FM / core

/usr/local/lib64/python3.6/site-packages/torch/include/ATen/core/

/usr/local/lib64/python3.6/site-packages/torch/include/ATen/core

Name	Size	Mode	Actions
boxing/	-	0755	rm
dispatch/	-	0755	rm
op_registration/	-	0755	rm
alias_info.h	2986	0644	edit dl rm
Array.h	768	0644	edit dl rm
ATenGeneral.h	45	0644	edit dl rm
ATenOpList.h	246	0644	edit dl rm
aten_interned_strings.h	25389	0644	edit dl rm
Backtrace.h	59	0644	edit dl rm
blob.h	5422	0644	edit dl rm
builtin_function.h	3649	0644	edit dl rm
DeprecatedTypeProperties.h	3773	0644	edit dl rm
DeprecatedTypePropertiesRegistry.h	795	0644	edit dl rm
Dict.h	13195	0644	edit dl rm
Dict_inl.h	7996	0644	edit dl rm
Dimname.h	1188	0644	edit dl rm
DimVector.h	247	0644	edit dl rm
DistributionsHelper.h	12594	0644	edit dl rm
Formatting.h	959	0644	edit dl rm
function.h	2145	0644	edit dl rm
functional.h	1460	0644	edit dl rm
function_schema.h	13577	0644	edit dl rm
function_schema_inl.h	9319	0644	edit dl rm
Generator.h	4935	0644	edit dl rm
grad_mode.h	210	0644	edit dl rm
interned_strings.h	25332	0644	edit dl rm
interned_strings_class.h	770	0644	edit dl rm
ivalue.h	38823	0644	edit dl rm
ivalue_inl.h	59963	0644	edit dl rm
ivalue_to.h	756	0644	edit dl rm
jit_type.h	75971	0644	edit dl rm
jit_type_base.h	6508	0644	edit dl rm
LegacyTypeDispatch.h	4626	0644	edit dl rm
List.h	15667	0644	edit dl rm
List_inl.h	11012	0644	edit dl rm
Macros.h	44	0644	edit dl rm
MT19937RNGEngine.h	6410	0644	edit dl rm
NamedTensor.h	5050	0644	edit dl rm
operator_name.h	3018	0644	edit dl rm
PhiloxRNGEngine.h	6496	0644	edit dl rm
PythonModeTLS.h	403	0644	edit dl rm
qualified_name.h	4358	0644	edit dl rm
QuantizerBase.h	2443	0644	edit dl rm
Range.h	418	0644	edit dl rm
Reduction.h	461	0644	edit dl rm
rref_interface.h	1144	0644	edit dl rm
Scalar.h	29	0644	edit dl rm
ScalarType.h	33	0644	edit dl rm
stack.h	6034	0644	edit dl rm
Tensor.h	1756	0644	edit dl rm
TensorAccessor.h	10296	0644	edit dl rm
TensorBase.h	32767	0644	edit dl rm
TensorBody.h	247555	0644	edit dl rm
TransformationHelper.h	6911	0644	edit dl rm
typeid.h	29	0644	edit dl rm
UndefinedTensorImpl.h	42	0644	edit dl rm
UnsafeFromTH.h	708	0644	edit dl rm
VariableHooksInterface.h	3312	0644	edit dl rm
Variadic.h	2257	0644	edit dl rm
Vitals.h	2305	0644	edit dl rm

Edit: /usr/local/lib64/python3.6/site-packages/torch/include/ATen/core/Generator.h (4935B)

#pragma once

#include <stdint.h>
#include <mutex>
#include <deque>
#include <atomic>
#include <typeinfo>
#include <utility>
#include <cstddef>

#include <c10/util/Exception.h>
#include <c10/util/C++17.h>
#include <c10/util/intrusive_ptr.h>
#include <c10/core/Device.h>
#include <c10/core/DispatchKeySet.h>

// For the record I don't think this is a correct pimpl idiom.
// Including Impl header in interface header defeats the purpose
// because you can't change Impl private members without forcing
// everything that included the interface to rebuild.
// Impl should be forward-declared in the interface header instead.
#include <c10/core/GeneratorImpl.h>

/**
 * Note [Generator]
 * ~~~~~~~~~~~~~~~~
 * A Pseudo Random Number Generator (PRNG) is an engine that uses an algorithm to
 * generate a seemingly random sequence of numbers, that may be later be used in creating
 * a random distribution. Such an engine almost always maintains a state and requires a
 * seed to start off the creation of random numbers. Often times, users have
 * found it beneficial to be able to explicitly create, retain, and destroy
 * PRNG states and also be able to have control over the seed value.
 *
 * A Generator in ATen gives users the ability to read, write and modify a PRNG engine.
 * For instance, it does so by letting users seed a PRNG engine, fork the state of the
 * engine, etc.
 *
 * By default, there is one generator per device, and a device's generator is
 * lazily created. A user can use the torch.Generator() api to create their own generator.
 */

/**
 * Note [Acquire lock when using random generators]
 * ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 * Generator and its derived classes are NOT thread-safe. Please note that most of the
 * places where we have inserted locking for generators are historically based, and we
 * haven't actually checked that everything is truly thread safe (and it probably isn't).
 * Please use the public mutex_ when using any methods from these classes, except for the
 * read-only methods. You can learn about the usage by looking into the unittests
 * (aten/src/ATen/cpu_generator_test.cpp) and other places where we have used lock_guard.
 *
 * TODO: Look into changing the threading semantics of Generators in ATen (e.g., making
 * them non-thread safe and instead making the generator state splittable, to accommodate
 * forks into other threads).
 */

namespace at {

class Tensor;

struct TORCH_API Generator {
  Generator() {}

explicit Generator(c10::intrusive_ptr<c10::GeneratorImpl> gen_impl)
   : impl_(std::move(gen_impl)) {
    if (impl_.get() == nullptr) {
      throw std::runtime_error("GeneratorImpl with nullptr is not supported");
    }
  }

bool operator==(const Generator& rhs) const {
    return this->impl_ == rhs.impl_;
  }

bool operator!=(const Generator& rhs) const {
    return !((*this) == rhs);
  }

bool defined() const {
    return static_cast<bool>(impl_);
  }

c10::GeneratorImpl* unsafeGetGeneratorImpl() const {
    return impl_.get();
  }

c10::GeneratorImpl* unsafeReleaseGeneratorImpl() {
    return impl_.release();
  }

const c10::intrusive_ptr<c10::GeneratorImpl>& getIntrusivePtr() const {
    return impl_;
  }

void set_current_seed(uint64_t seed) { impl_->set_current_seed(seed); }

uint64_t current_seed() const { return impl_->current_seed(); }

uint64_t seed() { return impl_->seed(); }

// Implementation not inlined to prevent cycle reference between
  // `ATen/core/Generator.h` and `ATen/core/Tensor.h`
  void set_state(const at::Tensor& new_state);

at::Tensor get_state() const;

std::mutex& mutex() {
    return impl_->mutex_;
  }

DispatchKeySet key_set() const {
    return impl_->key_set();
  }

Device device() const { return impl_->device(); }

inline void set_pyobj(PyObject* pyobj) const noexcept {
    impl_->set_pyobj(pyobj);
  }

inline PyObject* pyobj() const noexcept {
    return impl_->pyobj();
  }

template<typename T>
  T* get() const { return static_cast<T*>(impl_.get()); }

Generator clone() const {
    return Generator(impl_->clone());
  }

private:
  c10::intrusive_ptr<c10::GeneratorImpl> impl_;
};

template<class Impl, class... Args>
Generator make_generator(Args&&... args) {
  return Generator(c10::make_intrusive<Impl>(std::forward<Args>(args)...));
}

namespace detail {

/**
 * Helper function for checking the validity of new random generator
 * state. Right now following conditions are checked:
 *
 * - The new state tensor must be a torch.ByteTensor
 * - Data of the new state tensor must be contiguous
 */
static inline void check_rng_state(const c10::TensorImpl& new_state) {
  TORCH_CHECK_TYPE(
    new_state.layout() == kStrided && new_state.device().type() == kCPU && new_state.dtype() == kByte,
    "RNG state must be a torch.ByteTensor"
  );

TORCH_CHECK(new_state.is_contiguous(), "RNG state must be contiguous");
}

} // namespace detail

} // namespace at

cmd: