| /third_party/mindspore/tests/st/nccl/ |
| D | test_nccl_all_reduce_op.py | 45 self.all_reduce1 = P.AllReduce(self.op0, group=NCCL_WORLD_COMM_GROUP)
46 self.all_reduce2 = P.AllReduce(self.op1, group=NCCL_WORLD_COMM_GROUP)
47 self.all_reduce3 = P.AllReduce(self.op2, group=NCCL_WORLD_COMM_GROUP)
90 self.all_reduce1 = P.AllReduce(self.op0, group=NCCL_WORLD_COMM_GROUP)
91 self.all_reduce2 = P.AllReduce(self.op1, group=NCCL_WORLD_COMM_GROUP)
92 self.all_reduce3 = P.AllReduce(self.op2, group=NCCL_WORLD_COMM_GROUP)
131 self.all_reduce = P.AllReduce(self.op, group=NCCL_WORLD_COMM_GROUP)
|
| /third_party/mindspore/mindspore/ccsrc/backend/kernel_compiler/gpu/nccl/ |
| D | nccl_collective_gpu_kernel.cc | 22 …AllReduce, KernelAttr().AddAllSameAttr(true).AddInputAttr(kNumberTypeFloat32).AddOutputAttr(kNumbe…
25 …AllReduce, KernelAttr().AddAllSameAttr(true).AddInputAttr(kNumberTypeFloat16).AddOutputAttr(kNumbe…
27 MS_REG_GPU_KERNEL_ONE(AllReduce,
|
| D | nccl_gpu_kernel.h | 34 typedef ncclResult_t (*AllReduce)(const void *, void *, size_t, ncclDataType_t, ncclRedOp_t, cudaSt… typedef
|
| /third_party/mindspore/mindspore/ops/_grad/ |
| D | grad_comm_ops.py | 25 from ..operations.comm_ops import (AllGather, _MiniStepAllGather, _HostAllGather, AllReduce, Neighb…
33 @bprop_getters.register(AllReduce)
37 all_reduce_grad = AllReduce(ReduceOp.SUM, self.group)
165 all_reduce = AllReduce(group=group)
240 all_reduce = AllReduce(ReduceOp.SUM, self.group).add_prim_attr("fusion", fusion)
272 all_reduce = AllReduce(ReduceOp.SUM, self.group).add_prim_attr("fusion", fusion)
402 all_reduce = AllReduce(group=group)
453 all_reduce = AllReduce(group=group)
|
| /third_party/mindspore/mindspore/nn/wrap/ |
| D | grad_reducer.py | 21 from mindspore.ops.operations.comm_ops import AllReduce, AllGather
45 op = AllReduce('sum', group)
68 op = AllReduce('sum', group)
389 self.allreduce = AllReduce().add_prim_attr('fusion', fusion_type)
|
| /third_party/mindspore/mindspore/ccsrc/fl/server/ |
| D | collective_ops_impl.cc | 214 bool CollectiveOpsImpl::AllReduce(const void *sendbuff, void *recvbuff, size_t count) { in AllReduce() function in mindspore::fl::server::CollectiveOpsImpl
249 template bool CollectiveOpsImpl::AllReduce<float>(const void *sendbuff, void *recvbuff, size_t coun…
250 template bool CollectiveOpsImpl::AllReduce<size_t>(const void *sendbuff, void *recvbuff, size_t cou…
251 template bool CollectiveOpsImpl::AllReduce<int>(const void *sendbuff, void *recvbuff, size_t count);
|
| D | collective_ops_impl.h | 48 bool AllReduce(const void *sendbuff, void *recvbuff, size_t count);
|
| /third_party/mindspore/mindspore/ccsrc/runtime/device/gpu/distribution/ |
| D | collective_wrapper.cc | 37 ncclResult_t AllReduce(const void *input_addr, void *output_addr, size_t count, ncclDataType_t data… in AllReduce() function
39 …return NCCLWrapper::instance().AllReduce(input_addr, output_addr, count, data_type, reduce_type, s… in AllReduce()
