Skip to content

S
Serving

PaddlePaddle / Serving
大约 2 年 前同步成功

通知 187
Star 833
Fork 253
S
Serving
“7711d78cb3fa81eea5137c103ec49c04f3bdd109”上不存在“develop/doc/design/kernel_selection.html”
提交 f3953aa5 编写于 作者: T TeslaZhao 提交者： GitHub
浏览文件
操作

Merge branch 'develop' into develop

上级 a27774f2 7b7d650a
显示空白变更内容
内联 并排
Showing with 2342 addition and 972 deletion
README.md
浏览文件 @ f3953aa5
......@@ -172,6 +172,11 @@ Here, `client.predict` function has two arguments. `feed` is a `python dict` wit
- [An End-to-end tutorial from training to inference service deployment](doc/TRAIN_TO_SERVICE.md)
- [Write Bert-as-Service in 10 minutes](doc/BERT_10_MINS.md)
### Tutorial at AIStudio
- [Introduction to PaddleServing](https://aistudio.baidu.com/aistudio/projectdetail/605819)
- [Image Segmentation on Paddle Serving](https://aistudio.baidu.com/aistudio/projectdetail/457715)
- [Sentimental Analysis](https://aistudio.baidu.com/aistudio/projectdetail/509014)
### Developers
- [How to config Serving native operators on server side?](doc/SERVER_DAG.md)
- [How to develop a new Serving operator?](doc/NEW_OPERATOR.md)
......
README_CN.md
浏览文件 @ f3953aa5
......@@ -169,6 +169,11 @@ print(fetch_map)
- [端到端完成从训练到部署全流程](doc/TRAIN_TO_SERVICE_CN.md)
- [十分钟构建Bert-As-Service](doc/BERT_10_MINS_CN.md)
### AIStudio教程
- [PaddleServing作业](https://aistudio.baidu.com/aistudio/projectdetail/605819)
- [PaddleServing图像分割](https://aistudio.baidu.com/aistudio/projectdetail/457715)
- [PaddleServing情感分析](https://aistudio.baidu.com/aistudio/projectdetail/509014)
### 开发者教程
- [如何配置Server端的计算图?](doc/SERVER_DAG_CN.md)
- [如何开发一个新的General Op?](doc/NEW_OPERATOR_CN.md)
......
core/configure/proto/multi_lang_general_model_service.proto
浏览文件 @ f3953aa5
......@@ -37,6 +37,7 @@ message InferenceRequest {
repeated string feed_var_names = 2;
repeated string fetch_var_names = 3;
required bool is_python = 4 [ default = false ];
required uint64 log_id = 5 [ default = 0 ];
};
message InferenceResponse {
......
core/general-client/include/general_model.h
浏览文件 @ f3953aa5
......@@ -227,7 +227,8 @@ class PredictorClient {
const std::vector<std::vector<int>>& int_shape,
const std::vector<std::string>& fetch_name,
PredictorRes& predict_res_batch, // NOLINT
const int& pid);
const int& pid,
const uint64_t log_id);
int numpy_predict(
const std::vector<std::vector<py::array_t<float>>>& float_feed_batch,
......@@ -238,7 +239,8 @@ class PredictorClient {
const std::vector<std::vector<int>>& int_shape,
const std::vector<std::string>& fetch_name,
PredictorRes& predict_res_batch, // NOLINT
const int& pid);
const int& pid,
const uint64_t log_id);
private:
PredictorApi _api;
......
core/general-client/src/general_model.cpp
浏览文件 @ f3953aa5
......@@ -146,7 +146,8 @@ int PredictorClient::batch_predict(
const std::vector<std::vector<int>> &int_shape,
const std::vector<std::string> &fetch_name,
PredictorRes &predict_res_batch,
const int &pid) {
const int &pid,
const uint64_t log_id) {
int batch_size = std::max(float_feed_batch.size(), int_feed_batch.size());
predict_res_batch.clear();
......@@ -164,6 +165,7 @@ int PredictorClient::batch_predict(
VLOG(2) << "int feed name size: " << int_feed_name.size();
VLOG(2) << "max body size : " << brpc::fLU64::FLAGS_max_body_size;
Request req;
req.set_log_id(log_id);
for (auto &name : fetch_name) {
req.add_fetch_var_names(name);
}
......@@ -358,7 +360,8 @@ int PredictorClient::numpy_predict(
const std::vector<std::vector<int>> &int_shape,
const std::vector<std::string> &fetch_name,
PredictorRes &predict_res_batch,
const int &pid) {
const int &pid,
const uint64_t log_id) {
int batch_size = std::max(float_feed_batch.size(), int_feed_batch.size());
VLOG(2) << "batch size: " << batch_size;
predict_res_batch.clear();
......@@ -376,6 +379,7 @@ int PredictorClient::numpy_predict(
VLOG(2) << "int feed name size: " << int_feed_name.size();
VLOG(2) << "max body size : " << brpc::fLU64::FLAGS_max_body_size;
Request req;
req.set_log_id(log_id);
for (auto &name : fetch_name) {
req.add_fetch_var_names(name);
}
......
core/general-client/src/pybind_general_model.cpp
浏览文件 @ f3953aa5
......@@ -107,7 +107,8 @@ PYBIND11_MODULE(serving_client, m) {
const std::vector<std::vector<int>> &int_shape,
const std::vector<std::string> &fetch_name,
PredictorRes &predict_res_batch,
const int &pid) {
const int &pid,
const uint64_t log_id) {
return self.batch_predict(float_feed_batch,
float_feed_name,
float_shape,
......@@ -116,7 +117,8 @@ PYBIND11_MODULE(serving_client, m) {
int_shape,
fetch_name,
predict_res_batch,
pid);
pid,
log_id);
},
py::call_guard<py::gil_scoped_release>())
.def("numpy_predict",
......@@ -131,7 +133,8 @@ PYBIND11_MODULE(serving_client, m) {
const std::vector<std::vector<int>> &int_shape,
const std::vector<std::string> &fetch_name,
PredictorRes &predict_res_batch,
const int &pid) {
const int &pid,
const uint64_t log_id) {
return self.numpy_predict(float_feed_batch,
float_feed_name,
float_shape,
......@@ -140,7 +143,8 @@ PYBIND11_MODULE(serving_client, m) {
int_shape,
fetch_name,
predict_res_batch,
pid);
pid,
log_id);
},
py::call_guard<py::gil_scoped_release>());
}
......
core/general-server/op/general_copy_op.cpp
浏览文件 @ f3953aa5
......@@ -45,36 +45,41 @@ int GeneralCopyOp::inference() {
const std::string pre_name = pre_node_names[0];
const GeneralBlob *input_blob = get_depend_argument<GeneralBlob>(pre_name);
VLOG(2) << "precedent name: " << pre_name;
uint64_t log_id = input_blob->GetLogId();
VLOG(2) << "(logid=" << log_id << ") precedent name: " << pre_name;
const TensorVector *in = &input_blob->tensor_vector;
VLOG(2) << "input size: " << in->size();
VLOG(2) << "(logid=" << log_id << ") input size: " << in->size();
int batch_size = input_blob->GetBatchSize();
int input_var_num = 0;
GeneralBlob *res = mutable_data<GeneralBlob>();
res->SetLogId(log_id);
TensorVector *out = &res->tensor_vector;
VLOG(2) << "input batch size: " << batch_size;
VLOG(2) << "(logid=" << log_id << ") input batch size: " << batch_size;
res->SetBatchSize(batch_size);
if (!res) {
LOG(ERROR) << "Failed get op tls reader object output";
LOG(ERROR) << "(logid=" << log_id
<< ") Failed get op tls reader object output";
}
Timer timeline;
int64_t start = timeline.TimeStampUS();
VLOG(2) << "Going to init lod tensor";
VLOG(2) << "(logid=" << log_id << ") Going to init lod tensor";
for (int i = 0; i < in->size(); ++i) {
paddle::PaddleTensor lod_tensor;
CopyLod(&in->at(i), &lod_tensor);
lod_tensor.dtype = in->at(i).dtype;
lod_tensor.name = in->at(i).name;
VLOG(2) << "lod tensor [" << i << "].name = " << lod_tensor.name;
VLOG(2) << "(logid=" << log_id << ") lod tensor [" << i
<< "].name = " << lod_tensor.name;
out->push_back(lod_tensor);
}
VLOG(2) << "pack done.";
VLOG(2) << "(logid=" << log_id << ") pack done.";
for (int i = 0; i < out->size(); ++i) {
int64_t *src_ptr = static_cast<int64_t *>(in->at(i).data.data());
......@@ -86,7 +91,7 @@ int GeneralCopyOp::inference() {
}
}
VLOG(2) << "output done.";
VLOG(2) << "(logid=" << log_id << ") output done.";
timeline.Pause();
int64_t end = timeline.TimeStampUS();
......@@ -94,7 +99,7 @@ int GeneralCopyOp::inference() {
AddBlobInfo(res, start);
AddBlobInfo(res, end);
VLOG(2) << "read data from client success";
VLOG(2) << "(logid=" << log_id << ") read data from client success";
return 0;
}
......
core/general-server/op/general_dist_kv_infer_op.cpp
浏览文件 @ f3953aa5
......@@ -50,18 +50,20 @@ int GeneralDistKVInferOp::inference() {
const std::string pre_name = pre_node_names[0];
const GeneralBlob *input_blob = get_depend_argument<GeneralBlob>(pre_name);
VLOG(2) << "Get precedent op name: " << pre_name;
uint64_t log_id = input_blob->GetLogId();
VLOG(2) << "(logid=" << log_id << ") Get precedent op name: " << pre_name;
GeneralBlob *output_blob = mutable_data<GeneralBlob>();
if (!input_blob) {
LOG(ERROR) << "Failed mutable depended argument, op:" << pre_name;
LOG(ERROR) << "(logid=" << log_id
<< ") Failed mutable depended argument, op:" << pre_name;
return -1;
}
const TensorVector *in = &input_blob->tensor_vector;
TensorVector *out = &output_blob->tensor_vector;
int batch_size = input_blob->GetBatchSize();
VLOG(2) << "input batch size: " << batch_size;
VLOG(2) << "(logid=" << log_id << ") input batch size: " << batch_size;
std::vector<uint64_t> keys;
std::vector<rec::mcube::CubeValue> values;
int sparse_count = 0;
......@@ -96,13 +98,14 @@ int GeneralDistKVInferOp::inference() {
rec::mcube::CubeAPI *cube = rec::mcube::CubeAPI::instance();
std::vector<std::string> table_names = cube->get_table_names();
if (table_names.size() == 0) {
LOG(ERROR) << "cube init error or cube config not given.";
LOG(ERROR) << "(logid=" << log_id
<< ") cube init error or cube config not given.";
return -1;
}
int ret = cube->seek(table_names[0], keys, &values);
int64_t cube_end = timeline.TimeStampUS();
if (values.size() != keys.size() || values[0].buff.size() == 0) {
LOG(ERROR) << "cube value return null";
LOG(ERROR) << "(logid=" << log_id << ") cube value return null";
}
size_t EMBEDDING_SIZE = values[0].buff.size() / sizeof(float);
TensorVector sparse_out;
......@@ -153,14 +156,16 @@ int GeneralDistKVInferOp::inference() {
infer_in.insert(infer_in.end(), sparse_out.begin(), sparse_out.end());
output_blob->SetBatchSize(batch_size);
output_blob->SetLogId(log_id);
VLOG(2) << "infer batch size: " << batch_size;
VLOG(2) << "(logid=" << log_id << ") infer batch size: " << batch_size;
int64_t start = timeline.TimeStampUS();
if (InferManager::instance().infer(
engine_name().c_str(), &infer_in, out, batch_size)) {
LOG(ERROR) << "Failed do infer in fluid model: " << engine_name();
LOG(ERROR) << "(logid=" << log_id
<< ") Failed do infer in fluid model: " << engine_name();
return -1;
}
......
core/general-server/op/general_dist_kv_quant_infer_op.cpp
浏览文件 @ f3953aa5
......@@ -59,10 +59,13 @@ int GeneralDistKVQuantInferOp::inference() {
return -1;
}
uint64_t log_id = input_blob->GetLogId();
output_blob->SetLogId(log_id);
const TensorVector *in = &input_blob->tensor_vector;
TensorVector *out = &output_blob->tensor_vector;
int batch_size = input_blob->GetBatchSize();
VLOG(2) << "input batch size: " << batch_size;
VLOG(2) << "(logid=" << log_id << ") input batch size: " << batch_size;
std::vector<uint64_t> keys;
std::vector<rec::mcube::CubeValue> values;
int sparse_count = 0;
......@@ -94,13 +97,14 @@ int GeneralDistKVQuantInferOp::inference() {
rec::mcube::CubeAPI *cube = rec::mcube::CubeAPI::instance();
std::vector<std::string> table_names = cube->get_table_names();
if (table_names.size() == 0) {
LOG(ERROR) << "cube init error or cube config not given.";
LOG(ERROR) << "(logid=" << log_id
<< ") cube init error or cube config not given.";
return -1;
}
int ret = cube->seek(table_names[0], keys, &values);
if (values.size() != keys.size() || values[0].buff.size() == 0) {
LOG(ERROR) << "cube value return null";
LOG(ERROR) << "(logid=" << log_id << ") cube value return null";
}
TensorVector sparse_out;
......@@ -182,7 +186,7 @@ int GeneralDistKVQuantInferOp::inference() {
output_blob->SetBatchSize(batch_size);
VLOG(2) << "infer batch size: " << batch_size;
VLOG(2) << "(logid=" << log_id << ") infer batch size: " << batch_size;
Timer timeline;
int64_t start = timeline.TimeStampUS();
......@@ -190,7 +194,8 @@ int GeneralDistKVQuantInferOp::inference() {
if (InferManager::instance().infer(
engine_name().c_str(), &infer_in, out, batch_size)) {
LOG(ERROR) << "Failed do infer in fluid model: " << engine_name();
LOG(ERROR) << "(logid=" << log_id
<< ") Failed do infer in fluid model: " << engine_name();
return -1;
}
......
core/general-server/op/general_infer_helper.h
浏览文件 @ f3953aa5
......@@ -27,6 +27,7 @@ struct GeneralBlob {
std::vector<paddle::PaddleTensor> tensor_vector;
int64_t time_stamp[20];
int p_size = 0;
uint64_t _log_id = -1; // for logging
int _batch_size;
......@@ -38,9 +39,11 @@ struct GeneralBlob {
tensor_vector.clear();
}
int SetBatchSize(int batch_size) { _batch_size = batch_size; }
void SetBatchSize(int batch_size) { _batch_size = batch_size; }
void SetLogId(uint64_t log_id) { _log_id = log_id; }
int GetBatchSize() const { return _batch_size; }
uint64_t GetLogId() const { return _log_id; }
std::string ShortDebugString() const { return "Not implemented!"; }
};
......
core/general-server/op/general_infer_op.cpp
浏览文件 @ f3953aa5
......@@ -47,22 +47,26 @@ int GeneralInferOp::inference() {
const std::string pre_name = pre_node_names[0];
const GeneralBlob *input_blob = get_depend_argument<GeneralBlob>(pre_name);
VLOG(2) << "Get precedent op name: " << pre_name;
uint64_t log_id = input_blob->GetLogId();
VLOG(2) << "(logid=" << log_id << ") Get precedent op name: " << pre_name;
GeneralBlob *output_blob = mutable_data<GeneralBlob>();
output_blob->SetLogId(log_id);
if (!input_blob) {
LOG(ERROR) << "Failed mutable depended argument, op:" << pre_name;
LOG(ERROR) << "(logid=" << log_id
<< ") Failed mutable depended argument, op:" << pre_name;
return -1;
}
const TensorVector *in = &input_blob->tensor_vector;
TensorVector *out = &output_blob->tensor_vector;
int batch_size = input_blob->_batch_size;
VLOG(2) << "input batch size: " << batch_size;
int batch_size = input_blob->_batch_size();
VLOG(2) << "(logid=" << log_id << ") input batch size: " << batch_size;
output_blob->_batch_size = batch_size;
VLOG(2) << "infer batch size: " << batch_size;
VLOG(2) << "(logid=" << log_id << ") infer batch size: " << batch_size;
Timer timeline;
int64_t start = timeline.TimeStampUS();
......@@ -70,7 +74,8 @@ int GeneralInferOp::inference() {
if (InferManager::instance().infer(
engine_name().c_str(), in, out, batch_size)) {
LOG(ERROR) << "Failed do infer in fluid model: " << engine_name().c_str();
LOG(ERROR) << "(logid=" << log_id
<< ") Failed do infer in fluid model: " << engine_name().c_str();
return -1;
}
......
core/general-server/op/general_reader_op.cpp
浏览文件 @ f3953aa5
......@@ -37,9 +37,9 @@ int conf_check(const Request *req,
const std::shared_ptr<PaddleGeneralModelConfig> &model_config) {
int var_num = req->insts(0).tensor_array_size();
if (var_num != model_config->_feed_type.size()) {
VLOG(2) << "var num: " << var_num;
VLOG(2) << "model config var num: " << model_config->_feed_type.size();
LOG(ERROR) << "feed var number not match.";
LOG(ERROR) << "feed var number not match: model config["
<< model_config->_feed_type.size() << "] vs. actual[" << var_num
<< "]";
return -1;
}
......@@ -72,9 +72,8 @@ int conf_check(const Request *req,
int GeneralReaderOp::inference() {
// reade request from client
const Request *req = dynamic_cast<const Request *>(get_request_message());
VLOG(2) << "start to call load general model_conf op";
baidu::paddle_serving::predictor::Resource &resource =
baidu::paddle_serving::predictor::Resource::instance();
uint64_t log_id = req->log_id();
int batch_size = req->insts_size();
int input_var_num = 0;
std::vector<int64_t> elem_type;
......@@ -84,20 +83,30 @@ int GeneralReaderOp::inference() {
GeneralBlob *res = mutable_data<GeneralBlob>();
TensorVector *out = &res->tensor_vector;
res->SetBatchSize(batch_size);
res->SetLogId(log_id);
if (!res) {
LOG(ERROR) << "Failed get op tls reader object output";
LOG(ERROR) << "(logid=" << log_id
<< ") Failed get op tls reader object output";
}
Timer timeline;
int64_t start = timeline.TimeStampUS();
int var_num = req->insts(0).tensor_array_size();
VLOG(2) << "var num: " << var_num;
VLOG(2) << "(logid=" << log_id << ") var num: " << var_num;
VLOG(2) << "(logid=" << log_id
<< ") start to call load general model_conf op";
baidu::paddle_serving::predictor::Resource &resource =
baidu::paddle_serving::predictor::Resource::instance();
VLOG(2) << "(logid=" << log_id << ") get resource pointer done.";
VLOG(2) << "get resource pointer done.";
std::shared_ptr<PaddleGeneralModelConfig> model_config =
resource.get_general_model_config();
VLOG(2) << "print general model config done.";
VLOG(2) << "(logid=" << log_id << ") print general model config done.";
// TODO(guru4elephant): how to do conditional check?
