| /external/pytorch/torch/distributed/algorithms/ddp_comm_hooks/ |
| D | default_hooks.py | 34 process_group: dist.ProcessGroup, bucket: dist.GradBucket argument
57 bucket: dist.GradBucket, argument
105 bucket: dist.GradBucket, argument
152 hook: Callable[[Any, dist.GradBucket], torch.futures.Future[torch.Tensor]] argument
153 ) -> Callable[[Any, dist.GradBucket], torch.futures.Future[torch.Tensor]]:
169 hook_state, bucket: dist.GradBucket argument
190 hook: Callable[[Any, dist.GradBucket], torch.futures.Future[torch.Tensor]] argument
191 ) -> Callable[[Any, dist.GradBucket], torch.futures.Future[torch.Tensor]]:
208 hook_state, bucket: dist.GradBucket argument
|
| D | ddp_zero_hook.py | 22 bucket: dist.GradBucket, argument
102 bucket: dist.GradBucket, argument
135 bucket: dist.GradBucket, argument
175 hook: Callable[[Any, dist.GradBucket], torch.futures.Future], argument
179 ) -> Callable[[Any, dist.GradBucket], torch.futures.Future[torch.Tensor]]:
248 bucket: dist.GradBucket, argument
337 hook: Callable[[Any, dist.GradBucket], torch.futures.Future], argument
341 ) -> Callable[[Any, dist.GradBucket], torch.futures.Future[torch.Tensor]]:
410 bucket: dist.GradBucket, argument
|
| D | optimizer_overlap_hooks.py | 51 ) -> Callable[[Any, dist.GradBucket], torch.futures.Future[torch.Tensor]]:
66 bucket: dist.GradBucket, argument
129 hook: Callable[[Any, dist.GradBucket], torch.futures.Future[torch.Tensor]], argument
131 ) -> Callable[[Any, dist.GradBucket], torch.futures.Future[torch.Tensor]]:
139 hook_state, bucket: dist.GradBucket argument
|
| D | debugging_hooks.py | 4 from torch.distributed import GradBucket
10 def noop_hook(_: Any, bucket: GradBucket) -> torch.futures.Future[torch.Tensor]: argument
|
| D | quantization_hooks.py | 47 process_group: dist.ProcessGroup, bucket: dist.GradBucket argument
122 process_group: dist.ProcessGroup, bucket: dist.GradBucket, bucket_size=512 argument
|
| D | post_localSGD_hook.py | 70 state: PostLocalSGDState, bucket: dist.GradBucket argument
|
| D | mixed_precision_hooks.py | 26 hook_state: _AllreduceUpcastHookState, bucket: dist.GradBucket argument
|
| D | powerSGD_hook.py | 340 state: PowerSGDState, bucket: dist.GradBucket argument
651 state: PowerSGDState, bucket: dist.GradBucket argument
|
| /external/pytorch/torch/csrc/distributed/c10d/ |
| D | default_comm_hooks.hpp | 21 c10::intrusive_ptr<c10::ivalue::Future> runHook(GradBucket& bucket) override;
32 c10::intrusive_ptr<c10::ivalue::Future> runHook(GradBucket& bucket) override;
49 c10::intrusive_ptr<c10::ivalue::Future> runHook(GradBucket& bucket) override;
|
| D | default_comm_hooks.cpp | 12 GradBucket& bucket) { in runHook()
20 GradBucket& bucket) { in runHook()
47 GradBucket& bucket) { in runHook()
|
| D | comm.hpp | 19 class TORCH_API GradBucket { class
21 explicit GradBucket( in GradBucket() function in c10d::GradBucket
110 GradBucket& bucket) = 0;
|
| D | python_comm_hook.h | 24 c10::intrusive_ptr<c10::ivalue::Future> runHook(GradBucket& bucket) override;
|
| D | reducer.hpp | 110 GradBucket& grad_bucket);
114 GradBucket& grad_bucket);
119 std::vector<c10d::GradBucket> get_grad_buckets(
|
| D | python_comm_hook.cpp | 22 GradBucket& bucket) { in runHook()
|
| D | comm.cpp | 95 std::vector<at::Tensor> GradBucket::getGradients() const { in getGradients()
|
| D | init.cpp | 466 shared_ptr_class_<::c10d::GradBucket>( in c10d_init()
479 &::c10d::GradBucket::getIndex, in c10d_init()
491 &::c10d::GradBucket::getBuffer, in c10d_init()
500 &::c10d::GradBucket::getGradients, in c10d_init()
508 &::c10d::GradBucket::getParameters, in c10d_init()
517 &::c10d::GradBucket::isLast, in c10d_init()
526 &::c10d::GradBucket::setBuffer, in c10d_init()
646 [](::c10d::Reducer& reducer, ::c10d::GradBucket& bucket) in c10d_init()
655 [](::c10d::Reducer& reducer, ::c10d::GradBucket& bucket) in c10d_init()
|
| D | reducer.cpp | 485 std::vector<c10d::GradBucket> Reducer::get_grad_buckets( in get_grad_buckets()
488 std::vector<c10d::GradBucket> gradBuckets; in get_grad_buckets()
934 GradBucket& grad_bucket) { in run_comm_hook()
943 GradBucket& grad_bucket) { in run_allreduce_hook()
978 GradBucket grad_bucket( in all_reduce_bucket()
|
| /external/pytorch/docs/source/ |
| D | ddp_comm_hooks.rst | 26 Particularly, :class:`torch.distributed.GradBucket` represents a bucket of gradient tensors to be a…
28 .. autoclass:: torch.distributed.GradBucket
30 .. autofunction:: torch.distributed.GradBucket.index
31 .. autofunction:: torch.distributed.GradBucket.buffer
32 .. autofunction:: torch.distributed.GradBucket.gradients
33 .. autofunction:: torch.distributed.GradBucket.is_last
34 .. autofunction:: torch.distributed.GradBucket.set_buffer
35 .. autofunction:: torch.distributed.GradBucket.parameters
42 The input ``bucket`` is a :class:`torch.distributed.GradBucket` object.
|
| /external/pytorch/torch/distributed/ |
| D | __init__.py | 54 GradBucket,
|
| /external/pytorch/torch/_C/ |
| D | _distributed_c10d.pyi | 31 class GradBucket:
58 def _get_zeros_like_grad_buckets(self) -> list[GradBucket]: ...
68 def _run_comm_hook(self, bucket: GradBucket) -> Future: ...
|
| /external/pytorch/test/distributed/ |
| D | test_c10d_ucc.py | 895 def comm_hook(state: object, bucket: dist.GradBucket) -> int: argument
907 def comm_hook(state: object, bucket: dist.GradBucket): argument
963 state: object, bucket: dist.GradBucket argument
|
| D | test_c10d_gloo.py | 2067 def comm_hook(state: object, bucket: dist.GradBucket) -> int: argument
2079 def comm_hook(state: object, bucket: dist.GradBucket): argument
2133 state: object, bucket: dist.GradBucket argument
|
| D | test_c10d_common.py | 872 self, state: object, bucket: dist.GradBucket argument
|
| /external/pytorch/torch/distributed/optim/ |
| D | zero_redundancy_optimizer.py | 254 self.bucket_index_to_bucket: Dict[int, dist.GradBucket] = {}
|
| /external/pytorch/torch/nn/parallel/ |
| D | distributed.py | 2216 and sig.parameters["bucket"].annotation != dist.GradBucket
|