|
| D | nccl_wrapper.h | 38 …ncclResult_t AllReduce(const void *input_addr, void *output_addr, size_t count, ncclDataType_t dat…
|
| D | collective_wrapper.h | 39 extern "C" EXPORT_WRAPPER ncclResult_t AllReduce(const void *input_addr, void *output_addr, size_t …
|
| /third_party/mindspore/tests/ut/cpp/python_input/gtest_input/optimizer/ |
| D | opt_test.py | 1097 AllReduce = Primitive('AllReduce')
1105 return AddN((z, Mul(y, AllReduce(x))))
1109 return AddN((z, Mul(AllReduce(x), y)))
1113 return AddN((Mul(y, AllReduce(x)), z))
1117 return AddN((Mul(AllReduce(x), y), z))
1121 return Mul(AllReduce(AddN((z, x))), y)
1125 return AddN((Mul(AllReduce(x), y), Mul(z, z)))
1129 return AddN((Mul(z, z), Mul(AllReduce(x), y)))
1133 return Mul(AllReduce(AddN((Mul(z, z), x))), y)
|
| /third_party/mindspore/mindspore/ccsrc/backend/kernel_compiler/hccl/ |
| D | hcom_all_reduce.h | 38 MS_HCCL_REG_KERNEL(AllReduce, HcomAllReduceKernel);
|
| /third_party/mindspore/tests/ut/python/parallel/ |
| D | test_sparse_feature_bprop.py | 25 from mindspore.ops.operations.comm_ops import AllReduce
57 self.all_reduce = AllReduce()
|
| /third_party/mindspore/mindspore/ccsrc/fl/server/kernel/ |
| D | fed_avg_kernel.h | 92 …if (!CollectiveOpsImpl::GetInstance().AllReduce<T>(weight_addr, weight_addr, weight_size / sizeof(… in InitKernel()
96 if (!CollectiveOpsImpl::GetInstance().AllReduce<S>(data_size_addr, data_size_addr, 1)) { in InitKernel()
|
| /third_party/mindspore/tests/st/nontask_sink/ |
| D | test_allreduce.py | 35 self.all_reduce = P.AllReduce()
|
| /third_party/mindspore/tests/st/pynative/data_parallel/ |
| D | test_pynative_nccl_allreduce.py | 36 self.all_reduce = P.AllReduce()
|
| D | test_pynative_hccl_allreduce.py | 39 self.all_reduce = P.AllReduce()
|
| D | test_pynative_hccl.py | 38 self.all_reduce = P.AllReduce()
|
| /third_party/mindspore/tests/st/networks/models/resnet50/src_thor/ |
| D | metric.py | 52 self.allreduce = P.AllReduce(P.ReduceOp.SUM, GlobalComm.WORLD_COMM_GROUP)
|
| /third_party/mindspore/tests/st/networks/models/resnet50/src/ |
| D | metric.py | 52 self.allreduce = P.AllReduce(P.ReduceOp.SUM, GlobalComm.WORLD_COMM_GROUP)
|
| /third_party/mindspore/tests/ut/python/communication/ |
| D | test_comm.py | 28 from mindspore.ops.operations.comm_ops import AllReduce, AllGather, AlltoAll, ReduceOp, ReduceScatt…
51 self.reduce = AllReduce(op)
|
| /third_party/mindspore/tests/ut/python/pipeline/parse/ |
| D | test_for_stmt.py | 109 op = P.AllReduce().add_prim_attr('fusion', i)
|
| /third_party/mindspore/tests/ut/cpp/python_input/gtest_input/pre_activate/ |
| D | insert_tensor_move_for_hccl_op.py | 21 all_reduce = P.AllReduce()
|
| D | ir_fusion_test.py | 23 allreduce = P.AllReduce()
|
| /third_party/mindspore/mindspore/boost/ |
| D | adasum.py | 257 server_all_reduce = P.AllReduce("sum", group_name)
273 self.sync_barrier = P.AllReduce("sum", group=broadcast_group_name)
|