pytorch · ysiraichi · Nov 21, 2025 · Nov 21, 2025
diff --git a/.clang-format b/.clang-format
@@ -102,6 +102,10 @@ IncludeBlocks:   Regroup
 IncludeCategories:
   - Regex:           '^<torch/.*>'
     Priority:        3
+  - Regex:           '^<ATen/.*>'
+    Priority:        3
+  - Regex:           '^<c10/.*>'
+    Priority:        3
   - Regex:           '^<.*\.h>'
     Priority:        1
   - Regex:           '^<.*'

diff --git a/test/cpp/test_replication.cpp b/test/cpp/test_replication.cpp
@@ -1,8 +1,9 @@
-#include <ATen/ATen.h>
 #include <gtest/gtest.h>
 
 #include <iostream>
 
+#include <ATen/ATen.h>
+
 #include "absl/synchronization/blocking_counter.h"
 #include "xla/hlo/builder/xla_builder.h"
 #include "xla/shape_util.h"

diff --git a/test/cpp/test_status_common.h b/test/cpp/test_status_common.h
@@ -18,7 +18,6 @@
 #ifndef XLA_TEST_CPP_TEST_STATUS_COMMON_H_
 #define XLA_TEST_CPP_TEST_STATUS_COMMON_H_
 
-#include <c10/util/Exception.h>
 #include <gmock/gmock.h>
 #include <gtest/gtest.h>
 
@@ -27,6 +26,8 @@
 #include <stdexcept>
 #include <utility>
 
+#include <c10/util/Exception.h>
+
 #include "absl/status/status.h"
 #include "absl/status/statusor.h"
 

diff --git a/test/cpp/test_tensor.cpp b/test/cpp/test_tensor.cpp
@@ -1,9 +1,9 @@
-#include <ATen/ATen.h>
 #include <gtest/gtest.h>
 
 #include <limits>
 #include <vector>
 
+#include <ATen/ATen.h>
 #include <torch/csrc/autograd/variable.h>
 
 #include "absl/base/nullability.h"

diff --git a/test/cpp/test_xla_sharding.cpp b/test/cpp/test_xla_sharding.cpp
@@ -1,10 +1,10 @@
-#include <ATen/ATen.h>
 #include <google/protobuf/repeated_field.h>
 #include <gtest/gtest.h>
 #include <stdlib.h>
 
 #include <iostream>
 
+#include <ATen/ATen.h>
 #include <torch/csrc/lazy/core/lazy_graph_executor.h>
 
 #include "xla/protobuf_util.h"

diff --git a/torch_xla/csrc/aten_fallback.cpp b/torch_xla/csrc/aten_fallback.cpp
@@ -1,13 +1,12 @@
 #include "torch_xla/csrc/aten_fallback.h"
 
+#include <unordered_map>
+#include <vector>
+
 #include <ATen/DLConvertor.h>
 #include <ATen/native/CPUFallback.h>
 #include <ATen/ops/_copy_from_and_resize.h>
 #include <ATen/ops/_to_cpu.h>
-
-#include <unordered_map>
-#include <vector>
-
 #include <torch/csrc/utils/device_lazy_init.h>
 
 #include "torch_xla/csrc/function_call_tracker.h"

diff --git a/torch_xla/csrc/aten_xla_bridge.cpp b/torch_xla/csrc/aten_xla_bridge.cpp
@@ -1,11 +1,10 @@
 #include "torch_xla/csrc/aten_xla_bridge.h"
 
-#include <ATen/FunctionalTensorWrapper.h>
-
 #include <map>
 #include <string>
 #include <vector>
 
+#include <ATen/FunctionalTensorWrapper.h>
 #include <torch/csrc/lazy/core/tensor_util.h>
 
 #include "absl/log/absl_check.h"

diff --git a/torch_xla/csrc/aten_xla_bridge.h b/torch_xla/csrc/aten_xla_bridge.h
@@ -1,12 +1,12 @@
 #ifndef XLA_TORCH_XLA_CSRC_ATEN_XLA_BRIDGE_H_
 #define XLA_TORCH_XLA_CSRC_ATEN_XLA_BRIDGE_H_
 
+#include <vector>
+
 #include <ATen/Device.h>
 #include <ATen/Functions.h>
 #include <ATen/Tensor.h>
 
-#include <vector>
-
 #include "absl/base/nullability.h"
 #include "absl/status/statusor.h"
 #include "absl/types/span.h"

diff --git a/torch_xla/csrc/aten_xla_type.cpp b/torch_xla/csrc/aten_xla_type.cpp
@@ -1,3 +1,7 @@
+#include <iterator>
+#include <mutex>
+#include <optional>
+
 #include <ATen/ExpandUtils.h>
 #include <ATen/FunctionalTensorWrapper.h>
 #include <ATen/MetaFunctions.h>
@@ -10,11 +14,6 @@
 #include <ATen/ops/_embedding_bag_backward_native.h>
 #include <ATen/ops/expand_copy.h>
 #include <c10/core/Contiguity.h>
-
-#include <iterator>
-#include <mutex>
-#include <optional>
-
 #include <torch/csrc/lazy/core/helpers.h>
 #include <torch/csrc/lazy/core/shape_inference.h>
 #include <torch/csrc/lazy/core/tensor_util.h>

diff --git a/torch_xla/csrc/autocast_mode.cpp b/torch_xla/csrc/autocast_mode.cpp
@@ -4,7 +4,6 @@
 #include <ATen/autocast_mode.h>
 #include <c10/core/impl/LocalDispatchKeySet.h>
 #include <c10/util/intrusive_ptr.h>
-
 #include <torch/library.h>
 
 namespace at {

diff --git a/torch_xla/csrc/convert_ops.h b/torch_xla/csrc/convert_ops.h
@@ -1,10 +1,10 @@
 #ifndef XLA_TORCH_XLA_CSRC_CONVERT_OPS_H_
 #define XLA_TORCH_XLA_CSRC_CONVERT_OPS_H_
 
-#include <c10/core/ScalarType.h>
-
 #include <optional>
 
+#include <c10/core/ScalarType.h>
+
 #include "xla/hlo/builder/xla_builder.h"
 #include "xla/types.h"
 

diff --git a/torch_xla/csrc/data_ops.h b/torch_xla/csrc/data_ops.h
@@ -1,10 +1,10 @@
 #ifndef XLA_TORCH_XLA_CSRC_DATA_OPS_H_
 #define XLA_TORCH_XLA_CSRC_DATA_OPS_H_
 
-#include <c10/core/ScalarType.h>
-
 #include <vector>
 
+#include <c10/core/ScalarType.h>
+
 #include "absl/types/optional.h"
 #include "absl/types/span.h"
 #include "xla/hlo/builder/xla_builder.h"

diff --git a/torch_xla/csrc/dl_convertor.cpp b/torch_xla/csrc/dl_convertor.cpp
@@ -1,11 +1,11 @@
 #include "torch_xla/csrc/dl_convertor.h"
 
-#include <ATen/DLConvertor.h>
-
 #include <memory>
 #include <utility>
 #include <vector>
 
+#include <ATen/DLConvertor.h>
+
 #include "absl/log/absl_check.h"
 #include "absl/status/status.h"
 #include "absl/types/span.h"

diff --git a/torch_xla/csrc/helpers.h b/torch_xla/csrc/helpers.h
@@ -1,13 +1,12 @@
 #ifndef XLA_TORCH_XLA_CSRC_HELPERS_H_
 #define XLA_TORCH_XLA_CSRC_HELPERS_H_
 
-#include <c10/core/Scalar.h>
-
 #include <functional>
 #include <optional>
 #include <tuple>
 #include <vector>
 
+#include <c10/core/Scalar.h>
 #include <torch/csrc/lazy/core/shape.h>
 #include <torch/csrc/lazy/core/util.h>
 

diff --git a/torch_xla/csrc/init_python_bindings.cpp b/torch_xla/csrc/init_python_bindings.cpp
@@ -1,7 +1,4 @@
-#include <ATen/dlpack.h>
 #include <Python.h>
-#include <c10/core/Device.h>
-#include <c10/util/Metaprogramming.h>
 #include <google/protobuf/text_format.h>
 
 #include <algorithm>
@@ -17,6 +14,9 @@
 #include <unordered_map>
 #include <vector>
 
+#include <ATen/dlpack.h>
+#include <c10/core/Device.h>
+#include <c10/util/Metaprogramming.h>
 #include <torch/csrc/Exceptions.h>
 #include <torch/csrc/autograd/utils/wrap_outputs.h>
 #include <torch/csrc/autograd/variable.h>

diff --git a/torch_xla/csrc/ir.h b/torch_xla/csrc/ir.h
@@ -1,8 +1,6 @@
 #ifndef XLA_TORCH_XLA_CSRC_IR_H_
 #define XLA_TORCH_XLA_CSRC_IR_H_
 
-#include <ATen/core/interned_strings.h>
-
 #include <functional>
 #include <iostream>
 #include <memory>
@@ -12,6 +10,7 @@
 #include <utility>
 #include <vector>
 
+#include <ATen/core/interned_strings.h>
 #include <torch/csrc/lazy/core/hash.h>
 #include <torch/csrc/lazy/core/ir.h>
 #include <torch/csrc/lazy/core/ir_builder.h>

diff --git a/torch_xla/csrc/ops/cast.h b/torch_xla/csrc/ops/cast.h
@@ -1,10 +1,10 @@
 #ifndef XLA_TORCH_XLA_CSRC_OPS_CAST_H_
 #define XLA_TORCH_XLA_CSRC_OPS_CAST_H_
 
-#include <c10/core/ScalarType.h>
-
 #include <optional>
 
+#include <c10/core/ScalarType.h>
+
 #include "torch_xla/csrc/ir.h"
 
 namespace torch_xla {

diff --git a/torch_xla/csrc/ops/cumprod.h b/torch_xla/csrc/ops/cumprod.h
@@ -1,10 +1,10 @@
 #ifndef XLA_TORCH_XLA_CSRC_OPS_CUMPROD_H_
 #define XLA_TORCH_XLA_CSRC_OPS_CUMPROD_H_
 
-#include <c10/core/ScalarType.h>
-
 #include <optional>
 
+#include <c10/core/ScalarType.h>
+
 #include "torch_xla/csrc/ir.h"
 
 namespace torch_xla {

diff --git a/torch_xla/csrc/ops/cumsum.h b/torch_xla/csrc/ops/cumsum.h
@@ -1,10 +1,10 @@
 #ifndef XLA_TORCH_XLA_CSRC_OPS_CUMSUM_H_
 #define XLA_TORCH_XLA_CSRC_OPS_CUMSUM_H_
 
-#include <c10/core/ScalarType.h>
-
 #include <optional>
 
+#include <c10/core/ScalarType.h>
+
 #include "torch_xla/csrc/ir.h"
 
 namespace torch_xla {

diff --git a/torch_xla/csrc/ops/dot_general.cpp b/torch_xla/csrc/ops/dot_general.cpp
@@ -1,7 +1,6 @@
 #include "torch_xla/csrc/ops/dot_general.h"
 
 #include <c10/core/ScalarType.h>
-
 #include <torch/csrc/lazy/core/tensor_util.h>
 
 #include "torch_xla/csrc/dtype.h"

diff --git a/torch_xla/csrc/ops/index_ops.cpp b/torch_xla/csrc/ops/index_ops.cpp
@@ -3,7 +3,6 @@
 #include <ATen/ExpandUtils.h>
 #include <ATen/Functions.h>
 #include <ATen/ops/select_copy.h>
-
 #include <torch/csrc/lazy/core/util.h>
 
 #include "xla/permutation_util.h"

diff --git a/torch_xla/csrc/ops/log_softmax.h b/torch_xla/csrc/ops/log_softmax.h
@@ -1,10 +1,10 @@
 #ifndef XLA_TORCH_XLA_CSRC_OPS_LOG_SOFTMAX_H_
 #define XLA_TORCH_XLA_CSRC_OPS_LOG_SOFTMAX_H_
 
-#include <c10/core/ScalarType.h>
-
 #include <optional>
 
+#include <c10/core/ScalarType.h>
+
 #include "torch_xla/csrc/ir.h"
 
 namespace torch_xla {

diff --git a/torch_xla/csrc/ops/mean.h b/torch_xla/csrc/ops/mean.h
@@ -1,11 +1,11 @@
 #ifndef XLA_TORCH_XLA_CSRC_OPS_MEAN_H_
 #define XLA_TORCH_XLA_CSRC_OPS_MEAN_H_
 
