/usr/local/lib64/python3.6/site-packages/torch/include/caffe2/opt
NameSizeModeActions
annotations.h21410644editdlrm
backend_cutting.h5860644editdlrm
backend_transformer_base.h27730644editdlrm
bound_shape_inferencer.h58840644editdlrm
converter.h30130644editdlrm
device.h4420644editdlrm
distributed.h11260644editdlrm
fakefp16_transform.h5840644editdlrm
fusion.h41410644editdlrm
glow_net_transform.h16250644editdlrm
mobile.h3880644editdlrm
onnxifi_op.h194970644editdlrm
onnxifi_transformer.h68590644editdlrm
onnx_convert.h12790644editdlrm
optimizer.h4020644editdlrm
optimize_ideep.h3510644editdlrm
passes.h26420644editdlrm
shape_info.h47840644editdlrm
tvm_transformer.h29920644editdlrm
Edit: /usr/local/lib64/python3.6/site-packages/torch/include/caffe2/opt/distributed.h (1126B)
#pragma once #include "caffe2/core/common.h" #include "caffe2/core/logging.h" #include "caffe2/proto/caffe2_pb.h" #include "nomnigraph/Representations/NeuralNet.h" namespace caffe2 { /// \brief Convert to an NNModule and apply a mapping of /// tensor names to DeviceOptions to it. /// /// This *only* applies the map to Declare/Export /// nodes, which are representationally equivalent to /// external_input/external_output in caffe2 NetDefs. /// /// Throws an exception if the passed in blobMap contains /// blobs that are not present in the NNModule. TORCH_API nom::repr::NNModule convertToNNModule( caffe2::NetDef&, std::map); /// Helpers for the convertToNNModule for use /// if you already have an NNModule. /// You probably don't want to use these /// if you can use convertToNNModule instead. TORCH_API void addBlobDeviceOptions( std::map blobMap, nom::repr::NNModule* nn); TORCH_API void injectDataEdgeIndicators(nom::repr::NNModule* nn); TORCH_API void removeDataEdgeIndicators(nom::repr::NNModule* nn); } // namespace caffe2