Add files using upload-large-folder tool

lib/python3.12/site-packages/torch/include/ATen/ops/_adaptive_avg_pool2d_backward_ops.h

@@ -0,0 +1,39 @@
+#pragma once
+
+// @generated by torchgen/gen.py from Operator.h
+
+#include <tuple>
+#include <vector>
+
+// Forward declarations of any types needed in the operator signatures.
+// We can't directly include these classes because it will cause circular include dependencies.
+// This file is included by TensorBody.h, which defines the Tensor class.
+#include <ATen/core/ATen_fwd.h>
+
+namespace at {
+namespace _ops {
+
+
+struct TORCH_API _adaptive_avg_pool2d_backward {
+  using schema = at::Tensor (const at::Tensor &, const at::Tensor &);
+  using ptr_schema = schema*;
+  // See Note [static constexpr char* members for windows NVCC]
+  static constexpr const char* name = "aten::_adaptive_avg_pool2d_backward";
+  static constexpr const char* overload_name = "";
+  static constexpr const char* schema_str = "_adaptive_avg_pool2d_backward(Tensor grad_output, Tensor self) -> Tensor";
+  static at::Tensor call(const at::Tensor & grad_output, const at::Tensor & self);
+  static at::Tensor redispatch(c10::DispatchKeySet dispatchKeySet, const at::Tensor & grad_output, const at::Tensor & self);
+};
+
+struct TORCH_API _adaptive_avg_pool2d_backward_out {
+  using schema = at::Tensor & (const at::Tensor &, const at::Tensor &, at::Tensor &);
+  using ptr_schema = schema*;
+  // See Note [static constexpr char* members for windows NVCC]
+  static constexpr const char* name = "aten::_adaptive_avg_pool2d_backward";
+  static constexpr const char* overload_name = "out";
+  static constexpr const char* schema_str = "_adaptive_avg_pool2d_backward.out(Tensor grad_output, Tensor self, *, Tensor(a!) out) -> Tensor(a!)";
+  static at::Tensor & call(const at::Tensor & grad_output, const at::Tensor & self, at::Tensor & out);
+  static at::Tensor & redispatch(c10::DispatchKeySet dispatchKeySet, const at::Tensor & grad_output, const at::Tensor & self, at::Tensor & out);
+};
+
+}} // namespace at::_ops

lib/python3.12/site-packages/torch/include/ATen/ops/_adaptive_avg_pool3d_backward_ops.h

@@ -0,0 +1,39 @@
+#pragma once
+
+// @generated by torchgen/gen.py from Operator.h
+
+#include <tuple>
+#include <vector>
+
+// Forward declarations of any types needed in the operator signatures.
+// We can't directly include these classes because it will cause circular include dependencies.
+// This file is included by TensorBody.h, which defines the Tensor class.
+#include <ATen/core/ATen_fwd.h>
+
+namespace at {
+namespace _ops {
+
+
+struct TORCH_API _adaptive_avg_pool3d_backward {
+  using schema = at::Tensor (const at::Tensor &, const at::Tensor &);
+  using ptr_schema = schema*;
+  // See Note [static constexpr char* members for windows NVCC]
+  static constexpr const char* name = "aten::_adaptive_avg_pool3d_backward";
+  static constexpr const char* overload_name = "";
+  static constexpr const char* schema_str = "_adaptive_avg_pool3d_backward(Tensor grad_output, Tensor self) -> Tensor";
+  static at::Tensor call(const at::Tensor & grad_output, const at::Tensor & self);
+  static at::Tensor redispatch(c10::DispatchKeySet dispatchKeySet, const at::Tensor & grad_output, const at::Tensor & self);
+};
+
+struct TORCH_API _adaptive_avg_pool3d_backward_out {
+  using schema = at::Tensor & (const at::Tensor &, const at::Tensor &, at::Tensor &);
+  using ptr_schema = schema*;
+  // See Note [static constexpr char* members for windows NVCC]
+  static constexpr const char* name = "aten::_adaptive_avg_pool3d_backward";
+  static constexpr const char* overload_name = "out";
+  static constexpr const char* schema_str = "_adaptive_avg_pool3d_backward.out(Tensor grad_output, Tensor self, *, Tensor(a!) out) -> Tensor(a!)";
+  static at::Tensor & call(const at::Tensor & grad_output, const at::Tensor & self, at::Tensor & out);
+  static at::Tensor & redispatch(c10::DispatchKeySet dispatchKeySet, const at::Tensor & grad_output, const at::Tensor & self, at::Tensor & out);
+};
+
+}} // namespace at::_ops

lib/python3.12/site-packages/torch/include/ATen/ops/_addmm_activation_meta_dispatch.h

@@ -0,0 +1,25 @@
+#pragma once
+// @generated by torchgen/gen.py from DispatchKeyFunction.h
+
+// NB: The implementing C++ file is RegisterDispatchKey.cpp
+
+// The only #includes we need are for custom classes that have defaults in the C++ API
+#include <c10/core/MemoryFormat.h>
+#include <c10/core/Scalar.h>
+#include <ATen/core/Reduction.h>
+
+// Forward declarations of any types needed in the operator signatures.
+// We can't directly include these classes because it will cause circular include dependencies.
+// This file is included by TensorBody.h, which defines the Tensor class.
+#include <ATen/core/ATen_fwd.h>
+
+namespace at {
+
+namespace meta {
+
+TORCH_API at::Tensor _addmm_activation(const at::Tensor & self, const at::Tensor & mat1, const at::Tensor & mat2, const at::Scalar & beta=1, const at::Scalar & alpha=1, bool use_gelu=false);
+TORCH_API at::Tensor & _addmm_activation_out(at::Tensor & out, const at::Tensor & self, const at::Tensor & mat1, const at::Tensor & mat2, const at::Scalar & beta=1, const at::Scalar & alpha=1, bool use_gelu=false);
+TORCH_API at::Tensor & _addmm_activation_outf(const at::Tensor & self, const at::Tensor & mat1, const at::Tensor & mat2, const at::Scalar & beta, const at::Scalar & alpha, bool use_gelu, at::Tensor & out);
+
+} // namespace meta
+} // namespace at

lib/python3.12/site-packages/torch/include/ATen/ops/_addmm_activation_ops.h

@@ -0,0 +1,39 @@
+#pragma once
+
+// @generated by torchgen/gen.py from Operator.h
+
+#include <tuple>
+#include <vector>
+
+// Forward declarations of any types needed in the operator signatures.
+// We can't directly include these classes because it will cause circular include dependencies.
+// This file is included by TensorBody.h, which defines the Tensor class.
+#include <ATen/core/ATen_fwd.h>
+
+namespace at {
+namespace _ops {
+
+
+struct TORCH_API _addmm_activation_out {
+  using schema = at::Tensor & (const at::Tensor &, const at::Tensor &, const at::Tensor &, const at::Scalar &, const at::Scalar &, bool, at::Tensor &);
+  using ptr_schema = schema*;
+  // See Note [static constexpr char* members for windows NVCC]
+  static constexpr const char* name = "aten::_addmm_activation";
+  static constexpr const char* overload_name = "out";
+  static constexpr const char* schema_str = "_addmm_activation.out(Tensor self, Tensor mat1, Tensor mat2, *, Scalar beta=1, Scalar alpha=1, bool use_gelu=False, Tensor(a!) out) -> Tensor(a!)";
+  static at::Tensor & call(const at::Tensor & self, const at::Tensor & mat1, const at::Tensor & mat2, const at::Scalar & beta, const at::Scalar & alpha, bool use_gelu, at::Tensor & out);
+  static at::Tensor & redispatch(c10::DispatchKeySet dispatchKeySet, const at::Tensor & self, const at::Tensor & mat1, const at::Tensor & mat2, const at::Scalar & beta, const at::Scalar & alpha, bool use_gelu, at::Tensor & out);
+};
+
+struct TORCH_API _addmm_activation {
+  using schema = at::Tensor (const at::Tensor &, const at::Tensor &, const at::Tensor &, const at::Scalar &, const at::Scalar &, bool);
+  using ptr_schema = schema*;
+  // See Note [static constexpr char* members for windows NVCC]
+  static constexpr const char* name = "aten::_addmm_activation";
+  static constexpr const char* overload_name = "";
+  static constexpr const char* schema_str = "_addmm_activation(Tensor self, Tensor mat1, Tensor mat2, *, Scalar beta=1, Scalar alpha=1, bool use_gelu=False) -> Tensor";
+  static at::Tensor call(const at::Tensor & self, const at::Tensor & mat1, const at::Tensor & mat2, const at::Scalar & beta, const at::Scalar & alpha, bool use_gelu);
+  static at::Tensor redispatch(c10::DispatchKeySet dispatchKeySet, const at::Tensor & self, const at::Tensor & mat1, const at::Tensor & mat2, const at::Scalar & beta, const at::Scalar & alpha, bool use_gelu);
+};
+
+}} // namespace at::_ops

lib/python3.12/site-packages/torch/include/ATen/ops/_amp_foreach_non_finite_check_and_unscale.h

@@ -0,0 +1,44 @@
+#pragma once
+
+// @generated by torchgen/gen.py from Function.h
+
+#include <ATen/Context.h>
+#include <ATen/DeviceGuard.h>
+#include <ATen/TensorUtils.h>
+#include <ATen/TracerMode.h>
+#include <ATen/core/Generator.h>
+#include <ATen/core/Reduction.h>
+#include <ATen/core/Tensor.h>
+#include <c10/core/Scalar.h>
+#include <c10/core/Storage.h>
+#include <c10/core/TensorOptions.h>
+#include <c10/util/Deprecated.h>
+#include <optional>
+
+
+
+#include <ATen/ops/_amp_foreach_non_finite_check_and_unscale_ops.h>
+
+namespace at {
+
+
+// aten::_amp_foreach_non_finite_check_and_unscale_(Tensor(a!)[] self, Tensor(b!) found_inf, Tensor inv_scale) -> ()
+inline void _amp_foreach_non_finite_check_and_unscale_(at::TensorList self, at::Tensor & found_inf, const at::Tensor & inv_scale) {
+    return at::_ops::_amp_foreach_non_finite_check_and_unscale_::call(self, found_inf, inv_scale);
+}
+
+// aten::_amp_foreach_non_finite_check_and_unscale.out(Tensor[] self, Tensor(b!) found_inf, Tensor inv_scale, *, Tensor(a!)[] out) -> ()
+inline void _amp_foreach_non_finite_check_and_unscale_out(at::TensorList out, at::TensorList self, at::Tensor & found_inf, const at::Tensor & inv_scale) {
+    return at::_ops::_amp_foreach_non_finite_check_and_unscale_out::call(self, found_inf, inv_scale, out);
+}
+// aten::_amp_foreach_non_finite_check_and_unscale.out(Tensor[] self, Tensor(b!) found_inf, Tensor inv_scale, *, Tensor(a!)[] out) -> ()
+inline void _amp_foreach_non_finite_check_and_unscale_outf(at::TensorList self, at::Tensor & found_inf, const at::Tensor & inv_scale, at::TensorList out) {
+    return at::_ops::_amp_foreach_non_finite_check_and_unscale_out::call(self, found_inf, inv_scale, out);
+}
+
+// aten::_amp_foreach_non_finite_check_and_unscale(Tensor[] self, Tensor found_inf, Tensor inv_scale) -> (Tensor[] self_out, Tensor found_inf_out)
+inline ::std::tuple<::std::vector<at::Tensor>,at::Tensor> _amp_foreach_non_finite_check_and_unscale(at::TensorList self, const at::Tensor & found_inf, const at::Tensor & inv_scale) {
+    return at::_ops::_amp_foreach_non_finite_check_and_unscale::call(self, found_inf, inv_scale);
+}
+
+}

lib/python3.12/site-packages/torch/include/ATen/ops/_batch_norm_no_update_compositeexplicitautograd_dispatch.h

@@ -0,0 +1,25 @@
+#pragma once
+// @generated by torchgen/gen.py from DispatchKeyFunction.h
+
+// NB: The implementing C++ file is RegisterDispatchKey.cpp
+
+// The only #includes we need are for custom classes that have defaults in the C++ API
+#include <c10/core/MemoryFormat.h>
+#include <c10/core/Scalar.h>
+#include <ATen/core/Reduction.h>
+
+// Forward declarations of any types needed in the operator signatures.
+// We can't directly include these classes because it will cause circular include dependencies.
+// This file is included by TensorBody.h, which defines the Tensor class.
+#include <ATen/core/ATen_fwd.h>
+
+namespace at {
+
+namespace compositeexplicitautograd {
+
+TORCH_API ::std::tuple<at::Tensor,at::Tensor,at::Tensor,at::Tensor> _batch_norm_no_update(const at::Tensor & input, const ::std::optional<at::Tensor> & weight, const ::std::optional<at::Tensor> & bias, const ::std::optional<at::Tensor> & running_mean, const ::std::optional<at::Tensor> & running_var, double momentum, double eps);
+TORCH_API ::std::tuple<at::Tensor &,at::Tensor &,at::Tensor &,at::Tensor &> _batch_norm_no_update_out(at::Tensor & out0, at::Tensor & out1, at::Tensor & out2, at::Tensor & out3, const at::Tensor & input, const ::std::optional<at::Tensor> & weight, const ::std::optional<at::Tensor> & bias, const ::std::optional<at::Tensor> & running_mean, const ::std::optional<at::Tensor> & running_var, double momentum, double eps);
+TORCH_API ::std::tuple<at::Tensor &,at::Tensor &,at::Tensor &,at::Tensor &> _batch_norm_no_update_outf(const at::Tensor & input, const ::std::optional<at::Tensor> & weight, const ::std::optional<at::Tensor> & bias, const ::std::optional<at::Tensor> & running_mean, const ::std::optional<at::Tensor> & running_var, double momentum, double eps, at::Tensor & out0, at::Tensor & out1, at::Tensor & out2, at::Tensor & out3);
+
+} // namespace compositeexplicitautograd
+} // namespace at

lib/python3.12/site-packages/torch/include/ATen/ops/_batch_norm_no_update_native.h

