Add files using upload-large-folder tool

lib/python3.12/site-packages/torch/include/ATen/ops/_adaptive_avg_pool2d_compositeexplicitautograd_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 compositeexplicitautograd {
+
+TORCH_API at::Tensor & _adaptive_avg_pool2d_out(at::Tensor & out, const at::Tensor & self, at::IntArrayRef output_size);
+TORCH_API at::Tensor & _adaptive_avg_pool2d_outf(const at::Tensor & self, at::IntArrayRef output_size, at::Tensor & out);
+TORCH_API at::Tensor & _adaptive_avg_pool2d_symint_out(at::Tensor & out, const at::Tensor & self, c10::SymIntArrayRef output_size);
+TORCH_API at::Tensor & _adaptive_avg_pool2d_symint_outf(const at::Tensor & self, c10::SymIntArrayRef output_size, at::Tensor & out);
+
+} // namespace compositeexplicitautograd
+} // namespace at

lib/python3.12/site-packages/torch/include/ATen/ops/_adaptive_avg_pool3d_backward.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/_adaptive_avg_pool3d_backward_ops.h>
+
+namespace at {
+
+
+// aten::_adaptive_avg_pool3d_backward(Tensor grad_output, Tensor self) -> Tensor
+inline at::Tensor _adaptive_avg_pool3d_backward(const at::Tensor & grad_output, const at::Tensor & self) {
+    return at::_ops::_adaptive_avg_pool3d_backward::call(grad_output, self);
+}
+
+// aten::_adaptive_avg_pool3d_backward.out(Tensor grad_output, Tensor self, *, Tensor(a!) out) -> Tensor(a!)
+inline at::Tensor & _adaptive_avg_pool3d_backward_out(at::Tensor & out, const at::Tensor & grad_output, const at::Tensor & self) {
+    return at::_ops::_adaptive_avg_pool3d_backward_out::call(grad_output, self, out);
+}
+// aten::_adaptive_avg_pool3d_backward.out(Tensor grad_output, Tensor self, *, Tensor(a!) out) -> Tensor(a!)
+inline at::Tensor & _adaptive_avg_pool3d_backward_outf(const at::Tensor & grad_output, const at::Tensor & self, at::Tensor & out) {
+    return at::_ops::_adaptive_avg_pool3d_backward_out::call(grad_output, self, out);
+}
+
+}

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

@@ -0,0 +1,28 @@
+#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 _add_relu(const at::Tensor & self, const at::Tensor & other, const at::Scalar & alpha=1);
+TORCH_API at::Tensor & _add_relu_out(at::Tensor & out, const at::Tensor & self, const at::Tensor & other, const at::Scalar & alpha=1);
+TORCH_API at::Tensor & _add_relu_outf(const at::Tensor & self, const at::Tensor & other, const at::Scalar & alpha, at::Tensor & out);
+TORCH_API at::Tensor & _add_relu_(at::Tensor & self, const at::Tensor & other, const at::Scalar & alpha=1);
+TORCH_API at::Tensor _add_relu(const at::Tensor & self, const at::Scalar & other, const at::Scalar & alpha=1);
+TORCH_API at::Tensor & _add_relu_(at::Tensor & self, const at::Scalar & other, const at::Scalar & alpha=1);
+
+} // namespace cpu
+} // namespace at

lib/python3.12/site-packages/torch/include/ATen/ops/_amp_update_scale_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> _amp_update_scale(const at::Tensor & self, const at::Tensor & growth_tracker, const at::Tensor & found_inf, double scale_growth_factor, double scale_backoff_factor, int64_t growth_interval);
+TORCH_API at::Tensor & _amp_update_scale_out(at::Tensor & out, const at::Tensor & self, at::Tensor & growth_tracker, const at::Tensor & found_inf, double scale_growth_factor, double scale_backoff_factor, int64_t growth_interval);
+TORCH_API at::Tensor & _amp_update_scale_outf(const at::Tensor & self, at::Tensor & growth_tracker, const at::Tensor & found_inf, double scale_growth_factor, double scale_backoff_factor, int64_t growth_interval, at::Tensor & out);
+
+} // namespace compositeexplicitautograd
+} // namespace at

lib/python3.12/site-packages/torch/include/ATen/ops/_assert_async_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 void _assert_async_cpu(const at::Tensor & self);
+TORCH_API void _assert_async_cuda(const at::Tensor & self);
+TORCH_API void _assert_async_msg_cpu(const at::Tensor & self, c10::string_view assert_msg);
+TORCH_API void _assert_async_msg_cuda(const at::Tensor & self, c10::string_view assert_msg);
+} // namespace native
+} // namespace at

lib/python3.12/site-packages/torch/include/ATen/ops/_assert_tensor_metadata_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 _assert_tensor_metadata {
+  using schema = void (const at::Tensor &, at::OptionalSymIntArrayRef, at::OptionalSymIntArrayRef, ::std::optional<at::ScalarType>, ::std::optional<at::Device>, ::std::optional<at::Layout>);
+  using ptr_schema = schema*;
+  // See Note [static constexpr char* members for windows NVCC]
+  static constexpr const char* name = "aten::_assert_tensor_metadata";
+  static constexpr const char* overload_name = "";
+  static constexpr const char* schema_str = "_assert_tensor_metadata(Tensor a, SymInt[]? size=None, SymInt[]? stride=None, ScalarType? dtype=None, *, Device? device=None, Layout? layout=None) -> ()";
+  static void call(const at::Tensor & a, at::OptionalSymIntArrayRef size, at::OptionalSymIntArrayRef stride, ::std::optional<at::ScalarType> dtype, ::std::optional<at::Device> device, ::std::optional<at::Layout> layout);
+  static void redispatch(c10::DispatchKeySet dispatchKeySet, const at::Tensor & a, at::OptionalSymIntArrayRef size, at::OptionalSymIntArrayRef stride, ::std::optional<at::ScalarType> dtype, ::std::optional<at::Device> device, ::std::optional<at::Layout> layout);
+};
+
+}} // namespace at::_ops

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

@@ -0,0 +1,26 @@
+#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> _batch_norm_with_update_functional(const at::Tensor & input, const ::std::optional<at::Tensor> & weight, const ::std::optional<at::Tensor> & bias, const at::Tensor & running_mean, const at::Tensor & running_var, double momentum, double eps);
+TORCH_API ::std::tuple<at::Tensor &,at::Tensor &,at::Tensor &,at::Tensor &> _batch_norm_with_update_cpu_out(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_cpu(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_cuda_out(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_cuda(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_mkldnn(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 native
+} // namespace at

lib/python3.12/site-packages/torch/include/ATen/ops/_cast_Short_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 _cast_Short {
+  using schema = at::Tensor (const at::Tensor &, bool);
+  using ptr_schema = schema*;
+  // See Note [static constexpr char* members for windows NVCC]
+  static constexpr const char* name = "aten::_cast_Short";
+  static constexpr const char* overload_name = "";
+  static constexpr const char* schema_str = "_cast_Short(Tensor self, bool non_blocking=False) -> Tensor";
+  static at::Tensor call(const at::Tensor & self, bool non_blocking);
+  static at::Tensor redispatch(c10::DispatchKeySet dispatchKeySet, const at::Tensor & self, bool non_blocking);
+};
+
+}} // namespace at::_ops

lib/python3.12/site-packages/torch/include/ATen/ops/_cdist_forward_cpu_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 cpu {
+
+TORCH_API at::Tensor _cdist_forward(const at::Tensor & x1, const at::Tensor & x2, double p, ::std::optional<int64_t> compute_mode);
+
+} // namespace cpu
+} // namespace at

lib/python3.12/site-packages/torch/include/ATen/ops/_compute_linear_combination_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 _compute_linear_combination(const at::Tensor & input, const at::Tensor & coefficients);
+TORCH_API at::Tensor & _compute_linear_combination_out(at::Tensor & out, const at::Tensor & input, const at::Tensor & coefficients);
+TORCH_API at::Tensor & _compute_linear_combination_outf(const at::Tensor & input, const at::Tensor & coefficients, at::Tensor & out);
+
+} // namespace cpu
+} // namespace at

lib/python3.12/site-packages/torch/include/ATen/ops/_conj_copy.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_copy_ops.h>
+
+namespace at {
+
+
+// aten::_conj_copy(Tensor self) -> Tensor
+inline at::Tensor _conj_copy(const at::Tensor & self) {
+    return at::_ops::_conj_copy::call(self);
+}
+
+// aten::_conj_copy.out(Tensor self, *, Tensor(a!) out) -> Tensor(a!)
+inline at::Tensor & _conj_copy_out(at::Tensor & out, const at::Tensor & self) {
+    return at::_ops::_conj_copy_out::call(self, out);
+}
+// aten::_conj_copy.out(Tensor self, *, Tensor(a!) out) -> Tensor(a!)
+inline at::Tensor & _conj_copy_outf(const at::Tensor & self, at::Tensor & out) {
+    return at::_ops::_conj_copy_out::call(self, out);
+}
+
+}

lib/python3.12/site-packages/torch/include/ATen/ops/_convert_weight_to_int4pack_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 at::Tensor _convert_weight_to_int4pack(const at::Tensor & self, int64_t innerKTiles);
+
+} // namespace cuda
+} // namespace at

lib/python3.12/site-packages/torch/include/ATen/ops/_cufft_clear_plan_cache_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 _cufft_clear_plan_cache {
+  using schema = void (at::DeviceIndex);
+  using ptr_schema = schema*;
+  // See Note [static constexpr char* members for windows NVCC]
+  static constexpr const char* name = "aten::_cufft_clear_plan_cache";
+  static constexpr const char* overload_name = "";
+  static constexpr const char* schema_str = "_cufft_clear_plan_cache(DeviceIndex device_index) -> ()";
+  static void call(at::DeviceIndex device_index);
+  static void redispatch(c10::DispatchKeySet dispatchKeySet, at::DeviceIndex device_index);
+};
+
+}} // namespace at::_ops

lib/python3.12/site-packages/torch/include/ATen/ops/_cufft_get_plan_cache_max_size_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 int64_t _cufft_get_plan_cache_max_size(at::DeviceIndex device_index);
+} // namespace native
+} // namespace at

lib/python3.12/site-packages/torch/include/ATen/ops/_cummax_helper_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 void cummax_helper_cpu(const at::Tensor & self, at::Tensor & values, at::Tensor & indices, int64_t dim);
+TORCH_API void cummax_helper_cuda(const at::Tensor & self, at::Tensor & values, at::Tensor & indices, int64_t dim);
+} // namespace native
+} // namespace at

lib/python3.12/site-packages/torch/include/ATen/ops/_dirichlet_grad_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 at::Tensor & _dirichlet_grad_out(at::Tensor & out, const at::Tensor & x, const at::Tensor & alpha, const at::Tensor & total);
+TORCH_API at::Tensor & _dirichlet_grad_outf(const at::Tensor & x, const at::Tensor & alpha, const at::Tensor & total, at::Tensor & out);
+
+} // namespace compositeexplicitautograd
+} // namespace at

lib/python3.12/site-packages/torch/include/ATen/ops/_dirichlet_grad_cpu_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 cpu {
+
+TORCH_API at::Tensor _dirichlet_grad(const at::Tensor & x, const at::Tensor & alpha, const at::Tensor & total);
+
+} // namespace cpu
+} // namespace at

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

@@ -0,0 +1,47 @@
+#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/_efficient_attention_backward_ops.h>
+
+namespace at {
+
+
+// aten::_efficient_attention_backward(Tensor grad_out_, Tensor query, Tensor key, Tensor value, Tensor? bias, Tensor out, Tensor? cu_seqlens_q, Tensor? cu_seqlens_k, SymInt max_seqlen_q, SymInt max_seqlen_k, Tensor logsumexp, float dropout_p, Tensor philox_seed, Tensor philox_offset, int custom_mask_type, bool bias_requires_grad, *, float? scale=None, int? num_splits_key=None, int? window_size=None, bool shared_storage_dqdkdv=False) -> (Tensor, Tensor, Tensor, Tensor)
+inline ::std::tuple<at::Tensor,at::Tensor,at::Tensor,at::Tensor> _efficient_attention_backward(const at::Tensor & grad_out_, const at::Tensor & query, const at::Tensor & key, const at::Tensor & value, const ::std::optional<at::Tensor> & bias, const at::Tensor & out, const ::std::optional<at::Tensor> & cu_seqlens_q, const ::std::optional<at::Tensor> & cu_seqlens_k, int64_t max_seqlen_q, int64_t max_seqlen_k, const at::Tensor & logsumexp, double dropout_p, const at::Tensor & philox_seed, const at::Tensor & philox_offset, int64_t custom_mask_type, bool bias_requires_grad, ::std::optional<double> scale=::std::nullopt, ::std::optional<int64_t> num_splits_key=::std::nullopt, ::std::optional<int64_t> window_size=::std::nullopt, bool shared_storage_dqdkdv=false) {
+    return at::_ops::_efficient_attention_backward::call(grad_out_, query, key, value, bias, out, cu_seqlens_q, cu_seqlens_k, max_seqlen_q, max_seqlen_k, logsumexp, dropout_p, philox_seed, philox_offset, custom_mask_type, bias_requires_grad, scale, num_splits_key, window_size, shared_storage_dqdkdv);
+}
+namespace symint {
+  template <typename T, typename = std::enable_if_t<std::is_same_v<T, int64_t>>>
+  ::std::tuple<at::Tensor,at::Tensor,at::Tensor,at::Tensor> _efficient_attention_backward(const at::Tensor & grad_out_, const at::Tensor & query, const at::Tensor & key, const at::Tensor & value, const ::std::optional<at::Tensor> & bias, const at::Tensor & out, const ::std::optional<at::Tensor> & cu_seqlens_q, const ::std::optional<at::Tensor> & cu_seqlens_k, int64_t max_seqlen_q, int64_t max_seqlen_k, const at::Tensor & logsumexp, double dropout_p, const at::Tensor & philox_seed, const at::Tensor & philox_offset, int64_t custom_mask_type, bool bias_requires_grad, ::std::optional<double> scale=::std::nullopt, ::std::optional<int64_t> num_splits_key=::std::nullopt, ::std::optional<int64_t> window_size=::std::nullopt, bool shared_storage_dqdkdv=false) {
+    return at::_ops::_efficient_attention_backward::call(grad_out_, query, key, value, bias, out, cu_seqlens_q, cu_seqlens_k, max_seqlen_q, max_seqlen_k, logsumexp, dropout_p, philox_seed, philox_offset, custom_mask_type, bias_requires_grad, scale, num_splits_key, window_size, shared_storage_dqdkdv);
+  }
+}
+
+// aten::_efficient_attention_backward(Tensor grad_out_, Tensor query, Tensor key, Tensor value, Tensor? bias, Tensor out, Tensor? cu_seqlens_q, Tensor? cu_seqlens_k, SymInt max_seqlen_q, SymInt max_seqlen_k, Tensor logsumexp, float dropout_p, Tensor philox_seed, Tensor philox_offset, int custom_mask_type, bool bias_requires_grad, *, float? scale=None, int? num_splits_key=None, int? window_size=None, bool shared_storage_dqdkdv=False) -> (Tensor, Tensor, Tensor, Tensor)
+inline ::std::tuple<at::Tensor,at::Tensor,at::Tensor,at::Tensor> _efficient_attention_backward_symint(const at::Tensor & grad_out_, const at::Tensor & query, const at::Tensor & key, const at::Tensor & value, const ::std::optional<at::Tensor> & bias, const at::Tensor & out, const ::std::optional<at::Tensor> & cu_seqlens_q, const ::std::optional<at::Tensor> & cu_seqlens_k, c10::SymInt max_seqlen_q, c10::SymInt max_seqlen_k, const at::Tensor & logsumexp, double dropout_p, const at::Tensor & philox_seed, const at::Tensor & philox_offset, int64_t custom_mask_type, bool bias_requires_grad, ::std::optional<double> scale=::std::nullopt, ::std::optional<int64_t> num_splits_key=::std::nullopt, ::std::optional<int64_t> window_size=::std::nullopt, bool shared_storage_dqdkdv=false) {
+    return at::_ops::_efficient_attention_backward::call(grad_out_, query, key, value, bias, out, cu_seqlens_q, cu_seqlens_k, max_seqlen_q, max_seqlen_k, logsumexp, dropout_p, philox_seed, philox_offset, custom_mask_type, bias_requires_grad, scale, num_splits_key, window_size, shared_storage_dqdkdv);
+}
+namespace symint {
+  template <typename T, typename = std::enable_if_t<std::is_same_v<T, c10::SymInt>>>
+  ::std::tuple<at::Tensor,at::Tensor,at::Tensor,at::Tensor> _efficient_attention_backward(const at::Tensor & grad_out_, const at::Tensor & query, const at::Tensor & key, const at::Tensor & value, const ::std::optional<at::Tensor> & bias, const at::Tensor & out, const ::std::optional<at::Tensor> & cu_seqlens_q, const ::std::optional<at::Tensor> & cu_seqlens_k, c10::SymInt max_seqlen_q, c10::SymInt max_seqlen_k, const at::Tensor & logsumexp, double dropout_p, const at::Tensor & philox_seed, const at::Tensor & philox_offset, int64_t custom_mask_type, bool bias_requires_grad, ::std::optional<double> scale=::std::nullopt, ::std::optional<int64_t> num_splits_key=::std::nullopt, ::std::optional<int64_t> window_size=::std::nullopt, bool shared_storage_dqdkdv=false) {
+    return at::_ops::_efficient_attention_backward::call(grad_out_, query, key, value, bias, out, cu_seqlens_q, cu_seqlens_k, max_seqlen_q, max_seqlen_k, logsumexp, dropout_p, philox_seed, philox_offset, custom_mask_type, bias_requires_grad, scale, num_splits_key, window_size, shared_storage_dqdkdv);
+  }
+}
+
+}

lib/python3.12/site-packages/torch/include/ATen/ops/_efficientzerotensor_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 & _efficientzerotensor_out_symint(c10::SymIntArrayRef size, at::Tensor & out);
+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_cuda(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_meta_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 native
+} // namespace at

lib/python3.12/site-packages/torch/include/ATen/ops/_embedding_bag_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 {
+  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";
+  static constexpr const char* overload_name = "";
+  static constexpr const char* schema_str = "_embedding_bag(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_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";
+  static constexpr const char* overload_name = "out";
+  static constexpr const char* schema_str = "_embedding_bag.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.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/_empty_affine_quantized_ops.h>
+
+namespace at {
+
+
+// aten::_empty_affine_quantized(SymInt[] size, *, ScalarType? dtype=None, Layout? layout=None, Device? device=None, bool? pin_memory=None, float scale=1, int zero_point=0, MemoryFormat? memory_format=contiguous_format) -> Tensor
+inline 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) {
+    return at::_ops::_empty_affine_quantized::call(c10::fromIntArrayRefSlow(size), c10::optTypeMetaToScalarType(options.dtype_opt()), options.layout_opt(), options.device_opt(), options.pinned_memory_opt(), scale, zero_point, c10::impl::check_tensor_options_and_extract_memory_format(options, memory_format));
+}
+namespace symint {
+  template <typename T, typename = std::enable_if_t<std::is_same_v<T, int64_t>>>
+  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) {
+    return at::_ops::_empty_affine_quantized::call(c10::fromIntArrayRefSlow(size), c10::optTypeMetaToScalarType(options.dtype_opt()), options.layout_opt(), options.device_opt(), options.pinned_memory_opt(), scale, zero_point, c10::impl::check_tensor_options_and_extract_memory_format(options, memory_format));
+  }
+}
+
+// aten::_empty_affine_quantized(SymInt[] size, *, ScalarType? dtype=None, Layout? layout=None, Device? device=None, bool? pin_memory=None, float scale=1, int zero_point=0, MemoryFormat? memory_format=contiguous_format) -> Tensor
+inline 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) {
+    return at::_ops::_empty_affine_quantized::call(c10::fromIntArrayRefSlow(size), dtype, layout, device, pin_memory, scale, zero_point, memory_format);
+}
+namespace symint {
+  template <typename T, typename = std::enable_if_t<std::is_same_v<T, int64_t>>>
+  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) {
+    return at::_ops::_empty_affine_quantized::call(c10::fromIntArrayRefSlow(size), dtype, layout, device, pin_memory, scale, zero_point, memory_format);
+  }
+}
+
+// aten::_empty_affine_quantized(SymInt[] size, *, ScalarType? dtype=None, Layout? layout=None, Device? device=None, bool? pin_memory=None, float scale=1, int zero_point=0, MemoryFormat? memory_format=contiguous_format) -> Tensor
+inline 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) {
+    return at::_ops::_empty_affine_quantized::call(size, c10::optTypeMetaToScalarType(options.dtype_opt()), options.layout_opt(), options.device_opt(), options.pinned_memory_opt(), scale, zero_point, c10::impl::check_tensor_options_and_extract_memory_format(options, memory_format));
+}
+namespace symint {
+  template <typename T, typename = std::enable_if_t<std::is_same_v<T, c10::SymInt>>>
+  at::Tensor _empty_affine_quantized(c10::SymIntArrayRef size, at::TensorOptions options={}, double scale=1, int64_t zero_point=0, ::std::optional<at::MemoryFormat> memory_format=c10::MemoryFormat::Contiguous) {
+    return at::_ops::_empty_affine_quantized::call(size, c10::optTypeMetaToScalarType(options.dtype_opt()), options.layout_opt(), options.device_opt(), options.pinned_memory_opt(), scale, zero_point, c10::impl::check_tensor_options_and_extract_memory_format(options, memory_format));
+  }
+}
+
+// aten::_empty_affine_quantized(SymInt[] size, *, ScalarType? dtype=None, Layout? layout=None, Device? device=None, bool? pin_memory=None, float scale=1, int zero_point=0, MemoryFormat? memory_format=contiguous_format) -> Tensor
+inline 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) {
+    return at::_ops::_empty_affine_quantized::call(size, dtype, layout, device, pin_memory, scale, zero_point, memory_format);
+}
+namespace symint {
+  template <typename T, typename = std::enable_if_t<std::is_same_v<T, c10::SymInt>>>
+  at::Tensor _empty_affine_quantized(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) {
+    return at::_ops::_empty_affine_quantized::call(size, dtype, layout, device, pin_memory, scale, zero_point, memory_format);
+  }
+}
+
+// aten::_empty_affine_quantized.out(SymInt[] size, *, float scale=1, int zero_point=0, MemoryFormat? memory_format=contiguous_format, Tensor(a!) out) -> Tensor(a!)
+inline at::Tensor & _empty_affine_quantized_out(at::Tensor & out, at::IntArrayRef size, double scale=1, int64_t zero_point=0, ::std::optional<at::MemoryFormat> memory_format=c10::MemoryFormat::Contiguous) {
+    return at::_ops::_empty_affine_quantized_out::call(c10::fromIntArrayRefSlow(size), scale, zero_point, memory_format, out);
+}
+namespace symint {
+  template <typename T, typename = std::enable_if_t<std::is_same_v<T, int64_t>>>
+  at::Tensor & _empty_affine_quantized_out(at::Tensor & out, at::IntArrayRef size, double scale=1, int64_t zero_point=0, ::std::optional<at::MemoryFormat> memory_format=c10::MemoryFormat::Contiguous) {
+    return at::_ops::_empty_affine_quantized_out::call(c10::fromIntArrayRefSlow(size), scale, zero_point, memory_format, out);
+  }
+}
+
+// aten::_empty_affine_quantized.out(SymInt[] size, *, float scale=1, int zero_point=0, MemoryFormat? memory_format=contiguous_format, Tensor(a!) out) -> Tensor(a!)
+inline at::Tensor & _empty_affine_quantized_outf(at::IntArrayRef size, double scale, int64_t zero_point, ::std::optional<at::MemoryFormat> memory_format, at::Tensor & out) {
+    return at::_ops::_empty_affine_quantized_out::call(c10::fromIntArrayRefSlow(size), scale, zero_point, memory_format, out);
+}
+namespace symint {
+  template <typename T, typename = std::enable_if_t<std::is_same_v<T, int64_t>>>
+  at::Tensor & _empty_affine_quantized_outf(at::IntArrayRef size, double scale, int64_t zero_point, ::std::optional<at::MemoryFormat> memory_format, at::Tensor & out) {
+    return at::_ops::_empty_affine_quantized_out::call(c10::fromIntArrayRefSlow(size), scale, zero_point, memory_format, out);
+  }
+}
+
+// aten::_empty_affine_quantized.out(SymInt[] size, *, float scale=1, int zero_point=0, MemoryFormat? memory_format=contiguous_format, Tensor(a!) out) -> Tensor(a!)
+inline at::Tensor & _empty_affine_quantized_symint_out(at::Tensor & out, c10::SymIntArrayRef size, double scale=1, int64_t zero_point=0, ::std::optional<at::MemoryFormat> memory_format=c10::MemoryFormat::Contiguous) {
+    return at::_ops::_empty_affine_quantized_out::call(size, scale, zero_point, memory_format, out);
+}
+namespace symint {
+  template <typename T, typename = std::enable_if_t<std::is_same_v<T, c10::SymInt>>>
+  at::Tensor & _empty_affine_quantized_out(at::Tensor & out, c10::SymIntArrayRef size, double scale=1, int64_t zero_point=0, ::std::optional<at::MemoryFormat> memory_format=c10::MemoryFormat::Contiguous) {
+    return at::_ops::_empty_affine_quantized_out::call(size, scale, zero_point, memory_format, out);
+  }
+}
+
+// aten::_empty_affine_quantized.out(SymInt[] size, *, float scale=1, int zero_point=0, MemoryFormat? memory_format=contiguous_format, Tensor(a!) out) -> Tensor(a!)
+inline at::Tensor & _empty_affine_quantized_symint_outf(c10::SymIntArrayRef size, double scale, int64_t zero_point, ::std::optional<at::MemoryFormat> memory_format, at::Tensor & out) {
+    return at::_ops::_empty_affine_quantized_out::call(size, scale, zero_point, memory_format, out);
+}
+namespace symint {
+  template <typename T, typename = std::enable_if_t<std::is_same_v<T, c10::SymInt>>>
+  at::Tensor & _empty_affine_quantized_outf(c10::SymIntArrayRef size, double scale, int64_t zero_point, ::std::optional<at::MemoryFormat> memory_format, at::Tensor & out) {
+    return at::_ops::_empty_affine_quantized_out::call(size, scale, zero_point, memory_format, out);
+  }
+}
+
+}

lib/python3.12/site-packages/torch/include/ATen/ops/_empty_affine_quantized_compositeexplicitautograd_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 compositeexplicitautograd {
+
+TORCH_API at::Tensor & _empty_affine_quantized_out(at::Tensor & out, at::IntArrayRef size, double scale=1, int64_t zero_point=0, ::std::optional<at::MemoryFormat> memory_format=c10::MemoryFormat::Contiguous);
+TORCH_API at::Tensor & _empty_affine_quantized_outf(at::IntArrayRef size, double scale, int64_t zero_point, ::std::optional<at::MemoryFormat> memory_format, at::Tensor & out);
+TORCH_API at::Tensor & _empty_affine_quantized_symint_out(at::Tensor & out, c10::SymIntArrayRef size, 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_outf(c10::SymIntArrayRef size, double scale, int64_t zero_point, ::std::optional<at::MemoryFormat> memory_format, at::Tensor & out);
+
+} // namespace compositeexplicitautograd
+} // namespace at

lib/python3.12/site-packages/torch/include/ATen/ops/_fake_quantize_learnable_per_channel_affine.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/_fake_quantize_learnable_per_channel_affine_ops.h>
+
+namespace at {
+
+
+// aten::_fake_quantize_learnable_per_channel_affine(Tensor self, Tensor scale, Tensor zero_point, int axis, int quant_min, int quant_max, float grad_factor=1.0) -> Tensor
+inline at::Tensor _fake_quantize_learnable_per_channel_affine(const at::Tensor & self, const at::Tensor & scale, const at::Tensor & zero_point, int64_t axis, int64_t quant_min, int64_t quant_max, double grad_factor=1.0) {
+    return at::_ops::_fake_quantize_learnable_per_channel_affine::call(self, scale, zero_point, axis, quant_min, quant_max, grad_factor);
+}
+
+// aten::_fake_quantize_learnable_per_channel_affine.out(Tensor self, Tensor scale, Tensor zero_point, int axis, int quant_min, int quant_max, float grad_factor=1.0, *, Tensor(a!) out) -> Tensor(a!)
+inline at::Tensor & _fake_quantize_learnable_per_channel_affine_out(at::Tensor & out, const at::Tensor & self, const at::Tensor & scale, const at::Tensor & zero_point, int64_t axis, int64_t quant_min, int64_t quant_max, double grad_factor=1.0) {
+    return at::_ops::_fake_quantize_learnable_per_channel_affine_out::call(self, scale, zero_point, axis, quant_min, quant_max, grad_factor, out);
+}
+// aten::_fake_quantize_learnable_per_channel_affine.out(Tensor self, Tensor scale, Tensor zero_point, int axis, int quant_min, int quant_max, float grad_factor=1.0, *, Tensor(a!) out) -> Tensor(a!)
+inline at::Tensor & _fake_quantize_learnable_per_channel_affine_outf(const at::Tensor & self, const at::Tensor & scale, const at::Tensor & zero_point, int64_t axis, int64_t quant_min, int64_t quant_max, double grad_factor, at::Tensor & out) {
+    return at::_ops::_fake_quantize_learnable_per_channel_affine_out::call(self, scale, zero_point, axis, quant_min, quant_max, grad_factor, out);
+}
+
+}

lib/python3.12/site-packages/torch/include/ATen/ops/_fake_quantize_learnable_per_tensor_affine_cpu_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 cpu {
+
+TORCH_API at::Tensor _fake_quantize_learnable_per_tensor_affine(const at::Tensor & self, const at::Tensor & scale, const at::Tensor & zero_point, int64_t quant_min, int64_t quant_max, double grad_factor=1.0);
+
+} // namespace cpu
+} // namespace at

lib/python3.12/site-packages/torch/include/ATen/ops/_fft_c2c_cuda_dispatch.h

@@ -0,0 +1,28 @@
+#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 at::Tensor _fft_c2c(const at::Tensor & self, at::IntArrayRef dim, int64_t normalization, bool forward);
+TORCH_API at::Tensor _fft_c2c_symint(const at::Tensor & self, c10::SymIntArrayRef dim, int64_t normalization, bool forward);
+TORCH_API at::Tensor & _fft_c2c_out(at::Tensor & out, const at::Tensor & self, at::IntArrayRef dim, int64_t normalization, bool forward);
+TORCH_API at::Tensor & _fft_c2c_outf(const at::Tensor & self, at::IntArrayRef dim, int64_t normalization, bool forward, at::Tensor & out);
+TORCH_API at::Tensor & _fft_c2c_symint_out(at::Tensor & out, const at::Tensor & self, c10::SymIntArrayRef dim, int64_t normalization, bool forward);
+TORCH_API at::Tensor & _fft_c2c_symint_outf(const at::Tensor & self, c10::SymIntArrayRef dim, int64_t normalization, bool forward, at::Tensor & out);
+
+} // namespace cuda
+} // namespace at

lib/python3.12/site-packages/torch/include/ATen/ops/_foreach_ceil_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_ceil {
+  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_ceil";
+  static constexpr const char* overload_name = "";
+  static constexpr const char* schema_str = "_foreach_ceil(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_ceil_ {
+  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_ceil_";
+  static constexpr const char* overload_name = "";
+  static constexpr const char* schema_str = "_foreach_ceil_(Tensor(a!)[] self) -> ()";
+  static void call(at::TensorList self);
+  static void redispatch(c10::DispatchKeySet dispatchKeySet, at::TensorList self);
+};
+
+struct TORCH_API _foreach_ceil_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_ceil";
+  static constexpr const char* overload_name = "out";
+  static constexpr const char* schema_str = "_foreach_ceil.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_erf_compositeexplicitautograd_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 compositeexplicitautograd {
+
+TORCH_API ::std::vector<at::Tensor> _foreach_erf(at::TensorList self);
+TORCH_API void _foreach_erf_out(at::TensorList out, at::TensorList self);
+TORCH_API void _foreach_erf_outf(at::TensorList self, at::TensorList out);
+TORCH_API void _foreach_erf_(at::TensorList self);
+
+} // namespace compositeexplicitautograd
+} // namespace at

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

@@ -0,0 +1,25 @@
+#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_tensor_erf_slow(at::TensorList self);
+TORCH_API void _foreach_erf_out(at::TensorList self, at::TensorList out);
+TORCH_API void foreach_tensor_erf_slow_(at::TensorList self);
+TORCH_API ::std::vector<at::Tensor> foreach_tensor_erf_cuda(at::TensorList self);
+TORCH_API void foreach_tensor_erf_cuda_(at::TensorList self);
+} // namespace native
+} // namespace at

lib/python3.12/site-packages/torch/include/ATen/ops/_foreach_exp_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_exp {
+  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_exp";
+  static constexpr const char* overload_name = "";
+  static constexpr const char* schema_str = "_foreach_exp(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_exp_ {
+  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_exp_";
+  static constexpr const char* overload_name = "";
+  static constexpr const char* schema_str = "_foreach_exp_(Tensor(a!)[] self) -> ()";
+  static void call(at::TensorList self);
+  static void redispatch(c10::DispatchKeySet dispatchKeySet, at::TensorList self);
+};
+
+struct TORCH_API _foreach_exp_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_exp";
+  static constexpr const char* overload_name = "out";
+  static constexpr const char* schema_str = "_foreach_exp.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_expm1.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/_foreach_expm1_ops.h>
+
+namespace at {
+
+
+// aten::_foreach_expm1(Tensor[] self) -> Tensor[]
+inline ::std::vector<at::Tensor> _foreach_expm1(at::TensorList self) {
+    return at::_ops::_foreach_expm1::call(self);
+}
+
+// aten::_foreach_expm1_(Tensor(a!)[] self) -> ()
+inline void _foreach_expm1_(at::TensorList self) {
+    return at::_ops::_foreach_expm1_::call(self);
+}
+
+// aten::_foreach_expm1.out(Tensor[] self, *, Tensor(a!)[] out) -> ()
+inline void _foreach_expm1_out(at::TensorList out, at::TensorList self) {
+    return at::_ops::_foreach_expm1_out::call(self, out);
+}
+// aten::_foreach_expm1.out(Tensor[] self, *, Tensor(a!)[] out) -> ()
+inline void _foreach_expm1_outf(at::TensorList self, at::TensorList out) {
+    return at::_ops::_foreach_expm1_out::call(self, out);
+}
+
+}

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

@@ -0,0 +1,35 @@
+#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_tensor_clamp_min_scalar_kernel_slow(at::TensorList self, const at::Scalar & scalar);
+TORCH_API void _foreach_maximum_Scalar_out(at::TensorList self, const at::Scalar & scalar, at::TensorList out);
+TORCH_API void foreach_tensor_clamp_min_scalar_kernel_slow_(at::TensorList self, const at::Scalar & scalar);
+TORCH_API ::std::vector<at::Tensor> foreach_tensor_clamp_min_scalar_kernel_cuda(at::TensorList self, const at::Scalar & scalar);
+TORCH_API void foreach_tensor_clamp_min_scalar_kernel_cuda_(at::TensorList self, const at::Scalar & scalar);
+TORCH_API ::std::vector<at::Tensor> foreach_tensor_clamp_min_list_kernel_slow(at::TensorList self, at::TensorList other);
+TORCH_API void _foreach_maximum_List_out(at::TensorList self, at::TensorList other, at::TensorList out);
+TORCH_API void foreach_tensor_clamp_min_list_kernel_slow_(at::TensorList self, at::TensorList other);
+TORCH_API ::std::vector<at::Tensor> foreach_tensor_clamp_min_list_kernel_cuda(at::TensorList self, at::TensorList other);
+TORCH_API void foreach_tensor_clamp_min_list_kernel_cuda_(at::TensorList self, at::TensorList other);
+TORCH_API ::std::vector<at::Tensor> foreach_tensor_clamp_min_scalarlist_kernel_slow(at::TensorList self, at::ArrayRef<at::Scalar> scalars);
+TORCH_API void _foreach_maximum_ScalarList_out(at::TensorList self, at::ArrayRef<at::Scalar> scalars, at::TensorList out);
+TORCH_API void foreach_tensor_clamp_min_scalarlist_kernel_slow_(at::TensorList self, at::ArrayRef<at::Scalar> scalars);
+TORCH_API ::std::vector<at::Tensor> foreach_tensor_clamp_min_scalarlist_kernel_cuda(at::TensorList self, at::ArrayRef<at::Scalar> scalars);
+TORCH_API void foreach_tensor_clamp_min_scalarlist_kernel_cuda_(at::TensorList self, at::ArrayRef<at::Scalar> scalars);
+} // namespace native
+} // namespace at

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

@@ -0,0 +1,116 @@
+#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_maximum_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_maximum";
+  static constexpr const char* overload_name = "Scalar";
+  static constexpr const char* schema_str = "_foreach_maximum.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_maximum__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_maximum_";
+  static constexpr const char* overload_name = "Scalar";
+  static constexpr const char* schema_str = "_foreach_maximum_.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_maximum_List {
+  using schema = ::std::vector<at::Tensor> (at::TensorList, at::TensorList);
+  using ptr_schema = schema*;
+  // See Note [static constexpr char* members for windows NVCC]
+  static constexpr const char* name = "aten::_foreach_maximum";
+  static constexpr const char* overload_name = "List";
+  static constexpr const char* schema_str = "_foreach_maximum.List(Tensor[] self, Tensor[] other) -> Tensor[]";
+  static ::std::vector<at::Tensor> call(at::TensorList self, at::TensorList other);
+  static ::std::vector<at::Tensor> redispatch(c10::DispatchKeySet dispatchKeySet, at::TensorList self, at::TensorList other);
+};
+
+struct TORCH_API _foreach_maximum__List {
+  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_maximum_";
+  static constexpr const char* overload_name = "List";
+  static constexpr const char* schema_str = "_foreach_maximum_.List(Tensor(a!)[] self, Tensor[] other) -> ()";
+  static void call(at::TensorList self, at::TensorList other);
+  static void redispatch(c10::DispatchKeySet dispatchKeySet, at::TensorList self, at::TensorList other);
+};
+
+struct TORCH_API _foreach_maximum_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_maximum";
+  static constexpr const char* overload_name = "ScalarList";
+  static constexpr const char* schema_str = "_foreach_maximum.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_maximum__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_maximum_";
+  static constexpr const char* overload_name = "ScalarList";
+  static constexpr const char* schema_str = "_foreach_maximum_.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_maximum_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_maximum";
+  static constexpr const char* overload_name = "Scalar_out";
+  static constexpr const char* schema_str = "_foreach_maximum.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_maximum_List_out {
+  using schema = void (at::TensorList, at::TensorList, at::TensorList);
+  using ptr_schema = schema*;
+  // See Note [static constexpr char* members for windows NVCC]
+  static constexpr const char* name = "aten::_foreach_maximum";
+  static constexpr const char* overload_name = "List_out";
+  static constexpr const char* schema_str = "_foreach_maximum.List_out(Tensor[] self, Tensor[] other, *, Tensor(a!)[] out) -> ()";
+  static void call(at::TensorList self, at::TensorList other, at::TensorList out);
+  static void redispatch(c10::DispatchKeySet dispatchKeySet, at::TensorList self, at::TensorList other, at::TensorList out);
+};
+
+struct TORCH_API _foreach_maximum_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_maximum";
+  static constexpr const char* overload_name = "ScalarList_out";
+  static constexpr const char* schema_str = "_foreach_maximum.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);
+};
+
+}} // namespace at::_ops

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

@@ -0,0 +1,82 @@
+#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/_foreach_minimum_ops.h>
+
+namespace at {
+
+
+// aten::_foreach_minimum.Scalar(Tensor[] self, Scalar scalar) -> Tensor[]
+inline ::std::vector<at::Tensor> _foreach_minimum(at::TensorList self, const at::Scalar & scalar) {
+    return at::_ops::_foreach_minimum_Scalar::call(self, scalar);
+}
+
+// aten::_foreach_minimum_.Scalar(Tensor(a!)[] self, Scalar scalar) -> ()
+inline void _foreach_minimum_(at::TensorList self, const at::Scalar & scalar) {
+    return at::_ops::_foreach_minimum__Scalar::call(self, scalar);
+}
+
+// aten::_foreach_minimum.List(Tensor[] self, Tensor[] other) -> Tensor[]
+inline ::std::vector<at::Tensor> _foreach_minimum(at::TensorList self, at::TensorList other) {
+    return at::_ops::_foreach_minimum_List::call(self, other);
+}
+
+// aten::_foreach_minimum_.List(Tensor(a!)[] self, Tensor[] other) -> ()
+inline void _foreach_minimum_(at::TensorList self, at::TensorList other) {
+    return at::_ops::_foreach_minimum__List::call(self, other);
+}
+
+// aten::_foreach_minimum.ScalarList(Tensor[] self, Scalar[] scalars) -> Tensor[]
+inline ::std::vector<at::Tensor> _foreach_minimum(at::TensorList self, at::ArrayRef<at::Scalar> scalars) {
+    return at::_ops::_foreach_minimum_ScalarList::call(self, scalars);
+}
+
+// aten::_foreach_minimum_.ScalarList(Tensor(a!)[] self, Scalar[] scalars) -> ()
+inline void _foreach_minimum_(at::TensorList self, at::ArrayRef<at::Scalar> scalars) {
+    return at::_ops::_foreach_minimum__ScalarList::call(self, scalars);
+}
+
+// aten::_foreach_minimum.Scalar_out(Tensor[] self, Scalar scalar, *, Tensor(a!)[] out) -> ()
+inline void _foreach_minimum_out(at::TensorList out, at::TensorList self, const at::Scalar & scalar) {
+    return at::_ops::_foreach_minimum_Scalar_out::call(self, scalar, out);
+}
+// aten::_foreach_minimum.Scalar_out(Tensor[] self, Scalar scalar, *, Tensor(a!)[] out) -> ()
+inline void _foreach_minimum_outf(at::TensorList self, const at::Scalar & scalar, at::TensorList out) {
+    return at::_ops::_foreach_minimum_Scalar_out::call(self, scalar, out);
+}
+
+// aten::_foreach_minimum.List_out(Tensor[] self, Tensor[] other, *, Tensor(a!)[] out) -> ()
+inline void _foreach_minimum_out(at::TensorList out, at::TensorList self, at::TensorList other) {
+    return at::_ops::_foreach_minimum_List_out::call(self, other, out);
+}
+// aten::_foreach_minimum.List_out(Tensor[] self, Tensor[] other, *, Tensor(a!)[] out) -> ()
+inline void _foreach_minimum_outf(at::TensorList self, at::TensorList other, at::TensorList out) {
+    return at::_ops::_foreach_minimum_List_out::call(self, other, out);
+}
+
+// aten::_foreach_minimum.ScalarList_out(Tensor[] self, Scalar[] scalars, *, Tensor(a!)[] out) -> ()
+inline void _foreach_minimum_out(at::TensorList out, at::TensorList self, at::ArrayRef<at::Scalar> scalars) {
+    return at::_ops::_foreach_minimum_ScalarList_out::call(self, scalars, out);
+}
+// aten::_foreach_minimum.ScalarList_out(Tensor[] self, Scalar[] scalars, *, Tensor(a!)[] out) -> ()
+inline void _foreach_minimum_outf(at::TensorList self, at::ArrayRef<at::Scalar> scalars, at::TensorList out) {
+    return at::_ops::_foreach_minimum_ScalarList_out::call(self, scalars, out);
+}
+
+}

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

@@ -0,0 +1,34 @@
+#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_minimum(at::TensorList self, const at::Scalar & scalar);
+TORCH_API void _foreach_minimum_out(at::TensorList out, at::TensorList self, const at::Scalar & scalar);
+TORCH_API void _foreach_minimum_outf(at::TensorList self, const at::Scalar & scalar, at::TensorList out);
+TORCH_API void _foreach_minimum_(at::TensorList self, const at::Scalar & scalar);
+TORCH_API ::std::vector<at::Tensor> _foreach_minimum(at::TensorList self, at::TensorList other);
+TORCH_API void _foreach_minimum_out(at::TensorList out, at::TensorList self, at::TensorList other);
+TORCH_API void _foreach_minimum_outf(at::TensorList self, at::TensorList other, at::TensorList out);
+TORCH_API void _foreach_minimum_(at::TensorList self, at::TensorList other);
+TORCH_API ::std::vector<at::Tensor> _foreach_minimum(at::TensorList self, at::ArrayRef<at::Scalar> scalars);
+TORCH_API void _foreach_minimum_out(at::TensorList out, at::TensorList self, at::ArrayRef<at::Scalar> scalars);
+TORCH_API void _foreach_minimum_outf(at::TensorList self, at::ArrayRef<at::Scalar> scalars, at::TensorList out);
+TORCH_API void _foreach_minimum_(at::TensorList self, at::ArrayRef<at::Scalar> scalars);
+
+} // namespace compositeexplicitautograd
+} // namespace at

lib/python3.12/site-packages/torch/include/ATen/ops/_foreach_tanh_compositeexplicitautograd_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 compositeexplicitautograd {
+
+TORCH_API ::std::vector<at::Tensor> _foreach_tanh(at::TensorList self);
+TORCH_API void _foreach_tanh_out(at::TensorList out, at::TensorList self);
+TORCH_API void _foreach_tanh_outf(at::TensorList self, at::TensorList out);
+TORCH_API void _foreach_tanh_(at::TensorList self);
+
+} // namespace compositeexplicitautograd
+} // namespace at

lib/python3.12/site-packages/torch/include/ATen/ops/_functional_assert_scalar_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_assert_scalar(const at::Scalar & self, c10::string_view assert_msg, const at::Tensor & dep_token);
+
+} // namespace compositeexplicitautograd
+} // namespace at

lib/python3.12/site-packages/torch/include/ATen/ops/_functional_sym_constrain_range.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_sym_constrain_range_ops.h>
+
+namespace at {
+
+
+// aten::_functional_sym_constrain_range(Scalar size, int? min, int? max, Tensor dep_token) -> Tensor
+inline at::Tensor _functional_sym_constrain_range(const at::Scalar & size, ::std::optional<int64_t> min, ::std::optional<int64_t> max, const at::Tensor & dep_token) {
+    return at::_ops::_functional_sym_constrain_range::call(size, min, max, dep_token);
+}
+
+}

lib/python3.12/site-packages/torch/include/ATen/ops/_functional_sym_constrain_range_for_size_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 _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 native
+} // namespace at

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

@@ -0,0 +1,28 @@
+#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<::std::vector<at::Tensor>,::std::vector<at::Tensor>,::std::vector<at::Tensor>,::std::vector<at::Tensor>,::std::vector<at::Tensor>> _fused_adamw(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_adamw_out(at::TensorList 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={});
+TORCH_API void _fused_adamw_outf(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 ::std::tuple<::std::vector<at::Tensor>,::std::vector<at::Tensor>,::std::vector<at::Tensor>,::std::vector<at::Tensor>,::std::vector<at::Tensor>> _fused_adamw(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_adamw_out(at::TensorList 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={});
+TORCH_API void _fused_adamw_outf(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);
+
+} // namespace compositeexplicitautograd
+} // namespace at

lib/python3.12/site-packages/torch/include/ATen/ops/_fused_adamw_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_adamw_(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_adamw_(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_moving_avg_obs_fq_helper.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/_fused_moving_avg_obs_fq_helper_ops.h>
+
+namespace at {
+
+
+// aten::_fused_moving_avg_obs_fq_helper(Tensor self, Tensor observer_on, Tensor fake_quant_on, Tensor(a!) running_min, Tensor(b!) running_max, Tensor(c!) scale, Tensor(d!) zero_point, float averaging_const, int quant_min, int quant_max, int ch_axis, bool per_row_fake_quant=False, bool symmetric_quant=False) -> (Tensor output, Tensor mask)
+inline ::std::tuple<at::Tensor,at::Tensor> _fused_moving_avg_obs_fq_helper(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) {
+    return at::_ops::_fused_moving_avg_obs_fq_helper::call(self, observer_on, fake_quant_on, running_min, running_max, scale, zero_point, averaging_const, quant_min, quant_max, ch_axis, per_row_fake_quant, symmetric_quant);
+}
+
+// aten::_fused_moving_avg_obs_fq_helper.out(Tensor self, Tensor observer_on, Tensor fake_quant_on, Tensor(a!) running_min, Tensor(b!) running_max, Tensor(c!) scale, Tensor(d!) zero_point, float averaging_const, int quant_min, int quant_max, int ch_axis, bool per_row_fake_quant=False, bool symmetric_quant=False, *, Tensor(e!) out0, Tensor(f!) out1) -> (Tensor(e!), Tensor(f!))
+inline ::std::tuple<at::Tensor &,at::Tensor &> _fused_moving_avg_obs_fq_helper_out(at::Tensor & out0, at::Tensor & out1, 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) {
+    return at::_ops::_fused_moving_avg_obs_fq_helper_out::call(self, observer_on, fake_quant_on, running_min, running_max, scale, zero_point, averaging_const, quant_min, quant_max, ch_axis, per_row_fake_quant, symmetric_quant, out0, out1);
+}
+// aten::_fused_moving_avg_obs_fq_helper.out(Tensor self, Tensor observer_on, Tensor fake_quant_on, Tensor(a!) running_min, Tensor(b!) running_max, Tensor(c!) scale, Tensor(d!) zero_point, float averaging_const, int quant_min, int quant_max, int ch_axis, bool per_row_fake_quant=False, bool symmetric_quant=False, *, Tensor(e!) out0, Tensor(f!) out1) -> (Tensor(e!), Tensor(f!))
+inline ::std::tuple<at::Tensor &,at::Tensor &> _fused_moving_avg_obs_fq_helper_outf(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) {
+    return at::_ops::_fused_moving_avg_obs_fq_helper_out::call(self, observer_on, fake_quant_on, running_min, running_max, scale, zero_point, averaging_const, quant_min, quant_max, ch_axis, per_row_fake_quant, symmetric_quant, out0, out1);
+}
+
+// aten::_fused_moving_avg_obs_fq_helper_functional(Tensor self, Tensor observer_on, Tensor fake_quant_on, Tensor running_min, Tensor running_max, Tensor scale, Tensor zero_point, float averaging_const, int quant_min, int quant_max, int ch_axis, bool per_row_fake_quant=False, bool symmetric_quant=False) -> (Tensor output, Tensor mask, Tensor running_min_out, Tensor running_max_out, Tensor scale_out, Tensor zero_point_out)
+inline ::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) {
+    return at::_ops::_fused_moving_avg_obs_fq_helper_functional::call(self, observer_on, fake_quant_on, running_min, running_max, scale, zero_point, averaging_const, quant_min, quant_max, ch_axis, per_row_fake_quant, symmetric_quant);
+}
+
+}

lib/python3.12/site-packages/torch/include/ATen/ops/_fused_moving_avg_obs_fq_helper_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,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(at::Tensor & out0, at::Tensor & out1, 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_fq_helper_outf(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);
+
+} // namespace compositeexplicitautograd
+} // namespace at

lib/python3.12/site-packages/torch/include/ATen/ops/_fused_moving_avg_obs_fq_helper_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 _fused_moving_avg_obs_fq_helper {
+  using schema = ::std::tuple<at::Tensor,at::Tensor> (const at::Tensor &, const at::Tensor &, const at::Tensor &, at::Tensor &, at::Tensor &, at::Tensor &, at::Tensor &, double, int64_t, int64_t, int64_t, bool, bool);
+  using ptr_schema = schema*;
+  // See Note [static constexpr char* members for windows NVCC]
+  static constexpr const char* name = "aten::_fused_moving_avg_obs_fq_helper";
+  static constexpr const char* overload_name = "";
+  static constexpr const char* schema_str = "_fused_moving_avg_obs_fq_helper(Tensor self, Tensor observer_on, Tensor fake_quant_on, Tensor(a!) running_min, Tensor(b!) running_max, Tensor(c!) scale, Tensor(d!) zero_point, float averaging_const, int quant_min, int quant_max, int ch_axis, bool per_row_fake_quant=False, bool symmetric_quant=False) -> (Tensor output, Tensor mask)";
+  static ::std::tuple<at::Tensor,at::Tensor> call(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);
+  static ::std::tuple<at::Tensor,at::Tensor> redispatch(c10::DispatchKeySet dispatchKeySet, 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);
+};
+
+struct TORCH_API _fused_moving_avg_obs_fq_helper_out {
+  using schema = ::std::tuple<at::Tensor &,at::Tensor &> (const at::Tensor &, const at::Tensor &, const at::Tensor &, at::Tensor &, at::Tensor &, at::Tensor &, at::Tensor &, double, int64_t, int64_t, int64_t, bool, bool, at::Tensor &, at::Tensor &);
+  using ptr_schema = schema*;
+  // See Note [static constexpr char* members for windows NVCC]
+  static constexpr const char* name = "aten::_fused_moving_avg_obs_fq_helper";
+  static constexpr const char* overload_name = "out";
+  static constexpr const char* schema_str = "_fused_moving_avg_obs_fq_helper.out(Tensor self, Tensor observer_on, Tensor fake_quant_on, Tensor(a!) running_min, Tensor(b!) running_max, Tensor(c!) scale, Tensor(d!) zero_point, float averaging_const, int quant_min, int quant_max, int ch_axis, bool per_row_fake_quant=False, bool symmetric_quant=False, *, Tensor(e!) out0, Tensor(f!) out1) -> (Tensor(e!), Tensor(f!))";
+  static ::std::tuple<at::Tensor &,at::Tensor &> call(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);
+  static ::std::tuple<at::Tensor &,at::Tensor &> redispatch(c10::DispatchKeySet dispatchKeySet, 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);
+};
+
+struct TORCH_API _fused_moving_avg_obs_fq_helper_functional {
+  using schema = ::std::tuple<at::Tensor,at::Tensor,at::Tensor,at::Tensor,at::Tensor,at::Tensor> (const at::Tensor &, const at::Tensor &, const at::Tensor &, const at::Tensor &, const at::Tensor &, const at::Tensor &, const at::Tensor &, double, int64_t, int64_t, int64_t, bool, bool);
+  using ptr_schema = schema*;
+  // See Note [static constexpr char* members for windows NVCC]
+  static constexpr const char* name = "aten::_fused_moving_avg_obs_fq_helper_functional";
+  static constexpr const char* overload_name = "";
+  static constexpr const char* schema_str = "_fused_moving_avg_obs_fq_helper_functional(Tensor self, Tensor observer_on, Tensor fake_quant_on, Tensor running_min, Tensor running_max, Tensor scale, Tensor zero_point, float averaging_const, int quant_min, int quant_max, int ch_axis, bool per_row_fake_quant=False, bool symmetric_quant=False) -> (Tensor output, Tensor mask, Tensor running_min_out, Tensor running_max_out, Tensor scale_out, Tensor zero_point_out)";
+  static ::std::tuple<at::Tensor,at::Tensor,at::Tensor,at::Tensor,at::Tensor,at::Tensor> call(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, bool symmetric_quant);
+  static ::std::tuple<at::Tensor,at::Tensor,at::Tensor,at::Tensor,at::Tensor,at::Tensor> redispatch(c10::DispatchKeySet dispatchKeySet, 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, bool symmetric_quant);
+};
+
+}} // namespace at::_ops

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

@@ -0,0 +1,63 @@
+#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/_fused_sgd_ops.h>
+
+namespace at {
+
+
+// aten::_fused_sgd_(Tensor(a!)[] self, Tensor(b!)[] grads, Tensor(c!)[] momentum_buffer_list, *, float weight_decay, float momentum, float lr, float dampening, bool nesterov, bool maximize, bool is_first_step, Tensor? grad_scale=None, Tensor? found_inf=None) -> ()
+inline void _fused_sgd_(at::TensorList self, at::TensorList grads, at::TensorList momentum_buffer_list, double weight_decay, double momentum, double lr, double dampening, bool nesterov, bool maximize, bool is_first_step, const ::std::optional<at::Tensor> & grad_scale={}, const ::std::optional<at::Tensor> & found_inf={}) {
+    return at::_ops::_fused_sgd_::call(self, grads, momentum_buffer_list, weight_decay, momentum, lr, dampening, nesterov, maximize, is_first_step, grad_scale, found_inf);
+}
+
+// aten::_fused_sgd_.tensor_lr(Tensor(a!)[] self, Tensor(b!)[] grads, Tensor(c!)[] momentum_buffer_list, *, float weight_decay, float momentum, Tensor lr, float dampening, bool nesterov, bool maximize, bool is_first_step, Tensor? grad_scale=None, Tensor? found_inf=None) -> ()
+inline void _fused_sgd_(at::TensorList self, at::TensorList grads, at::TensorList momentum_buffer_list, double weight_decay, double momentum, const at::Tensor & lr, double dampening, bool nesterov, bool maximize, bool is_first_step, const ::std::optional<at::Tensor> & grad_scale={}, const ::std::optional<at::Tensor> & found_inf={}) {
+    return at::_ops::_fused_sgd__tensor_lr::call(self, grads, momentum_buffer_list, weight_decay, momentum, lr, dampening, nesterov, maximize, is_first_step, grad_scale, found_inf);
+}
+
+// aten::_fused_sgd.out(Tensor[] self, Tensor(b!)[] grads, Tensor(c!)[] momentum_buffer_list, *, float weight_decay, float momentum, float lr, float dampening, bool nesterov, bool maximize, bool is_first_step, Tensor? grad_scale=None, Tensor? found_inf=None, Tensor(a!)[] out) -> ()
+inline void _fused_sgd_out(at::TensorList out, at::TensorList self, at::TensorList grads, at::TensorList momentum_buffer_list, double weight_decay, double momentum, double lr, double dampening, bool nesterov, bool maximize, bool is_first_step, const ::std::optional<at::Tensor> & grad_scale={}, const ::std::optional<at::Tensor> & found_inf={}) {
+    return at::_ops::_fused_sgd_out::call(self, grads, momentum_buffer_list, weight_decay, momentum, lr, dampening, nesterov, maximize, is_first_step, grad_scale, found_inf, out);
+}
+// aten::_fused_sgd.out(Tensor[] self, Tensor(b!)[] grads, Tensor(c!)[] momentum_buffer_list, *, float weight_decay, float momentum, float lr, float dampening, bool nesterov, bool maximize, bool is_first_step, Tensor? grad_scale=None, Tensor? found_inf=None, Tensor(a!)[] out) -> ()
+inline void _fused_sgd_outf(at::TensorList self, at::TensorList grads, at::TensorList momentum_buffer_list, double weight_decay, double momentum, double lr, double dampening, bool nesterov, bool maximize, bool is_first_step, const ::std::optional<at::Tensor> & grad_scale, const ::std::optional<at::Tensor> & found_inf, at::TensorList out) {
+    return at::_ops::_fused_sgd_out::call(self, grads, momentum_buffer_list, weight_decay, momentum, lr, dampening, nesterov, maximize, is_first_step, grad_scale, found_inf, out);
+}
+
+// aten::_fused_sgd(Tensor[] self, Tensor[] grads, Tensor[] momentum_buffer_list, *, float weight_decay, float momentum, float lr, float dampening, bool nesterov, bool maximize, bool is_first_step, Tensor? grad_scale=None, Tensor? found_inf=None) -> (Tensor[] self_out, Tensor[] grads_out, Tensor[] momentum_buffer_list_out)
+inline ::std::tuple<::std::vector<at::Tensor>,::std::vector<at::Tensor>,::std::vector<at::Tensor>> _fused_sgd(at::TensorList self, at::TensorList grads, at::TensorList momentum_buffer_list, double weight_decay, double momentum, double lr, double dampening, bool nesterov, bool maximize, bool is_first_step, const ::std::optional<at::Tensor> & grad_scale={}, const ::std::optional<at::Tensor> & found_inf={}) {
+    return at::_ops::_fused_sgd::call(self, grads, momentum_buffer_list, weight_decay, momentum, lr, dampening, nesterov, maximize, is_first_step, grad_scale, found_inf);
+}
+
+// aten::_fused_sgd.tensor_lr_out(Tensor[] self, Tensor(b!)[] grads, Tensor(c!)[] momentum_buffer_list, *, float weight_decay, float momentum, Tensor lr, float dampening, bool nesterov, bool maximize, bool is_first_step, Tensor? grad_scale=None, Tensor? found_inf=None, Tensor(a!)[] out) -> ()
+inline void _fused_sgd_out(at::TensorList out, at::TensorList self, at::TensorList grads, at::TensorList momentum_buffer_list, double weight_decay, double momentum, const at::Tensor & lr, double dampening, bool nesterov, bool maximize, bool is_first_step, const ::std::optional<at::Tensor> & grad_scale={}, const ::std::optional<at::Tensor> & found_inf={}) {
+    return at::_ops::_fused_sgd_tensor_lr_out::call(self, grads, momentum_buffer_list, weight_decay, momentum, lr, dampening, nesterov, maximize, is_first_step, grad_scale, found_inf, out);
+}
+// aten::_fused_sgd.tensor_lr_out(Tensor[] self, Tensor(b!)[] grads, Tensor(c!)[] momentum_buffer_list, *, float weight_decay, float momentum, Tensor lr, float dampening, bool nesterov, bool maximize, bool is_first_step, Tensor? grad_scale=None, Tensor? found_inf=None, Tensor(a!)[] out) -> ()
+inline void _fused_sgd_outf(at::TensorList self, at::TensorList grads, at::TensorList momentum_buffer_list, double weight_decay, double momentum, const at::Tensor & lr, double dampening, bool nesterov, bool maximize, bool is_first_step, const ::std::optional<at::Tensor> & grad_scale, const ::std::optional<at::Tensor> & found_inf, at::TensorList out) {
+    return at::_ops::_fused_sgd_tensor_lr_out::call(self, grads, momentum_buffer_list, weight_decay, momentum, lr, dampening, nesterov, maximize, is_first_step, grad_scale, found_inf, out);
+}
+
+// aten::_fused_sgd.tensor_lr(Tensor[] self, Tensor[] grads, Tensor[] momentum_buffer_list, *, float weight_decay, float momentum, Tensor lr, float dampening, bool nesterov, bool maximize, bool is_first_step, Tensor? grad_scale=None, Tensor? found_inf=None) -> (Tensor[] self_out, Tensor[] grads_out, Tensor[] momentum_buffer_list_out)
+inline ::std::tuple<::std::vector<at::Tensor>,::std::vector<at::Tensor>,::std::vector<at::Tensor>> _fused_sgd(at::TensorList self, at::TensorList grads, at::TensorList momentum_buffer_list, double weight_decay, double momentum, const at::Tensor & lr, double dampening, bool nesterov, bool maximize, bool is_first_step, const ::std::optional<at::Tensor> & grad_scale={}, const ::std::optional<at::Tensor> & found_inf={}) {
+    return at::_ops::_fused_sgd_tensor_lr::call(self, grads, momentum_buffer_list, weight_decay, momentum, lr, dampening, nesterov, maximize, is_first_step, grad_scale, found_inf);
+}
+
+}

lib/python3.12/site-packages/torch/include/ATen/ops/_grid_sampler_2d_cpu_fallback_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 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(const at::Tensor & input, const at::Tensor & grid, int64_t interpolation_mode, int64_t padding_mode, bool align_corners, at::Tensor & out);
+} // namespace native
+} // namespace at

lib/python3.12/site-packages/torch/include/ATen/ops/_has_compatible_shallow_copy_type.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/_has_compatible_shallow_copy_type_ops.h>
+
+namespace at {
+
+
+// aten::_has_compatible_shallow_copy_type(Tensor self, Tensor from) -> bool
+inline bool _has_compatible_shallow_copy_type(const at::Tensor & self, const at::Tensor & from) {
+    return at::_ops::_has_compatible_shallow_copy_type::call(self, from);
+}
+
+}

lib/python3.12/site-packages/torch/include/ATen/ops/_histogramdd_bin_edges_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 void _histogramdd_bin_edges_out(at::TensorList out, const at::Tensor & self, at::IntArrayRef bins, ::std::optional<at::ArrayRef<double>> range=::std::nullopt, const ::std::optional<at::Tensor> & weight={}, bool density=false);
+TORCH_API void _histogramdd_bin_edges_outf(const at::Tensor & self, at::IntArrayRef bins, ::std::optional<at::ArrayRef<double>> range, const ::std::optional<at::Tensor> & weight, bool density, at::TensorList out);
+
+} // namespace compositeexplicitautograd
+} // namespace at

lib/python3.12/site-packages/torch/include/ATen/ops/_histogramdd_bin_edges_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 void _histogramdd_bin_edges_out(const at::Tensor & self, at::IntArrayRef bins, ::std::optional<at::ArrayRef<double>> range, const ::std::optional<at::Tensor> & weight, bool density, at::TensorList out);
+TORCH_API ::std::vector<at::Tensor> histogramdd_bin_edges(const at::Tensor & self, at::IntArrayRef bins, ::std::optional<at::ArrayRef<double>> range=::std::nullopt, const ::std::optional<at::Tensor> & weight={}, bool density=false);
+} // namespace native
+} // namespace at

lib/python3.12/site-packages/torch/include/ATen/ops/_histogramdd_from_bin_cts_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 at::Tensor & _histogramdd_from_bin_cts_out(const at::Tensor & self, at::IntArrayRef bins, ::std::optional<at::ArrayRef<double>> range, const ::std::optional<at::Tensor> & weight, bool density, at::Tensor & out);
+TORCH_API at::Tensor _histogramdd(const at::Tensor & self, at::IntArrayRef bins, ::std::optional<at::ArrayRef<double>> range=::std::nullopt, const ::std::optional<at::Tensor> & weight={}, bool density=false);
+} // namespace native
+} // namespace at

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

@@ -0,0 +1,25 @@
+#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 _index_put_impl(const at::Tensor & self, const c10::List<::std::optional<at::Tensor>> & indices, const at::Tensor & values, bool accumulate=false, bool unsafe=false);
+TORCH_API at::Tensor & _index_put_impl_out(const at::Tensor & self, const c10::List<::std::optional<at::Tensor>> & indices, const at::Tensor & values, bool accumulate, bool unsafe, at::Tensor & out);
+TORCH_API at::Tensor & _index_put_impl_(at::Tensor & self, const c10::List<::std::optional<at::Tensor>> & indices, const at::Tensor & values, bool accumulate=false, bool unsafe=false);
+TORCH_API at::Tensor & _index_put_impl_quantized_cpu_(at::Tensor & self, const c10::List<::std::optional<at::Tensor>> & indices, const at::Tensor & values, bool accumulate=false, bool unsafe=false);
+TORCH_API at::Tensor & _index_put_impl_quantized_cuda_(at::Tensor & self, const c10::List<::std::optional<at::Tensor>> & indices, const at::Tensor & values, bool accumulate=false, bool unsafe=false);
+} // namespace native
+} // namespace at