/*
......@@ -118,7 +127,8 @@ int GeneralReaderOp::inference() {
for (int i = 0; i < var_num; ++i) {
paddle::PaddleTensor lod_tensor;
elem_type[i] = req->insts(0).tensor_array(i).elem_type();
VLOG(2) << "var[" << i << "] has elem type: " << elem_type[i];
VLOG(2) << "(logid=" << log_id << ") var[" << i
<< "] has elem type: " << elem_type[i];
if (elem_type[i] == 0) { // int64
elem_size[i] = sizeof(int64_t);
lod_tensor.dtype = paddle::PaddleDType::INT64;
......@@ -133,17 +143,19 @@ int GeneralReaderOp::inference() {
if (model_config->_is_lod_feed[i]) {
lod_tensor.lod.resize(1);
lod_tensor.lod[0].push_back(0);
VLOG(2) << "var[" << i << "] is lod_tensor";
VLOG(2) << "(logid=" << log_id << ") var[" << i << "] is lod_tensor";
} else {
lod_tensor.shape.push_back(batch_size);
capacity[i] = 1;
for (int k = 0; k < req->insts(0).tensor_array(i).shape_size(); ++k) {
int dim = req->insts(0).tensor_array(i).shape(k);
VLOG(2) << "shape for var[" << i << "]: " << dim;
VLOG(2) << "(logid=" << log_id << ") shape for var[" << i
<< "]: " << dim;
capacity[i] *= dim;
lod_tensor.shape.push_back(dim);
}
VLOG(2) << "var[" << i << "] is tensor, capacity: " << capacity[i];
VLOG(2) << "(logid=" << log_id << ") var[" << i
<< "] is tensor, capacity: " << capacity[i];
}
lod_tensor.name = model_config->_feed_name[i];
out->push_back(lod_tensor);
......@@ -163,11 +175,12 @@ int GeneralReaderOp::inference() {
} else if (tensor.int_data_size() > 0) {
data_len = tensor.int_data_size();
}
VLOG(2) << "tensor size for var[" << i << "]: " << data_len;
VLOG(2) << "(logid=" << log_id << ") tensor size for var[" << i
<< "]: " << data_len;
tensor_size += data_len;
int cur_len = out->at(i).lod[0].back();
VLOG(2) << "current len: " << cur_len;
VLOG(2) << "(logid=" << log_id << ") current len: " << cur_len;
int sample_len = 0;
if (tensor.shape_size() == 1) {
......@@ -176,7 +189,7 @@ int GeneralReaderOp::inference() {
sample_len = tensor.shape(0);
}
out->at(i).lod[0].push_back(cur_len + sample_len);
VLOG(2) << "new len: " << cur_len + sample_len;
VLOG(2) << "(logid=" << log_id << ") new len: " << cur_len + sample_len;
}
out->at(i).data.Resize(tensor_size * elem_size[i]);
out->at(i).shape = {out->at(i).lod[0].back()};
......@@ -186,11 +199,11 @@ int GeneralReaderOp::inference() {
if (out->at(i).shape.size() == 1) {
out->at(i).shape.push_back(1);
}
VLOG(2) << "var[" << i
VLOG(2) << "(logid=" << log_id << ") var[" << i
<< "] is lod_tensor and len=" << out->at(i).lod[0].back();
} else {
out->at(i).data.Resize(batch_size * capacity[i] * elem_size[i]);
VLOG(2) << "var[" << i
VLOG(2) << "(logid=" << log_id << ") var[" << i
<< "] is tensor and capacity=" << batch_size * capacity[i];
}
}
......@@ -199,8 +212,8 @@ int GeneralReaderOp::inference() {
for (int i = 0; i < var_num; ++i) {
if (elem_type[i] == 0) {
int64_t *dst_ptr = static_cast<int64_t *>(out->at(i).data.data());
VLOG(2) << "first element data in var[" << i << "] is "
<< req->insts(0).tensor_array(i).int64_data(0);
VLOG(2) << "(logid=" << log_id << ") first element data in var[" << i
<< "] is " << req->insts(0).tensor_array(i).int64_data(0);
int offset = 0;
for (int j = 0; j < batch_size; ++j) {
int elem_num = req->insts(j).tensor_array(i).int64_data_size();
......@@ -215,8 +228,8 @@ int GeneralReaderOp::inference() {
}
} else if (elem_type[i] == 1) {
float *dst_ptr = static_cast<float *>(out->at(i).data.data());
VLOG(2) << "first element data in var[" << i << "] is "
<< req->insts(0).tensor_array(i).float_data(0);
VLOG(2) << "(logid=" << log_id << ") first element data in var[" << i
<< "] is " << req->insts(0).tensor_array(i).float_data(0);
int offset = 0;
for (int j = 0; j < batch_size; ++j) {
int elem_num = req->insts(j).tensor_array(i).float_data_size();
......@@ -231,8 +244,8 @@ int GeneralReaderOp::inference() {
}
} else if (elem_type[i] == 2) {
int32_t *dst_ptr = static_cast<int32_t *>(out->at(i).data.data());
VLOG(2) << "first element data in var[" << i << "] is "
<< req->insts(0).tensor_array(i).int_data(0);
VLOG(2) << "(logid=" << log_id << ") first element data in var[" << i
<< "] is " << req->insts(0).tensor_array(i).int_data(0);
int offset = 0;
for (int j = 0; j < batch_size; ++j) {
int elem_num = req->insts(j).tensor_array(i).int_data_size();
......@@ -248,7 +261,7 @@ int GeneralReaderOp::inference() {
}
}
VLOG(2) << "output size: " << out->size();
VLOG(2) << "(logid=" << log_id << ") output size: " << out->size();
timeline.Pause();
int64_t end = timeline.TimeStampUS();
......@@ -257,7 +270,7 @@ int GeneralReaderOp::inference() {
AddBlobInfo(res, start);
AddBlobInfo(res, end);
VLOG(2) << "read data from client success";
VLOG(2) << "(logid=" << log_id << ") read data from client success";
return 0;
}
DEFINE_OP(GeneralReaderOp);
......
core/general-server/op/general_response_op.cpp
浏览文件 @ f3953aa5
......@@ -42,6 +42,9 @@ using baidu::paddle_serving::predictor::PaddleGeneralModelConfig;
int GeneralResponseOp::inference() {
const std::vector<std::string> pre_node_names = pre_names();
VLOG(2) << "pre node names size: " << pre_node_names.size();
const GeneralBlob *input_blob;
uint64_t log_id =
get_depend_argument<GeneralBlob>(pre_node_names[0])->GetLogId();
const Request *req = dynamic_cast<const Request *>(get_request_message());
// response inst with only fetch_var_names
......@@ -52,15 +55,17 @@ int GeneralResponseOp::inference() {
// timeline.Start();
int64_t start = timeline.TimeStampUS();
VLOG(2) << "start to call load general model_conf op";
VLOG(2) << "(logid=" << log_id
<< ") start to call load general model_conf op";
baidu::paddle_serving::predictor::Resource &resource =
baidu::paddle_serving::predictor::Resource::instance();
VLOG(2) << "get resource pointer done.";
VLOG(2) << "(logid=" << log_id << ") get resource pointer done.";
std::shared_ptr<PaddleGeneralModelConfig> model_config =
resource.get_general_model_config();
VLOG(2) << "max body size : " << brpc::fLU64::FLAGS_max_body_size;
VLOG(2) << "(logid=" << log_id
<< ") max body size : " << brpc::fLU64::FLAGS_max_body_size;
std::vector<int> fetch_index;
fetch_index.resize(req->fetch_var_names_size());
......@@ -69,16 +74,16 @@ int GeneralResponseOp::inference() {
model_config->_fetch_alias_name_to_index[req->fetch_var_names(i)];
}
const GeneralBlob *input_blob;
for (uint32_t pi = 0; pi < pre_node_names.size(); ++pi) {
const std::string &pre_name = pre_node_names[pi];
VLOG(2) << "pre names[" << pi << "]: " << pre_name << " ("
<< pre_node_names.size() << ")";
VLOG(2) << "(logid=" << log_id << ") pre names[" << pi << "]: " << pre_name
<< " (" << pre_node_names.size() << ")";
input_blob = get_depend_argument<GeneralBlob>(pre_name);
// fprintf(stderr, "input(%s) blob address %x\n", pre_names.c_str(),
// input_blob);
if (!input_blob) {
LOG(ERROR) << "Failed mutable depended argument, op: " << pre_name;
LOG(ERROR) << "(logid=" << log_id
<< ") Failed mutable depended argument, op: " << pre_name;
return -1;
}
......@@ -92,17 +97,19 @@ int GeneralResponseOp::inference() {
for (auto &idx : fetch_index) {
Tensor *tensor = fetch_inst->add_tensor_array();
if (model_config->_is_lod_fetch[idx]) {
VLOG(2) << "out[" << idx << "] " << model_config->_fetch_name[idx]
<< " is lod_tensor";
VLOG(2) << "(logid=" << log_id << ") out[" << idx << "] "
<< model_config->_fetch_name[idx] << " is lod_tensor";
for (int k = 0; k < in->at(idx).shape.size(); ++k) {
VLOG(2) << "shape[" << k << "]: " << in->at(idx).shape[k];
VLOG(2) << "(logid=" << log_id << ") shape[" << k
<< "]: " << in->at(idx).shape[k];
tensor->add_shape(in->at(idx).shape[k]);
}
} else {
VLOG(2) << "out[" << idx << "] " << model_config->_fetch_name[idx]
<< " is tensor";
VLOG(2) << "(logid=" << log_id << ") out[" << idx << "] "
<< model_config->_fetch_name[idx] << " is tensor";
for (int k = 0; k < in->at(idx).shape.size(); ++k) {
VLOG(2) << "shape[" << k << "]: " << in->at(idx).shape[k];
VLOG(2) << "(logid=" << log_id << ") shape[" << k
<< "]: " << in->at(idx).shape[k];
tensor->add_shape(in->at(idx).shape[k]);
}
}
......@@ -119,8 +126,8 @@ int GeneralResponseOp::inference() {
auto dtype = in->at(idx).dtype;
if (dtype == paddle::PaddleDType::INT64) {
VLOG(2) << "Prepare int64 var [" << model_config->_fetch_name[idx]
<< "].";
VLOG(2) << "(logid=" << log_id << ") Prepare int64 var ["
<< model_config->_fetch_name[idx] << "].";
int64_t *data_ptr = static_cast<int64_t *>(in->at(idx).data.data());
// from
// https://stackoverflow.com/questions/15499641/copy-a-stdvector-to-a-repeated-field-from-protobuf-with-memcpy
......@@ -130,16 +137,16 @@ int GeneralResponseOp::inference() {
fetch_p->mutable_tensor_array(var_idx)->mutable_int64_data()->Swap(
&tmp_data);
} else if (dtype == paddle::PaddleDType::FLOAT32) {
VLOG(2) << "Prepare float var [" << model_config->_fetch_name[idx]
<< "].";
VLOG(2) << "(logid=" << log_id << ") Prepare float var ["
<< model_config->_fetch_name[idx] << "].";
float *data_ptr = static_cast<float *>(in->at(idx).data.data());
google::protobuf::RepeatedField<float> tmp_data(data_ptr,
data_ptr + cap);
fetch_p->mutable_tensor_array(var_idx)->mutable_float_data()->Swap(
&tmp_data);
} else if (dtype == paddle::PaddleDType::INT32) {
VLOG(2) << "Prepare int32 var [" << model_config->_fetch_name[idx]
<< "].";
VLOG(2) << "(logid=" << log_id << ")Prepare int32 var ["
<< model_config->_fetch_name[idx] << "].";
int32_t *data_ptr = static_cast<int32_t *>(in->at(idx).data.data());
google::protobuf::RepeatedField<int32_t> tmp_data(data_ptr,
data_ptr + cap);
......@@ -154,7 +161,8 @@ int GeneralResponseOp::inference() {
}
}
VLOG(2) << "fetch var [" << model_config->_fetch_name[idx] << "] ready";
VLOG(2) << "(logid=" << log_id << ") fetch var ["
<< model_config->_fetch_name[idx] << "] ready";
var_idx++;
}
}
......@@ -167,7 +175,8 @@ int GeneralResponseOp::inference() {
// a more elegant way.
for (uint32_t pi = 0; pi < pre_node_names.size(); ++pi) {
input_blob = get_depend_argument<GeneralBlob>(pre_node_names[pi]);
VLOG(2) << "p size for input blob: " << input_blob->p_size;
VLOG(2) << "(logid=" << log_id
<< ") p size for input blob: " << input_blob->p_size;
int profile_time_idx = -1;
if (pi == 0) {
profile_time_idx = 0;
......
core/general-server/op/general_text_reader_op.cpp
浏览文件 @ f3953aa5
......@@ -35,6 +35,7 @@ using baidu::paddle_serving::predictor::PaddleGeneralModelConfig;
int GeneralTextReaderOp::inference() {
// reade request from client
const Request *req = dynamic_cast<const Request *>(get_request_message());
uint64_t log_id = req->log_id();
int batch_size = req->insts_size();
int input_var_num = 0;
......@@ -44,16 +45,18 @@ int GeneralTextReaderOp::inference() {
std::vector<int64_t> capacity;
GeneralBlob *res = mutable_data<GeneralBlob>();
TensorVector *out = &res->tensor_vector;
res->SetBatchSize(batch_size);
if (!res) {
LOG(ERROR) << "Failed get op tls reader object output";
LOG(ERROR) << "(logid=" << log_id
<< ") Failed get op tls reader object output";
}
TensorVector *out = &res->tensor_vector;
res->SetBatchSize(batch_size);
res->SetLogId(log_id);
if (batch_size <= 0) {
LOG(ERROR) << "Batch size < 0";
LOG(ERROR) << "(logid=" << log_id << ") Batch size < 0";
return -1;
}
......@@ -61,17 +64,18 @@ int GeneralTextReaderOp::inference() {
int64_t start = timeline.TimeStampUS();
int var_num = req->insts(0).tensor_array_size();
VLOG(2) << "var num: " << var_num;
VLOG(2) << "(logid=" << log_id << ") var num: " << var_num;
VLOG(2) << "start to call load general model_conf op";
VLOG(2) << "(logid=" << log_id
<< ") start to call load general model_conf op";
baidu::paddle_serving::predictor::Resource &resource =
baidu::paddle_serving::predictor::Resource::instance();
VLOG(2) << "get resource pointer done.";
VLOG(2) << "(logid=" << log_id << ") get resource pointer done.";
std::shared_ptr<PaddleGeneralModelConfig> model_config =
resource.get_general_model_config();
VLOG(2) << "print general model config done.";
VLOG(2) << "(logid=" << log_id << ") print general model config done.";
elem_type.resize(var_num);
elem_size.resize(var_num);
......@@ -79,7 +83,8 @@ int GeneralTextReaderOp::inference() {
for (int i = 0; i < var_num; ++i) {
paddle::PaddleTensor lod_tensor;
elem_type[i] = req->insts(0).tensor_array(i).elem_type();
VLOG(2) << "var[" << i << "] has elem type: " << elem_type[i];
VLOG(2) << "(logid=" << log_id << ") var[" << i
<< "] has elem type: " << elem_type[i];
if (elem_type[i] == 0) { // int64
elem_size[i] = sizeof(int64_t);
lod_tensor.dtype = paddle::PaddleDType::INT64;
......@@ -91,17 +96,19 @@ int GeneralTextReaderOp::inference() {
if (req->insts(0).tensor_array(i).shape(0) == -1) {
lod_tensor.lod.resize(1);
lod_tensor.lod[0].push_back(0);
VLOG(2) << "var[" << i << "] is lod_tensor";
VLOG(2) << "(logid=" << log_id << ") var[" << i << "] is lod_tensor";
} else {
lod_tensor.shape.push_back(batch_size);
capacity[i] = 1;
for (int k = 0; k < req->insts(0).tensor_array(i).shape_size(); ++k) {
int dim = req->insts(0).tensor_array(i).shape(k);
VLOG(2) << "shape for var[" << i << "]: " << dim;
VLOG(2) << "(logid=" << log_id << ") shape for var[" << i
<< "]: " << dim;
capacity[i] *= dim;
lod_tensor.shape.push_back(dim);
}
VLOG(2) << "var[" << i << "] is tensor, capacity: " << capacity[i];
VLOG(2) << "(logid=" << log_id << ") var[" << i
<< "] is tensor, capacity: " << capacity[i];
}
lod_tensor.name = model_config->_feed_name[i];
out->push_back(lod_tensor);
......@@ -117,11 +124,11 @@ int GeneralTextReaderOp::inference() {
}
out->at(i).data.Resize(out->at(i).lod[0].back() * elem_size[i]);
out->at(i).shape = {out->at(i).lod[0].back(), 1};
VLOG(2) << "var[" << i
VLOG(2) << "(logid=" << log_id << ") var[" << i
<< "] is lod_tensor and len=" << out->at(i).lod[0].back();
} else {
out->at(i).data.Resize(batch_size * capacity[i] * elem_size[i]);
VLOG(2) << "var[" << i
VLOG(2) << "(logid=" << log_id << ") var[" << i
<< "] is tensor and capacity=" << batch_size * capacity[i];
}
}
......@@ -163,7 +170,7 @@ int GeneralTextReaderOp::inference() {
AddBlobInfo(res, start);
AddBlobInfo(res, end);
VLOG(2) << "read data from client success";
VLOG(2) << "(logid=" << log_id << ") read data from client success";
return 0;
}
DEFINE_OP(GeneralTextReaderOp);
......
core/general-server/op/general_text_response_op.cpp
浏览文件 @ f3953aa5
......@@ -40,6 +40,9 @@ int GeneralTextResponseOp::inference() {
VLOG(2) << "Going to run inference";
const std::vector<std::string> pre_node_names = pre_names();
VLOG(2) << "pre node names size: " << pre_node_names.size();
const GeneralBlob *input_blob;
uint64_t log_id =
get_depend_argument<GeneralBlob>(pre_node_names[0])->GetLogId();
const Request *req = dynamic_cast<const Request *>(get_request_message());
// response inst with only fetch_var_names
......@@ -48,11 +51,12 @@ int GeneralTextResponseOp::inference() {
Timer timeline;
int64_t start = timeline.TimeStampUS();
VLOG(2) << "start to call load general model_conf op";
VLOG(2) << "(logid=" << log_id
<< ") start to call load general model_conf op";
baidu::paddle_serving::predictor::Resource &resource =
baidu::paddle_serving::predictor::Resource::instance();
VLOG(2) << "get resource pointer done.";
VLOG(2) << "(logid=" << log_id << ") get resource pointer done.";
std::shared_ptr<PaddleGeneralModelConfig> model_config =
resource.get_general_model_config();
......@@ -63,20 +67,20 @@ int GeneralTextResponseOp::inference() {
model_config->_fetch_alias_name_to_index[req->fetch_var_names(i)];
}
const GeneralBlob *input_blob;
for (uint32_t pi = 0; pi < pre_node_names.size(); ++pi) {
const std::string &pre_name = pre_node_names[pi];
VLOG(2) << "pre names[" << pi << "]: " << pre_name << " ("
<< pre_node_names.size() << ")";
VLOG(2) << "(logid=" << log_id << ") pre names[" << pi << "]: " << pre_name
<< " (" << pre_node_names.size() << ")";
input_blob = get_depend_argument<GeneralBlob>(pre_name);
if (!input_blob) {
LOG(ERROR) << "Failed mutable depended argument, op: " << pre_name;
LOG(ERROR) << "(logid=" << log_id
<< ") Failed mutable depended argument, op: " << pre_name;
return -1;
}
const TensorVector *in = &input_blob->tensor_vector;
int batch_size = input_blob->GetBatchSize();
VLOG(2) << "input batch size: " << batch_size;
VLOG(2) << "(logid=" << log_id << ") input batch size: " << batch_size;
ModelOutput *output = res->add_outputs();
output->set_engine_name(
......@@ -88,12 +92,13 @@ int GeneralTextResponseOp::inference() {
// currently only response float tensor or lod_tensor
tensor->set_elem_type(1);
if (model_config->_is_lod_fetch[idx]) {
VLOG(2) << "out[" << idx << " is lod_tensor";
VLOG(2) << "(logid=" << log_id << ") out[" << idx << " is lod_tensor";
tensor->add_shape(-1);
} else {
VLOG(2) << "out[" << idx << "] is tensor";
VLOG(2) << "(logid=" << log_id << ") out[" << idx << "] is tensor";
for (int k = 1; k < in->at(idx).shape.size(); ++k) {
VLOG(2) << "shape[" << k - 1 << "]: " << in->at(idx).shape[k];
VLOG(2) << "(logid=" << log_id << ") shape[" << k - 1
<< "]: " << in->at(idx).shape[k];
tensor->add_shape(in->at(idx).shape[k]);
}
}
......@@ -137,7 +142,8 @@ int GeneralTextResponseOp::inference() {
// a more elegant way.
for (uint32_t pi = 0; pi < pre_node_names.size(); ++pi) {
input_blob = get_depend_argument<GeneralBlob>(pre_node_names[pi]);
VLOG(2) << "p size for input blob: " << input_blob->p_size;
VLOG(2) << "(logid=" << log_id
<< ") p size for input blob: " << input_blob->p_size;
int profile_time_idx = -1;
if (pi == 0) {
profile_time_idx = 0;
......
core/general-server/proto/general_model_service.proto
浏览文件 @ f3953aa5
......@@ -37,6 +37,7 @@ message Request {
repeated FeedInst insts = 1;
repeated string fetch_var_names = 2;
optional bool profile_server = 3 [ default = false ];
required uint64 log_id = 4 [ default = 0 ];
};
message Response {
......