@@ -0,0 +1,22 @@
+#pragma once
+
+// @generated by torchgen/gen.py from NativeFunction.h
+
+#include <c10/core/Scalar.h>
+#include <c10/core/Storage.h>
+#include <c10/core/TensorOptions.h>
+#include <c10/util/Deprecated.h>
+#include <optional>
+#include <c10/core/QScheme.h>
+#include <ATen/core/Reduction.h>
+#include <ATen/core/Tensor.h>
+#include <tuple>
+#include <vector>
+
+
+namespace at {
+namespace native {
+TORCH_API ::std::tuple<at::Tensor,at::Tensor,at::Tensor,at::Tensor> _batch_norm_no_update(const at::Tensor & input, const ::std::optional<at::Tensor> & weight, const ::std::optional<at::Tensor> & bias, const ::std::optional<at::Tensor> & running_mean, const ::std::optional<at::Tensor> & running_var, double momentum, double eps);
+TORCH_API ::std::tuple<at::Tensor &,at::Tensor &,at::Tensor &,at::Tensor &> _batch_norm_no_update_out(const at::Tensor & input, const ::std::optional<at::Tensor> & weight, const ::std::optional<at::Tensor> & bias, const ::std::optional<at::Tensor> & running_mean, const ::std::optional<at::Tensor> & running_var, double momentum, double eps, at::Tensor & out0, at::Tensor & out1, at::Tensor & out2, at::Tensor & out3);
+} // namespace native
+} // namespace at

lib/python3.12/site-packages/torch/include/ATen/ops/_batch_norm_with_update_cuda_dispatch.h

@@ -0,0 +1,25 @@
+#pragma once
+// @generated by torchgen/gen.py from DispatchKeyFunction.h
+
+// NB: The implementing C++ file is RegisterDispatchKey.cpp
+
+// The only #includes we need are for custom classes that have defaults in the C++ API
+#include <c10/core/MemoryFormat.h>
+#include <c10/core/Scalar.h>
+#include <ATen/core/Reduction.h>
+
+// Forward declarations of any types needed in the operator signatures.
+// We can't directly include these classes because it will cause circular include dependencies.
+// This file is included by TensorBody.h, which defines the Tensor class.
+#include <ATen/core/ATen_fwd.h>
+
+namespace at {
+
+namespace cuda {
+
+TORCH_API ::std::tuple<at::Tensor &,at::Tensor &,at::Tensor &,at::Tensor &> _batch_norm_with_update_out(at::Tensor & out, at::Tensor & save_mean, at::Tensor & save_invstd, at::Tensor & reserve, const at::Tensor & input, const ::std::optional<at::Tensor> & weight, const ::std::optional<at::Tensor> & bias, at::Tensor & running_mean, at::Tensor & running_var, double momentum, double eps);
+TORCH_API ::std::tuple<at::Tensor &,at::Tensor &,at::Tensor &,at::Tensor &> _batch_norm_with_update_outf(const at::Tensor & input, const ::std::optional<at::Tensor> & weight, const ::std::optional<at::Tensor> & bias, at::Tensor & running_mean, at::Tensor & running_var, double momentum, double eps, at::Tensor & out, at::Tensor & save_mean, at::Tensor & save_invstd, at::Tensor & reserve);
+TORCH_API ::std::tuple<at::Tensor,at::Tensor,at::Tensor,at::Tensor> _batch_norm_with_update(const at::Tensor & input, const ::std::optional<at::Tensor> & weight, const ::std::optional<at::Tensor> & bias, at::Tensor & running_mean, at::Tensor & running_var, double momentum, double eps);
+
+} // namespace cuda
+} // namespace at

lib/python3.12/site-packages/torch/include/ATen/ops/_cast_Float.h

@@ -0,0 +1,30 @@
+#pragma once
+
+// @generated by torchgen/gen.py from Function.h
+
+#include <ATen/Context.h>
+#include <ATen/DeviceGuard.h>
+#include <ATen/TensorUtils.h>
+#include <ATen/TracerMode.h>
+#include <ATen/core/Generator.h>
+#include <ATen/core/Reduction.h>
+#include <ATen/core/Tensor.h>
+#include <c10/core/Scalar.h>
+#include <c10/core/Storage.h>
+#include <c10/core/TensorOptions.h>
+#include <c10/util/Deprecated.h>
+#include <optional>
+
+
+
+#include <ATen/ops/_cast_Float_ops.h>
+
+namespace at {
+
+
+// aten::_cast_Float(Tensor self, bool non_blocking=False) -> Tensor
+inline at::Tensor _cast_Float(const at::Tensor & self, bool non_blocking=false) {
+    return at::_ops::_cast_Float::call(self, non_blocking);
+}
+
+}

lib/python3.12/site-packages/torch/include/ATen/ops/_cast_Int_native.h

@@ -0,0 +1,21 @@
+#pragma once
+
+// @generated by torchgen/gen.py from NativeFunction.h
+
+#include <c10/core/Scalar.h>
+#include <c10/core/Storage.h>
+#include <c10/core/TensorOptions.h>
+#include <c10/util/Deprecated.h>
+#include <optional>
+#include <c10/core/QScheme.h>
+#include <ATen/core/Reduction.h>
+#include <ATen/core/Tensor.h>
+#include <tuple>
+#include <vector>
+
+
+namespace at {
+namespace native {
+TORCH_API at::Tensor _cast_Int(const at::Tensor & self, bool non_blocking=false);
+} // namespace native
+} // namespace at

lib/python3.12/site-packages/torch/include/ATen/ops/_cast_Long_native.h

@@ -0,0 +1,21 @@
+#pragma once
+
+// @generated by torchgen/gen.py from NativeFunction.h
+
+#include <c10/core/Scalar.h>
+#include <c10/core/Storage.h>
+#include <c10/core/TensorOptions.h>
+#include <c10/util/Deprecated.h>
+#include <optional>
+#include <c10/core/QScheme.h>
+#include <ATen/core/Reduction.h>
+#include <ATen/core/Tensor.h>
+#include <tuple>
+#include <vector>
+
+
+namespace at {
+namespace native {
+TORCH_API at::Tensor _cast_Long(const at::Tensor & self, bool non_blocking=false);
+} // namespace native
+} // namespace at

lib/python3.12/site-packages/torch/include/ATen/ops/_choose_qparams_per_tensor_ops.h

@@ -0,0 +1,28 @@
+#pragma once
+
+// @generated by torchgen/gen.py from Operator.h
+
+#include <tuple>
+#include <vector>
+
+// Forward declarations of any types needed in the operator signatures.
+// We can't directly include these classes because it will cause circular include dependencies.
+// This file is included by TensorBody.h, which defines the Tensor class.
+#include <ATen/core/ATen_fwd.h>
+
+namespace at {
+namespace _ops {
+
+
+struct TORCH_API _choose_qparams_per_tensor {
+  using schema = ::std::tuple<double,int64_t> (const at::Tensor &, bool);
+  using ptr_schema = schema*;
+  // See Note [static constexpr char* members for windows NVCC]
+  static constexpr const char* name = "aten::_choose_qparams_per_tensor";
+  static constexpr const char* overload_name = "";
+  static constexpr const char* schema_str = "_choose_qparams_per_tensor(Tensor self, bool reduce_range=False) -> (float, int)";
+  static ::std::tuple<double,int64_t> call(const at::Tensor & self, bool reduce_range);
+  static ::std::tuple<double,int64_t> redispatch(c10::DispatchKeySet dispatchKeySet, const at::Tensor & self, bool reduce_range);
+};
+
+}} // namespace at::_ops

lib/python3.12/site-packages/torch/include/ATen/ops/_chunk_cat_native.h

@@ -0,0 +1,24 @@
+#pragma once
+
+// @generated by torchgen/gen.py from NativeFunction.h
+
+#include <c10/core/Scalar.h>
+#include <c10/core/Storage.h>
+#include <c10/core/TensorOptions.h>
+#include <c10/util/Deprecated.h>
+#include <optional>
+#include <c10/core/QScheme.h>
+#include <ATen/core/Reduction.h>
+#include <ATen/core/Tensor.h>
+#include <tuple>
+#include <vector>
+
+
+namespace at {
+namespace native {
+TORCH_API at::Tensor _chunk_cat(at::TensorList tensors, int64_t dim, int64_t num_chunks);
+TORCH_API at::Tensor & _chunk_cat_out(at::TensorList tensors, int64_t dim, int64_t num_chunks, at::Tensor & out);
+TORCH_API at::Tensor _chunk_cat_cuda(at::TensorList tensors, int64_t dim, int64_t num_chunks);
+TORCH_API at::Tensor & _chunk_cat_out_cuda(at::TensorList tensors, int64_t dim, int64_t num_chunks, at::Tensor & out);
+} // namespace native
+} // namespace at

lib/python3.12/site-packages/torch/include/ATen/ops/_conj_physical.h

@@ -0,0 +1,39 @@
+#pragma once
+
+// @generated by torchgen/gen.py from Function.h
+
+#include <ATen/Context.h>
+#include <ATen/DeviceGuard.h>
+#include <ATen/TensorUtils.h>
+#include <ATen/TracerMode.h>
+#include <ATen/core/Generator.h>
+#include <ATen/core/Reduction.h>
+#include <ATen/core/Tensor.h>
+#include <c10/core/Scalar.h>
+#include <c10/core/Storage.h>
+#include <c10/core/TensorOptions.h>
+#include <c10/util/Deprecated.h>
+#include <optional>
+
+
+
+#include <ATen/ops/_conj_physical_ops.h>
+
+namespace at {
+
+
+// aten::_conj_physical(Tensor self) -> Tensor
+inline at::Tensor _conj_physical(const at::Tensor & self) {
+    return at::_ops::_conj_physical::call(self);
+}
+
+// aten::_conj_physical.out(Tensor self, *, Tensor(a!) out) -> Tensor(a!)
+inline at::Tensor & _conj_physical_out(at::Tensor & out, const at::Tensor & self) {
+    return at::_ops::_conj_physical_out::call(self, out);
+}
+// aten::_conj_physical.out(Tensor self, *, Tensor(a!) out) -> Tensor(a!)
+inline at::Tensor & _conj_physical_outf(const at::Tensor & self, at::Tensor & out) {
+    return at::_ops::_conj_physical_out::call(self, out);
+}
+
+}

lib/python3.12/site-packages/torch/include/ATen/ops/_convert_indices_from_coo_to_csr_cpu_dispatch.h

@@ -0,0 +1,25 @@
+#pragma once
+// @generated by torchgen/gen.py from DispatchKeyFunction.h
+
+// NB: The implementing C++ file is RegisterDispatchKey.cpp
+
+// The only #includes we need are for custom classes that have defaults in the C++ API
+#include <c10/core/MemoryFormat.h>
+#include <c10/core/Scalar.h>
+#include <ATen/core/Reduction.h>
+
+// Forward declarations of any types needed in the operator signatures.
+// We can't directly include these classes because it will cause circular include dependencies.
+// This file is included by TensorBody.h, which defines the Tensor class.
+#include <ATen/core/ATen_fwd.h>
+
+namespace at {
+
+namespace cpu {
+
+TORCH_API at::Tensor _convert_indices_from_coo_to_csr(const at::Tensor & self, int64_t size, bool out_int32=false);
+TORCH_API at::Tensor & _convert_indices_from_coo_to_csr_out(at::Tensor & out, const at::Tensor & self, int64_t size, bool out_int32=false);
+TORCH_API at::Tensor & _convert_indices_from_coo_to_csr_outf(const at::Tensor & self, int64_t size, bool out_int32, at::Tensor & out);
+
+} // namespace cpu
+} // namespace at

lib/python3.12/site-packages/torch/include/ATen/ops/_convert_indices_from_csr_to_coo_compositeexplicitautogradnonfunctional_dispatch.h

@@ -0,0 +1,23 @@
+#pragma once
+// @generated by torchgen/gen.py from DispatchKeyFunction.h
+
+// NB: The implementing C++ file is RegisterDispatchKey.cpp
+
+// The only #includes we need are for custom classes that have defaults in the C++ API
+#include <c10/core/MemoryFormat.h>
+#include <c10/core/Scalar.h>
+#include <ATen/core/Reduction.h>
+
+// Forward declarations of any types needed in the operator signatures.
+// We can't directly include these classes because it will cause circular include dependencies.
+// This file is included by TensorBody.h, which defines the Tensor class.
+#include <ATen/core/ATen_fwd.h>
+
+namespace at {
+
+namespace compositeexplicitautogradnonfunctional {
+
+TORCH_API at::Tensor _convert_indices_from_csr_to_coo(const at::Tensor & crow_indices, const at::Tensor & col_indices, bool out_int32=false, bool transpose=false);
+
+} // namespace compositeexplicitautogradnonfunctional
+} // namespace at

lib/python3.12/site-packages/torch/include/ATen/ops/_convolution_compositeimplicitautograd_dispatch.h

