Home
last modified time | relevance | path

Searched refs:optimizer_weight_shard_size (Results 1 – 8 of 8) sorted by relevance

/third_party/mindspore/mindspore/parallel/
D_auto_parallel_context.py549 def set_optimizer_weight_shard_size(self, optimizer_weight_shard_size): argument
558 …if not isinstance(optimizer_weight_shard_size, int) or isinstance(optimizer_weight_shard_size, boo…
560 but got {type(optimizer_weight_shard_size)}.")
561 if optimizer_weight_shard_size <= 1:
565 self._context_handle.set_optimizer_weight_shard_size(optimizer_weight_shard_size)
685 communi_parallel_mode=str, optimizer_weight_shard_size=int,
/third_party/mindspore/mindspore/ccsrc/frontend/parallel/tensor_layout/
Dtensor_layout.cc436 …int64_t optimizer_weight_shard_size = ParallelContext::GetInstance()->optimizer_weight_shard_size(… in GenerateOptShardSliceShape() local
437 if (optimizer_weight_shard_size != -1) { in GenerateOptShardSliceShape()
438 repeated_num = optimizer_weight_shard_size; in GenerateOptShardSliceShape()
/third_party/mindspore/mindspore/ccsrc/frontend/parallel/
Dcontext.h102 void set_optimizer_weight_shard_size(int64_t optimizer_weight_shard_size);
103 int64_t optimizer_weight_shard_size() const { return optimizer_weight_shard_size_; } in optimizer_weight_shard_size() function
Dcontext.cc145 void ParallelContext::set_optimizer_weight_shard_size(int64_t optimizer_weight_shard_size) { in set_optimizer_weight_shard_size() argument
146 optimizer_weight_shard_size_ = optimizer_weight_shard_size; in set_optimizer_weight_shard_size()
/third_party/mindspore/mindspore/ccsrc/frontend/parallel/ops_info/
Doperator_info.cc565 …int64_t optimizer_weight_shard_size = ParallelContext::GetInstance()->optimizer_weight_shard_size(… in CreateGroupForOptShard() local
566 if (optimizer_weight_shard_size != -1) { in CreateGroupForOptShard()
570 if (repeated_size % optimizer_weight_shard_size != 0) { in CreateGroupForOptShard()
571 …S_LOG(WARNING) << "Parallel optimizer: optimizer_weight_shard_size " << optimizer_weight_shard_size in CreateGroupForOptShard()
575 repeated_size = repeated_size / optimizer_weight_shard_size; in CreateGroupForOptShard()
579 group_devices.begin() + index / optimizer_weight_shard_size * optimizer_weight_shard_size, in CreateGroupForOptShard()
580 … group_devices.begin() + (index / optimizer_weight_shard_size + 1) * optimizer_weight_shard_size); in CreateGroupForOptShard()
609 tensor_layout->set_opt_weight_shard_size(LongToInt(optimizer_weight_shard_size)); in CreateGroupForOptShard()
/third_party/mindspore/tests/ut/python/parallel/
Dtest_shared_param_and_mix_precision.py100 context.set_auto_parallel_context(optimizer_weight_shard_size=2)
Dtest_parallel_optimizer.py137 context.set_auto_parallel_context(optimizer_weight_shard_size=2)
/third_party/mindspore/mindspore/ccsrc/pipeline/jit/
Dinit.cc193 .def("get_optimizer_weight_shard_size", &ParallelContext::optimizer_weight_shard_size, in PYBIND11_MODULE()