core/general-server/proto/load_general_model_service.proto
浏览文件 @ f3953aa5
......@@ -21,6 +21,7 @@ option cc_generic_services = true;
message RequestAndResponse {
required int32 a = 1;
required float b = 2;
required uint64 log_id = 3 [ default = 0 ];
};
service LoadGeneralModelService {
......
core/pdcodegen/src/pdcodegen.cpp
浏览文件 @ f3953aa5
......@@ -280,25 +280,29 @@ class PdsCodeGenerator : public CodeGenerator {
" baidu::rpc::ClosureGuard done_guard(done);\n"
" baidu::rpc::Controller* cntl = \n"
" static_cast<baidu::rpc::Controller*>(cntl_base);\n"
" uint64_t log_id = request->log_id();\n"
" cntl->set_log_id(log_id);\n"
" ::baidu::paddle_serving::predictor::InferService* svr = \n"
" "
"::baidu::paddle_serving::predictor::InferServiceManager::instance("
").item(\"$service$\");\n"
" if (svr == NULL) {\n"
" LOG(ERROR) << \"Not found service: $service$\";\n"
" LOG(ERROR) << \"(logid=\" << log_id << \") Not found service: "
"$service$\";\n"
" cntl->SetFailed(404, \"Not found service: $service$\");\n"
" return ;\n"
" }\n"
" LOG(INFO) << \" remote_side=\[\" << cntl->remote_side() << " // NOLINT
"\"\]\";\n"
" LOG(INFO) << \" local_side=\[\" << cntl->local_side() << " // NOLINT
"\"\]\";\n"
" LOG(INFO) << \" service_name=\[\" << \"$name$\" << \"\]\";\n" // NOLINT
" LOG(INFO) << \" log_id=\[\" << cntl->log_id() << \"\]\";\n" // NOLINT
" int err_code = svr->inference(request, response);\n"
" LOG(INFO) << \"(logid=\" << log_id << \") remote_side=\[\" " // NOLINT
"<< cntl->remote_side() << \"\]\";\n"
" LOG(INFO) << \"(logid=\" << log_id << \") local_side=\[\" " // NOLINT
"<< cntl->local_side() << \"\]\";\n"
" LOG(INFO) << \"(logid=\" << log_id << \") service_name=\[\" " // NOLINT
"<< \"$name$\" << \"\]\";\n"
" int err_code = svr->inference(request, response, log_id);\n"
" if (err_code != 0) {\n"
" LOG(WARNING)\n"
" << \"Failed call inferservice[$name$], name[$service$]\"\n"
" << \"(logid=\" << log_id << \") Failed call "
"inferservice[$name$], name[$service$]\"\n"
" << \", error_code: \" << err_code;\n"
" cntl->SetFailed(err_code, \"InferService inference "
"failed!\");\n"
......@@ -306,7 +310,8 @@ class PdsCodeGenerator : public CodeGenerator {
" gettimeofday(&tv, NULL);\n"
" long end = tv.tv_sec * 1000000 + tv.tv_usec;\n"
" // flush notice log\n"
" LOG(INFO) << \" tc=\[\" << (end - start) << \"\]\";\n", // NOLINT
" LOG(INFO) << \"(logid=\" << log_id << \") tc=\[\" << (end - " // NOLINT
"start) << \"\]\";\n", // NOLINT
"name",
class_name,
"service",
......@@ -317,26 +322,31 @@ class PdsCodeGenerator : public CodeGenerator {
" baidu::rpc::ClosureGuard done_guard(done);\n"
" baidu::rpc::Controller* cntl = \n"
" static_cast<baidu::rpc::Controller*>(cntl_base);\n"
" uint64_t log_id = equest->log_id();\n"
" cntl->set_log_id(log_id);\n"
" ::baidu::paddle_serving::predictor::InferService* svr = \n"
" "
"::baidu::paddle_serving::predictor::InferServiceManager::instance("
").item(\"$service$\");\n"
" if (svr == NULL) {\n"
" LOG(ERROR) << \"Not found service: $service$\";\n"
" LOG(ERROR) << \"(logid=\" << log_id << \") Not found service: "
"$service$\";\n"
" cntl->SetFailed(404, \"Not found service: $service$\");\n"
" return ;\n"
" }\n"
" LOG(INFO) << \" remote_side=\[\" << cntl->remote_side() << " // NOLINT
"\"\]\";\n"
" LOG(INFO) << \" local_side=\[\" << cntl->local_side() << " // NOLINT
"\"\]\";\n"
" LOG(INFO) << \" service_name=\[\" << \"$name$\" << \"\]\";\n" // NOLINT
" LOG(INFO) << \" log_id=\[\" << cntl->log_id() << \"\]\";\n" // NOLINT
" LOG(INFO) << \"(logid=\" << log_id << \") remote_side=\[\" " // NOLINT
"<< cntl->remote_side() << \"\]\";\n"
" LOG(INFO) << \"(logid=\" << log_id << \") local_side=\[\" " // NOLINT
"<< cntl->local_side() << \"\]\";\n"
" LOG(INFO) << \"(logid=\" << log_id << \") service_name=\[\" " // NOLINT
"<< \"$name$\" << \"\]\";\n"
" butil::IOBufBuilder debug_os;\n"
" int err_code = svr->inference(request, response, &debug_os);\n"
" int err_code = svr->inference(request, response, log_id, "
"&debug_os);\n"
" if (err_code != 0) {\n"
" LOG(WARNING)\n"
" << \"Failed call inferservice[$name$], name[$service$]\"\n"
" << \"(logid=\" << log_id << \") Failed call "
"inferservice[$name$], name[$service$]\"\n"
" << \", error_code: \" << err_code;\n"
" cntl->SetFailed(err_code, \"InferService inference "
"failed!\");\n"
......@@ -345,9 +355,11 @@ class PdsCodeGenerator : public CodeGenerator {
" gettimeofday(&tv, NULL);\n"
" long end = tv.tv_sec * 1000000 + tv.tv_usec;\n"
" // flush notice log\n"
" LOG(INFO) << \" tc=\[\" << (end - start) << \"\]\";\n" // NOLINT
" LOG(INFO) << \"(logid=\" << log_id << \") tc=\[\" << (end - " // NOLINT
"start) << \"\]\";\n"
" LOG(INFO)\n"
" << \"TC=[\" << (end - start) << \"] Received debug "
" << \"(logid=\" << log_id << \") TC=[\" << (end - start) << "
"\"] Received debug "
"request[log_id=\" << cntl->log_id()\n"
" << \"] from \" << cntl->remote_side()\n"
" << \" to \" << cntl->local_side();\n",
......@@ -1011,25 +1023,31 @@ class PdsCodeGenerator : public CodeGenerator {
" brpc::ClosureGuard done_guard(done);\n"
" brpc::Controller* cntl = \n"
" static_cast<brpc::Controller*>(cntl_base);\n"
" uint64_t log_id = request->log_id();\n"
" cntl->set_log_id(log_id);\n"
" ::baidu::paddle_serving::predictor::InferService* svr = \n"
" "
"::baidu::paddle_serving::predictor::InferServiceManager::instance("
").item(\"$service$\");\n"
" if (svr == NULL) {\n"
" LOG(ERROR) << \"Not found service: $service$\";\n"
" LOG(ERROR) << \"(logid=\" << log_id << \") Not found service: "
"$service$\";\n"
" cntl->SetFailed(404, \"Not found service: $service$\");\n"
" return ;\n"
" }\n"
" LOG(INFO) << \" remote_side=\[\" << cntl->remote_side() << " // NOLINT
" LOG(INFO) << \"(logid=\" << log_id << \") "
"remote_side=\[\" << cntl->remote_side() << " // NOLINT
"\"\]\";\n"
" LOG(INFO) << \" local_side=\[\" << cntl->local_side() << " // NOLINT
" LOG(INFO) << \"(logid=\" << log_id << \") "
"local_side=\[\" << cntl->local_side() << " // NOLINT
"\"\]\";\n"
" LOG(INFO) << \" service_name=\[\" << \"$name$\" << \"\]\";\n" // NOLINT
" LOG(INFO) << \" log_id=\[\" << cntl->log_id() << \"\]\";\n" // NOLINT
" int err_code = svr->inference(request, response);\n"
" LOG(INFO) << \"(logid=\" << log_id << \") "
"service_name=\[\" << \"$name$\" << \"\]\";\n" // NOLINT
" int err_code = svr->inference(request, response, log_id);\n"
" if (err_code != 0) {\n"
" LOG(WARNING)\n"
" << \"Failed call inferservice[$name$], name[$service$]\"\n"
" << \"(logid=\" << log_id << \") Failed call "
"inferservice[$name$], name[$service$]\"\n"
" << \", error_code: \" << err_code;\n"
" cntl->SetFailed(err_code, \"InferService inference "
"failed!\");\n"
......@@ -1037,7 +1055,8 @@ class PdsCodeGenerator : public CodeGenerator {
" gettimeofday(&tv, NULL);\n"
" long end = tv.tv_sec * 1000000 + tv.tv_usec;\n"
" // flush notice log\n"
" LOG(INFO) << \" tc=\[\" << (end - start) << \"\]\";\n", // NOLINT
" LOG(INFO) << \"(logid=\" << log_id << \") tc=\[\" << (end - " // NOLINT
"start) << \"\]\";\n", // NOLINT
"name",
class_name,
"service",
......@@ -1048,26 +1067,31 @@ class PdsCodeGenerator : public CodeGenerator {
" brpc::ClosureGuard done_guard(done);\n"
" brpc::Controller* cntl = \n"
" static_cast<brpc::Controller*>(cntl_base);\n"
" uint64_t log_id = request->log_id();\n"
" cntl->set_log_id(log_id);\n"
" ::baidu::paddle_serving::predictor::InferService* svr = \n"
" "
"::baidu::paddle_serving::predictor::InferServiceManager::instance("
").item(\"$service$\");\n"
" if (svr == NULL) {\n"
" LOG(ERROR) << \"Not found service: $service$\";\n"
" LOG(ERROR) << \"(logid=\" << log_id << \") Not found service: "
"$service$\";\n"
" cntl->SetFailed(404, \"Not found service: $service$\");\n"
" return ;\n"
" }\n"
" LOG(INFO) << \" remote_side=\[\" << cntl->remote_side() << " // NOLINT
"\"\]\";\n"
" LOG(INFO) << \" local_side=\[\" << cntl->local_side() << " // NOLINT
"\"\]\";\n"
" LOG(INFO) << \" service_name=\[\" << \"$name$\" << \"\]\";\n" // NOLINT
" LOG(INFO) << \" log_id=\[\" << cntl->log_id() << \"\]\";\n" // NOLINT
" LOG(INFO) << \"(logid=\" << log_id << \") remote_side=\[\" " // NOLINT
" << cntl->remote_side() << \"\]\";\n"
" LOG(INFO) << \"(logid=\" << log_id << \") local_side=\[\" " // NOLINT
"<< cntl->local_side() << \"\]\";\n"
" LOG(INFO) << \"(logid=\" << log_id << \") service_name=\[\" " // NOLINT
"<< \"$name$\" << \"\]\";\n"
" butil::IOBufBuilder debug_os;\n"
" int err_code = svr->inference(request, response, &debug_os);\n"
" int err_code = svr->inference(request, response, log_id, "
"&debug_os);\n"
" if (err_code != 0) {\n"
" LOG(WARNING)\n"
" << \"Failed call inferservice[$name$], name[$service$]\"\n"
" << \"(logid=\" << log_id << \") Failed call "
"inferservice[$name$], name[$service$]\"\n"
" << \", error_code: \" << err_code;\n"
" cntl->SetFailed(err_code, \"InferService inference "
"failed!\");\n"
......@@ -1076,9 +1100,11 @@ class PdsCodeGenerator : public CodeGenerator {
" gettimeofday(&tv, NULL);\n"
" long end = tv.tv_sec * 1000000 + tv.tv_usec;\n"
" // flush notice log\n"
" LOG(INFO) << \" tc=\[\" << (end - start) << \"\]\";\n" // NOLINT
" LOG(INFO) << \"(logid=\" << log_id << \") tc=\[\" << (end - " // NOLINT
"start) << \"\]\";\n" // NOLINT
" LOG(INFO)\n"
" << \"TC=[\" << (end - start) << \"] Received debug "
" << \"(logid=\" << log_id << \") TC=[\" << (end - start) << "
"\"] Received debug "
"request[log_id=\" << cntl->log_id()\n"
" << \"] from \" << cntl->remote_side()\n"
" << \" to \" << cntl->local_side();\n",
......
core/predictor/framework/channel.h
浏览文件 @ f3953aa5
......@@ -72,9 +72,10 @@ class Channel {
const std::string& op() { return _op; }
int share_to_bus(Bus* bus) {
int share_to_bus(Bus* bus, const uint64_t log_id) {
if (bus->regist(_op, this) != 0) {
LOG(ERROR) << "Failed regist channel[" << _op << "] to bus!";
LOG(ERROR) << "(logid=" << log_id << ") Failed regist channel[" << _op
<< "] to bus!";
return -1;
}
......
core/predictor/framework/dag.cpp
浏览文件 @ f3953aa5
......@@ -155,13 +155,11 @@ int Dag::init(const configure::Workflow& conf, const std::string& name) {
}
if (FLAGS_el_log_level == 16) {
LOG(INFO) << "DAG: " << _dag_name;
LOG(INFO) << ", Op Num: " << _index_nodes.size();
LOG(INFO) << "DAG: " << _dag_name << ", Op Num: " << _index_nodes.size();
for (uint32_t nid = 0; nid < _index_nodes.size(); nid++) {
DagNode* node = _index_nodes[nid];
LOG(INFO) << ", OP-" << node->id << "-" << node->name << "-"
<< node->type;
LOG(INFO) << " depends: " << node->depends.size();
LOG(INFO) << "OP-" << node->id << "-" << node->name << "-" << node->type
<< " depends: " << node->depends.size();
boost::unordered_map<std::string, EdgeMode>::iterator it;
for (it = node->depends.begin(); it != node->depends.end(); it++) {
......@@ -214,7 +212,7 @@ int Dag::topo_sort() {
}
}
for (int i = 0; i < in_degree.size(); ++i) {
LOG(INFO) << "(" << _index_nodes[i]->name << ") in_degree[" << i
VLOG(2) << "(" << _index_nodes[i]->name << ") in_degree[" << i
<< "]: " << in_degree[i];
}
int sorted_num = 0;
......
core/predictor/framework/dag_view.cpp
浏览文件 @ f3953aa5
......@@ -26,7 +26,9 @@ namespace baidu {
namespace paddle_serving {
namespace predictor {
int DagView::init(Dag* dag, const std::string& service_name) {
int DagView::init(Dag* dag,
const std::string& service_name,
const uint64_t log_id) {
_name = dag->name();
_full_name = service_name + NAME_DELIMITER + dag->name();
_bus = butil::get_object<Bus>();
......@@ -36,17 +38,20 @@ int DagView::init(Dag* dag, const std::string& service_name) {
for (uint32_t si = 0; si < stage_size; si++) {
const DagStage* stage = dag->stage_by_index(si);
if (stage == NULL) {
LOG(ERROR) << "Failed get stage by index:" << si;
LOG(ERROR) << "(logid=" << log_id << ") Failed get stage by index:" << si;
return ERR_INTERNAL_FAILURE;
}
ViewStage* vstage = butil::get_object<ViewStage>();
if (vstage == NULL) {
LOG(ERROR) << "Failed get vstage from object pool"
LOG(ERROR) << "(logid=" << log_id
<< ") Failed get vstage from object pool"
<< "at:" << si;
return ERR_MEM_ALLOC_FAILURE;
}
VLOG(2) << "stage[" << si << "] name: " << stage->full_name;
VLOG(2) << "stage[" << si << "] node size: " << stage->nodes.size();
VLOG(2) << "(logid=" << log_id << ") stage[" << si
<< "] name: " << stage->full_name;
VLOG(2) << "(logid=" << log_id << ") stage[" << si
<< "] node size: " << stage->nodes.size();
vstage->full_name = service_name + NAME_DELIMITER + stage->full_name;
uint32_t node_size = stage->nodes.size();
// create tls view node
......@@ -54,31 +59,39 @@ int DagView::init(Dag* dag, const std::string& service_name) {
DagNode* node = stage->nodes[ni];
ViewNode* vnode = butil::get_object<ViewNode>();
if (vnode == NULL) {
LOG(ERROR) << "Failed get vnode at:" << ni;
LOG(ERROR) << "(logid=" << log_id << ") Failed get vnode at:" << ni;
return ERR_MEM_ALLOC_FAILURE;
}
// factory type
Op* op = OpRepository::instance().get_op(node->type);
if (op == NULL) {
LOG(ERROR) << "Failed get op with type:" << node->type;
LOG(ERROR) << "(logid=" << log_id
<< ") Failed get op with type:" << node->type;
return ERR_INTERNAL_FAILURE;
}
// initialize a TLS op object
VLOG(2) << "dag view initialized: \n"
VLOG(2) << "(logid=" << log_id << ") dag view initialized: \n"
<< "node id: " << node->id << "\n"
<< "node name: " << node->name << "\n"
<< "node type: " << node->type;
if (op->init(_bus, dag, node->id, node->name, node->type, node->conf) !=
0) {
LOG(WARNING) << "Failed init op, type:" << node->type;
if (op->init(_bus,
dag,
node->id,
node->name,
node->type,
node->conf,
log_id) != 0) {
LOG(WARNING) << "(logid=" << log_id
<< ") Failed init op, type:" << node->type;
return ERR_INTERNAL_FAILURE;
}
op->set_full_name(service_name + NAME_DELIMITER + node->full_name);
// Set the name of the Op as the key of the matching engine.
VLOG(2) << "op->set_engine_name(" << node->name.c_str() << ")";
VLOG(2) << "(logid=" << log_id << ") op->set_engine_name("
<< node->name.c_str() << ")";
op->set_engine_name(node->name);
vnode->conf = node;
......@@ -88,7 +101,7 @@ int DagView::init(Dag* dag, const std::string& service_name) {
it != vnode->conf->depends.end();
++it) {
std::string pre_node_name = it->first;
VLOG(2) << "add op pre name: \n"
VLOG(2) << "(logid=" << log_id << ") add op pre name: \n"
<< "current op name: " << vnode->op->op_name()
<< ", previous op name: " << pre_node_name;
vnode->op->add_pre_node_name(pre_node_name);
......@@ -102,7 +115,7 @@ int DagView::init(Dag* dag, const std::string& service_name) {
//<< " previous op name: "
//<< _view[si - 1]->nodes.back()->op->op_name();
// vstage->nodes.back()->op->set_pre_node_name(
//_view[si - 1]->nodes.back()->op->op_name());
// _view[si - 1]->nodes.back()->op->op_name());
/*}*/
_view.push_back(vstage);
}
......@@ -133,14 +146,15 @@ int DagView::deinit() {
return ERR_OK;
}
int DagView::execute(butil::IOBufBuilder* debug_os) {
int DagView::execute(const uint64_t log_id, butil::IOBufBuilder* debug_os) {
uint32_t stage_size = _view.size();
for (uint32_t si = 0; si < stage_size; si++) {
TRACEPRINTF("start to execute stage[%u]", si);
int errcode = execute_one_stage(_view[si], debug_os);
TRACEPRINTF("finish to execute stage[%u]", si);
TRACEPRINTF("(logid=%" PRIu64 ") start to execute stage[%u]", log_id, si);
int errcode = execute_one_stage(_view[si], log_id, debug_os);
TRACEPRINTF("(logid=%" PRIu64 ") finish to execute stage[%u]", log_id, si);
if (errcode < 0) {
LOG(ERROR) << "failed execute stage[" << _view[si]->debug();
LOG(ERROR) << "(logid=" << log_id << ") Failed execute stage["
<< _view[si]->debug();
return errcode;
}
}
......@@ -151,29 +165,34 @@ int DagView::execute(butil::IOBufBuilder* debug_os) {
// You can derive a subclass to implement this func.
// ParallelDagView maybe the one you want.
int DagView::execute_one_stage(ViewStage* vstage,
const uint64_t log_id,
butil::IOBufBuilder* debug_os) {
butil::Timer stage_time(butil::Timer::STARTED);
uint32_t node_size = vstage->nodes.size();
VLOG(2) << "vstage->nodes.size(): " << node_size;
VLOG(2) << "(logid=" << log_id << ") vstage->nodes.size(): " << node_size;
for (uint32_t ni = 0; ni < node_size; ni++) {
ViewNode* vnode = vstage->nodes[ni];
DagNode* conf = vnode->conf;
Op* op = vnode->op;
TRACEPRINTF("start to execute op[%s]", op->name());
int errcode = op->process(debug_os != NULL);
TRACEPRINTF("finish to execute op[%s]", op->name());
TRACEPRINTF(
"(logid=%" PRIu64 ") start to execute op[%s]", log_id, op->name());
int errcode = op->process(log_id, debug_os != NULL);
TRACEPRINTF(
"(logid=%" PRIu64 ") finish to execute op[%s]", log_id, op->name());
if (errcode < 0) {
LOG(ERROR) << "Execute failed, Op:" << op->debug_string();
LOG(ERROR) << "(logid=" << log_id
<< ") Execute failed, Op:" << op->debug_string();
return errcode;
}
if (errcode > 0) {
LOG(INFO) << "Execute ignore, Op:" << op->debug_string();
LOG(INFO) << "(logid=" << log_id
<< ") Execute ignore, Op:" << op->debug_string();
continue;
}
if (debug_os) {
(*debug_os) << "{\"op_name\": \"" << op->name()
(*debug_os) << "(logid=" << log_id << ") {\"op_name\": \"" << op->name()
<< "\", \"debug_str:\": \"" << op->debug_string()
<< "\", \"time_info\": \"" << op->time_info() << "\"}";
}
......@@ -186,34 +205,34 @@ int DagView::execute_one_stage(ViewStage* vstage,
return ERR_OK;
}
int DagView::set_request_channel(Channel& request) {
int DagView::set_request_channel(Channel& request, const uint64_t log_id) {
// Each workflow should get the very beginning
// request (channel), and commit it to bus, for
// the first stage ops consuming.
request.share_to_bus(_bus);
request.share_to_bus(_bus, log_id);
return ERR_OK;
}
const Channel* DagView::get_response_channel() const {
const Channel* DagView::get_response_channel(const uint64_t log_id) const {
// Caller obtains response channel from bus, and
// writes it to rpc response(protbuf/json)
if (_view.size() < 1) {
LOG(ERROR) << "invalid empty view stage!";
LOG(ERROR) << "(logid=" << log_id << ") invalid empty view stage!";
return NULL;
}
ViewStage* last_stage = _view[_view.size() - 1];
if (last_stage->nodes.size() != 1 || last_stage->nodes[0] == NULL) {
LOG(ERROR) << "Invalid last stage, size[" << last_stage->nodes.size()
<< "] != 1";
LOG(ERROR) << "(logid=" << log_id << ") Invalid last stage, size["
<< last_stage->nodes.size() << "] != 1";
return NULL;
}
Op* last_op = last_stage->nodes[0]->op;
if (last_op == NULL) {
LOG(ERROR) << "Last op is NULL";
LOG(ERROR) << "(logid=" << log_id << ") Last op is NULL";
return NULL;
}
return last_op->mutable_channel();
......
core/predictor/framework/dag_view.h
浏览文件 @ f3953aa5
......@@ -47,21 +47,22 @@ class DagView {
~DagView() {}
int init(Dag* dag, const std::string& service_name);
int init(Dag* dag, const std::string& service_name, const uint64_t log_id);
int deinit();
int execute(butil::IOBufBuilder* debug_os);
int execute(const uint64_t log_id, butil::IOBufBuilder* debug_os);
// The default execution strategy is in sequencing
// You can derive a subclass to implement this func.
// ParallelDagView maybe the one you want.
virtual int execute_one_stage(ViewStage* vstage,
const uint64_t log_id,
butil::IOBufBuilder* debug_os);
int set_request_channel(Channel& request); // NOLINT
int set_request_channel(Channel& request, const uint64_t log_id); // NOLINT
const Channel* get_response_channel() const;
const Channel* get_response_channel(const uint64_t log_id) const;
const std::string& name() const { return _name; }
......
core/predictor/framework/service.cpp
浏览文件 @ f3953aa5
......@@ -19,6 +19,7 @@
#include <butil/time.h> // butil::Timer
#endif
#include <inttypes.h>
#include <list>
#include <string>
#include <vector>
......@@ -135,50 +136,63 @@ const std::string& InferService::name() const { return _infer_service_format; }
// ´®ÐÐÖ´ÐÐÿ¸öworkflow
int InferService::inference(const google::protobuf::Message* request,
google::protobuf::Message* response,
const uint64_t log_id,
butil::IOBufBuilder* debug_os) {
TRACEPRINTF("start to inference");
TRACEPRINTF("(logid=%" PRIu64 ") start to inference", log_id);
// when funtion call begins, framework will reset
// thread local variables&resources automatically.
if (Resource::instance().thread_clear() != 0) {
LOG(ERROR) << "Failed thread clear whole resource";
LOG(ERROR) << "(logid=" << log_id << ") Failed thread clear whole resource";
return ERR_INTERNAL_FAILURE;
}
TRACEPRINTF("finish to thread clear");
TRACEPRINTF("(logid=%" PRIu64 ") finish to thread clear", log_id);
if (_enable_map_request_to_workflow) {
LOG(INFO) << "enable map request == True";
std::vector<Workflow*>* workflows = _map_request_to_workflow(request);
VLOG(2) << "(logid=" << log_id << ") enable map request == True";
std::vector<Workflow*>* workflows =
_map_request_to_workflow(request, log_id);
if (!workflows || workflows->size() == 0) {
LOG(ERROR) << "Failed to map request to workflow";
LOG(ERROR) << "(logid=" << log_id
<< ") Failed to map request to workflow";
return ERR_INTERNAL_FAILURE;
}
size_t fsize = workflows->size();
for (size_t fi = 0; fi < fsize; ++fi) {
Workflow* workflow = (*workflows)[fi];
if (workflow == NULL) {
LOG(ERROR) << "Failed to get valid workflow at: " << fi;
LOG(ERROR) << "(logid=" << log_id
<< ") Failed to get valid workflow at: " << fi;
return ERR_INTERNAL_FAILURE;
}
TRACEPRINTF("start to execute workflow[%s]", workflow->name().c_str());
int errcode = _execute_workflow(workflow, request, response, debug_os);
TRACEPRINTF("finish to execute workflow[%s]", workflow->name().c_str());
TRACEPRINTF("(logid=%" PRIu64 ") start to execute workflow[%s]",
log_id,
workflow->name().c_str());
int errcode =
_execute_workflow(workflow, request, response, log_id, debug_os);
TRACEPRINTF("(logid=%" PRIu64 ") finish to execute workflow[%s]",
log_id,
workflow->name().c_str());
if (errcode < 0) {
LOG(ERROR) << "Failed execute workflow[" << workflow->name()
<< "] in:" << name();
LOG(ERROR) << "(logid=" << log_id << ") Failed execute workflow["
<< workflow->name() << "] in:" << name();
return errcode;
}
}
} else {
LOG(INFO) << "enable map request == False";
TRACEPRINTF("start to execute one workflow");
VLOG(2) << "(logid=" << log_id << ") enable map request == False";
TRACEPRINTF("(logid=%" PRIu64 ") start to execute one workflow", log_id);
size_t fsize = _flows.size();
for (size_t fi = 0; fi < fsize; ++fi) {
TRACEPRINTF("start to execute one workflow-%lu", fi);
int errcode = execute_one_workflow(fi, request, response, debug_os);
TRACEPRINTF("finish to execute one workflow-%lu", fi);
TRACEPRINTF(
"(logid=%" PRIu64 ") start to execute one workflow-%lu", log_id, fi);
int errcode =
execute_one_workflow(fi, request, response, log_id, debug_os);
TRACEPRINTF(
"(logid=%" PRIu64 ") finish to execute one workflow-%lu", log_id, fi);
if (errcode < 0) {
LOG(ERROR) << "Failed execute 0-th workflow in:" << name();
LOG(ERROR) << "(logid=" << log_id
<< ") Failed execute 0-th workflow in:" << name();
return errcode;
}
}
......@@ -188,26 +202,30 @@ int InferService::inference(const google::protobuf::Message* request,
int InferService::debug(const google::protobuf::Message* request,
google::protobuf::Message* response,
const uint64_t log_id,
butil::IOBufBuilder* debug_os) {
return inference(request, response, debug_os);
return inference(request, response, log_id, debug_os);
}
int InferService::execute_one_workflow(uint32_t index,
const google::protobuf::Message* request,
google::protobuf::Message* response,
const uint64_t log_id,
butil::IOBufBuilder* debug_os) {
if (index >= _flows.size()) {
LOG(ERROR) << "Faield execute workflow, index: " << index
LOG(ERROR) << "(logid=" << log_id
<< ") Faield execute workflow, index: " << index
<< " >= max:" << _flows.size();
return ERR_OVERFLOW_FAILURE;
}
Workflow* workflow = _flows[index];
return _execute_workflow(workflow, request, response, debug_os);
return _execute_workflow(workflow, request, response, log_id, debug_os);
}
int InferService::_execute_workflow(Workflow* workflow,
const google::protobuf::Message* request,
google::protobuf::Message* response,
const uint64_t log_id,
butil::IOBufBuilder* debug_os) {
butil::Timer workflow_time(butil::Timer::STARTED);
// create and submit beginer channel
......@@ -215,54 +233,62 @@ int InferService::_execute_workflow(Workflow* workflow,
req_channel.init(0, START_OP_NAME);
req_channel = request;
DagView* dv = workflow->fetch_dag_view(full_name());
dv->set_request_channel(req_channel);
DagView* dv = workflow->fetch_dag_view(full_name(), log_id);
dv->set_request_channel(req_channel, log_id);
// call actual inference interface
int errcode = dv->execute(debug_os);
int errcode = dv->execute(log_id, debug_os);
if (errcode < 0) {
LOG(ERROR) << "Failed execute dag for workflow:" << workflow->name();
LOG(ERROR) << "(logid=" << log_id
<< ") Failed execute dag for workflow:" << workflow->name();
return errcode;
}
TRACEPRINTF("finish to dv execute");
TRACEPRINTF("(logid=%" PRIu64 ") finish to dv execute", log_id);
// create ender channel and copy
const Channel* res_channel = dv->get_response_channel();
const Channel* res_channel = dv->get_response_channel(log_id);
if (res_channel == NULL) {
LOG(ERROR) << "(logid=" << log_id << ") Failed get response channel";
return ERR_INTERNAL_FAILURE;
}
if (!_merger || !_merger->merge(res_channel->message(), response)) {
LOG(ERROR) << "Failed merge channel res to response";
LOG(ERROR) << "(logid=" << log_id
<< ") Failed merge channel res to response";
return ERR_INTERNAL_FAILURE;
}
TRACEPRINTF("finish to copy from");
TRACEPRINTF("(logid=%" PRIu64 ") finish to copy from", log_id);
workflow_time.stop();
LOG(INFO) << "workflow total time: " << workflow_time.u_elapsed();
LOG(INFO) << "(logid=" << log_id
<< ") workflow total time: " << workflow_time.u_elapsed();
PredictorMetric::GetInstance()->update_latency_metric(
WORKFLOW_METRIC_PREFIX + dv->full_name(), workflow_time.u_elapsed());
// return tls data to object pool
workflow->return_dag_view(dv);
TRACEPRINTF("finish to return dag view");
TRACEPRINTF("(logid=%" PRIu64 ") finish to return dag view", log_id);
return ERR_OK;
}
std::vector<Workflow*>* InferService::_map_request_to_workflow(
const google::protobuf::Message* request) {
const google::protobuf::Message* request, const uint64_t log_id) {
const google::protobuf::Descriptor* desc = request->GetDescriptor();
const google::protobuf::FieldDescriptor* field =
desc->FindFieldByName(_request_field_key);
if (field == NULL) {
LOG(ERROR) << "No field[" << _request_field_key << "] in ["
<< desc->full_name() << "].";
LOG(ERROR) << "(logid=" << log_id << ") No field[" << _request_field_key
<< "] in [" << desc->full_name() << "].";
return NULL;
}
if (field->is_repeated()) {
LOG(ERROR) << "field[" << desc->full_name() << "." << _request_field_key
<< "] is repeated.";
LOG(ERROR) << "(logid=" << log_id << ") field[" << desc->full_name() << "."
<< _request_field_key << "] is repeated.";
return NULL;
}
if (field->cpp_type() != google::protobuf::FieldDescriptor::CPPTYPE_STRING) {
LOG(ERROR) << "field[" << desc->full_name() << "." << _request_field_key
<< "] should be string";
LOG(ERROR) << "(logid=" << log_id << ") field[" << desc->full_name() << "."
<< _request_field_key << "] should be string";
return NULL;
}
const std::string& field_value =
......@@ -270,7 +296,7 @@ std::vector<Workflow*>* InferService::_map_request_to_workflow(
std::vector<Workflow*>* p_workflow =
_request_to_workflow_map.seek(field_value);
if (p_workflow == NULL) {
LOG(ERROR) << "cannot find key[" << field_value
LOG(ERROR) << "(logid=" << log_id << ") cannot find key[" << field_value
<< "] in _request_to_workflow_map";
return NULL;
}
......
core/predictor/framework/service.h
浏览文件 @ f3953aa5
......@@ -52,25 +52,29 @@ class InferService {
// Execute each workflow serially
virtual int inference(const google::protobuf::Message* request,
google::protobuf::Message* response,
const uint64_t log_id,
butil::IOBufBuilder* debug_os = NULL);
int debug(const google::protobuf::Message* request,
google::protobuf::Message* response,
const uint64_t log_id,
butil::IOBufBuilder* debug_os);
int execute_one_workflow(uint32_t index,
const google::protobuf::Message* request,
google::protobuf::Message* response,
const uint64_t log_id,
butil::IOBufBuilder* debug_os);
private:
int _execute_workflow(Workflow* workflow,
const google::protobuf::Message* request,
google::protobuf::Message* response,
const uint64_t log_id,
butil::IOBufBuilder* debug_os);
std::vector<Workflow*>* _map_request_to_workflow(
const google::protobuf::Message* request);
const google::protobuf::Message* request, const uint64_t log_id);
private:
std::vector<Workflow*> _flows;
......@@ -88,6 +92,7 @@ class ParallelInferService : public InferService {
// Execute workflows in parallel
int inference(const google::protobuf::Message* request,
google::protobuf::Message* response,
const uint64_t log_id,
butil::IOBufBuilder* debug_os) {
return 0;
}
......
core/predictor/framework/workflow.cpp
浏览文件 @ f3953aa5
......@@ -32,21 +32,22 @@ int Workflow::init(const configure::Workflow& conf) {
return 0;
}
DagView* Workflow::fetch_dag_view(const std::string& service_name) {
DagView* Workflow::fetch_dag_view(const std::string& service_name,
const uint64_t log_id) {
DagView* view = NULL;
if (_type == "Sequence") {
view = butil::get_object<DagView>();
} else if (_type == "Parallel") {
view = butil::get_object<ParallelDagView>();
} else {
LOG(ERROR) << "Unknown dag type:" << _type << "!";
LOG(ERROR) << "(logid=" << log_id << ") Unknown dag type:" << _type << "!";
return NULL;
}
if (view == NULL) {
LOG(ERROR) << "create dag view from pool failed!";
LOG(ERROR) << "(logid=" << log_id << ") create dag view from pool failed!";
return NULL;
}
view->init(&_dag, service_name);
view->init(&_dag, service_name, log_id);
return view;
}
......
core/predictor/framework/workflow.h
浏览文件 @ f3953aa5
......@@ -36,7 +36,8 @@ class Workflow {
// different apps.
int init(const configure::Workflow& conf);
DagView* fetch_dag_view(const std::string& service_name);
DagView* fetch_dag_view(const std::string& service_name,
const uint64_t log_id);
int deinit() { return 0; }
......
core/predictor/op/op.cpp
浏览文件 @ f3953aa5
......@@ -35,7 +35,8 @@ int Op::init(Bus* bus,
uint32_t id,
const std::string& name,
const std::string& type,
void* conf) {
void* conf,
const uint64_t log_id) {
_bus = bus;
_dag = dag;
_id = id;
......@@ -45,7 +46,8 @@ int Op::init(Bus* bus,
_timer = butil::get_object<TimerFlow>();
if (!_timer) {
LOG(ERROR) << "Invalid timerflow in op:" << this->name();
LOG(ERROR) << "(logid=" << log_id
<< ") Invalid timerflow in op:" << this->name();
return -1;
}
......@@ -55,7 +57,8 @@ int Op::init(Bus* bus,
Channel* channel = mutable_channel();
if (channel == NULL) {
LOG(ERROR) << "Failed mutable channel in op: " << this->id() << ", "
LOG(ERROR) << "(logid=" << log_id
<< ") Failed mutable channel in op: " << this->id() << ", "
<< this->name() << "!";
return -1;
}
......@@ -96,18 +99,20 @@ int Op::check_time(const char* tag) {
return 0;
}
int Op::process(bool debug) {
int Op::process(const uint64_t log_id, bool debug) {
butil::Timer op_time(butil::Timer::STARTED);
if (debug && _timer) {
_timer->start();
}
if (!_has_init) {
LOG(ERROR) << "Make sure op has been init before inference";
LOG(ERROR) << "(logid=" << log_id
<< ") Make sure op has been init before inference";
return ERR_INTERNAL_FAILURE;
}
if (_has_calc) {
LOG(INFO) << "Op: " << _name << " already processed before";
LOG(INFO) << "(logid=" << log_id << ") Op: " << _name
<< " already processed before";
return ERR_OK;
}
......@@ -143,7 +148,7 @@ int Op::process(bool debug) {
// 3. share output to bus
Channel* channel = mutable_channel();
channel->share_to_bus(_bus);
channel->share_to_bus(_bus, log_id);
// 4. mark has calculated
_has_calc = true;
......@@ -156,7 +161,8 @@ int Op::process(bool debug) {
op_time.stop();
PredictorMetric::GetInstance()->update_latency_metric(
OP_METRIC_PREFIX + full_name(), op_time.u_elapsed());
LOG(INFO) << " " << name() << "_time=[" << op_time.u_elapsed() << "]";
LOG(INFO) << "(logid=" << log_id << ") " << name() << "_time=["
<< op_time.u_elapsed() << "]";
return ERR_OK;
}
......
core/predictor/op/op.h
浏览文件 @ f3953aa5
......@@ -113,13 +113,14 @@ class Op {
uint32_t id,
const std::string& name,
const std::string& type,
void* conf);
void* conf,
const uint64_t log_id);
int deinit();
int check_time(const char* tag);
int process(bool debug);
int process(const uint64_t log_id, bool debug);
std::string time_info();
......
core/sdk-cpp/proto/general_model_service.proto
浏览文件 @ f3953aa5
......@@ -37,6 +37,7 @@ message Request {
repeated FeedInst insts = 1;
repeated string fetch_var_names = 2;
optional bool profile_server = 3 [ default = false ];
required uint64 log_id = 4 [ default = 0 ];
};
message Response {
......
doc/PIPELINE_SERVING.md
浏览文件 @ f3953aa5
......@@ -33,6 +33,7 @@ The graph execution engine consists of OPs and Channels, and the connected OPs s
- The default function of a single OP is to access a single Paddle Serving Service based on the input Channel data and put the result into the output Channel.
- OP supports user customization, including preprocess, process, postprocess functions that can be inherited and implemented by the user.
- OP can set the number of concurrencies to increase the number of concurrencies processed.
- OP can obtain data from multiple different RPC requests for Auto-Batching.
- OP can be started by a thread or process.
### Channel Design
......@@ -46,6 +47,7 @@ The graph execution engine consists of OPs and Channels, and the connected OPs s
</center>
### Extreme Case Consideration
- Request timeout
......@@ -59,9 +61,9 @@ The graph execution engine consists of OPs and Channels, and the connected OPs s
- Whether input buffers and output buffers in Channel will increase indefinitely
- It will not increase indefinitely. The input to the entire graph execution engine is placed inside a Channel's internal queue, directly acting as a traffic control buffer queue for the entire service.
- For input buffer, adjust the number of concurrencies of OP1 and OP2 according to the amount of computation, so that the number of input buffers from each input OP is relatively balanced.
- For output buffer, you can use a similar process as input buffer, which adjusts the concurrency of OP3 and OP4 to control the buffer length of output buffer.
- Note: The length of the input buffer depends on the speed at which each item in the internal queue is ready, and the length of the output buffer depends on the speed at which downstream OPs obtain data from the output buffer.
- For input buffer, adjust the number of concurrencies of OP1 and OP2 according to the amount of computation, so that the number of input buffers from each input OP is relatively balanced. (The length of the input buffer depends on the speed at which each item in the internal queue is ready)
- For output buffer, you can use a similar process as input buffer, which adjusts the concurrency of OP3 and OP4 to control the buffer length of output buffer. (The length of the output buffer depends on the speed at which downstream OPs obtain data from the output buffer)
- The amount of data in the Channel will not exceed `worker_num` of gRPC, that is, it will not exceed the thread pool size.
## Detailed Design
......@@ -79,31 +81,36 @@ def __init__(name=None,
client_config=None,
concurrency=1,
timeout=-1,
retry=1)
retry=1,
batch_size=1,
auto_batching_timeout=None)
```
The meaning of each parameter is as follows:
| Parameter | Meaning |
| :--------------: | :----------------------------------------------------------: |
| :-------------------: | :----------------------------------------------------------: |
| name | (str) String used to identify the OP type, which must be globally unique. |
| input_ops | (list) A list of all previous OPs of the current Op. |
| server_endpoints | (list) List of endpoints for remote Paddle Serving Service. If this parameter is not set, the OP will not access the remote Paddle Serving Service, that is, the process operation will not be performed. |
| fetch_list | (list) List of fetch variable names for remote Paddle Serving Service. |
| client_config | (str) The path of the client configuration file corresponding to the Paddle Serving Service. |
| concurrency | (int) The number of concurrent OPs. |
| timeout | (int) The timeout time of the process operation, in seconds. If the value is less than zero, no timeout is considered. |
| timeout | (int) The timeout time of the process operation, in ms. If the value is less than zero, no timeout is considered. |
| retry | (int) Timeout number of retries. When the value is 1, no retries are made. |
| batch_size | (int) The expected batch_size of Auto-Batching, since building batches may time out, the actual batch_size may be less than the set value. |
| auto_batching_timeout | (float) Timeout for building batches of Auto-Batching (the unit is ms). |
#### 2. General OP Secondary Development Interface
| Interface or Variable | Explain |
| :--------------------------------------------: | :----------------------------------------------------------: |
| def preprocess(self, input_dicts) | Process the data obtained from the channel, and the processed data will be used as the input of the **process** function. |
| def process(self, feed_dict) | The RPC prediction process is based on the Paddle Serving Client, and the processed data will be used as the input of the **postprocess** function. |
| def postprocess(self, input_dicts, fetch_dict) | After processing the prediction results, the processed data will be put into the subsequent Channel to be obtained by the subsequent OP. |
| :----------------------------------------------: | :----------------------------------------------------------: |
| def preprocess(self, input_dicts) | Process the data obtained from the channel, and the processed data will be used as the input of the **process** function. (This function handles a **sample**) |
| def process(self, feed_dict_list, typical_logid) | The RPC prediction process is based on the Paddle Serving Client, and the processed data will be used as the input of the **postprocess** function. (This function handles a **batch**) |
| def postprocess(self, input_dicts, fetch_dict) | After processing the prediction results, the processed data will be put into the subsequent Channel to be obtained by the subsequent OP. (This function handles a **sample**) |
| def init_op(self) | Used to load resources (such as word dictionary). |
| self.concurrency_idx | Concurrency index of current thread / process (different kinds of OP are calculated separately). |
| self.concurrency_idx | Concurrency index of current process(not thread) (different kinds of OP are calculated separately). |
In a running cycle, OP will execute three operations: preprocess, process, and postprocess (when the `server_endpoints` parameter is not set, the process operation is not executed). Users can rewrite these three functions. The default implementation is as follows:
......@@ -117,24 +124,28 @@ def preprocess(self, input_dicts):
(_, input_dict), = input_dicts.items()
return input_dict
def process(self, feed_dict):
err, err_info = ChannelData.check_npdata(feed_dict)
def process(self, feed_dict_list, typical_logid):
err, err_info = ChannelData.check_batch_npdata(feed_dict_list)
if err != 0:
raise NotImplementedError(
"{} Please override preprocess func.".format(err_info))
call_result = self.client.predict(
feed=feed_dict, fetch=self._fetch_names)
feed=feed_dict_list, fetch=self._fetch_names, log_id=typical_logid)
if isinstance(self.client, MultiLangClient):
if call_result is None or call_result["serving_status_code"] != 0:
return None
call_result.pop("serving_status_code")
return call_result
def postprocess(self, input_dicts, fetch_dict):
return fetch_dict
```
The parameter of **preprocess** is the data `input_dicts` in the previous Channel. This variable is a dictionary with the name of the previous OP as key and the output of the corresponding OP as value.
The parameter of **preprocess** is the data `input_dicts` in the previous Channel. This variable (as a **sample**) is a dictionary with the name of the previous OP as key and the output of the corresponding OP as value.
The parameter of **process** is the input variable `fetch_dict` (the return value of the preprocess function) of the Paddle Serving Client prediction interface. This variable is a dictionary with feed_name as the key and the data in the ndarray format as the value.
The parameter of **process** is the input variable `fetch_dict_list` (a list of the return value of the preprocess function) of the Paddle Serving Client prediction interface. This variable (as a **batch**) is a list of dictionaries with feed_name as the key and the data in the ndarray format as the value. `typical_logid` is used as the logid that penetrates to PaddleServingService.
The parameters of **postprocess** are `input_dicts` and `fetch_dict`. `input_dicts` is consistent with the parameter of preprocess, and `fetch_dict` is the return value of the process function (if process is not executed, this value is the return value of preprocess).
The parameters of **postprocess** are `input_dicts` and `fetch_dict`. `input_dicts` is consistent with the parameter of preprocess, and `fetch_dict` (as a **sample**) is a sample of the return batch of the process function (if process is not executed, this value is the return value of preprocess).
Users can also rewrite the **init_op** function to load some custom resources (such as word dictionary). The default implementation is as follows:
......@@ -143,7 +154,7 @@ def init_op(self):
pass
```
It should be noted that in the threaded version of OP, each OP will only call this function once, so the loaded resources must be thread safe.
It should be **noted** that in the threaded version of OP, each OP will only call this function once, so the loaded resources must be thread safe.
#### 3. RequestOp Definition
......@@ -248,6 +259,8 @@ dag:
client_type: brpc # Use brpc or grpc client. The default is brpc
retry: 1 # The number of times DAG executor retries after failure. The default value is 1, that is, no retrying
use_profile: false # Whether to print the log on the server side. The default is false
tracer:
interval_s: 600 # Monitoring time interval of Tracer (in seconds). Do not start monitoring when the value is less than 1. The default value is -1
```
......@@ -282,14 +295,8 @@ from paddle_serving_server.pipeline import PipelineServer
from paddle_serving_server.pipeline.proto import pipeline_service_pb2
from paddle_serving_server.pipeline.channel import ChannelDataEcode
import numpy as np
import logging
from paddle_serving_app.reader import IMDBDataset
logging.basicConfig(level=logging.DEBUG)
_LOGGER = logging.getLogger()
class ImdbRequestOp(RequestOp):
def init_op(self):
self.imdb_dataset = IMDBDataset()
......@@ -390,15 +397,23 @@ dag:
use_profile: true
```
After the function is enabled, the server will print the corresponding log information to the standard output in the process of prediction. In order to show the time consumption of each stage more intuitively, scripts are provided for further analysis and processing of log files.
After the function is enabled, the server will print the corresponding log information to the standard output in the process of prediction. In order to show the time consumption of each stage more intuitively, Analyst module is provided for further analysis and processing of log files.
The output of the server is first saved to a file. Taking profile as an example, the script converts the time monitoring information in the log into JSON format and saves it to the trace file. The trace file can be visualized through the tracing function of Chrome browser.
The output of the server is first saved to a file. Taking `profile.txt` as an example, the script converts the time monitoring information in the log into JSON format and saves it to the `trace` file. The `trace` file can be visualized through the tracing function of Chrome browser.
```shell
python timeline_trace.py profile trace
from paddle_serving_server.pipeline import Analyst
import json
import sys
if __name__ == "__main__":
log_filename = "profile.txt"
trace_filename = "trace"
analyst = Analyst(log_filename)
analyst.save_trace(trace_filename)
```
Specific operation: open Chrome browser, input in the address bar `chrome://tracing/` , jump to the tracing page, click the load button, open the saved trace file, and then visualize the time information of each stage of the prediction service.
Specific operation: open Chrome browser, input in the address bar `chrome://tracing/` , jump to the tracing page, click the load button, open the saved `trace` file, and then visualize the time information of each stage of the prediction service.
### Output profile information on client side
......
doc/PIPELINE_SERVING_CN.md
浏览文件 @ f3953aa5
......@@ -6,6 +6,7 @@ Paddle Serving 通常用于单模型的一键部署，但端到端的深度学
Paddle Serving 提供了用户友好的多模型组合服务编程框架，Pipeline Serving，旨在降低编程门槛，提高资源使用率（尤其是GPU设备），提升整体的预估效率。
## 整体架构设计
Server端基于 gRPC 和图执行引擎构建，两者的关系如下图所示。
......@@ -30,9 +31,10 @@ Server端基于 gRPC 和图执行引擎构建，两者的关系如下图所示
### OP的设计
- 单个OP默认的功能是根据输入的 Channel 数据，访问一个 Paddle Serving 的单模型服务，并将结果存在输出的 Channel
- 单个 OP 默认的功能是根据输入的 Channel 数据，访问一个 Paddle Serving 的单模型服务，并将结果存在输出的 Channel
- 单个 OP 可以支持用户自定义，包括 preprocess，process，postprocess 三个函数都可以由用户继承和实现
- 单个 OP 可以控制并发数，从而增加处理并发数
- 单个 OP 可以获取多个不同 RPC 请求的数据，以实现 Auto-Batching
- OP 可以由线程或进程启动
### Channel的设计
......@@ -59,11 +61,9 @@ Server端基于 gRPC 和图执行引擎构建，两者的关系如下图所示
- Channel 设计中的 input buffer 和 output buffer 是否会无限增加
- 不会。整个图执行引擎的输入会放到一个 Channel 的 internal queue 里面，直接作为整个服务的流量控制缓冲队列
- 对于 input buffer，根据计算量的情况调整 OP1 和 OP2 的并发数，使得 input buffer 来自各个输入 OP 的数量相对平衡
- 对于 output buffer，可以采用和 input buffer 类似的处理方法，即调整 OP3 和 OP4 的并发数，使得 output buffer 的缓冲长度得到控制
- 注：input buffer 的长度取决于 internal queue 中每个 item 完全 ready 的速度，output buffer 的长度取决于下游 OP 从 output buffer 获取数据的速度
## 详细设计
- 对于 input buffer，根据计算量的情况调整 OP1 和 OP2 的并发数，使得 input buffer 来自各个输入 OP 的数量相对平衡（input buffer 的长度取决于 internal queue 中每个 item 完全 ready 的速度）
- 对于 output buffer，可以采用和 input buffer 类似的处理方法，即调整 OP3 和 OP4 的并发数，使得 output buffer 的缓冲长度得到控制（output buffer 的长度取决于下游 OP 从 output buffer 获取数据的速度）
- 同时 Channel 中数据量不会超过 gRPC 的 `worker_num`，即线程池大小
### 用户接口设计
......@@ -79,31 +79,36 @@ def __init__(name=None,
client_config=None,
concurrency=1,
timeout=-1,
retry=1)
retry=1,
batch_size=1,
auto_batching_timeout=None)
```
各参数含义如下
| 参数名 | 含义 |
| :--------------: | :----------------------------------------------------------: |
| :-------------------: | :----------------------------------------------------------: |
| name | （str）用于标识 OP 类型的字符串，该字段必须全局唯一。 |
| input_ops | （list）当前 OP 的所有前继 OP 的列表。 |
| server_endpoints | （list）远程 Paddle Serving Service 的 endpoints 列表。如果不设置该参数，则不访问远程 Paddle Serving Service，即 不会执行 process 操作。 |
| fetch_list | （list）远程 Paddle Serving Service 的 fetch 列表。 |
| client_config | （str）Paddle Serving Service 对应的 Client 端配置文件路径。 |
| concurrency | （int）OP 的并发数。 |
| timeout | （int）process 操作的超时时间，单位为秒。若该值小于零，则视作不超时。 |
| timeout | （int）process 操作的超时时间，单位为毫秒。若该值小于零，则视作不超时。 |
| retry | （int）超时重试次数。当该值为 1 时，不进行重试。 |
| batch_size | （int）进行 Auto-Batching 的期望 batch_size 大小，由于构建 batch 可能超时，实际 batch_size 可能小于设定值。 |
| auto_batching_timeout | （float）进行 Auto-Batching 构建 batch 的超时时间，单位为毫秒。 |
#### 2. 普通 OP二次开发接口
| 变量或接口 | 说明 |
| :--------------------------------------------: | :----------------------------------------------------------: |
| def preprocess(self, input_dicts) | 对从 Channel 中获取的数据进行处理，处理完的数据将作为 **process** 函数的输入。 |
| def process(self, feed_dict) | 基于 Paddle Serving Client 进行 RPC 预测，处理完的数据将作为 **postprocess** 函数的输入。 |
| def postprocess(self, input_dicts, fetch_dict) | 处理预测结果，处理完的数据将被放入后继 Channel 中，以被后继 OP 获取。 |
| :----------------------------------------------: | :----------------------------------------------------------: |
| def preprocess(self, input_dicts) | 对从 Channel 中获取的数据进行处理，处理完的数据将作为 **process** 函数的输入。（该函数对一个 **sample** 进行处理） |
| def process(self, feed_dict_list, typical_logid) | 基于 Paddle Serving Client 进行 RPC 预测，处理完的数据将作为 **postprocess** 函数的输入。（该函数对一个 **batch** 进行处理） |
| def postprocess(self, input_dicts, fetch_dict) | 处理预测结果，处理完的数据将被放入后继 Channel 中，以被后继 OP 获取。（该函数对一个 **sample** 进行处理） |
| def init_op(self) | 用于加载资源（如字典等）。 |
| self.concurrency_idx | 当前线程（进程）的并发数索引（不同种类的 OP 单独计算）。 |
| self.concurrency_idx | 当前进程（非线程）的并发数索引（不同种类的 OP 单独计算）。 |
OP 在一个运行周期中会依次执行 preprocess，process，postprocess 三个操作（当不设置 `server_endpoints` 参数时，不执行 process 操作），用户可以对这三个函数进行重写，默认实现如下：
......@@ -117,25 +122,28 @@ def preprocess(self, input_dicts):
(_, input_dict), = input_dicts.items()
return input_dict
def process(self, feed_dict):
err, err_info = ChannelData.check_npdata(feed_dict)
def process(self, feed_dict_list, typical_logid):
err, err_info = ChannelData.check_batch_npdata(feed_dict_list)
if err != 0:
raise NotImplementedError(
"{} Please override preprocess func.".format(err_info))
call_result = self.client.predict(
feed=feed_dict, fetch=self._fetch_names)
feed=feed_dict_list, fetch=self._fetch_names, log_id=typical_logid)
if isinstance(self.client, MultiLangClient):
if call_result is None or call_result["serving_status_code"] != 0:
return None
call_result.pop("serving_status_code")
return call_result
def postprocess(self, input_dicts, fetch_dict):
return fetch_dict
```
**preprocess** 的参数是前继 Channel 中的数据 `input_dicts`，该变量（作为一个 **sample**）是一个以前继 OP 的 name 为 Key，对应 OP 的输出为 Value 的字典。
**preprocess** 的参数是前继 Channel 中的数据 `input_dicts`，该变量是一个以前继 OP 的 name 为 Key，对应 OP 的输出为 Value 的字典。
**process** 的参数是 Paddle Serving Client 预测接口的输入变量 `fetch_dict`（preprocess 函数的返回值），该变量是一个以 feed_name 为 Key，对应 ndarray 格式的数据为 Value 的字典。
**process** 的参数是 Paddle Serving Client 预测接口的输入变量 `fetch_dict_list`（preprocess 函数的返回值的列表），该变量（作为一个 **batch**）是一个列表，列表中的元素为以 feed_name 为 Key，对应 ndarray 格式的数据为 Value 的字典。`typical_logid` 作为向 PaddleServingService 穿透的 logid。
**postprocess** 的参数是 `input_dicts` 和 `fetch_dict`，`input_dicts` 与 preprocess 的参数一致，`fetch_dict` 是 process 函数的返回值（如果没有执行 process ，则该值为 preprocess 的返回值）。
**postprocess** 的参数是 `input_dicts` 和 `fetch_dict`，`input_dicts` 与 preprocess 的参数一致，`fetch_dict` （作为一个 **sample**）是 process 函数的返回 batch 中的一个 sample（如果没有执行 process ，则该值为 preprocess 的返回值）。
用户还可以对 **init_op** 函数进行重写，已加载自定义的一些资源（比如字典等），默认实现如下：
......@@ -144,7 +152,7 @@ def init_op(self):
pass
```
需要注意的是，在线程版 OP 中，每个 OP 只会调用一次该函数，故加载的资源必须要求是线程安全的。
需要**注意**的是，在线程版 OP 中，每个 OP 只会调用一次该函数，故加载的资源必须要求是线程安全的。
#### 3. RequestOp 定义
......@@ -249,6 +257,8 @@ dag:
client_type: brpc # 使用 brpc 或 grpc client，默认为 brpc
retry: 1 # DAG Executor 在失败后重试次数，默认为 1，即不重试
use_profile: false # 是否在 Server 端打印日志，默认为 false
tracer:
interval_s: 600 # Tracer 监控的时间间隔，单位为秒。当该值小于 1 时不启动监控，默认为 -1
```
......@@ -283,14 +293,8 @@ from paddle_serving_server.pipeline import PipelineServer
from paddle_serving_server.pipeline.proto import pipeline_service_pb2
from paddle_serving_server.pipeline.channel import ChannelDataEcode
import numpy as np
import logging
from paddle_serving_app.reader import IMDBDataset
logging.basicConfig(level=logging.DEBUG)
_LOGGER = logging.getLogger()
class ImdbRequestOp(RequestOp):
def init_op(self):
self.imdb_dataset = IMDBDataset()
......@@ -311,7 +315,6 @@ class CombineOp(Op):
def preprocess(self, input_data):
combined_prediction = 0
for op_name, data in input_data.items():
_LOGGER.info("{}: {}".format(op_name, data["prediction"]))
combined_prediction += data["prediction"]
data = {"prediction": combined_prediction / 2}
return data
......@@ -391,15 +394,23 @@ dag:
use_profile: true
```
开启该功能后，Server 端在预测的过程中会将对应的日志信息打印到标准输出，为了更直观地展现各阶段的耗时，提供脚本对日志文件做进一步的分析处理。
开启该功能后，Server 端在预测的过程中会将对应的日志信息打印到标准输出，为了更直观地展现各阶段的耗时，提供 Analyst 模块对日志文件做进一步的分析处理。
使用时先将 Server 的输出保存到文件，以 profile 为例，脚本将日志中的时间打点信息转换成 json 格式保存到trace 文件，trace 文件可以通过 chrome 浏览器的 tracing 功能进行可视化。
使用时先将 Server 的输出保存到文件，以 `profile.txt` 为例，脚本将日志中的时间打点信息转换成 json 格式保存到 `trace` 文件，`trace` 文件可以通过 chrome 浏览器的 tracing 功能进行可视化。
```shell
python timeline_trace.py profile trace
```python
from paddle_serving_server.pipeline import Analyst
import json
import sys
if __name__ == "__main__":
log_filename = "profile.txt"
trace_filename = "trace"
analyst = Analyst(log_filename)
analyst.save_trace(trace_filename)
```
具体操作：打开 chrome 浏览器，在地址栏输入 chrome://tracing/ ，跳转至 tracing 页面，点击 load 按钮，打开保存的 trace 文件，即可将预测服务的各阶段时间信息可视化。
具体操作：打开 chrome 浏览器，在地址栏输入 `chrome://tracing/` ，跳转至 tracing 页面，点击 load 按钮，打开保存的 `trace` 文件，即可将预测服务的各阶段时间信息可视化。
### 在 Client 端输出 Profile 信息
......
doc/pipeline_serving-image1.png
查看替换文件 @ a27774f2 浏览文件 @ f3953aa5
doc/pipeline_serving-image1.png

96.0 KB | W: | H:

doc/pipeline_serving-image1.png

107.7 KB | W: | H:

doc/pipeline_serving-image1.png
doc/pipeline_serving-image1.png
doc/pipeline_serving-image1.png
doc/pipeline_serving-image1.png
  • 2-up
  • Swipe
  • Onion skin
java/src/main/java/io/paddle/serving/client/Client.java
浏览文件 @ f3953aa5
......@@ -192,14 +192,16 @@ public class Client {
private InferenceRequest _packInferenceRequest(
List<HashMap<String, INDArray>> feed_batch,
Iterable<String> fetch) throws IllegalArgumentException {
Iterable<String> fetch,
long log_id) throws IllegalArgumentException {
List<String> feed_var_names = new ArrayList<String>();
feed_var_names.addAll(feed_batch.get(0).keySet());
InferenceRequest.Builder req_builder = InferenceRequest.newBuilder()
.addAllFeedVarNames(feed_var_names)
.addAllFetchVarNames(fetch)
.setIsPython(false);
.setIsPython(false)
.setLogId(log_id);
for (HashMap<String, INDArray> feed_data: feed_batch) {
FeedInst.Builder inst_builder = FeedInst.newBuilder();
for (String name: feed_var_names) {
......@@ -332,76 +334,151 @@ public class Client {
public Map<String, INDArray> predict(
HashMap<String, INDArray> feed,
Iterable<String> fetch) {
return predict(feed, fetch, false);
return predict(feed, fetch, false, 0);
}
public Map<String, INDArray> predict(
HashMap<String, INDArray> feed,
Iterable<String> fetch,
long log_id) {
return predict(feed, fetch, false, log_id);
}
public Map<String, HashMap<String, INDArray>> ensemble_predict(
HashMap<String, INDArray> feed,
Iterable<String> fetch) {
return ensemble_predict(feed, fetch, false);
return ensemble_predict(feed, fetch, false, 0);
}
public Map<String, HashMap<String, INDArray>> ensemble_predict(
HashMap<String, INDArray> feed,
Iterable<String> fetch,
long log_id) {
return ensemble_predict(feed, fetch, false, log_id);
}
public PredictFuture asyn_predict(
HashMap<String, INDArray> feed,
Iterable<String> fetch) {
return asyn_predict(feed, fetch, false);
return asyn_predict(feed, fetch, false, 0);
}
public PredictFuture asyn_predict(
HashMap<String, INDArray> feed,
Iterable<String> fetch,
long log_id) {
return asyn_predict(feed, fetch, false, log_id);
}
public Map<String, INDArray> predict(
HashMap<String, INDArray> feed,
Iterable<String> fetch,
Boolean need_variant_tag) {
return predict(feed, fetch, need_variant_tag, 0);
}
public Map<String, INDArray> predict(
HashMap<String, INDArray> feed,
Iterable<String> fetch,
Boolean need_variant_tag,
long log_id) {
List<HashMap<String, INDArray>> feed_batch
= new ArrayList<HashMap<String, INDArray>>();
feed_batch.add(feed);
return predict(feed_batch, fetch, need_variant_tag);
return predict(feed_batch, fetch, need_variant_tag, log_id);
}
public Map<String, HashMap<String, INDArray>> ensemble_predict(
HashMap<String, INDArray> feed,
Iterable<String> fetch,
Boolean need_variant_tag) {
return ensemble_predict(feed, fetch, need_variant_tag, 0);
}
public Map<String, HashMap<String, INDArray>> ensemble_predict(
HashMap<String, INDArray> feed,
Iterable<String> fetch,
Boolean need_variant_tag,
long log_id) {
List<HashMap<String, INDArray>> feed_batch
= new ArrayList<HashMap<String, INDArray>>();
feed_batch.add(feed);
return ensemble_predict(feed_batch, fetch, need_variant_tag);
return ensemble_predict(feed_batch, fetch, need_variant_tag, log_id);
}
public PredictFuture asyn_predict(
HashMap<String, INDArray> feed,
Iterable<String> fetch,
Boolean need_variant_tag) {
return asyn_predict(feed, fetch, need_variant_tag, 0);
}
public PredictFuture asyn_predict(
HashMap<String, INDArray> feed,
Iterable<String> fetch,
Boolean need_variant_tag,
long log_id) {
List<HashMap<String, INDArray>> feed_batch
= new ArrayList<HashMap<String, INDArray>>();
feed_batch.add(feed);
return asyn_predict(feed_batch, fetch, need_variant_tag);
return asyn_predict(feed_batch, fetch, need_variant_tag, log_id);
}
public Map<String, INDArray> predict(
List<HashMap<String, INDArray>> feed_batch,
Iterable<String> fetch) {
return predict(feed_batch, fetch, false);
return predict(feed_batch, fetch, false, 0);
}
public Map<String, INDArray> predict(
List<HashMap<String, INDArray>> feed_batch,
Iterable<String> fetch,
long log_id) {
return predict(feed_batch, fetch, false, log_id);
}
public Map<String, HashMap<String, INDArray>> ensemble_predict(
List<HashMap<String, INDArray>> feed_batch,
Iterable<String> fetch) {
return ensemble_predict(feed_batch, fetch, false);
return ensemble_predict(feed_batch, fetch, false, 0);
}
public Map<String, HashMap<String, INDArray>> ensemble_predict(
List<HashMap<String, INDArray>> feed_batch,
Iterable<String> fetch,
long log_id) {
return ensemble_predict(feed_batch, fetch, false, log_id);
}
public PredictFuture asyn_predict(
List<HashMap<String, INDArray>> feed_batch,
Iterable<String> fetch) {
return asyn_predict(feed_batch, fetch, false);
return asyn_predict(feed_batch, fetch, false, 0);
}
public PredictFuture asyn_predict(
List<HashMap<String, INDArray>> feed_batch,
Iterable<String> fetch,
long log_id) {
return asyn_predict(feed_batch, fetch, false, log_id);
}
public Map<String, INDArray> predict(
List<HashMap<String, INDArray>> feed_batch,
Iterable<String> fetch,
Boolean need_variant_tag) {
return predict(feed_batch, fetch, need_variant_tag, 0);
}
public Map<String, INDArray> predict(
List<HashMap<String, INDArray>> feed_batch,
Iterable<String> fetch,
Boolean need_variant_tag,
long log_id) {
try {
profiler_.record("java_prepro_0");
InferenceRequest req = _packInferenceRequest(feed_batch, fetch);
InferenceRequest req = _packInferenceRequest(
feed_batch, fetch, log_id);
profiler_.record("java_prepro_1");
profiler_.record("java_client_infer_0");
......@@ -415,7 +492,7 @@ public class Client {
= new ArrayList<Map.Entry<String, HashMap<String, INDArray>>>(
ensemble_result.entrySet());
if (list.size() != 1) {
System.out.format("predict failed: please use ensemble_predict impl.\n");
System.out.format("Failed to predict: please use ensemble_predict impl.\n");
return null;
}
profiler_.record("java_postpro_1");
......@@ -423,7 +500,7 @@ public class Client {
return list.get(0).getValue();
} catch (StatusRuntimeException e) {
System.out.format("predict failed: %s\n", e.toString());
System.out.format("Failed to predict: %s\n", e.toString());
return null;
}
}
......@@ -432,9 +509,18 @@ public class Client {
List<HashMap<String, INDArray>> feed_batch,
Iterable<String> fetch,
Boolean need_variant_tag) {
return ensemble_predict(feed_batch, fetch, need_variant_tag, 0);
}
public Map<String, HashMap<String, INDArray>> ensemble_predict(
List<HashMap<String, INDArray>> feed_batch,
Iterable<String> fetch,
Boolean need_variant_tag,
long log_id) {
try {
profiler_.record("java_prepro_0");
InferenceRequest req = _packInferenceRequest(feed_batch, fetch);
InferenceRequest req = _packInferenceRequest(
feed_batch, fetch, log_id);
profiler_.record("java_prepro_1");
profiler_.record("java_client_infer_0");
......@@ -449,7 +535,7 @@ public class Client {
return ensemble_result;
} catch (StatusRuntimeException e) {
System.out.format("predict failed: %s\n", e.toString());
System.out.format("Failed to predict: %s\n", e.toString());
return null;
}
}
......@@ -458,7 +544,16 @@ public class Client {
List<HashMap<String, INDArray>> feed_batch,
Iterable<String> fetch,
Boolean need_variant_tag) {
InferenceRequest req = _packInferenceRequest(feed_batch, fetch);
return asyn_predict(feed_batch, fetch, need_variant_tag, 0);
}
public PredictFuture asyn_predict(
List<HashMap<String, INDArray>> feed_batch,
Iterable<String> fetch,
Boolean need_variant_tag,
long log_id) {
InferenceRequest req = _packInferenceRequest(
feed_batch, fetch, log_id);
ListenableFuture<InferenceResponse> future = futureStub_.inference(req);
PredictFuture predict_future = new PredictFuture(future,
(InferenceResponse resp) -> {
......
java/src/main/proto/multi_lang_general_model_service.proto
浏览文件 @ f3953aa5
......@@ -37,6 +37,7 @@ message InferenceRequest {
repeated string feed_var_names = 2;
repeated string fetch_var_names = 3;
required bool is_python = 4 [ default = false ];
required uint64 log_id = 5 [ default = 0 ];
};
message InferenceResponse {
......
python/examples/blazeface/README.md
浏览文件 @ f3953aa5
......@@ -3,7 +3,7 @@
## Get Model
```
python -m paddle_serving_app.package --get_model blazeface
tar -xzvf blazeface.tar.gz
tar -xf blazeface.tar.gz
```
## RPC Service
......
python/examples/ocr/README.md
浏览文件 @ f3953aa5
......@@ -21,8 +21,13 @@ tar xf test_imgs.tar
### Start Service
```
#choose one of cpu/gpu commands as following
#for cpu user
python -m paddle_serving_server.serve --model ocr_det_model --port 9293
python ocr_web_server.py cpu
#for gpu user
python -m paddle_serving_server_gpu.serve --model ocr_det_model --port 9293 --gpu_id 0
python ocr_web_server.py
python ocr_web_server.py gpu
```
### Client Prediction
......@@ -33,7 +38,11 @@ If you want a faster web service, please try Web Debugger Service
## Web Debugger Service
```
python ocr_debugger_server.py
#choose one of cpu/gpu commands as following
#for cpu user
python ocr_debugger_server.py cpu
#for gpu user
python ocr_debugger_server.py gpu
```
## Web Debugger Client Prediction
......@@ -54,15 +63,17 @@ Dataset: RCTW 500 sample images
| Serving web service | 8.69 | 13.41 | 109.97 | 2.82 | 87.76 | 4.29 | 3.98 | 78.51 | 3.66 | 4.12 | 181.02 | 136.49 | 317.51 |
| Serving Debugger web service | 8.73 | 16.42 | 115.27 | 2.93 | 20.63 | 3.97 | 4.48 | 13.84 | 3.60 | 6.91 | 49.45 | 147.33 | 196.78 |
## Appendix: Det or Rec only
## Appendix: For Users who want to launch Det or Rec only
if you are going to detect images not recognize it or directly recognize the words from images. We also provide Det and Rec server for you.
### Det Server
```
python det_web_server.py
python det_web_server.py cpu #for cpu user
python det_web_server.py gpu #for gpu user
#or
python det_debugger_server.py
python det_debugger_server.py cpu #for cpu user
python det_debugger_server.py gpu #for gpu user
```
### Det Client
......@@ -75,9 +86,11 @@ python ocr_web_client.py
### Rec Server
```
python rec_web_server.py
python rec_web_server.py cpu #for cpu user
python rec_web_server.py gpu #for gpu user
#or
python rec_debugger_server.py
python rec_debugger_server.py cpu #for cpu user
python rec_debugger_server.py gpu #for gpu user
```
### Rec Client
......
python/examples/ocr/README_CN.md
浏览文件 @ f3953aa5
......@@ -15,19 +15,18 @@ wget --no-check-certificate https://paddle-serving.bj.bcebos.com/ocr/test_imgs.t
tar xf test_imgs.tar
```
### 客户端预测
```
python ocr_rpc_client.py
```
## Web Service服务
### 启动服务
```
#根据CPU/GPU设备选择一种启动方式
#for cpu user
python -m paddle_serving_server.serve --model ocr_det_model --port 9293
python ocr_web_server.py cpu
#for gpu user
python -m paddle_serving_server_gpu.serve --model ocr_det_model --port 9293 --gpu_id 0
python ocr_web_server.py
python ocr_web_server.py gpu
```
### 启动客户端
......@@ -38,7 +37,11 @@ python ocr_web_client.py
如果用户需要更快的执行速度，请尝试Debugger版Web服务
## 启动Debugger版Web服务
```
python ocr_debugger_server.py
#根据CPU/GPU设备选择一种启动方式
#for cpu user
python ocr_debugger_server.py cpu
#for gpu user
python ocr_debugger_server.py gpu
```
## 启动客户端
......@@ -66,9 +69,11 @@ GPU: Nvidia Tesla V100单卡
### 启动检测服务
```
python det_web_server.py
python det_web_server.py cpu #for cpu user
python det_web_server.py gpu #for gpu user
#or
python det_debugger_server.py
python det_debugger_server.py cpu #for cpu user
python det_debugger_server.py gpu #for gpu user
```
### 检测服务客户端
......@@ -81,9 +86,11 @@ python ocr_web_client.py
### 启动识别服务
```
python rec_web_server.py
python rec_web_server.py cpu #for cpu user
python rec_web_server.py gpu #for gpu user
#or
python rec_debugger_server.py
python rec_debugger_server.py cpu #for cpu user
python rec_debugger_server.py gpu #for gpu user
```
### 识别服务客户端
......
python/examples/ocr/det_debugger_server.py
浏览文件 @ f3953aa5
......@@ -21,7 +21,10 @@ from paddle_serving_client import Client
from paddle_serving_app.reader import Sequential, ResizeByFactor
from paddle_serving_app.reader import Div, Normalize, Transpose
from paddle_serving_app.reader import DBPostProcess, FilterBoxes
from paddle_serving_server_gpu.web_service import WebService
if sys.argv[1] == 'gpu':
from paddle_serving_server_gpu.web_service import WebService
elif sys.argv[1] == 'cpu':
from paddle_serving_server.web_service import WebService
import time
import re
import base64
......@@ -64,8 +67,12 @@ class OCRService(WebService):
ocr_service = OCRService(name="ocr")
ocr_service.load_model_config("ocr_det_model")
ocr_service.set_gpus("0")
ocr_service.prepare_server(workdir="workdir", port=9292, device="gpu", gpuid=0)
if sys.argv[1] == 'gpu':
ocr_service.set_gpus("0")
ocr_service.prepare_server(
workdir="workdir", port=9292, device="gpu", gpuid=0)
elif sys.argv[1] == 'cpu':
ocr_service.prepare_server(workdir="workdir", port=9292)
ocr_service.init_det()
ocr_service.run_debugger_service()
ocr_service.run_web_service()
python/examples/ocr/det_web_server.py
浏览文件 @ f3953aa5
......@@ -21,7 +21,10 @@ from paddle_serving_client import Client
from paddle_serving_app.reader import Sequential, ResizeByFactor
from paddle_serving_app.reader import Div, Normalize, Transpose
from paddle_serving_app.reader import DBPostProcess, FilterBoxes
from paddle_serving_server_gpu.web_service import WebService
if sys.argv[1] == 'gpu':
from paddle_serving_server_gpu.web_service import WebService
elif sys.argv[1] == 'cpu':
from paddle_serving_server.web_service import WebService
import time
import re
import base64
......@@ -65,8 +68,12 @@ class OCRService(WebService):
ocr_service = OCRService(name="ocr")
ocr_service.load_model_config("ocr_det_model")
ocr_service.set_gpus("0")
ocr_service.prepare_server(workdir="workdir", port=9292, device="gpu", gpuid=0)
if sys.argv[1] == 'gpu':
ocr_service.set_gpus("0")
ocr_service.prepare_server(
workdir="workdir", port=9292, device="gpu", gpuid=0)
elif sys.argv[1] == 'cpu':
ocr_service.prepare_server(workdir="workdir", port=9292, device="cpu")
ocr_service.init_det()
ocr_service.run_rpc_service()
ocr_service.run_web_service()
python/examples/ocr/ocr_debugger_server.py
浏览文件 @ f3953aa5
......@@ -22,7 +22,10 @@ from paddle_serving_client import Client
from paddle_serving_app.reader import Sequential, URL2Image, ResizeByFactor
from paddle_serving_app.reader import Div, Normalize, Transpose
from paddle_serving_app.reader import DBPostProcess, FilterBoxes, GetRotateCropImage, SortedBoxes
from paddle_serving_server_gpu.web_service import WebService
if sys.argv[1] == 'gpu':
from paddle_serving_server_gpu.web_service import WebService
elif sys.argv[1] == 'cpu':
from paddle_serving_server.web_service import WebService
from paddle_serving_app.local_predict import Debugger
import time
import re
......@@ -37,8 +40,12 @@ class OCRService(WebService):
(2, 0, 1))
])
self.det_client = Debugger()
if sys.argv[1] == 'gpu':
self.det_client.load_model_config(
det_model_config, gpu=True, profile=False)
elif sys.argv[1] == 'cpu':
self.det_client.load_model_config(
det_model_config, gpu=False, profile=False)
self.ocr_reader = OCRReader()
def preprocess(self, feed=[], fetch=[]):
......@@ -99,5 +106,8 @@ ocr_service = OCRService(name="ocr")
ocr_service.load_model_config("ocr_rec_model")
ocr_service.prepare_server(workdir="workdir", port=9292)
ocr_service.init_det_debugger(det_model_config="ocr_det_model")
ocr_service.run_debugger_service(gpu=True)
if sys.argv[1] == 'gpu':
ocr_service.run_debugger_service(gpu=True)
elif sys.argv[1] == 'cpu':
ocr_service.run_debugger_service()
ocr_service.run_web_service()
python/examples/ocr/ocr_web_server.py
浏览文件 @ f3953aa5
......@@ -22,7 +22,10 @@ from paddle_serving_client import Client
from paddle_serving_app.reader import Sequential, URL2Image, ResizeByFactor
from paddle_serving_app.reader import Div, Normalize, Transpose
from paddle_serving_app.reader import DBPostProcess, FilterBoxes, GetRotateCropImage, SortedBoxes
from paddle_serving_server_gpu.web_service import WebService
if sys.argv[1] == 'gpu':
from paddle_serving_server_gpu.web_service import WebService
elif sys.argv[1] == 'cpu':
from paddle_serving_server.web_service import WebService
import time
import re
import base64
......@@ -90,8 +93,12 @@ class OCRService(WebService):
ocr_service = OCRService(name="ocr")
ocr_service.load_model_config("ocr_rec_model")
ocr_service.set_gpus("0")
ocr_service.prepare_server(workdir="workdir", port=9292, device="gpu", gpuid=0)
if sys.argv[1] == 'gpu':
ocr_service.set_gpus("0")
ocr_service.prepare_server(
workdir="workdir", port=9292, device="gpu", gpuid=0)
elif sys.argv[1] == 'cpu':
ocr_service.prepare_server(workdir="workdir", port=9292)
ocr_service.init_det_client(
det_port=9293,
det_client_config="ocr_det_client/serving_client_conf.prototxt")
......
python/examples/ocr/rec_web_server.py
浏览文件 @ f3953aa5
......@@ -22,7 +22,10 @@ from paddle_serving_client import Client
from paddle_serving_app.reader import Sequential, URL2Image, ResizeByFactor
from paddle_serving_app.reader import Div, Normalize, Transpose
from paddle_serving_app.reader import DBPostProcess, FilterBoxes, GetRotateCropImage, SortedBoxes
from paddle_serving_server_gpu.web_service import WebService
if sys.argv[1] == 'gpu':
from paddle_serving_server_gpu.web_service import WebService
elif sys.argv[1] == 'cpu':
from paddle_serving_server.web_service import WebService
import time
import re
import base64
......@@ -64,8 +67,12 @@ class OCRService(WebService):
ocr_service = OCRService(name="ocr")
ocr_service.load_model_config("ocr_rec_model")
ocr_service.set_gpus("0")
ocr_service.init_rec()
ocr_service.prepare_server(workdir="workdir", port=9292, device="gpu", gpuid=0)
if sys.argv[1] == 'gpu':
ocr_service.set_gpus("0")
ocr_service.prepare_server(
workdir="workdir", port=9292, device="gpu", gpuid=0)
elif sys.argv[1] == 'cpu':
ocr_service.prepare_server(workdir="workdir", port=9292, device="cpu")
ocr_service.run_rpc_service()
ocr_service.run_web_service()
python/examples/pipeline/imdb_model_ensemble/config.yml
浏览文件 @ f3953aa5
port: 18080
worker_num: 1
worker_num: 4
build_dag_each_worker: false
dag:
is_thread_op: true
is_thread_op: false
client_type: brpc
retry: 1
use_profile: false
tracer:
interval_s: 10
python/examples/pipeline/imdb_model_ensemble/test_pipeline_client.py
浏览文件 @ f3953aa5
......@@ -11,7 +11,7 @@
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
from paddle_serving_client.pipeline import PipelineClient
from paddle_serving_server.pipeline import PipelineClient
import numpy as np
client = PipelineClient()
......@@ -20,12 +20,16 @@ client.connect(['127.0.0.1:18080'])
words = 'i am very sad | 0'
futures = []
for i in range(100):
for i in range(4):
futures.append(
client.predict(
feed_dict={"words": words}, fetch=["prediction"], asyn=True))
feed_dict={"words": words},
fetch=["prediction"],
asyn=True,
profile=False))
for f in futures:
res = f.result()
if res["ecode"] != 0:
print("predict failed: {}".format(res))
print(res)
python/examples/pipeline/imdb_model_ensemble/test_pipeline_server.py
浏览文件 @ f3953aa5
......@@ -12,18 +12,21 @@
# See the License for the specific language governing permissions and
# limitations under the License.
# pylint: disable=doc-string-missing
from paddle_serving_server.pipeline import Op, RequestOp, ResponseOp
from paddle_serving_server.pipeline import PipelineServer
from paddle_serving_server.pipeline.proto import pipeline_service_pb2
from paddle_serving_server.pipeline.channel import ChannelDataEcode
import numpy as np
import logging
from paddle_serving_app.reader import IMDBDataset
logging.basicConfig(level=logging.DEBUG)
import logging
_LOGGER = logging.getLogger()
user_handler = logging.StreamHandler()
user_handler.setLevel(logging.INFO)
user_handler.setFormatter(
logging.Formatter(
"%(levelname)s %(asctime)s [%(filename)s:%(lineno)d] %(message)s"))
_LOGGER.addHandler(user_handler)
class ImdbRequestOp(RequestOp):
......@@ -76,7 +79,9 @@ bow_op = Op(name="bow",
client_config="imdb_bow_client_conf/serving_client_conf.prototxt",
concurrency=1,
timeout=-1,
retry=1)
retry=1,
batch_size=3,
auto_batching_timeout=1000)
cnn_op = Op(name="cnn",
input_ops=[read_op],
server_endpoints=["127.0.0.1:9292"],
......@@ -84,13 +89,17 @@ cnn_op = Op(name="cnn",
client_config="imdb_cnn_client_conf/serving_client_conf.prototxt",
concurrency=1,
timeout=-1,
retry=1)
retry=1,
batch_size=1,
auto_batching_timeout=None)
combine_op = CombineOp(
name="combine",
input_ops=[bow_op, cnn_op],
concurrency=5,
concurrency=1,
timeout=-1,
retry=1)
retry=1,
batch_size=2,
auto_batching_timeout=None)
# fetch output of bow_op
# response_op = ImdbResponseOp(input_ops=[bow_op])
......
python/paddle_serving_client/__init__.py
浏览文件 @ f3953aa5
......@@ -233,7 +233,7 @@ class Client(object):
# key))
pass
def predict(self, feed=None, fetch=None, need_variant_tag=False):
def predict(self, feed=None, fetch=None, need_variant_tag=False, log_id=0):
self.profile_.record('py_prepro_0')
if feed is None or fetch is None:
......@@ -319,12 +319,12 @@ class Client(object):
res = self.client_handle_.numpy_predict(
float_slot_batch, float_feed_names, float_shape, int_slot_batch,
int_feed_names, int_shape, fetch_names, result_batch_handle,
self.pid)
self.pid, log_id)
elif self.has_numpy_input == False:
res = self.client_handle_.batch_predict(
float_slot_batch, float_feed_names, float_shape, int_slot_batch,
int_feed_names, int_shape, fetch_names, result_batch_handle,
self.pid)
self.pid, log_id)
else:
raise ValueError(
"Please make sure the inputs are all in list type or all in numpy.array type"
......@@ -347,6 +347,11 @@ class Client(object):
result_map[name] = result_batch_handle.get_int64_by_name(
mi, name)
shape = result_batch_handle.get_shape(mi, name)
if result_map[name].size == 0:
raise ValueError(
"Failed to fetch, maybe the type of [{}]"
" is wrong, please check the model file".format(
name))
result_map[name].shape = shape
if name in self.lod_tensor_set:
result_map["{}.lod".format(
......@@ -354,6 +359,11 @@ class Client(object):
elif self.fetch_names_to_type_[name] == float32_type:
result_map[name] = result_batch_handle.get_float_by_name(
mi, name)
if result_map[name].size == 0:
raise ValueError(
"Failed to fetch, maybe the type of [{}]"
" is wrong, please check the model file".format(
name))
shape = result_batch_handle.get_shape(mi, name)
result_map[name].shape = shape
if name in self.lod_tensor_set:
......@@ -364,6 +374,11 @@ class Client(object):
# result_map[name] will be py::array(numpy array)
result_map[name] = result_batch_handle.get_int32_by_name(
mi, name)
if result_map[name].size == 0:
raise ValueError(
"Failed to fetch, maybe the type of [{}]"
" is wrong, please check the model file".format(
name))
shape = result_batch_handle.get_shape(mi, name)
result_map[name].shape = shape
if name in self.lod_tensor_set:
......@@ -466,10 +481,11 @@ class MultiLangClient(object):
if var.is_lod_tensor:
self.lod_tensor_set_.add(var.alias_name)
def _pack_inference_request(self, feed, fetch, is_python):
def _pack_inference_request(self, feed, fetch, is_python, log_id):
req = multi_lang_general_model_service_pb2.InferenceRequest()
req.fetch_var_names.extend(fetch)
req.is_python = is_python
req.log_id = log_id
feed_batch = None
if isinstance(feed, dict):
feed_batch = [feed]
......@@ -602,12 +618,13 @@ class MultiLangClient(object):
fetch,
need_variant_tag=False,
asyn=False,
is_python=True):
is_python=True,
log_id=0):
if not asyn:
try:
self.profile_.record('py_prepro_0')
req = self._pack_inference_request(
feed, fetch, is_python=is_python)
feed, fetch, is_python=is_python, log_id=log_id)
self.profile_.record('py_prepro_1')
self.profile_.record('py_client_infer_0')
......@@ -626,7 +643,8 @@ class MultiLangClient(object):
except grpc.RpcError as e:
return {"serving_status_code": e.code()}
else:
req = self._pack_inference_request(feed, fetch, is_python=is_python)
req = self._pack_inference_request(
feed, fetch, is_python=is_python, log_id=log_id)
call_future = self.stub_.Inference.future(
req, timeout=self.rpc_timeout_s_)
return MultiLangPredictFuture(
......
python/paddle_serving_server/__init__.py
浏览文件 @ f3953aa5
......@@ -518,6 +518,7 @@ class MultiLangServerServiceServicer(multi_lang_general_model_service_pb2_grpc.
feed_names = list(request.feed_var_names)
fetch_names = list(request.fetch_var_names)
is_python = request.is_python
log_id = request.log_id
feed_batch = []
for feed_inst in request.insts:
feed_dict = {}
......@@ -546,7 +547,7 @@ class MultiLangServerServiceServicer(multi_lang_general_model_service_pb2_grpc.
data.shape = list(feed_inst.tensor_array[idx].shape)
feed_dict[name] = data
feed_batch.append(feed_dict)
return feed_batch, fetch_names, is_python
return feed_batch, fetch_names, is_python, log_id
def _pack_inference_response(self, ret, fetch_names, is_python):
resp = multi_lang_general_model_service_pb2.InferenceResponse()
......@@ -598,10 +599,13 @@ class MultiLangServerServiceServicer(multi_lang_general_model_service_pb2_grpc.
return resp
def Inference(self, request, context):
feed_dict, fetch_names, is_python = self._unpack_inference_request(
request)
feed_dict, fetch_names, is_python, log_id = \
self._unpack_inference_request(request)
ret = self.bclient_.predict(
feed=feed_dict, fetch=fetch_names, need_variant_tag=True)
feed=feed_dict,
fetch=fetch_names,
need_variant_tag=True,
log_id=log_id)
return self._pack_inference_response(ret, fetch_names, is_python)
def GetClientConfig(self, request, context):
......
python/paddle_serving_server/web_service.py
浏览文件 @ f3953aa5
......@@ -91,6 +91,8 @@ class WebService(object):
request.json["fetch"])
if isinstance(feed, dict) and "fetch" in feed:
del feed["fetch"]
if len(feed) == 0:
raise ValueError("empty input")
fetch_map = self.client.predict(feed=feed, fetch=fetch)
result = self.postprocess(
feed=request.json["feed"], fetch=fetch, fetch_map=fetch_map)
......@@ -122,6 +124,32 @@ class WebService(object):
self.app_instance = app_instance
def run_debugger_service(self):
import socket
localIP = socket.gethostbyname(socket.gethostname())
print("web service address:")
print("http://{}:{}/{}/prediction".format(localIP, self.port,
self.name))
app_instance = Flask(__name__)
@app_instance.before_first_request
def init():
self._launch_local_predictor()
service_name = "/" + self.name + "/prediction"
@app_instance.route(service_name, methods=["POST"])
def run():
return self.get_prediction(request)
self.app_instance = app_instance
def _launch_local_predictor(self):
from paddle_serving_app.local_predict import Debugger
self.client = Debugger()
self.client.load_model_config(
"{}".format(self.model_config), gpu=False, profile=False)
def run_web_service(self):
self.app_instance.run(host="0.0.0.0",
port=self.port,
......
python/paddle_serving_server_gpu/__init__.py
浏览文件 @ f3953aa5
......@@ -578,6 +578,7 @@ class MultiLangServerServiceServicer(multi_lang_general_model_service_pb2_grpc.
feed_names = list(request.feed_var_names)
fetch_names = list(request.fetch_var_names)
is_python = request.is_python
log_id = request.log_id
feed_batch = []
for feed_inst in request.insts:
feed_dict = {}
......@@ -606,7 +607,7 @@ class MultiLangServerServiceServicer(multi_lang_general_model_service_pb2_grpc.
data.shape = list(feed_inst.tensor_array[idx].shape)
feed_dict[name] = data
feed_batch.append(feed_dict)
return feed_batch, fetch_names, is_python
return feed_batch, fetch_names, is_python, log_id
def _pack_inference_response(self, ret, fetch_names, is_python):
resp = multi_lang_general_model_service_pb2.InferenceResponse()
......@@ -659,10 +660,13 @@ class MultiLangServerServiceServicer(multi_lang_general_model_service_pb2_grpc.
return resp
def Inference(self, request, context):
feed_dict, fetch_names, is_python = self._unpack_inference_request(
request)
feed_dict, fetch_names, is_python, log_id \
= self._unpack_inference_request(request)
ret = self.bclient_.predict(
feed=feed_dict, fetch=fetch_names, need_variant_tag=True)
feed=feed_dict,
fetch=fetch_names,
need_variant_tag=True,
log_id=log_id)
return self._pack_inference_response(ret, fetch_names, is_python)
def GetClientConfig(self, request, context):
......
python/pipeline/__init__.py
浏览文件 @ f3953aa5
......@@ -11,7 +11,7 @@
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
import logger # this module must be the first to import
from operator import Op, RequestOp, ResponseOp
from pipeline_server import PipelineServer
from pipeline_client import PipelineClient
......
python/pipeline/analyse.py
浏览文件 @ f3953aa5
......@@ -17,7 +17,7 @@ import copy
import re
import logging
_LOGGER = logging.getLogger()
_LOGGER = logging.getLogger(__name__)
class Analyst(object):
......@@ -69,7 +69,7 @@ class Analyst(object):
with open(self._profile_file) as f:
for line in f.readlines():
line = line.strip().split("\t")
if line[0] == "PROFILE":
if line[0] == "PROFILE" and len(line) >= 3:
trace_list = self._prase_line(line[1], line[2], counter)
counter += 1
for trace in trace_list:
......@@ -164,7 +164,7 @@ class OpAnalyst(object):
def add(self, name_str, ts_list):
if self._close:
_LOGGER.error("OpAnalyst is closed.")
_LOGGER.error("Failed to add item: OpAnalyst is closed.")
return
op_name, curr_idx, step = self._parse(name_str)
if op_name not in self.op_time_list_dict:
......
python/pipeline/channel.py
浏览文件 @ f3953aa5
......@@ -12,6 +12,7 @@
# See the License for the specific language governing permissions and
# limitations under the License.
# pylint: disable=doc-string-missing
from time import time as _time
import threading
import multiprocessing
import multiprocessing.queues
......@@ -25,9 +26,10 @@ else:
import numpy as np
import logging
import enum
import os
import copy
_LOGGER = logging.getLogger()
_LOGGER = logging.getLogger(__name__)
class ChannelDataEcode(enum.Enum):
......@@ -68,21 +70,25 @@ class ChannelData(object):
'''
if ecode is not None:
if data_id is None or error_info is None:
raise ValueError("data_id and error_info cannot be None")
_LOGGER.critical("Failed to generate ChannelData: data_id"
" and error_info cannot be None")
os._exit(-1)
datatype = ChannelDataType.ERROR.value
else:
if datatype == ChannelDataType.CHANNEL_NPDATA.value:
ecode, error_info = ChannelData.check_npdata(npdata)
if ecode != ChannelDataEcode.OK.value:
datatype = ChannelDataType.ERROR.value
_LOGGER.error(error_info)
_LOGGER.error("(logid={}) {}".format(data_id, error_info))
elif datatype == ChannelDataType.DICT.value:
ecode, error_info = ChannelData.check_dictdata(dictdata)
if ecode != ChannelDataEcode.OK.value:
datatype = ChannelDataType.ERROR.value
_LOGGER.error(error_info)
_LOGGER.error("(logid={}) {}".format(data_id, error_info))
else:
raise ValueError("datatype not match")
_LOGGER.critical("(logid={}) datatype not match".format(
data_id))
os._exit(-1)
self.datatype = datatype
self.npdata = npdata
self.dictdata = dictdata
......@@ -106,16 +112,26 @@ class ChannelData(object):
for sample in dictdata:
if not isinstance(sample, dict):
ecode = ChannelDataEcode.TYPE_ERROR.value
error_info = "the value of data must " \
"be dict, but get {}.".format(type(sample))
error_info = "Failed to check data: the type of " \
"data must be dict, but get {}.".format(type(sample))
break
elif not isinstance(dictdata, dict):
# batch size = 1
ecode = ChannelDataEcode.TYPE_ERROR.value
error_info = "the value of data must " \
error_info = "Failed to check data: the type of data must " \
"be dict, but get {}.".format(type(dictdata))
return ecode, error_info
@staticmethod
def check_batch_npdata(batch):
ecode = ChannelDataEcode.OK.value
error_info = None
for npdata in batch:
ecode, error_info = ChannelData.check_npdata(npdata)
if ecode != ChannelDataEcode.OK.value:
break
return ecode, error_info
@staticmethod
def check_npdata(npdata):
ecode = ChannelDataEcode.OK.value
......@@ -125,27 +141,30 @@ class ChannelData(object):
for sample in npdata:
if not isinstance(sample, dict):
ecode = ChannelDataEcode.TYPE_ERROR.value
error_info = "the value of data must " \
"be dict, but get {}.".format(type(sample))
error_info = "Failed to check data: the " \
"value of data must be dict, but get {}.".format(
type(sample))
break
for _, value in sample.items():
if not isinstance(value, np.ndarray):
ecode = ChannelDataEcode.TYPE_ERROR.value
error_info = "the value of data must " \
"be np.ndarray, but get {}.".format(type(value))
error_info = "Failed to check data: the" \
" value of data must be np.ndarray, but get {}.".format(
type(value))
return ecode, error_info
elif isinstance(npdata, dict):
# batch_size = 1
for _, value in npdata.items():
if not isinstance(value, np.ndarray):
ecode = ChannelDataEcode.TYPE_ERROR.value
error_info = "the value of data must " \
"be np.ndarray, but get {}.".format(type(value))
error_info = "Failed to check data: the value " \
"of data must be np.ndarray, but get {}.".format(
type(value))
break
else:
ecode = ChannelDataEcode.TYPE_ERROR.value
error_info = "the value of data must " \
"be dict, but get {}.".format(type(npdata))
error_info = "Failed to check data: the value of data " \
"must be dict, but get {}.".format(type(npdata))
return ecode, error_info
def parse(self):
......@@ -157,9 +176,19 @@ class ChannelData(object):
# return dict
feed = self.dictdata
else:
raise TypeError("Error type({}) in datatype.".format(self.datatype))
_LOGGER.critical("Failed to parse channeldata: error " \
"type({}) in datatype.".format(self.datatype))
os._exit(-1)
return feed
def __cmp__(self, other):
if self.id < other.id:
return -1
elif self.id == other.id:
return 0
else:
return 1
def __str__(self):
return "type[{}], ecode[{}], id[{}]".format(
ChannelDataType(self.datatype).name, self.ecode, self.id)
......@@ -175,7 +204,7 @@ class ProcessChannel(object):
Only when all types of Ops get the data of the same ID,
the data will be poped; The Op of the same type will not
get the data of the same ID.
3. (TODO) Timeout and BatchSize are not fully supported.
3. Function front support timeout param to make auto-batching.
Note:
1. The ID of the data in the channel must be different.
......@@ -194,16 +223,15 @@ class ProcessChannel(object):
maintains the data obtained from queue.
"""
def __init__(self, manager, name=None, maxsize=0, timeout=None):
def __init__(self, manager, name=None, maxsize=0):
# For queue multiprocess: after putting an object on
# an empty queue there may be an infinitessimal delay
# before the queue's :meth:`~Queue.empty`
# see more:
# - https://bugs.python.org/issue18277
# - https://hg.python.org/cpython/rev/860fc6a2bd21
self._que = manager.Queue(maxsize=maxsize)
self._que = manager.PriorityQueue(maxsize=maxsize)
self._maxsize = maxsize
self._timeout = timeout
self.name = name
self._stop = manager.Value('i', 0)
......@@ -219,6 +247,12 @@ class ProcessChannel(object):
self._base_cursor = manager.Value('i', 0)
self._output_buf = manager.list()
def get_maxsize(self):
return self._maxsize
def size(self):
return self._que.qsize()
def get_producers(self):
return self._producers
......@@ -228,37 +262,41 @@ class ProcessChannel(object):
def _log(self, info_str):
return "[{}] {}".format(self.name, info_str)
def debug(self):
return self._log("p: {}, c: {}".format(self.get_producers(),
self.get_consumers()))
def add_producer(self, op_name):
""" not thread safe, and can only be called during initialization. """
if op_name in self._producers:
raise ValueError(
self._log("producer({}) is already in channel".format(op_name)))
_LOGGER.critical(
self._log("Failed to add producer: producer({})" \
" is already in channel".format(op_name)))
os._exit(-1)
self._producers.append(op_name)
_LOGGER.debug(self._log("Succ add a producer: {}".format(op_name)))
def add_consumer(self, op_name):
""" not thread safe, and can only be called during initialization. """
if op_name in self._consumer_cursors:
raise ValueError(
self._log("consumer({}) is already in channel".format(op_name)))
_LOGGER.critical(
self._log("Failed to add consumer: consumer({})" \
" is already in channel".format(op_name)))
os._exit(-1)
self._consumer_cursors[op_name] = 0
if self._cursor_count.get(0) is None:
self._cursor_count[0] = 0
self._cursor_count[0] += 1
_LOGGER.debug(self._log("Succ add a consumer: {}".format(op_name)))
def push(self, channeldata, op_name=None):
_LOGGER.debug(
self._log("{} try to push data: {}".format(op_name,
channeldata.__str__())))
self._log("(logid={}) Op({}) Pushing data".format(channeldata.id,
op_name)))
if len(self._producers) == 0:
raise Exception(
_LOGGER.critical(
self._log(
"expected number of producers to be greater than 0, but the it is 0."
