RpcTestCommon):
             )
             futs.append(fut)
 
-        j = 0
-        for val in torch.futures.wait_all(futs):
+        for j, val in enumerate(torch.futures.wait_all(futs)):
             self.assertEqual(
                 val, my_tensor_function(torch.ones(j, j), torch.ones(j, j))
             )
-            j += 1
 
     @dist_init
     def test_py_tensors_in_container(self):

torch/utils/_strobelight/examples/cli_function_profiler_example.py

@@ -30,6 +30,6 @@ if __name__ == "__main__":
     def work2():
         sum = 0
         for _ in range(100000000):
-            sum += 1
+            sum += 1  # noqa: SIM113
 
     work2()

torchgen/_autoheuristic/train_decision.py

@@ -861,9 +861,8 @@ class DecisionEvaluator:
         """
 
         y_true = self.df["actual_winner"] if self.ranking else self.df["winner"]
-        i = 0
-        for pred, true, prob, leaf_id in zip(
-            self.predictions, y_true, self.probas, self.leaf_ids
+        for i, (pred, true, prob, leaf_id) in enumerate(
+            zip(self.predictions, y_true, self.probas, self.leaf_ids)
         ):
             avail_choices = self.df["avail_choices"].iloc[i]
             top_k_choices = self.top_k_classes(
@@ -884,7 +883,6 @@ class DecisionEvaluator:
                 i,
             )
             self.eval_ranking_prediction(true, top_k_choices, i)
-            i += 1
 
         total = len(self.predictions)
         if return_safe_proba: