pytorch/caffe2/operators/load_save_op_util.h
Sean Lynch f73e32cd04 Reduce amount of work done within a global lock within ParallelLoadOp (#43508)
Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/43508

Differential Revision: D22952007

fbshipit-source-id: 11e28d20175271e6068edce8cb36f9fcf867a02a
2020-08-26 18:19:40 -07:00

61 lines
1.6 KiB
C++

#ifndef CAFFE2_OPERATORS_LOAD_SAVE_OP_UTIL_H_
#define CAFFE2_OPERATORS_LOAD_SAVE_OP_UTIL_H_
#include <set>
#include <string>
#include <unordered_map>
#include "caffe2/core/blob.h"
#include "caffe2/core/blob_serialization.h"
namespace caffe2 {
namespace load_save_op_util {
struct BlobState {
int64_t total_size;
int64_t current_size;
bool is_tensor;
std::set<int32_t> seen_chunks_ids;
explicit BlobState(
int64_t total_size = 0,
int64_t current_size = 0,
bool is_tensor = false)
: total_size(total_size),
current_size(current_size),
is_tensor(is_tensor) {}
};
CAFFE2_API std::string buildBlobNameFromDbKey(
const std::string& dbKey,
const std::string& strip_prefix = "",
const std::string& add_prefix = "");
// We are tracking sizes of already read tensor parts while reading data
// chunks. This way we can make sure that all chunks were loaded in the end.
CAFFE2_API void ProcessBlob(
Blob* blob,
const BlobProto& proto,
std::unordered_map<std::string, BlobState>* blob_states_ptr,
const std::string& key,
int* loaded_blobs);
CAFFE2_API void prepareBlob(
Blob* blob,
std::unordered_map<std::string, BlobState>* blob_states_ptr,
const std::string& key);
CAFFE2_API void updateBlobStates(
const BlobProto& proto,
std::unordered_map<std::string, BlobState>* blob_states_ptr,
const std::string& key,
int* loaded_blobs);
CAFFE2_API void validateBlobStates(
const std::unordered_map<std::string, BlobState>& blob_states);
} // namespace load_save_op_util
} // namespace caffe2
#endif // CAFFE2_OPERATORS_LOAD_SAVE_OP_UTIL_H_