-#include <c10/core/ScalarType.h>
-
 #include <optional>
 #include <vector>
 
+#include <c10/core/ScalarType.h>
+
 #include "xla/types.h"
 
 #include "torch_xla/csrc/ir.h"

diff --git a/torch_xla/csrc/ops/mse_loss.cpp b/torch_xla/csrc/ops/mse_loss.cpp
@@ -1,7 +1,6 @@
 #include "torch_xla/csrc/ops/mse_loss.h"
 
 #include <ATen/core/Reduction.h>
-
 #include <torch/csrc/lazy/core/util.h>
 
 #include "torch_xla/csrc/lowering_context.h"

diff --git a/torch_xla/csrc/ops/prod.h b/torch_xla/csrc/ops/prod.h
@@ -1,10 +1,10 @@
 #ifndef XLA_TORCH_XLA_CSRC_OPS_PROD_H_
 #define XLA_TORCH_XLA_CSRC_OPS_PROD_H_
 
-#include <c10/core/ScalarType.h>
-
 #include <optional>
 
+#include <c10/core/ScalarType.h>
+
 #include "absl/types/span.h"
 
 #include "torch_xla/csrc/ir.h"

diff --git a/torch_xla/csrc/ops/scalar.h b/torch_xla/csrc/ops/scalar.h
@@ -1,11 +1,11 @@
 #ifndef XLA_TORCH_XLA_CSRC_OPS_SCALAR_H_
 #define XLA_TORCH_XLA_CSRC_OPS_SCALAR_H_
 
+#include <iostream>
+
 #include <ATen/core/Formatting.h>
 #include <c10/core/Scalar.h>
 
-#include <iostream>
-
 #include "torch_xla/csrc/ir.h"
 #include "torch_xla/csrc/runtime/types.h"
 

diff --git a/torch_xla/csrc/ops/softmax.h b/torch_xla/csrc/ops/softmax.h
@@ -1,10 +1,10 @@
 #ifndef XLA_TORCH_XLA_CSRC_OPS_SOFTMAX_H_
 #define XLA_TORCH_XLA_CSRC_OPS_SOFTMAX_H_
 
-#include <c10/core/ScalarType.h>
-
 #include <optional>
 
+#include <c10/core/ScalarType.h>
+
 #include "torch_xla/csrc/ir.h"
 
 namespace torch_xla {

diff --git a/torch_xla/csrc/ops/sum.h b/torch_xla/csrc/ops/sum.h
@@ -1,10 +1,10 @@
 #ifndef XLA_TORCH_XLA_CSRC_OPS_SUM_H_
 #define XLA_TORCH_XLA_CSRC_OPS_SUM_H_
 
-#include <c10/core/ScalarType.h>
-
 #include <optional>
 
+#include <c10/core/ScalarType.h>
+
 #include "absl/types/span.h"
 
 #include "torch_xla/csrc/ir.h"

diff --git a/torch_xla/csrc/reduction.cpp b/torch_xla/csrc/reduction.cpp
@@ -1,10 +1,9 @@
 #include "torch_xla/csrc/reduction.h"
 
-#include <ATen/core/Reduction.h>
-
 #include <cmath>
 #include <unordered_set>
 
+#include <ATen/core/Reduction.h>
 #include <torch/csrc/lazy/core/helpers.h>
 #include <torch/csrc/lazy/core/util.h>
 

diff --git a/torch_xla/csrc/runtime/computation_client.h b/torch_xla/csrc/runtime/computation_client.h
@@ -1,15 +1,14 @@
 #ifndef XLA_CLIENT_COMPUTATION_CLIENT_H_
 #define XLA_CLIENT_COMPUTATION_CLIENT_H_
 
-#include <ATen/Tensor.h>
-
 #include <algorithm>
 #include <cmath>
 #include <map>
 #include <memory>
 #include <string>
 #include <vector>
 
+#include <ATen/Tensor.h>
 #include <torch/csrc/lazy/backend/backend_data.h>
 #include <torch/csrc/lazy/backend/lowering_context.h>
 #include <torch/csrc/lazy/core/hash.h>