@@ -0,0 +1,24 @@
+#pragma once
+// @generated by torchgen/gen.py from DispatchKeyFunction.h
+
+// NB: The implementing C++ file is RegisterDispatchKey.cpp
+
+// The only #includes we need are for custom classes that have defaults in the C++ API
+#include <c10/core/MemoryFormat.h>
+#include <c10/core/Scalar.h>
+#include <ATen/core/Reduction.h>
+
+// Forward declarations of any types needed in the operator signatures.
+// We can't directly include these classes because it will cause circular include dependencies.
+// This file is included by TensorBody.h, which defines the Tensor class.
+#include <ATen/core/ATen_fwd.h>
+
+namespace at {
+
+namespace compositeimplicitautograd {
+
+TORCH_API at::Tensor _convolution(const at::Tensor & input, const at::Tensor & weight, const ::std::optional<at::Tensor> & bias, at::IntArrayRef stride, at::IntArrayRef padding, at::IntArrayRef dilation, bool transposed, at::IntArrayRef output_padding, int64_t groups, bool benchmark, bool deterministic, bool cudnn_enabled);
+TORCH_API at::Tensor _convolution_symint(const at::Tensor & input, const at::Tensor & weight, const ::std::optional<at::Tensor> & bias, c10::SymIntArrayRef stride, c10::SymIntArrayRef padding, c10::SymIntArrayRef dilation, bool transposed, at::IntArrayRef output_padding, c10::SymInt groups, bool benchmark, bool deterministic, bool cudnn_enabled);
+
+} // namespace compositeimplicitautograd
+} // namespace at

lib/python3.12/site-packages/torch/include/ATen/ops/_convolution_double_backward_compositeimplicitautograd_dispatch.h

@@ -0,0 +1,24 @@
+#pragma once
+// @generated by torchgen/gen.py from DispatchKeyFunction.h
+
+// NB: The implementing C++ file is RegisterDispatchKey.cpp
+
+// The only #includes we need are for custom classes that have defaults in the C++ API
+#include <c10/core/MemoryFormat.h>
+#include <c10/core/Scalar.h>
+#include <ATen/core/Reduction.h>
+
+// Forward declarations of any types needed in the operator signatures.
+// We can't directly include these classes because it will cause circular include dependencies.
+// This file is included by TensorBody.h, which defines the Tensor class.
+#include <ATen/core/ATen_fwd.h>
+
+namespace at {
+
+namespace compositeimplicitautograd {
+
+TORCH_API ::std::tuple<at::Tensor,at::Tensor,at::Tensor> _convolution_double_backward(const ::std::optional<at::Tensor> & ggI, const ::std::optional<at::Tensor> & ggW, const ::std::optional<at::Tensor> & ggb, const at::Tensor & gO, const at::Tensor & weight, const at::Tensor & self, at::IntArrayRef stride, at::IntArrayRef padding, at::IntArrayRef dilation, bool transposed, at::IntArrayRef output_padding, int64_t groups, ::std::array<bool,3> output_mask);
+TORCH_API ::std::tuple<at::Tensor,at::Tensor,at::Tensor> _convolution_double_backward_symint(const ::std::optional<at::Tensor> & ggI, const ::std::optional<at::Tensor> & ggW, const ::std::optional<at::Tensor> & ggb, const at::Tensor & gO, const at::Tensor & weight, const at::Tensor & self, c10::SymIntArrayRef stride, c10::SymIntArrayRef padding, c10::SymIntArrayRef dilation, bool transposed, c10::SymIntArrayRef output_padding, c10::SymInt groups, ::std::array<bool,3> output_mask);
+
+} // namespace compositeimplicitautograd
+} // namespace at

lib/python3.12/site-packages/torch/include/ATen/ops/_copy_from_native.h

@@ -0,0 +1,21 @@
+#pragma once
+
+// @generated by torchgen/gen.py from NativeFunction.h
+
+#include <c10/core/Scalar.h>
+#include <c10/core/Storage.h>
+#include <c10/core/TensorOptions.h>
+#include <c10/util/Deprecated.h>
+#include <optional>
+#include <c10/core/QScheme.h>
+#include <ATen/core/Reduction.h>
+#include <ATen/core/Tensor.h>
+#include <tuple>
+#include <vector>
+
+
+namespace at {
+namespace native {
+TORCH_API at::Tensor & _copy_from_out(const at::Tensor & self, const at::Tensor & dst, bool non_blocking, at::Tensor & out);
+} // namespace native
+} // namespace at

lib/python3.12/site-packages/torch/include/ATen/ops/_ctc_loss_backward.h

@@ -0,0 +1,44 @@
+#pragma once
+
+// @generated by torchgen/gen.py from Function.h
+
+#include <ATen/Context.h>
+#include <ATen/DeviceGuard.h>
+#include <ATen/TensorUtils.h>
+#include <ATen/TracerMode.h>
+#include <ATen/core/Generator.h>
+#include <ATen/core/Reduction.h>
+#include <ATen/core/Tensor.h>
+#include <c10/core/Scalar.h>
+#include <c10/core/Storage.h>
+#include <c10/core/TensorOptions.h>
+#include <c10/util/Deprecated.h>
+#include <optional>
+
+
+
+#include <ATen/ops/_ctc_loss_backward_ops.h>
+
+namespace at {
+
+
+// aten::_ctc_loss_backward(Tensor grad, Tensor log_probs, Tensor targets, int[] input_lengths, int[] target_lengths, Tensor neg_log_likelihood, Tensor log_alpha, int blank, bool zero_infinity=False) -> Tensor
+inline at::Tensor _ctc_loss_backward(const at::Tensor & grad, const at::Tensor & log_probs, const at::Tensor & targets, at::IntArrayRef input_lengths, at::IntArrayRef target_lengths, const at::Tensor & neg_log_likelihood, const at::Tensor & log_alpha, int64_t blank, bool zero_infinity=false) {
+    return at::_ops::_ctc_loss_backward::call(grad, log_probs, targets, input_lengths, target_lengths, neg_log_likelihood, log_alpha, blank, zero_infinity);
+}
+
+// aten::_ctc_loss_backward.Tensor(Tensor grad, Tensor log_probs, Tensor targets, Tensor input_lengths, Tensor target_lengths, Tensor neg_log_likelihood, Tensor log_alpha, int blank, bool zero_infinity=False) -> Tensor
+inline at::Tensor _ctc_loss_backward(const at::Tensor & grad, const at::Tensor & log_probs, const at::Tensor & targets, const at::Tensor & input_lengths, const at::Tensor & target_lengths, const at::Tensor & neg_log_likelihood, const at::Tensor & log_alpha, int64_t blank, bool zero_infinity=false) {
+    return at::_ops::_ctc_loss_backward_Tensor::call(grad, log_probs, targets, input_lengths, target_lengths, neg_log_likelihood, log_alpha, blank, zero_infinity);
+}
+
+// aten::_ctc_loss_backward.out(Tensor grad, Tensor log_probs, Tensor targets, int[] input_lengths, int[] target_lengths, Tensor neg_log_likelihood, Tensor log_alpha, int blank, bool zero_infinity=False, *, Tensor(a!) out) -> Tensor(a!)
+inline at::Tensor & _ctc_loss_backward_out(at::Tensor & out, const at::Tensor & grad, const at::Tensor & log_probs, const at::Tensor & targets, at::IntArrayRef input_lengths, at::IntArrayRef target_lengths, const at::Tensor & neg_log_likelihood, const at::Tensor & log_alpha, int64_t blank, bool zero_infinity=false) {
+    return at::_ops::_ctc_loss_backward_out::call(grad, log_probs, targets, input_lengths, target_lengths, neg_log_likelihood, log_alpha, blank, zero_infinity, out);
+}
+// aten::_ctc_loss_backward.out(Tensor grad, Tensor log_probs, Tensor targets, int[] input_lengths, int[] target_lengths, Tensor neg_log_likelihood, Tensor log_alpha, int blank, bool zero_infinity=False, *, Tensor(a!) out) -> Tensor(a!)
+inline at::Tensor & _ctc_loss_backward_outf(const at::Tensor & grad, const at::Tensor & log_probs, const at::Tensor & targets, at::IntArrayRef input_lengths, at::IntArrayRef target_lengths, const at::Tensor & neg_log_likelihood, const at::Tensor & log_alpha, int64_t blank, bool zero_infinity, at::Tensor & out) {
+    return at::_ops::_ctc_loss_backward_out::call(grad, log_probs, targets, input_lengths, target_lengths, neg_log_likelihood, log_alpha, blank, zero_infinity, out);
+}
+
+}

lib/python3.12/site-packages/torch/include/ATen/ops/_cudnn_ctc_loss_compositeexplicitautograd_dispatch.h

@@ -0,0 +1,24 @@
+#pragma once
+// @generated by torchgen/gen.py from DispatchKeyFunction.h
+
+// NB: The implementing C++ file is RegisterDispatchKey.cpp
+
+// The only #includes we need are for custom classes that have defaults in the C++ API
+#include <c10/core/MemoryFormat.h>
+#include <c10/core/Scalar.h>
+#include <ATen/core/Reduction.h>
+
+// Forward declarations of any types needed in the operator signatures.
+// We can't directly include these classes because it will cause circular include dependencies.
+// This file is included by TensorBody.h, which defines the Tensor class.
+#include <ATen/core/ATen_fwd.h>
+
+namespace at {
+
+namespace compositeexplicitautograd {
+
+TORCH_API ::std::tuple<at::Tensor &,at::Tensor &> _cudnn_ctc_loss_out(at::Tensor & out0, at::Tensor & out1, const at::Tensor & log_probs, const at::Tensor & targets, at::IntArrayRef input_lengths, at::IntArrayRef target_lengths, int64_t blank, bool deterministic, bool zero_infinity);
+TORCH_API ::std::tuple<at::Tensor &,at::Tensor &> _cudnn_ctc_loss_outf(const at::Tensor & log_probs, const at::Tensor & targets, at::IntArrayRef input_lengths, at::IntArrayRef target_lengths, int64_t blank, bool deterministic, bool zero_infinity, at::Tensor & out0, at::Tensor & out1);
+
+} // namespace compositeexplicitautograd
+} // namespace at

lib/python3.12/site-packages/torch/include/ATen/ops/_cudnn_ctc_loss_native.h

@@ -0,0 +1,23 @@
+#pragma once
+
+// @generated by torchgen/gen.py from NativeFunction.h
+
+#include <c10/core/Scalar.h>
+#include <c10/core/Storage.h>
+#include <c10/core/TensorOptions.h>
+#include <c10/util/Deprecated.h>
+#include <optional>
+#include <c10/core/QScheme.h>
+#include <ATen/core/Reduction.h>
+#include <ATen/core/Tensor.h>
+#include <tuple>
+#include <vector>
+
+
+namespace at {
+namespace native {
+TORCH_API ::std::tuple<at::Tensor &,at::Tensor &> _cudnn_ctc_loss_out(const at::Tensor & log_probs, const at::Tensor & targets, at::IntArrayRef input_lengths, at::IntArrayRef target_lengths, int64_t blank, bool deterministic, bool zero_infinity, at::Tensor & out0, at::Tensor & out1);
+TORCH_API ::std::tuple<at::Tensor,at::Tensor> _cudnn_ctc_loss(const at::Tensor & log_probs, const at::Tensor & targets, at::IntArrayRef input_lengths, at::IntArrayRef target_lengths, int64_t blank, bool deterministic, bool zero_infinity);
+TORCH_API ::std::tuple<at::Tensor,at::Tensor> _cudnn_ctc_loss_tensor(const at::Tensor & log_probs, const at::Tensor & targets, const at::Tensor & input_lengths, const at::Tensor & target_lengths, int64_t blank, bool deterministic, bool zero_infinity);
+} // namespace native
+} // namespace at

lib/python3.12/site-packages/torch/include/ATen/ops/_cudnn_rnn_flatten_weight_ops.h

@@ -0,0 +1,39 @@
+#pragma once
+
+// @generated by torchgen/gen.py from Operator.h
+
+#include <tuple>
+#include <vector>
+
+// Forward declarations of any types needed in the operator signatures.
+// We can't directly include these classes because it will cause circular include dependencies.
+// This file is included by TensorBody.h, which defines the Tensor class.
+#include <ATen/core/ATen_fwd.h>
+
+namespace at {
+namespace _ops {
+
+
+struct TORCH_API _cudnn_rnn_flatten_weight {
+  using schema = at::Tensor (at::TensorList, int64_t, c10::SymInt, int64_t, c10::SymInt, c10::SymInt, int64_t, bool, bool);
+  using ptr_schema = schema*;
+  // See Note [static constexpr char* members for windows NVCC]
+  static constexpr const char* name = "aten::_cudnn_rnn_flatten_weight";
+  static constexpr const char* overload_name = "";
+  static constexpr const char* schema_str = "_cudnn_rnn_flatten_weight(Tensor[] weight_arr, int weight_stride0, SymInt input_size, int mode, SymInt hidden_size, SymInt proj_size, int num_layers, bool batch_first, bool bidirectional) -> Tensor";
+  static at::Tensor call(at::TensorList weight_arr, int64_t weight_stride0, c10::SymInt input_size, int64_t mode, c10::SymInt hidden_size, c10::SymInt proj_size, int64_t num_layers, bool batch_first, bool bidirectional);
+  static at::Tensor redispatch(c10::DispatchKeySet dispatchKeySet, at::TensorList weight_arr, int64_t weight_stride0, c10::SymInt input_size, int64_t mode, c10::SymInt hidden_size, c10::SymInt proj_size, int64_t num_layers, bool batch_first, bool bidirectional);
+};
+
+struct TORCH_API _cudnn_rnn_flatten_weight_out {
+  using schema = at::Tensor & (at::TensorList, int64_t, c10::SymInt, int64_t, c10::SymInt, c10::SymInt, int64_t, bool, bool, at::Tensor &);
+  using ptr_schema = schema*;
+  // See Note [static constexpr char* members for windows NVCC]
+  static constexpr const char* name = "aten::_cudnn_rnn_flatten_weight";
+  static constexpr const char* overload_name = "out";
+  static constexpr const char* schema_str = "_cudnn_rnn_flatten_weight.out(Tensor[] weight_arr, int weight_stride0, SymInt input_size, int mode, SymInt hidden_size, SymInt proj_size, int num_layers, bool batch_first, bool bidirectional, *, Tensor(a!) out) -> Tensor(a!)";
+  static at::Tensor & call(at::TensorList weight_arr, int64_t weight_stride0, c10::SymInt input_size, int64_t mode, c10::SymInt hidden_size, c10::SymInt proj_size, int64_t num_layers, bool batch_first, bool bidirectional, at::Tensor & out);
+  static at::Tensor & redispatch(c10::DispatchKeySet dispatchKeySet, at::TensorList weight_arr, int64_t weight_stride0, c10::SymInt input_size, int64_t mode, c10::SymInt hidden_size, c10::SymInt proj_size, int64_t num_layers, bool batch_first, bool bidirectional, at::Tensor & out);
+};
+
+}} // namespace at::_ops

lib/python3.12/site-packages/torch/include/ATen/ops/_cufft_clear_plan_cache_native.h

@@ -0,0 +1,21 @@
+#pragma once
+
+// @generated by torchgen/gen.py from NativeFunction.h
+
+#include <c10/core/Scalar.h>
+#include <c10/core/Storage.h>
+#include <c10/core/TensorOptions.h>
+#include <c10/util/Deprecated.h>
+#include <optional>
+#include <c10/core/QScheme.h>
+#include <ATen/core/Reduction.h>
+#include <ATen/core/Tensor.h>
+#include <tuple>
+#include <vector>
+
+
+namespace at {
+namespace native {
+TORCH_API void _cufft_clear_plan_cache(at::DeviceIndex device_index);
+} // namespace native
+} // namespace at

lib/python3.12/site-packages/torch/include/ATen/ops/_cufft_get_plan_cache_size_compositeimplicitautograd_dispatch.h

@@ -0,0 +1,23 @@
+#pragma once
+// @generated by torchgen/gen.py from DispatchKeyFunction.h
+
+// NB: The implementing C++ file is RegisterDispatchKey.cpp
+
+// The only #includes we need are for custom classes that have defaults in the C++ API
+#include <c10/core/MemoryFormat.h>
+#include <c10/core/Scalar.h>
+#include <ATen/core/Reduction.h>
+
+// Forward declarations of any types needed in the operator signatures.
+// We can't directly include these classes because it will cause circular include dependencies.
+// This file is included by TensorBody.h, which defines the Tensor class.
+#include <ATen/core/ATen_fwd.h>
+
+namespace at {
+
+namespace compositeimplicitautograd {
+
+TORCH_API int64_t _cufft_get_plan_cache_size(at::DeviceIndex device_index);
+
+} // namespace compositeimplicitautograd
+} // namespace at

lib/python3.12/site-packages/torch/include/ATen/ops/_efficientzerotensor.h

@@ -0,0 +1,113 @@
+#pragma once
+
+// @generated by torchgen/gen.py from Function.h
+
+#include <ATen/Context.h>
+#include <ATen/DeviceGuard.h>
+#include <ATen/TensorUtils.h>
+#include <ATen/TracerMode.h>
+#include <ATen/core/Generator.h>
+#include <ATen/core/Reduction.h>
+#include <ATen/core/Tensor.h>
+#include <c10/core/Scalar.h>
+#include <c10/core/Storage.h>
+#include <c10/core/TensorOptions.h>
+#include <c10/util/Deprecated.h>
+#include <optional>
+
+
+
+#include <ATen/ops/_efficientzerotensor_ops.h>
+
+namespace at {
+
+
+// aten::_efficientzerotensor(SymInt[] size, *, ScalarType? dtype=None, Layout? layout=None, Device? device=None, bool? pin_memory=None) -> Tensor
+inline at::Tensor _efficientzerotensor(at::IntArrayRef size, at::TensorOptions options={}) {
+    return at::_ops::_efficientzerotensor::call(c10::fromIntArrayRefSlow(size), c10::optTypeMetaToScalarType(options.dtype_opt()), options.layout_opt(), options.device_opt(), options.pinned_memory_opt());
+}
+namespace symint {
+  template <typename T, typename = std::enable_if_t<std::is_same_v<T, int64_t>>>
+  at::Tensor _efficientzerotensor(at::IntArrayRef size, at::TensorOptions options={}) {
+    return at::_ops::_efficientzerotensor::call(c10::fromIntArrayRefSlow(size), c10::optTypeMetaToScalarType(options.dtype_opt()), options.layout_opt(), options.device_opt(), options.pinned_memory_opt());
+  }
+}
+
+// aten::_efficientzerotensor(SymInt[] size, *, ScalarType? dtype=None, Layout? layout=None, Device? device=None, bool? pin_memory=None) -> Tensor
+inline at::Tensor _efficientzerotensor(at::IntArrayRef size, ::std::optional<at::ScalarType> dtype, ::std::optional<at::Layout> layout, ::std::optional<at::Device> device, ::std::optional<bool> pin_memory) {
+    return at::_ops::_efficientzerotensor::call(c10::fromIntArrayRefSlow(size), dtype, layout, device, pin_memory);
+}
+namespace symint {
+  template <typename T, typename = std::enable_if_t<std::is_same_v<T, int64_t>>>
+  at::Tensor _efficientzerotensor(at::IntArrayRef size, ::std::optional<at::ScalarType> dtype, ::std::optional<at::Layout> layout, ::std::optional<at::Device> device, ::std::optional<bool> pin_memory) {
+    return at::_ops::_efficientzerotensor::call(c10::fromIntArrayRefSlow(size), dtype, layout, device, pin_memory);
+  }
+}
+
+// aten::_efficientzerotensor(SymInt[] size, *, ScalarType? dtype=None, Layout? layout=None, Device? device=None, bool? pin_memory=None) -> Tensor
+inline at::Tensor _efficientzerotensor_symint(c10::SymIntArrayRef size, at::TensorOptions options={}) {
+    return at::_ops::_efficientzerotensor::call(size, c10::optTypeMetaToScalarType(options.dtype_opt()), options.layout_opt(), options.device_opt(), options.pinned_memory_opt());
+}
+namespace symint {
+  template <typename T, typename = std::enable_if_t<std::is_same_v<T, c10::SymInt>>>
+  at::Tensor _efficientzerotensor(c10::SymIntArrayRef size, at::TensorOptions options={}) {
+    return at::_ops::_efficientzerotensor::call(size, c10::optTypeMetaToScalarType(options.dtype_opt()), options.layout_opt(), options.device_opt(), options.pinned_memory_opt());
+  }
+}
+
+// aten::_efficientzerotensor(SymInt[] size, *, ScalarType? dtype=None, Layout? layout=None, Device? device=None, bool? pin_memory=None) -> Tensor
+inline at::Tensor _efficientzerotensor_symint(c10::SymIntArrayRef size, ::std::optional<at::ScalarType> dtype, ::std::optional<at::Layout> layout, ::std::optional<at::Device> device, ::std::optional<bool> pin_memory) {
+    return at::_ops::_efficientzerotensor::call(size, dtype, layout, device, pin_memory);
+}
+namespace symint {
+  template <typename T, typename = std::enable_if_t<std::is_same_v<T, c10::SymInt>>>
+  at::Tensor _efficientzerotensor(c10::SymIntArrayRef size, ::std::optional<at::ScalarType> dtype, ::std::optional<at::Layout> layout, ::std::optional<at::Device> device, ::std::optional<bool> pin_memory) {
+    return at::_ops::_efficientzerotensor::call(size, dtype, layout, device, pin_memory);
+  }
+}
+
+// aten::_efficientzerotensor.out(SymInt[] size, *, Tensor(a!) out) -> Tensor(a!)
+inline at::Tensor & _efficientzerotensor_out(at::Tensor & out, at::IntArrayRef size) {
+    return at::_ops::_efficientzerotensor_out::call(c10::fromIntArrayRefSlow(size), out);
+}
+namespace symint {
+  template <typename T, typename = std::enable_if_t<std::is_same_v<T, int64_t>>>
+  at::Tensor & _efficientzerotensor_out(at::Tensor & out, at::IntArrayRef size) {
+    return at::_ops::_efficientzerotensor_out::call(c10::fromIntArrayRefSlow(size), out);
+  }
+}
+
+// aten::_efficientzerotensor.out(SymInt[] size, *, Tensor(a!) out) -> Tensor(a!)
+inline at::Tensor & _efficientzerotensor_outf(at::IntArrayRef size, at::Tensor & out) {
+    return at::_ops::_efficientzerotensor_out::call(c10::fromIntArrayRefSlow(size), out);
+}
+namespace symint {
+  template <typename T, typename = std::enable_if_t<std::is_same_v<T, int64_t>>>
+  at::Tensor & _efficientzerotensor_outf(at::IntArrayRef size, at::Tensor & out) {
+    return at::_ops::_efficientzerotensor_out::call(c10::fromIntArrayRefSlow(size), out);
+  }
+}
+
+// aten::_efficientzerotensor.out(SymInt[] size, *, Tensor(a!) out) -> Tensor(a!)
+inline at::Tensor & _efficientzerotensor_symint_out(at::Tensor & out, c10::SymIntArrayRef size) {
+    return at::_ops::_efficientzerotensor_out::call(size, out);
+}
+namespace symint {
+  template <typename T, typename = std::enable_if_t<std::is_same_v<T, c10::SymInt>>>
+  at::Tensor & _efficientzerotensor_out(at::Tensor & out, c10::SymIntArrayRef size) {
+    return at::_ops::_efficientzerotensor_out::call(size, out);
+  }
+}
+
+// aten::_efficientzerotensor.out(SymInt[] size, *, Tensor(a!) out) -> Tensor(a!)
+inline at::Tensor & _efficientzerotensor_symint_outf(c10::SymIntArrayRef size, at::Tensor & out) {
+    return at::_ops::_efficientzerotensor_out::call(size, out);
+}
+namespace symint {
+  template <typename T, typename = std::enable_if_t<std::is_same_v<T, c10::SymInt>>>
+  at::Tensor & _efficientzerotensor_outf(c10::SymIntArrayRef size, at::Tensor & out) {
+    return at::_ops::_efficientzerotensor_out::call(size, out);
+  }
+}
+
+}

lib/python3.12/site-packages/torch/include/ATen/ops/_efficientzerotensor_meta_dispatch.h

@@ -0,0 +1,26 @@
+#pragma once
+// @generated by torchgen/gen.py from DispatchKeyFunction.h
+
+// NB: The implementing C++ file is RegisterDispatchKey.cpp
+
+// The only #includes we need are for custom classes that have defaults in the C++ API
+#include <c10/core/MemoryFormat.h>
+#include <c10/core/Scalar.h>
+#include <ATen/core/Reduction.h>
+
+// Forward declarations of any types needed in the operator signatures.
+// We can't directly include these classes because it will cause circular include dependencies.
+// This file is included by TensorBody.h, which defines the Tensor class.
+#include <ATen/core/ATen_fwd.h>
+
+namespace at {
+
+namespace meta {
+
+TORCH_API at::Tensor _efficientzerotensor(at::IntArrayRef size, at::TensorOptions options={});
+TORCH_API at::Tensor _efficientzerotensor(at::IntArrayRef size, ::std::optional<at::ScalarType> dtype, ::std::optional<at::Layout> layout, ::std::optional<at::Device> device, ::std::optional<bool> pin_memory);
+TORCH_API at::Tensor _efficientzerotensor_symint(c10::SymIntArrayRef size, at::TensorOptions options={});
+TORCH_API at::Tensor _efficientzerotensor_symint(c10::SymIntArrayRef size, ::std::optional<at::ScalarType> dtype, ::std::optional<at::Layout> layout, ::std::optional<at::Device> device, ::std::optional<bool> pin_memory);
+
+} // namespace meta
+} // namespace at

lib/python3.12/site-packages/torch/include/ATen/ops/_embedding_bag_cuda_dispatch.h

@@ -0,0 +1,23 @@
+#pragma once
+// @generated by torchgen/gen.py from DispatchKeyFunction.h
+
+// NB: The implementing C++ file is RegisterDispatchKey.cpp
+
+// The only #includes we need are for custom classes that have defaults in the C++ API
+#include <c10/core/MemoryFormat.h>
+#include <c10/core/Scalar.h>
+#include <ATen/core/Reduction.h>
+
+// Forward declarations of any types needed in the operator signatures.
+// We can't directly include these classes because it will cause circular include dependencies.
+// This file is included by TensorBody.h, which defines the Tensor class.
+#include <ATen/core/ATen_fwd.h>
+
+namespace at {
+
+namespace cuda {
+
+TORCH_API ::std::tuple<at::Tensor,at::Tensor,at::Tensor,at::Tensor> _embedding_bag(const at::Tensor & weight, const at::Tensor & indices, const at::Tensor & offsets, bool scale_grad_by_freq=false, int64_t mode=0, bool sparse=false, const ::std::optional<at::Tensor> & per_sample_weights={}, bool include_last_offset=false, int64_t padding_idx=-1);
+
+} // namespace cuda
+} // namespace at

lib/python3.12/site-packages/torch/include/ATen/ops/_embedding_bag_forward_only_ops.h

@@ -0,0 +1,39 @@
+#pragma once
+
+// @generated by torchgen/gen.py from Operator.h
+
+#include <tuple>
+#include <vector>
+
+// Forward declarations of any types needed in the operator signatures.
+// We can't directly include these classes because it will cause circular include dependencies.
+// This file is included by TensorBody.h, which defines the Tensor class.
+#include <ATen/core/ATen_fwd.h>
+
+namespace at {
+namespace _ops {
+
+
+struct TORCH_API _embedding_bag_forward_only {
+  using schema = ::std::tuple<at::Tensor,at::Tensor,at::Tensor,at::Tensor> (const at::Tensor &, const at::Tensor &, const at::Tensor &, bool, int64_t, bool, const ::std::optional<at::Tensor> &, bool, int64_t);
+  using ptr_schema = schema*;
+  // See Note [static constexpr char* members for windows NVCC]
+  static constexpr const char* name = "aten::_embedding_bag_forward_only";
+  static constexpr const char* overload_name = "";
+  static constexpr const char* schema_str = "_embedding_bag_forward_only(Tensor weight, Tensor indices, Tensor offsets, bool scale_grad_by_freq=False, int mode=0, bool sparse=False, Tensor? per_sample_weights=None, bool include_last_offset=False, int padding_idx=-1) -> (Tensor, Tensor, Tensor, Tensor)";
+  static ::std::tuple<at::Tensor,at::Tensor,at::Tensor,at::Tensor> call(const at::Tensor & weight, const at::Tensor & indices, const at::Tensor & offsets, bool scale_grad_by_freq, int64_t mode, bool sparse, const ::std::optional<at::Tensor> & per_sample_weights, bool include_last_offset, int64_t padding_idx);
+  static ::std::tuple<at::Tensor,at::Tensor,at::Tensor,at::Tensor> redispatch(c10::DispatchKeySet dispatchKeySet, const at::Tensor & weight, const at::Tensor & indices, const at::Tensor & offsets, bool scale_grad_by_freq, int64_t mode, bool sparse, const ::std::optional<at::Tensor> & per_sample_weights, bool include_last_offset, int64_t padding_idx);
+};
+
+struct TORCH_API _embedding_bag_forward_only_out {
+  using schema = ::std::tuple<at::Tensor &,at::Tensor &,at::Tensor &,at::Tensor &> (const at::Tensor &, const at::Tensor &, const at::Tensor &, bool, int64_t, bool, const ::std::optional<at::Tensor> &, bool, int64_t, at::Tensor &, at::Tensor &, at::Tensor &, at::Tensor &);
+  using ptr_schema = schema*;
+  // See Note [static constexpr char* members for windows NVCC]
+  static constexpr const char* name = "aten::_embedding_bag_forward_only";
+  static constexpr const char* overload_name = "out";
+  static constexpr const char* schema_str = "_embedding_bag_forward_only.out(Tensor weight, Tensor indices, Tensor offsets, bool scale_grad_by_freq=False, int mode=0, bool sparse=False, Tensor? per_sample_weights=None, bool include_last_offset=False, int padding_idx=-1, *, Tensor(a!) out0, Tensor(b!) out1, Tensor(c!) out2, Tensor(d!) out3) -> (Tensor(a!), Tensor(b!), Tensor(c!), Tensor(d!))";
+  static ::std::tuple<at::Tensor &,at::Tensor &,at::Tensor &,at::Tensor &> call(const at::Tensor & weight, const at::Tensor & indices, const at::Tensor & offsets, bool scale_grad_by_freq, int64_t mode, bool sparse, const ::std::optional<at::Tensor> & per_sample_weights, bool include_last_offset, int64_t padding_idx, at::Tensor & out0, at::Tensor & out1, at::Tensor & out2, at::Tensor & out3);
+  static ::std::tuple<at::Tensor &,at::Tensor &,at::Tensor &,at::Tensor &> redispatch(c10::DispatchKeySet dispatchKeySet, const at::Tensor & weight, const at::Tensor & indices, const at::Tensor & offsets, bool scale_grad_by_freq, int64_t mode, bool sparse, const ::std::optional<at::Tensor> & per_sample_weights, bool include_last_offset, int64_t padding_idx, at::Tensor & out0, at::Tensor & out1, at::Tensor & out2, at::Tensor & out3);
+};
+
+}} // namespace at::_ops

lib/python3.12/site-packages/torch/include/ATen/ops/_empty_affine_quantized_cpu_dispatch.h

@@ -0,0 +1,26 @@
+#pragma once
+// @generated by torchgen/gen.py from DispatchKeyFunction.h
+
+// NB: The implementing C++ file is RegisterDispatchKey.cpp
+
+// The only #includes we need are for custom classes that have defaults in the C++ API
+#include <c10/core/MemoryFormat.h>
+#include <c10/core/Scalar.h>
+#include <ATen/core/Reduction.h>
+
+// Forward declarations of any types needed in the operator signatures.
+// We can't directly include these classes because it will cause circular include dependencies.
+// This file is included by TensorBody.h, which defines the Tensor class.
+#include <ATen/core/ATen_fwd.h>
+
+namespace at {
+
+namespace cpu {
+
+TORCH_API at::Tensor _empty_affine_quantized(at::IntArrayRef size, at::TensorOptions options={}, double scale=1, int64_t zero_point=0, ::std::optional<at::MemoryFormat> memory_format=c10::MemoryFormat::Contiguous);
+TORCH_API at::Tensor _empty_affine_quantized(at::IntArrayRef size, ::std::optional<at::ScalarType> dtype, ::std::optional<at::Layout> layout, ::std::optional<at::Device> device, ::std::optional<bool> pin_memory, double scale, int64_t zero_point, ::std::optional<at::MemoryFormat> memory_format);
+TORCH_API at::Tensor _empty_affine_quantized_symint(c10::SymIntArrayRef size, at::TensorOptions options={}, double scale=1, int64_t zero_point=0, ::std::optional<at::MemoryFormat> memory_format=c10::MemoryFormat::Contiguous);
+TORCH_API at::Tensor _empty_affine_quantized_symint(c10::SymIntArrayRef size, ::std::optional<at::ScalarType> dtype, ::std::optional<at::Layout> layout, ::std::optional<at::Device> device, ::std::optional<bool> pin_memory, double scale, int64_t zero_point, ::std::optional<at::MemoryFormat> memory_format);
+
+} // namespace cpu
+} // namespace at

lib/python3.12/site-packages/torch/include/ATen/ops/_fake_quantize_learnable_per_tensor_affine_backward_ops.h

@@ -0,0 +1,28 @@
+#pragma once
+
+// @generated by torchgen/gen.py from Operator.h
+
+#include <tuple>
+#include <vector>
+
+// Forward declarations of any types needed in the operator signatures.
+// We can't directly include these classes because it will cause circular include dependencies.
+// This file is included by TensorBody.h, which defines the Tensor class.
+#include <ATen/core/ATen_fwd.h>
+
+namespace at {
+namespace _ops {
+
+
+struct TORCH_API _fake_quantize_learnable_per_tensor_affine_backward {
+  using schema = ::std::tuple<at::Tensor,at::Tensor,at::Tensor> (const at::Tensor &, const at::Tensor &, const at::Tensor &, const at::Tensor &, int64_t, int64_t, double);
+  using ptr_schema = schema*;
+  // See Note [static constexpr char* members for windows NVCC]
+  static constexpr const char* name = "aten::_fake_quantize_learnable_per_tensor_affine_backward";
+  static constexpr const char* overload_name = "";
+  static constexpr const char* schema_str = "_fake_quantize_learnable_per_tensor_affine_backward(Tensor grad, Tensor self, Tensor scale, Tensor zero_point, int quant_min, int quant_max, float grad_factor=1.0) -> (Tensor, Tensor, Tensor)";
+  static ::std::tuple<at::Tensor,at::Tensor,at::Tensor> call(const at::Tensor & grad, const at::Tensor & self, const at::Tensor & scale, const at::Tensor & zero_point, int64_t quant_min, int64_t quant_max, double grad_factor);
+  static ::std::tuple<at::Tensor,at::Tensor,at::Tensor> redispatch(c10::DispatchKeySet dispatchKeySet, const at::Tensor & grad, const at::Tensor & self, const at::Tensor & scale, const at::Tensor & zero_point, int64_t quant_min, int64_t quant_max, double grad_factor);
+};
+
+}} // namespace at::_ops

lib/python3.12/site-packages/torch/include/ATen/ops/_fake_quantize_per_tensor_affine_cachemask_tensor_qparams_native.h

@@ -0,0 +1,22 @@
+#pragma once
+
+// @generated by torchgen/gen.py from NativeFunction.h
+
+#include <c10/core/Scalar.h>
+#include <c10/core/Storage.h>
+#include <c10/core/TensorOptions.h>
+#include <c10/util/Deprecated.h>
+#include <optional>
+#include <c10/core/QScheme.h>
+#include <ATen/core/Reduction.h>
+#include <ATen/core/Tensor.h>
+#include <tuple>
+#include <vector>
+
+
+namespace at {
+namespace native {
+TORCH_API ::std::tuple<at::Tensor &,at::Tensor &> _fake_quantize_per_tensor_affine_cachemask_tensor_qparams_out(const at::Tensor & self, const at::Tensor & scale, const at::Tensor & zero_point, const at::Tensor & fake_quant_enabled, int64_t quant_min, int64_t quant_max, at::Tensor & out0, at::Tensor & out1);
+TORCH_API ::std::tuple<at::Tensor,at::Tensor> _fake_quantize_per_tensor_affine_cachemask_tensor_qparams(const at::Tensor & self, const at::Tensor & scale, const at::Tensor & zero_point, const at::Tensor & fake_quant_enabled, int64_t quant_min, int64_t quant_max);
+} // namespace native
+} // namespace at

lib/python3.12/site-packages/torch/include/ATen/ops/_fft_r2c_cpu_dispatch.h

@@ -0,0 +1,25 @@
+#pragma once
+// @generated by torchgen/gen.py from DispatchKeyFunction.h
+
+// NB: The implementing C++ file is RegisterDispatchKey.cpp
+
+// The only #includes we need are for custom classes that have defaults in the C++ API
+#include <c10/core/MemoryFormat.h>
+#include <c10/core/Scalar.h>
+#include <ATen/core/Reduction.h>
+
+// Forward declarations of any types needed in the operator signatures.
+// We can't directly include these classes because it will cause circular include dependencies.
+// This file is included by TensorBody.h, which defines the Tensor class.
+#include <ATen/core/ATen_fwd.h>
+
+namespace at {
+
+namespace cpu {
+
+TORCH_API at::Tensor _fft_r2c(const at::Tensor & self, at::IntArrayRef dim, int64_t normalization, bool onesided);
+TORCH_API at::Tensor & _fft_r2c_out(at::Tensor & out, const at::Tensor & self, at::IntArrayRef dim, int64_t normalization, bool onesided);
+TORCH_API at::Tensor & _fft_r2c_outf(const at::Tensor & self, at::IntArrayRef dim, int64_t normalization, bool onesided, at::Tensor & out);
+
+} // namespace cpu
+} // namespace at

lib/python3.12/site-packages/torch/include/ATen/ops/_flash_attention_backward_native.h

@@ -0,0 +1,21 @@
+#pragma once
+
+// @generated by torchgen/gen.py from NativeFunction.h
+
+#include <c10/core/Scalar.h>
+#include <c10/core/Storage.h>
+#include <c10/core/TensorOptions.h>
+#include <c10/util/Deprecated.h>
+#include <optional>
+#include <c10/core/QScheme.h>
+#include <ATen/core/Reduction.h>
+#include <ATen/core/Tensor.h>
+#include <tuple>
+#include <vector>
+
+
+namespace at {
+namespace native {
+TORCH_API ::std::tuple<at::Tensor,at::Tensor,at::Tensor> _flash_attention_backward(const at::Tensor & grad_out, const at::Tensor & query, const at::Tensor & key, const at::Tensor & value, const at::Tensor & out, const at::Tensor & logsumexp, const at::Tensor & cum_seq_q, const at::Tensor & cum_seq_k, int64_t max_q, int64_t max_k, double dropout_p, bool is_causal, const at::Tensor & philox_seed, const at::Tensor & philox_offset, ::std::optional<double> scale=::std::nullopt, ::std::optional<int64_t> window_size_left=::std::nullopt, ::std::optional<int64_t> window_size_right=::std::nullopt);
+} // namespace native
+} // namespace at

lib/python3.12/site-packages/torch/include/ATen/ops/_flash_attention_forward_native.h

@@ -0,0 +1,21 @@
+#pragma once
+
+// @generated by torchgen/gen.py from NativeFunction.h
+
+#include <c10/core/Scalar.h>
+#include <c10/core/Storage.h>
+#include <c10/core/TensorOptions.h>
+#include <c10/util/Deprecated.h>
+#include <optional>
+#include <c10/core/QScheme.h>
+#include <ATen/core/Reduction.h>
+#include <ATen/core/Tensor.h>
+#include <tuple>
+#include <vector>
+
+
+namespace at {
+namespace native {
+TORCH_API ::std::tuple<at::Tensor,at::Tensor,at::Tensor,at::Tensor,at::Tensor> _flash_attention_forward(const at::Tensor & query, const at::Tensor & key, const at::Tensor & value, const ::std::optional<at::Tensor> & cum_seq_q, const ::std::optional<at::Tensor> & cum_seq_k, int64_t max_q, int64_t max_k, double dropout_p, bool is_causal, bool return_debug_mask, ::std::optional<double> scale=::std::nullopt, ::std::optional<int64_t> window_size_left=::std::nullopt, ::std::optional<int64_t> window_size_right=::std::nullopt, const ::std::optional<at::Tensor> & seqused_k={}, const ::std::optional<at::Tensor> & alibi_slopes={});
+} // namespace native
+} // namespace at

lib/python3.12/site-packages/torch/include/ATen/ops/_foreach_add_ops.h

@@ -0,0 +1,149 @@
+#pragma once
+
+// @generated by torchgen/gen.py from Operator.h
+
+#include <tuple>
+#include <vector>
+
+// Forward declarations of any types needed in the operator signatures.
+// We can't directly include these classes because it will cause circular include dependencies.
+// This file is included by TensorBody.h, which defines the Tensor class.
+#include <ATen/core/ATen_fwd.h>
+
+namespace at {
+namespace _ops {
+
+
+struct TORCH_API _foreach_add_Scalar {
+  using schema = ::std::vector<at::Tensor> (at::TensorList, const at::Scalar &);
+  using ptr_schema = schema*;
+  // See Note [static constexpr char* members for windows NVCC]
+  static constexpr const char* name = "aten::_foreach_add";
+  static constexpr const char* overload_name = "Scalar";
+  static constexpr const char* schema_str = "_foreach_add.Scalar(Tensor[] self, Scalar scalar) -> Tensor[]";
+  static ::std::vector<at::Tensor> call(at::TensorList self, const at::Scalar & scalar);
+  static ::std::vector<at::Tensor> redispatch(c10::DispatchKeySet dispatchKeySet, at::TensorList self, const at::Scalar & scalar);
+};
+
+struct TORCH_API _foreach_add__Scalar {
+  using schema = void (at::TensorList, const at::Scalar &);
+  using ptr_schema = schema*;
+  // See Note [static constexpr char* members for windows NVCC]
+  static constexpr const char* name = "aten::_foreach_add_";
+  static constexpr const char* overload_name = "Scalar";
+  static constexpr const char* schema_str = "_foreach_add_.Scalar(Tensor(a!)[] self, Scalar scalar) -> ()";
+  static void call(at::TensorList self, const at::Scalar & scalar);
+  static void redispatch(c10::DispatchKeySet dispatchKeySet, at::TensorList self, const at::Scalar & scalar);
+};
+
+struct TORCH_API _foreach_add_List {
+  using schema = ::std::vector<at::Tensor> (at::TensorList, at::TensorList, const at::Scalar &);
+  using ptr_schema = schema*;
+  // See Note [static constexpr char* members for windows NVCC]
+  static constexpr const char* name = "aten::_foreach_add";
+  static constexpr const char* overload_name = "List";
+  static constexpr const char* schema_str = "_foreach_add.List(Tensor[] self, Tensor[] other, *, Scalar alpha=1) -> Tensor[]";
+  static ::std::vector<at::Tensor> call(at::TensorList self, at::TensorList other, const at::Scalar & alpha);
+  static ::std::vector<at::Tensor> redispatch(c10::DispatchKeySet dispatchKeySet, at::TensorList self, at::TensorList other, const at::Scalar & alpha);
+};
+
+struct TORCH_API _foreach_add__List {
+  using schema = void (at::TensorList, at::TensorList, const at::Scalar &);
+  using ptr_schema = schema*;
+  // See Note [static constexpr char* members for windows NVCC]
+  static constexpr const char* name = "aten::_foreach_add_";
+  static constexpr const char* overload_name = "List";
+  static constexpr const char* schema_str = "_foreach_add_.List(Tensor(a!)[] self, Tensor[] other, *, Scalar alpha=1) -> ()";
+  static void call(at::TensorList self, at::TensorList other, const at::Scalar & alpha);
+  static void redispatch(c10::DispatchKeySet dispatchKeySet, at::TensorList self, at::TensorList other, const at::Scalar & alpha);
+};
+
+struct TORCH_API _foreach_add_ScalarList {
+  using schema = ::std::vector<at::Tensor> (at::TensorList, at::ArrayRef<at::Scalar>);
+  using ptr_schema = schema*;
+  // See Note [static constexpr char* members for windows NVCC]
+  static constexpr const char* name = "aten::_foreach_add";
+  static constexpr const char* overload_name = "ScalarList";
+  static constexpr const char* schema_str = "_foreach_add.ScalarList(Tensor[] self, Scalar[] scalars) -> Tensor[]";
+  static ::std::vector<at::Tensor> call(at::TensorList self, at::ArrayRef<at::Scalar> scalars);
+  static ::std::vector<at::Tensor> redispatch(c10::DispatchKeySet dispatchKeySet, at::TensorList self, at::ArrayRef<at::Scalar> scalars);
+};
+
+struct TORCH_API _foreach_add__ScalarList {
+  using schema = void (at::TensorList, at::ArrayRef<at::Scalar>);
+  using ptr_schema = schema*;
+  // See Note [static constexpr char* members for windows NVCC]
+  static constexpr const char* name = "aten::_foreach_add_";
+  static constexpr const char* overload_name = "ScalarList";
+  static constexpr const char* schema_str = "_foreach_add_.ScalarList(Tensor(a!)[] self, Scalar[] scalars) -> ()";
+  static void call(at::TensorList self, at::ArrayRef<at::Scalar> scalars);
+  static void redispatch(c10::DispatchKeySet dispatchKeySet, at::TensorList self, at::ArrayRef<at::Scalar> scalars);
+};
+
+struct TORCH_API _foreach_add_Tensor {
+  using schema = ::std::vector<at::Tensor> (at::TensorList, const at::Tensor &, const at::Scalar &);
+  using ptr_schema = schema*;
+  // See Note [static constexpr char* members for windows NVCC]
+  static constexpr const char* name = "aten::_foreach_add";
+  static constexpr const char* overload_name = "Tensor";
+  static constexpr const char* schema_str = "_foreach_add.Tensor(Tensor[] self, Tensor other, *, Scalar alpha=1) -> Tensor[]";
+  static ::std::vector<at::Tensor> call(at::TensorList self, const at::Tensor & other, const at::Scalar & alpha);
+  static ::std::vector<at::Tensor> redispatch(c10::DispatchKeySet dispatchKeySet, at::TensorList self, const at::Tensor & other, const at::Scalar & alpha);
+};
+
+struct TORCH_API _foreach_add__Tensor {
+  using schema = void (at::TensorList, const at::Tensor &, const at::Scalar &);
+  using ptr_schema = schema*;
+  // See Note [static constexpr char* members for windows NVCC]
+  static constexpr const char* name = "aten::_foreach_add_";
+  static constexpr const char* overload_name = "Tensor";
+  static constexpr const char* schema_str = "_foreach_add_.Tensor(Tensor(a!)[] self, Tensor other, *, Scalar alpha=1) -> ()";
+  static void call(at::TensorList self, const at::Tensor & other, const at::Scalar & alpha);
+  static void redispatch(c10::DispatchKeySet dispatchKeySet, at::TensorList self, const at::Tensor & other, const at::Scalar & alpha);
+};
+
+struct TORCH_API _foreach_add_Scalar_out {
+  using schema = void (at::TensorList, const at::Scalar &, at::TensorList);
+  using ptr_schema = schema*;
+  // See Note [static constexpr char* members for windows NVCC]
+  static constexpr const char* name = "aten::_foreach_add";
+  static constexpr const char* overload_name = "Scalar_out";
+  static constexpr const char* schema_str = "_foreach_add.Scalar_out(Tensor[] self, Scalar scalar, *, Tensor(a!)[] out) -> ()";
+  static void call(at::TensorList self, const at::Scalar & scalar, at::TensorList out);
+  static void redispatch(c10::DispatchKeySet dispatchKeySet, at::TensorList self, const at::Scalar & scalar, at::TensorList out);
+};
+
+struct TORCH_API _foreach_add_List_out {
+  using schema = void (at::TensorList, at::TensorList, const at::Scalar &, at::TensorList);
+  using ptr_schema = schema*;
+  // See Note [static constexpr char* members for windows NVCC]
+  static constexpr const char* name = "aten::_foreach_add";
+  static constexpr const char* overload_name = "List_out";
+  static constexpr const char* schema_str = "_foreach_add.List_out(Tensor[] self, Tensor[] other, *, Scalar alpha=1, Tensor(a!)[] out) -> ()";
+  static void call(at::TensorList self, at::TensorList other, const at::Scalar & alpha, at::TensorList out);
+  static void redispatch(c10::DispatchKeySet dispatchKeySet, at::TensorList self, at::TensorList other, const at::Scalar & alpha, at::TensorList out);
+};
+
+struct TORCH_API _foreach_add_ScalarList_out {
+  using schema = void (at::TensorList, at::ArrayRef<at::Scalar>, at::TensorList);
+  using ptr_schema = schema*;
+  // See Note [static constexpr char* members for windows NVCC]
+  static constexpr const char* name = "aten::_foreach_add";
+  static constexpr const char* overload_name = "ScalarList_out";
+  static constexpr const char* schema_str = "_foreach_add.ScalarList_out(Tensor[] self, Scalar[] scalars, *, Tensor(a!)[] out) -> ()";
+  static void call(at::TensorList self, at::ArrayRef<at::Scalar> scalars, at::TensorList out);
+  static void redispatch(c10::DispatchKeySet dispatchKeySet, at::TensorList self, at::ArrayRef<at::Scalar> scalars, at::TensorList out);
+};
+
+struct TORCH_API _foreach_add_Tensor_out {
+  using schema = void (at::TensorList, const at::Tensor &, const at::Scalar &, at::TensorList);
+  using ptr_schema = schema*;
+  // See Note [static constexpr char* members for windows NVCC]
+  static constexpr const char* name = "aten::_foreach_add";
+  static constexpr const char* overload_name = "Tensor_out";
+  static constexpr const char* schema_str = "_foreach_add.Tensor_out(Tensor[] self, Tensor other, *, Scalar alpha=1, Tensor(a!)[] out) -> ()";
+  static void call(at::TensorList self, const at::Tensor & other, const at::Scalar & alpha, at::TensorList out);
+  static void redispatch(c10::DispatchKeySet dispatchKeySet, at::TensorList self, const at::Tensor & other, const at::Scalar & alpha, at::TensorList out);
+};
+
+}} // namespace at::_ops

lib/python3.12/site-packages/torch/include/ATen/ops/_foreach_copy_native.h

@@ -0,0 +1,24 @@
+#pragma once
+
+// @generated by torchgen/gen.py from NativeFunction.h
+
+#include <c10/core/Scalar.h>
+#include <c10/core/Storage.h>
+#include <c10/core/TensorOptions.h>
+#include <c10/util/Deprecated.h>
+#include <optional>
+#include <c10/core/QScheme.h>
+#include <ATen/core/Reduction.h>
+#include <ATen/core/Tensor.h>
+#include <tuple>
+#include <vector>
+
+
+namespace at {
+namespace native {
+TORCH_API ::std::vector<at::Tensor> _foreach_copy(at::TensorList self, at::TensorList src, bool non_blocking=false);
+TORCH_API void _foreach_copy_out(at::TensorList self, at::TensorList src, bool non_blocking, at::TensorList out);
+TORCH_API void foreach_tensor_copy_list_kernel_slow_(at::TensorList self, at::TensorList src, bool non_blocking=false);
+TORCH_API void foreach_tensor_copy_list_kernel_cuda_(at::TensorList self, at::TensorList src, bool non_blocking=false);
+} // namespace native
+} // namespace at

lib/python3.12/site-packages/torch/include/ATen/ops/_foreach_copy_ops.h

@@ -0,0 +1,50 @@
+#pragma once
+
+// @generated by torchgen/gen.py from Operator.h
+
+#include <tuple>
+#include <vector>
+
+// Forward declarations of any types needed in the operator signatures.
+// We can't directly include these classes because it will cause circular include dependencies.
+// This file is included by TensorBody.h, which defines the Tensor class.
+#include <ATen/core/ATen_fwd.h>
+
+namespace at {
+namespace _ops {
+
+
+struct TORCH_API _foreach_copy_ {
+  using schema = void (at::TensorList, at::TensorList, bool);
+  using ptr_schema = schema*;
+  // See Note [static constexpr char* members for windows NVCC]
+  static constexpr const char* name = "aten::_foreach_copy_";
+  static constexpr const char* overload_name = "";
+  static constexpr const char* schema_str = "_foreach_copy_(Tensor(a!)[] self, Tensor[] src, bool non_blocking=False) -> ()";
+  static void call(at::TensorList self, at::TensorList src, bool non_blocking);
+  static void redispatch(c10::DispatchKeySet dispatchKeySet, at::TensorList self, at::TensorList src, bool non_blocking);
+};
+
+struct TORCH_API _foreach_copy {
+  using schema = ::std::vector<at::Tensor> (at::TensorList, at::TensorList, bool);
+  using ptr_schema = schema*;
+  // See Note [static constexpr char* members for windows NVCC]
+  static constexpr const char* name = "aten::_foreach_copy";
+  static constexpr const char* overload_name = "";
+  static constexpr const char* schema_str = "_foreach_copy(Tensor[] self, Tensor[] src, bool non_blocking=False) -> Tensor[] self_out";
+  static ::std::vector<at::Tensor> call(at::TensorList self, at::TensorList src, bool non_blocking);
+  static ::std::vector<at::Tensor> redispatch(c10::DispatchKeySet dispatchKeySet, at::TensorList self, at::TensorList src, bool non_blocking);
+};
+
+struct TORCH_API _foreach_copy_out {
+  using schema = void (at::TensorList, at::TensorList, bool, at::TensorList);
+  using ptr_schema = schema*;
+  // See Note [static constexpr char* members for windows NVCC]
+  static constexpr const char* name = "aten::_foreach_copy";
+  static constexpr const char* overload_name = "out";
+  static constexpr const char* schema_str = "_foreach_copy.out(Tensor[] self, Tensor[] src, bool non_blocking=False, *, Tensor(a!)[] out) -> ()";
+  static void call(at::TensorList self, at::TensorList src, bool non_blocking, at::TensorList out);
+  static void redispatch(c10::DispatchKeySet dispatchKeySet, at::TensorList self, at::TensorList src, bool non_blocking, at::TensorList out);
+};
+
+}} // namespace at::_ops

lib/python3.12/site-packages/torch/include/ATen/ops/_foreach_div_compositeexplicitautograd_dispatch.h

@@ -0,0 +1,38 @@
+#pragma once
+// @generated by torchgen/gen.py from DispatchKeyFunction.h
+
+// NB: The implementing C++ file is RegisterDispatchKey.cpp
+
+// The only #includes we need are for custom classes that have defaults in the C++ API
+#include <c10/core/MemoryFormat.h>
+#include <c10/core/Scalar.h>
+#include <ATen/core/Reduction.h>
+
+// Forward declarations of any types needed in the operator signatures.
+// We can't directly include these classes because it will cause circular include dependencies.
+// This file is included by TensorBody.h, which defines the Tensor class.
+#include <ATen/core/ATen_fwd.h>
+
+namespace at {
+
+namespace compositeexplicitautograd {
+
+TORCH_API ::std::vector<at::Tensor> _foreach_div(at::TensorList self, const at::Scalar & scalar);
+TORCH_API void _foreach_div_out(at::TensorList out, at::TensorList self, const at::Scalar & scalar);
+TORCH_API void _foreach_div_outf(at::TensorList self, const at::Scalar & scalar, at::TensorList out);
+TORCH_API void _foreach_div_(at::TensorList self, const at::Scalar & scalar);
+TORCH_API ::std::vector<at::Tensor> _foreach_div(at::TensorList self, at::TensorList other);
+TORCH_API void _foreach_div_out(at::TensorList out, at::TensorList self, at::TensorList other);
+TORCH_API void _foreach_div_outf(at::TensorList self, at::TensorList other, at::TensorList out);
+TORCH_API void _foreach_div_(at::TensorList self, at::TensorList other);
+TORCH_API ::std::vector<at::Tensor> _foreach_div(at::TensorList self, at::ArrayRef<at::Scalar> scalars);
+TORCH_API void _foreach_div_out(at::TensorList out, at::TensorList self, at::ArrayRef<at::Scalar> scalars);
+TORCH_API void _foreach_div_outf(at::TensorList self, at::ArrayRef<at::Scalar> scalars, at::TensorList out);
+TORCH_API void _foreach_div_(at::TensorList self, at::ArrayRef<at::Scalar> scalars);
+TORCH_API ::std::vector<at::Tensor> _foreach_div(at::TensorList self, const at::Tensor & other);
+TORCH_API void _foreach_div_out(at::TensorList out, at::TensorList self, const at::Tensor & other);
+TORCH_API void _foreach_div_outf(at::TensorList self, const at::Tensor & other, at::TensorList out);
+TORCH_API void _foreach_div_(at::TensorList self, const at::Tensor & other);
+
+} // namespace compositeexplicitautograd
+} // namespace at

lib/python3.12/site-packages/torch/include/ATen/ops/_foreach_expm1_ops.h

@@ -0,0 +1,50 @@
+#pragma once
+
+// @generated by torchgen/gen.py from Operator.h
+
+#include <tuple>
+#include <vector>
+
+// Forward declarations of any types needed in the operator signatures.
+// We can't directly include these classes because it will cause circular include dependencies.
+// This file is included by TensorBody.h, which defines the Tensor class.
+#include <ATen/core/ATen_fwd.h>
+
+namespace at {
+namespace _ops {
+
+
+struct TORCH_API _foreach_expm1 {
+  using schema = ::std::vector<at::Tensor> (at::TensorList);
+  using ptr_schema = schema*;
+  // See Note [static constexpr char* members for windows NVCC]
+  static constexpr const char* name = "aten::_foreach_expm1";
+  static constexpr const char* overload_name = "";
+  static constexpr const char* schema_str = "_foreach_expm1(Tensor[] self) -> Tensor[]";
+  static ::std::vector<at::Tensor> call(at::TensorList self);
+  static ::std::vector<at::Tensor> redispatch(c10::DispatchKeySet dispatchKeySet, at::TensorList self);
+};
+
+struct TORCH_API _foreach_expm1_ {
+  using schema = void (at::TensorList);
+  using ptr_schema = schema*;
+  // See Note [static constexpr char* members for windows NVCC]
+  static constexpr const char* name = "aten::_foreach_expm1_";
+  static constexpr const char* overload_name = "";
+  static constexpr const char* schema_str = "_foreach_expm1_(Tensor(a!)[] self) -> ()";
+  static void call(at::TensorList self);
+  static void redispatch(c10::DispatchKeySet dispatchKeySet, at::TensorList self);
+};
+
+struct TORCH_API _foreach_expm1_out {
+  using schema = void (at::TensorList, at::TensorList);
+  using ptr_schema = schema*;
+  // See Note [static constexpr char* members for windows NVCC]
+  static constexpr const char* name = "aten::_foreach_expm1";
+  static constexpr const char* overload_name = "out";
+  static constexpr const char* schema_str = "_foreach_expm1.out(Tensor[] self, *, Tensor(a!)[] out) -> ()";
+  static void call(at::TensorList self, at::TensorList out);
+  static void redispatch(c10::DispatchKeySet dispatchKeySet, at::TensorList self, at::TensorList out);
+};
+
+}} // namespace at::_ops

lib/python3.12/site-packages/torch/include/ATen/ops/_foreach_log1p_ops.h

@@ -0,0 +1,50 @@
+#pragma once
+
+// @generated by torchgen/gen.py from Operator.h
+
+#include <tuple>
+#include <vector>
+
+// Forward declarations of any types needed in the operator signatures.
+// We can't directly include these classes because it will cause circular include dependencies.
+// This file is included by TensorBody.h, which defines the Tensor class.
+#include <ATen/core/ATen_fwd.h>
+
+namespace at {
+namespace _ops {
+
+
+struct TORCH_API _foreach_log1p {
+  using schema = ::std::vector<at::Tensor> (at::TensorList);
+  using ptr_schema = schema*;
+  // See Note [static constexpr char* members for windows NVCC]
+  static constexpr const char* name = "aten::_foreach_log1p";
+  static constexpr const char* overload_name = "";
+  static constexpr const char* schema_str = "_foreach_log1p(Tensor[] self) -> Tensor[]";
+  static ::std::vector<at::Tensor> call(at::TensorList self);
+  static ::std::vector<at::Tensor> redispatch(c10::DispatchKeySet dispatchKeySet, at::TensorList self);
+};
+
+struct TORCH_API _foreach_log1p_ {
+  using schema = void (at::TensorList);
+  using ptr_schema = schema*;
+  // See Note [static constexpr char* members for windows NVCC]
+  static constexpr const char* name = "aten::_foreach_log1p_";
+  static constexpr const char* overload_name = "";
+  static constexpr const char* schema_str = "_foreach_log1p_(Tensor(a!)[] self) -> ()";
+  static void call(at::TensorList self);
+  static void redispatch(c10::DispatchKeySet dispatchKeySet, at::TensorList self);
+};
+
+struct TORCH_API _foreach_log1p_out {
+  using schema = void (at::TensorList, at::TensorList);
+  using ptr_schema = schema*;
+  // See Note [static constexpr char* members for windows NVCC]
+  static constexpr const char* name = "aten::_foreach_log1p";
+  static constexpr const char* overload_name = "out";
+  static constexpr const char* schema_str = "_foreach_log1p.out(Tensor[] self, *, Tensor(a!)[] out) -> ()";
+  static void call(at::TensorList self, at::TensorList out);
+  static void redispatch(c10::DispatchKeySet dispatchKeySet, at::TensorList self, at::TensorList out);
+};
+
+}} // namespace at::_ops

lib/python3.12/site-packages/torch/include/ATen/ops/_foreach_sin_cuda_dispatch.h

@@ -0,0 +1,24 @@
+#pragma once
+// @generated by torchgen/gen.py from DispatchKeyFunction.h
+
+// NB: The implementing C++ file is RegisterDispatchKey.cpp
+
+// The only #includes we need are for custom classes that have defaults in the C++ API
+#include <c10/core/MemoryFormat.h>
+#include <c10/core/Scalar.h>
+#include <ATen/core/Reduction.h>
+
+// Forward declarations of any types needed in the operator signatures.
+// We can't directly include these classes because it will cause circular include dependencies.
+// This file is included by TensorBody.h, which defines the Tensor class.
+#include <ATen/core/ATen_fwd.h>
+
+namespace at {
+
+namespace cuda {
+
+TORCH_API ::std::vector<at::Tensor> _foreach_sin(at::TensorList self);
+TORCH_API void _foreach_sin_(at::TensorList self);
+
+} // namespace cuda
+} // namespace at

lib/python3.12/site-packages/torch/include/ATen/ops/_foreach_trunc_cuda_dispatch.h

@@ -0,0 +1,24 @@
+#pragma once
+// @generated by torchgen/gen.py from DispatchKeyFunction.h
+
+// NB: The implementing C++ file is RegisterDispatchKey.cpp
+
+// The only #includes we need are for custom classes that have defaults in the C++ API
+#include <c10/core/MemoryFormat.h>
+#include <c10/core/Scalar.h>
+#include <ATen/core/Reduction.h>
+
+// Forward declarations of any types needed in the operator signatures.
+// We can't directly include these classes because it will cause circular include dependencies.
+// This file is included by TensorBody.h, which defines the Tensor class.
+#include <ATen/core/ATen_fwd.h>
+
+namespace at {
+
+namespace cuda {
+
+TORCH_API ::std::vector<at::Tensor> _foreach_trunc(at::TensorList self);
+TORCH_API void _foreach_trunc_(at::TensorList self);
+
+} // namespace cuda
+} // namespace at

lib/python3.12/site-packages/torch/include/ATen/ops/_functional_assert_async.h

@@ -0,0 +1,30 @@
+#pragma once
+
+// @generated by torchgen/gen.py from Function.h
+
+#include <ATen/Context.h>
+#include <ATen/DeviceGuard.h>
+#include <ATen/TensorUtils.h>
+#include <ATen/TracerMode.h>
+#include <ATen/core/Generator.h>
+#include <ATen/core/Reduction.h>
+#include <ATen/core/Tensor.h>
+#include <c10/core/Scalar.h>
+#include <c10/core/Storage.h>
+#include <c10/core/TensorOptions.h>
+#include <c10/util/Deprecated.h>
+#include <optional>
+
+
+
+#include <ATen/ops/_functional_assert_async_ops.h>
+
+namespace at {
+
+
+// aten::_functional_assert_async.msg(Tensor self, str assert_msg, Tensor dep_token) -> Tensor
+inline at::Tensor _functional_assert_async(const at::Tensor & self, c10::string_view assert_msg, const at::Tensor & dep_token) {
+    return at::_ops::_functional_assert_async_msg::call(self, assert_msg, dep_token);
+}
+
+}

lib/python3.12/site-packages/torch/include/ATen/ops/_functional_sym_constrain_range_for_size_compositeexplicitautograd_dispatch.h

@@ -0,0 +1,23 @@
+#pragma once
+// @generated by torchgen/gen.py from DispatchKeyFunction.h
+
+// NB: The implementing C++ file is RegisterDispatchKey.cpp
+
+// The only #includes we need are for custom classes that have defaults in the C++ API
+#include <c10/core/MemoryFormat.h>
+#include <c10/core/Scalar.h>
+#include <ATen/core/Reduction.h>
+
+// Forward declarations of any types needed in the operator signatures.
+// We can't directly include these classes because it will cause circular include dependencies.
+// This file is included by TensorBody.h, which defines the Tensor class.
+#include <ATen/core/ATen_fwd.h>
+
+namespace at {
+
+namespace compositeexplicitautograd {
+
+TORCH_API at::Tensor _functional_sym_constrain_range_for_size(const at::Scalar & size, ::std::optional<int64_t> min, ::std::optional<int64_t> max, const at::Tensor & dep_token);
+
+} // namespace compositeexplicitautograd
+} // namespace at

lib/python3.12/site-packages/torch/include/ATen/ops/_fused_adam_cuda_dispatch.h

@@ -0,0 +1,24 @@
+#pragma once
+// @generated by torchgen/gen.py from DispatchKeyFunction.h
+
+// NB: The implementing C++ file is RegisterDispatchKey.cpp
+
+// The only #includes we need are for custom classes that have defaults in the C++ API
+#include <c10/core/MemoryFormat.h>
+#include <c10/core/Scalar.h>
+#include <ATen/core/Reduction.h>
+
+// Forward declarations of any types needed in the operator signatures.
+// We can't directly include these classes because it will cause circular include dependencies.
+// This file is included by TensorBody.h, which defines the Tensor class.
+#include <ATen/core/ATen_fwd.h>
+
+namespace at {
+
+namespace cuda {
+
+TORCH_API void _fused_adam_(at::TensorList self, at::TensorList grads, at::TensorList exp_avgs, at::TensorList exp_avg_sqs, at::TensorList max_exp_avg_sqs, at::TensorList state_steps, double lr, double beta1, double beta2, double weight_decay, double eps, bool amsgrad, bool maximize, const ::std::optional<at::Tensor> & grad_scale={}, const ::std::optional<at::Tensor> & found_inf={});
+TORCH_API void _fused_adam_(at::TensorList self, at::TensorList grads, at::TensorList exp_avgs, at::TensorList exp_avg_sqs, at::TensorList max_exp_avg_sqs, at::TensorList state_steps, const at::Tensor & lr, double beta1, double beta2, double weight_decay, double eps, bool amsgrad, bool maximize, const ::std::optional<at::Tensor> & grad_scale={}, const ::std::optional<at::Tensor> & found_inf={});
+
+} // namespace cuda
+} // namespace at

lib/python3.12/site-packages/torch/include/ATen/ops/_fused_adam_native.h

@@ -0,0 +1,28 @@
+#pragma once
+
+// @generated by torchgen/gen.py from NativeFunction.h
+
+#include <c10/core/Scalar.h>
+#include <c10/core/Storage.h>
+#include <c10/core/TensorOptions.h>
+#include <c10/util/Deprecated.h>
+#include <optional>
+#include <c10/core/QScheme.h>
+#include <ATen/core/Reduction.h>
+#include <ATen/core/Tensor.h>
+#include <tuple>
+#include <vector>
+
+
+namespace at {
+namespace native {
+TORCH_API ::std::tuple<::std::vector<at::Tensor>,::std::vector<at::Tensor>,::std::vector<at::Tensor>,::std::vector<at::Tensor>,::std::vector<at::Tensor>> _fused_adam(at::TensorList self, at::TensorList grads, at::TensorList exp_avgs, at::TensorList exp_avg_sqs, at::TensorList max_exp_avg_sqs, at::TensorList state_steps, double lr, double beta1, double beta2, double weight_decay, double eps, bool amsgrad, bool maximize, const ::std::optional<at::Tensor> & grad_scale={}, const ::std::optional<at::Tensor> & found_inf={});
+TORCH_API void _fused_adam_out(at::TensorList self, at::TensorList grads, at::TensorList exp_avgs, at::TensorList exp_avg_sqs, at::TensorList max_exp_avg_sqs, at::TensorList state_steps, double lr, double beta1, double beta2, double weight_decay, double eps, bool amsgrad, bool maximize, const ::std::optional<at::Tensor> & grad_scale, const ::std::optional<at::Tensor> & found_inf, at::TensorList out);
+TORCH_API void _fused_adam_kernel_cpu_(at::TensorList self, at::TensorList grads, at::TensorList exp_avgs, at::TensorList exp_avg_sqs, at::TensorList max_exp_avg_sqs, at::TensorList state_steps, double lr, double beta1, double beta2, double weight_decay, double eps, bool amsgrad, bool maximize, const ::std::optional<at::Tensor> & grad_scale={}, const ::std::optional<at::Tensor> & found_inf={});
+TORCH_API void _fused_adam_kernel_cuda_(at::TensorList self, at::TensorList grads, at::TensorList exp_avgs, at::TensorList exp_avg_sqs, at::TensorList max_exp_avg_sqs, at::TensorList state_steps, double lr, double beta1, double beta2, double weight_decay, double eps, bool amsgrad, bool maximize, const ::std::optional<at::Tensor> & grad_scale={}, const ::std::optional<at::Tensor> & found_inf={});
+TORCH_API ::std::tuple<::std::vector<at::Tensor>,::std::vector<at::Tensor>,::std::vector<at::Tensor>,::std::vector<at::Tensor>,::std::vector<at::Tensor>> _fused_adam(at::TensorList self, at::TensorList grads, at::TensorList exp_avgs, at::TensorList exp_avg_sqs, at::TensorList max_exp_avg_sqs, at::TensorList state_steps, const at::Tensor & lr, double beta1, double beta2, double weight_decay, double eps, bool amsgrad, bool maximize, const ::std::optional<at::Tensor> & grad_scale={}, const ::std::optional<at::Tensor> & found_inf={});
+TORCH_API void _fused_adam_tensor_lr_out(at::TensorList self, at::TensorList grads, at::TensorList exp_avgs, at::TensorList exp_avg_sqs, at::TensorList max_exp_avg_sqs, at::TensorList state_steps, const at::Tensor & lr, double beta1, double beta2, double weight_decay, double eps, bool amsgrad, bool maximize, const ::std::optional<at::Tensor> & grad_scale, const ::std::optional<at::Tensor> & found_inf, at::TensorList out);
+TORCH_API void _fused_adam_kernel_cpu_(at::TensorList self, at::TensorList grads, at::TensorList exp_avgs, at::TensorList exp_avg_sqs, at::TensorList max_exp_avg_sqs, at::TensorList state_steps, const at::Tensor & lr, double beta1, double beta2, double weight_decay, double eps, bool amsgrad, bool maximize, const ::std::optional<at::Tensor> & grad_scale={}, const ::std::optional<at::Tensor> & found_inf={});
+TORCH_API void _fused_adam_kernel_cuda_(at::TensorList self, at::TensorList grads, at::TensorList exp_avgs, at::TensorList exp_avg_sqs, at::TensorList max_exp_avg_sqs, at::TensorList state_steps, const at::Tensor & lr, double beta1, double beta2, double weight_decay, double eps, bool amsgrad, bool maximize, const ::std::optional<at::Tensor> & grad_scale={}, const ::std::optional<at::Tensor> & found_inf={});
+} // namespace native
+} // namespace at

lib/python3.12/site-packages/torch/include/ATen/ops/_fused_dropout_native.h

@@ -0,0 +1,22 @@
+#pragma once
+
+// @generated by torchgen/gen.py from NativeFunction.h
+
+#include <c10/core/Scalar.h>
+#include <c10/core/Storage.h>
+#include <c10/core/TensorOptions.h>
+#include <c10/util/Deprecated.h>
+#include <optional>
+#include <c10/core/QScheme.h>
+#include <ATen/core/Reduction.h>
+#include <ATen/core/Tensor.h>
+#include <tuple>
+#include <vector>
+
+
+namespace at {
+namespace native {
+TORCH_API ::std::tuple<at::Tensor &,at::Tensor &> _fused_dropout_out(const at::Tensor & self, double p, ::std::optional<at::Generator> generator, at::Tensor & out0, at::Tensor & out1);
+TORCH_API ::std::tuple<at::Tensor,at::Tensor> fused_dropout_cuda(const at::Tensor & self, double p, ::std::optional<at::Generator> generator=::std::nullopt);
+} // namespace native
+} // namespace at

lib/python3.12/site-packages/torch/include/ATen/ops/_fused_moving_avg_obs_fq_helper_native.h

@@ -0,0 +1,24 @@
+#pragma once
+
+// @generated by torchgen/gen.py from NativeFunction.h
+
+#include <c10/core/Scalar.h>
+#include <c10/core/Storage.h>
+#include <c10/core/TensorOptions.h>
+#include <c10/util/Deprecated.h>
+#include <optional>
+#include <c10/core/QScheme.h>
+#include <ATen/core/Reduction.h>
+#include <ATen/core/Tensor.h>
+#include <tuple>
+#include <vector>
+
+
+namespace at {
+namespace native {
+TORCH_API ::std::tuple<at::Tensor,at::Tensor,at::Tensor,at::Tensor,at::Tensor,at::Tensor> _fused_moving_avg_obs_fq_helper_functional(const at::Tensor & self, const at::Tensor & observer_on, const at::Tensor & fake_quant_on, const at::Tensor & running_min, const at::Tensor & running_max, const at::Tensor & scale, const at::Tensor & zero_point, double averaging_const, int64_t quant_min, int64_t quant_max, int64_t ch_axis, bool per_row_fake_quant=false, bool symmetric_quant=false);
+TORCH_API ::std::tuple<at::Tensor &,at::Tensor &> _fused_moving_avg_obs_fq_helper_out(const at::Tensor & self, const at::Tensor & observer_on, const at::Tensor & fake_quant_on, at::Tensor & running_min, at::Tensor & running_max, at::Tensor & scale, at::Tensor & zero_point, double averaging_const, int64_t quant_min, int64_t quant_max, int64_t ch_axis, bool per_row_fake_quant, bool symmetric_quant, at::Tensor & out0, at::Tensor & out1);
+TORCH_API ::std::tuple<at::Tensor,at::Tensor> fused_moving_avg_obs_fake_quant_cpu(const at::Tensor & self, const at::Tensor & observer_on, const at::Tensor & fake_quant_on, at::Tensor & running_min, at::Tensor & running_max, at::Tensor & scale, at::Tensor & zero_point, double averaging_const, int64_t quant_min, int64_t quant_max, int64_t ch_axis, bool per_row_fake_quant=false, bool symmetric_quant=false);
+TORCH_API ::std::tuple<at::Tensor,at::Tensor> fused_moving_avg_obs_fake_quant_cuda(const at::Tensor & self, const at::Tensor & observer_on, const at::Tensor & fake_quant_on, at::Tensor & running_min, at::Tensor & running_max, at::Tensor & scale, at::Tensor & zero_point, double averaging_const, int64_t quant_min, int64_t quant_max, int64_t ch_axis, bool per_row_fake_quant=false, bool symmetric_quant=false);
+} // namespace native
+} // namespace at

lib/python3.12/site-packages/torch/include/ATen/ops/_grid_sampler_2d_cpu_fallback_compositeexplicitautograd_dispatch.h

@@ -0,0 +1,25 @@
+#pragma once
+// @generated by torchgen/gen.py from DispatchKeyFunction.h
+
+// NB: The implementing C++ file is RegisterDispatchKey.cpp
+
+// The only #includes we need are for custom classes that have defaults in the C++ API
+#include <c10/core/MemoryFormat.h>
+#include <c10/core/Scalar.h>
+#include <ATen/core/Reduction.h>
+
+// Forward declarations of any types needed in the operator signatures.
+// We can't directly include these classes because it will cause circular include dependencies.
+// This file is included by TensorBody.h, which defines the Tensor class.
+#include <ATen/core/ATen_fwd.h>
+
+namespace at {
+
+namespace compositeexplicitautograd {
+
+TORCH_API at::Tensor _grid_sampler_2d_cpu_fallback(const at::Tensor & input, const at::Tensor & grid, int64_t interpolation_mode, int64_t padding_mode, bool align_corners);
+TORCH_API at::Tensor & _grid_sampler_2d_cpu_fallback_out(at::Tensor & out, const at::Tensor & input, const at::Tensor & grid, int64_t interpolation_mode, int64_t padding_mode, bool align_corners);
+TORCH_API at::Tensor & _grid_sampler_2d_cpu_fallback_outf(const at::Tensor & input, const at::Tensor & grid, int64_t interpolation_mode, int64_t padding_mode, bool align_corners, at::Tensor & out);
+
+} // namespace compositeexplicitautograd
+} // namespace at