))
"(logid={}) Op({}) Failed to push data: expected number"
" of producers to be greater than 0, but the it is 0.".
format(channeldata.id, op_name)))
os._exit(-1)
elif len(self._producers) == 1:
with self._cv:
while self._stop.value == 0:
......@@ -269,23 +307,23 @@ class ProcessChannel(object):
self._cv.wait()
if self._stop.value == 1:
raise ChannelStopError()
_LOGGER.debug(
self._log("{} channel size: {}".format(op_name,
self._que.qsize())))
self._cv.notify_all()
_LOGGER.debug(self._log("{} notify all".format(op_name)))
_LOGGER.debug(self._log("{} push data succ!".format(op_name)))
_LOGGER.debug(
self._log("(logid={}) Op({}) Pushed data into internal queue.".
format(channeldata.id, op_name)))
return True
elif op_name is None:
raise Exception(
_LOGGER.critical(
self._log(
"There are multiple producers, so op_name cannot be None."))
"(logid={}) Op({}) Failed to push data: there are multiple "
"producers, so op_name cannot be None.".format(
channeldata.id, op_name)))
os._exit(-1)
producer_num = len(self._producers)
data_id = channeldata.id
put_data = None
with self._cv:
_LOGGER.debug(self._log("{} get lock".format(op_name)))
if data_id not in self._input_buf:
self._input_buf[data_id] = {
name: None
......@@ -307,14 +345,12 @@ class ProcessChannel(object):
if put_data is None:
_LOGGER.debug(
self._log("{} push data succ, but not push to queue.".
format(op_name)))
self._log(
"(logid={}) Op({}) Pushed data into input_buffer.".
format(data_id, op_name)))
else:
while self._stop.value == 0:
try:
_LOGGER.debug(
self._log("{} push data succ: {}".format(
op_name, put_data.__str__())))
self._que.put(put_data, timeout=0)
break
except Queue.Empty:
......@@ -323,43 +359,59 @@ class ProcessChannel(object):
raise ChannelStopError()
_LOGGER.debug(
self._log("multi | {} push data succ!".format(op_name)))
self._log(
"(logid={}) Op({}) Pushed data into internal_queue.".
format(data_id, op_name)))
self._cv.notify_all()
return True
def front(self, op_name=None):
_LOGGER.debug(self._log("{} try to get data...".format(op_name)))
def front(self, op_name=None, timeout=None):
_LOGGER.debug(
self._log("Op({}) Getting data[?]; timeout(s)={}".format(op_name,
timeout)))
endtime = None
if timeout is not None:
if timeout <= 0:
timeout = None
else:
endtime = _time() + timeout
if len(self._consumer_cursors) == 0:
raise Exception(
_LOGGER.critical(
self._log(
"expected number of consumers to be greater than 0, but the it is 0."
))
"Op({}) Failed to get data: expected number of consumers to be " \
"greater than 0, but the it is 0.".format(op_name)))
os._exit(-1)
elif len(self._consumer_cursors) == 1:
resp = None
with self._cv:
while self._stop.value == 0 and resp is None:
try:
_LOGGER.debug(
self._log("{} try to get(with channel empty: {})".
format(op_name, self._que.empty())))
resp = self._que.get(timeout=0)
break
except Queue.Empty:
if timeout is not None:
remaining = endtime - _time()
if remaining <= 0.0:
_LOGGER.debug(
self._log(
"{} wait for empty queue(with channel empty: {})".
format(op_name, self._que.empty())))
self._log("Op({}) Failed to get data: "
"timeout".format(op_name)))
raise ChannelTimeoutError()
self._cv.wait(remaining)
else:
self._cv.wait()
if self._stop.value == 1:
raise ChannelStopError()
_LOGGER.debug(
self._log("{} get data succ: {}".format(op_name, resp.__str__(
))))
self._log("(logid={}) Op({}) Got data".format(resp.values()[0]
.id, op_name)))
return resp
elif op_name is None:
raise Exception(
_LOGGER.critical(
self._log(
"There are multiple consumers, so op_name cannot be None."))
"Op({}) Failed to get data: there are multiple consumers, "
"so op_name cannot be None.".format(op_name)))
os._exit(-1)
# In output_buf, different Ops (according to op_name) have different
# cursors. In addition, there is a base_cursor. Their difference is
......@@ -376,23 +428,24 @@ class ProcessChannel(object):
# it is necessary to obtain a data from queue and add it to output_buf.
while self._stop.value == 0 and self._consumer_cursors[
op_name] - self._base_cursor.value >= len(self._output_buf):
_LOGGER.debug(
self._log(
"({}) B self._consumer_cursors: {}, self._base_cursor: {}, len(self._output_buf): {}".
format(op_name, self._consumer_cursors,
self._base_cursor.value, len(self._output_buf))))
try:
_LOGGER.debug(
self._log("{} try to get(with channel size: {})".format(
op_name, self._que.qsize())))
channeldata = self._que.get(timeout=0)
self._output_buf.append(channeldata)
_LOGGER.debug(
self._log(
"(logid={}) Op({}) Pop ready item into output_buffer".
format(channeldata.values()[0].id, op_name)))
break
except Queue.Empty:
if timeout is not None:
remaining = endtime - _time()
if remaining <= 0.0:
_LOGGER.debug(
self._log(
"{} wait for empty queue(with channel size: {})".
format(op_name, self._que.qsize())))
self._log("Op({}) Failed to get data: timeout".
format(op_name)))
raise ChannelTimeoutError()
self._cv.wait(remaining)
else:
self._cv.wait()
if self._stop.value == 1:
raise ChannelStopError()
......@@ -401,7 +454,6 @@ class ProcessChannel(object):
base_cursor = self._base_cursor.value
data_idx = consumer_cursor - base_cursor
resp = self._output_buf[data_idx]
_LOGGER.debug(self._log("{} get data: {}".format(op_name, resp)))
self._cursor_count[consumer_cursor] -= 1
if consumer_cursor == base_cursor and self._cursor_count[
......@@ -413,6 +465,7 @@ class ProcessChannel(object):
self._base_cursor.value += 1
# to avoid cursor overflow
if self._base_cursor.value >= self._reset_max_cursor:
_LOGGER.info(self._log("Reset cursor in Channel"))
self._base_cursor.value -= self._reset_max_cursor
for name in self._consumer_cursors.keys():
self._consumer_cursors[name] -= self._reset_max_cursor
......@@ -430,25 +483,21 @@ class ProcessChannel(object):
self._cursor_count[new_consumer_cursor] = 0
self._cursor_count[new_consumer_cursor] += 1
_LOGGER.debug(
self._log(
"({}) A self._consumer_cursors: {}, self._base_cursor: {}, len(self._output_buf): {}".
format(op_name, self._consumer_cursors,
self._base_cursor.value, len(self._output_buf))))
_LOGGER.debug(self._log("{} notify all".format(op_name)))
self._cv.notify_all()
_LOGGER.debug(self._log("multi | {} get data succ!".format(op_name)))
return resp # reference, read only
_LOGGER.debug(
self._log("(logid={}) Op({}) Got data from output_buffer".format(
resp.values()[0].id, op_name)))
return resp
def stop(self):
_LOGGER.debug(self._log("stop."))
_LOGGER.info(self._log("stop."))
self._stop.value = 1
with self._cv:
self._cv.notify_all()
class ThreadChannel(Queue.Queue):
class ThreadChannel(Queue.PriorityQueue):
"""
(Thread version)The channel used for communication between Ops.
......@@ -458,7 +507,7 @@ class ThreadChannel(Queue.Queue):
Only when all types of Ops get the data of the same ID,
the data will be poped; The Op of the same type will not
get the data of the same ID.
3. (TODO) Timeout and BatchSize are not fully supported.
3. Function front support timeout param to make auto-batching.
Note:
1. The ID of the data in the channel must be different.
......@@ -477,10 +526,9 @@ class ThreadChannel(Queue.Queue):
maintains the data obtained from queue.
"""
def __init__(self, name=None, maxsize=-1, timeout=None):
def __init__(self, name=None, maxsize=-1):
Queue.Queue.__init__(self, maxsize=maxsize)
self._maxsize = maxsize
self._timeout = timeout
self.name = name
self._stop = False
......@@ -496,6 +544,12 @@ class ThreadChannel(Queue.Queue):
self._base_cursor = 0
self._output_buf = []
def get_maxsize(self):
return self._maxsize
def size(self):
return self.qsize()
def get_producers(self):
return self._producers
......@@ -505,37 +559,41 @@ class ThreadChannel(Queue.Queue):
def _log(self, info_str):
return "[{}] {}".format(self.name, info_str)
def debug(self):
return self._log("p: {}, c: {}".format(self.get_producers(),
self.get_consumers()))
def add_producer(self, op_name):
""" not thread safe, and can only be called during initialization. """
if op_name in self._producers:
raise ValueError(
self._log("producer({}) is already in channel".format(op_name)))
_LOGGER.critical(
self._log("Failed to add producer: producer({}) is "
"already in channel".format(op_name)))
os._exit(-1)
self._producers.append(op_name)
_LOGGER.debug(self._log("Succ add a producer: {}".format(op_name)))
def add_consumer(self, op_name):
""" not thread safe, and can only be called during initialization. """
if op_name in self._consumer_cursors:
raise ValueError(
self._log("consumer({}) is already in channel".format(op_name)))
_LOGGER.critical(
self._log("Failed to add consumer: consumer({}) is "
"already in channel".format(op_name)))
os._exit(-1)
self._consumer_cursors[op_name] = 0
if self._cursor_count.get(0) is None:
self._cursor_count[0] = 0
self._cursor_count[0] += 1
_LOGGER.debug(self._log("Succ add a consumer: {}".format(op_name)))
def push(self, channeldata, op_name=None):
_LOGGER.debug(
self._log("{} try to push data: {}".format(op_name,
channeldata.__str__())))
self._log("(logid={}) Op({}) Pushing data".format(channeldata.id,
op_name)))
if len(self._producers) == 0:
raise Exception(
_LOGGER.critical(
self._log(
"expected number of producers to be greater than 0, but the it is 0."
))
"(logid={}) Op({}) Failed to push data: expected number of "
"producers to be greater than 0, but the it is 0.".format(
channeldata.id, op_name)))
os._exit(-1)
elif len(self._producers) == 1:
with self._cv:
while self._stop is False:
......@@ -547,18 +605,22 @@ class ThreadChannel(Queue.Queue):
if self._stop:
raise ChannelStopError()
self._cv.notify_all()
_LOGGER.debug(self._log("{} push data succ!".format(op_name)))
_LOGGER.debug(
self._log("(logid={}) Op({}) Pushed data into internal_queue.".
format(channeldata.id, op_name)))
return True
elif op_name is None:
raise Exception(
_LOGGER.critical(
self._log(
"There are multiple producers, so op_name cannot be None."))
"(logid={}) Op({}) Failed to push data: there are multiple"
" producers, so op_name cannot be None.".format(
channeldata.id, op_name)))
os._exit(-1)
producer_num = len(self._producers)
data_id = channeldata.id
put_data = None
with self._cv:
_LOGGER.debug(self._log("{} get lock".format(op_name)))
if data_id not in self._input_buf:
self._input_buf[data_id] = {
name: None
......@@ -575,8 +637,9 @@ class ThreadChannel(Queue.Queue):
if put_data is None:
_LOGGER.debug(
self._log("{} push data succ, but not push to queue.".
format(op_name)))
self._log(
"(logid={}) Op({}) Pushed data into input_buffer.".
format(data_id, op_name)))
else:
while self._stop is False:
try:
......@@ -588,17 +651,29 @@ class ThreadChannel(Queue.Queue):
raise ChannelStopError()
_LOGGER.debug(
self._log("multi | {} push data succ!".format(op_name)))
self._log(
"(logid={}) Op({}) Pushed data into internal_queue.".
format(data_id, op_name)))
self._cv.notify_all()
return True
def front(self, op_name=None):
_LOGGER.debug(self._log("{} try to get data".format(op_name)))
def front(self, op_name=None, timeout=None):
_LOGGER.debug(
self._log("Op({}) Getting data[?]; timeout(s)={}".format(op_name,
timeout)))
endtime = None
if timeout is not None:
if timeout <= 0:
timeout = None
else:
endtime = _time() + timeout
if len(self._consumer_cursors) == 0:
raise Exception(
_LOGGER.critical(
self._log(
"expected number of consumers to be greater than 0, but the it is 0."
))
"Op({}) Failed to get data: expected number of consumers to be "
"greater than 0, but the it is 0.".format(op_name)))
os._exit(-1)
elif len(self._consumer_cursors) == 1:
resp = None
with self._cv:
......@@ -607,17 +682,29 @@ class ThreadChannel(Queue.Queue):
resp = self.get(timeout=0)
break
except Queue.Empty:
if timeout is not None:
remaining = endtime - _time()
if remaining <= 0.0:
_LOGGER.debug(
self._log(
"Op({}) Failed to get data: timeout".
format(op_name)))
raise ChannelTimeoutError()
self._cv.wait(remaining)
else:
self._cv.wait()
if self._stop:
raise ChannelStopError()
_LOGGER.debug(
self._log("{} get data succ: {}".format(op_name, resp.__str__(
))))
self._log("(logid={}) Op({}) Got data".format(resp.values()[0]
.id, op_name)))
return resp
elif op_name is None:
raise Exception(
self._log(
"There are multiple consumers, so op_name cannot be None."))
_LOGGER.critical(
self._log("Op({}) Failed to get data: there are multiple "
"consumers, so op_name cannot be None.".format(
op_name)))
os._exit(-1)
# In output_buf, different Ops (according to op_name) have different
# cursors. In addition, there is a base_cursor. Their difference is
......@@ -637,8 +724,21 @@ class ThreadChannel(Queue.Queue):
try:
channeldata = self.get(timeout=0)
self._output_buf.append(channeldata)
_LOGGER.debug(
self._log(
"(logid={}) Op({}) Pop ready item into output_buffer".
format(channeldata.values()[0].id, op_name)))
break
except Queue.Empty:
if timeout is not None:
remaining = endtime - _time()
if remaining <= 0.0:
_LOGGER.debug(
self._log("Op({}) Failed to get data: timeout".
format(op_name)))
raise ChannelTimeoutError()
self._cv.wait(remaining)
else:
self._cv.wait()
if self._stop:
raise ChannelStopError()
......@@ -659,6 +759,7 @@ class ThreadChannel(Queue.Queue):
self._base_cursor += 1
# to avoid cursor overflow
if self._base_cursor >= self._reset_max_cursor:
_LOGGER.info(self._log("Reset cursor in Channel"))
self._base_cursor -= self._reset_max_cursor
for name in self._consumer_cursors:
self._consumer_cursors[name] -= self._reset_max_cursor
......@@ -668,7 +769,6 @@ class ThreadChannel(Queue.Queue):
}
else:
resp = copy.deepcopy(self._output_buf[data_idx])
_LOGGER.debug(self._log("{} get data: {}".format(op_name, resp)))
self._consumer_cursors[op_name] += 1
new_consumer_cursor = self._consumer_cursors[op_name]
......@@ -678,16 +778,23 @@ class ThreadChannel(Queue.Queue):
self._cv.notify_all()
_LOGGER.debug(self._log("multi | {} get data succ!".format(op_name)))
_LOGGER.debug(
self._log("(logid={}) Op({}) Got data from output_buffer".format(
resp.values()[0].id, op_name)))
return resp
def stop(self):
_LOGGER.debug(self._log("stop."))
_LOGGER.info(self._log("stop."))
self._stop = True
with self._cv:
self._cv.notify_all()
class ChannelTimeoutError(RuntimeError):
def __init__(self):
pass
class ChannelStopError(RuntimeError):
def __init__(self):
pass
python/pipeline/dag.py
浏览文件 @ f3953aa5
......@@ -24,42 +24,45 @@ else:
raise Exception("Error Python version")
import os
import logging
import collections
from .operator import Op, RequestOp, ResponseOp, VirtualOp
from .channel import (ThreadChannel, ProcessChannel, ChannelData,
ChannelDataEcode, ChannelDataType, ChannelStopError)
from .profiler import TimeProfiler
from .util import NameGenerator
from .profiler import TimeProfiler, PerformanceTracer
from .util import NameGenerator, ThreadIdGenerator, PipelineProcSyncManager
from .proto import pipeline_service_pb2
_LOGGER = logging.getLogger()
_LOGGER = logging.getLogger(__name__)
class DAGExecutor(object):
def __init__(self, response_op, dag_config, show_info):
self._retry = dag_config.get('retry', 1)
def __init__(self, response_op, server_conf, worker_idx):
build_dag_each_worker = server_conf["build_dag_each_worker"]
server_worker_num = server_conf["worker_num"]
dag_conf = server_conf["dag"]
client_type = dag_config.get('client_type', 'brpc')
self._server_use_profile = dag_config.get('use_profile', False)
channel_size = dag_config.get('channel_size', 0)
self._is_thread_op = dag_config.get('is_thread_op', True)
self._retry = dag_conf["retry"]
client_type = dag_conf["client_type"]
self._server_use_profile = dag_conf["use_profile"]
channel_size = dag_conf["channel_size"]
self._is_thread_op = dag_conf["is_thread_op"]
if show_info and self._server_use_profile:
_LOGGER.info("================= PROFILER ================")
if self._is_thread_op:
_LOGGER.info("op: thread")
_LOGGER.info("profile mode: sync")
else:
_LOGGER.info("op: process")
_LOGGER.info("profile mode: asyn")
_LOGGER.info("-------------------------------------------")
tracer_conf = dag_conf["tracer"]
tracer_interval_s = tracer_conf["interval_s"]
self.name = "@G"
self.name = "@DAGExecutor"
self._profiler = TimeProfiler()
self._profiler.enable(True)
self._tracer = None
if tracer_interval_s >= 1:
self._tracer = PerformanceTracer(
self._is_thread_op, tracer_interval_s, server_worker_num)
self._dag = DAG(self.name, response_op, self._server_use_profile,
self._is_thread_op, client_type, channel_size,
show_info)
build_dag_each_worker, self._tracer)
(in_channel, out_channel, pack_rpc_func,
unpack_rpc_func) = self._dag.build()
self._dag.start()
......@@ -69,15 +72,23 @@ class DAGExecutor(object):
self._pack_rpc_func = pack_rpc_func
self._unpack_rpc_func = unpack_rpc_func
_LOGGER.debug(self._log(in_channel.debug()))
_LOGGER.debug(self._log(out_channel.debug()))
if self._tracer is not None:
self._tracer.start()
# generate id: data_id == request_id == log_id
base_counter = 0
gen_id_step = 1
if build_dag_each_worker:
base_counter = worker_idx
gen_id_step = server_worker_num
self._id_generator = ThreadIdGenerator(
max_id=1000000000000000000,
base_counter=base_counter,
step=gen_id_step)
self._id_lock = threading.Lock()
self._id_counter = 0
self._reset_max_id = 1000000000000000000
self._cv_pool = {}
self._cv_for_cv_pool = threading.Condition()
self._fetch_buffer = None
self._fetch_buffer = {}
self._recive_func = None
self._client_profile_key = "pipeline.profile"
......@@ -86,32 +97,38 @@ class DAGExecutor(object):
def start(self):
self._recive_func = threading.Thread(
target=DAGExecutor._recive_out_channel_func, args=(self, ))
self._recive_func.daemon = True
self._recive_func.start()
_LOGGER.debug("[DAG Executor] Start recive thread")
def stop(self):
self._dag.stop()
self._dag.join()
_LOGGER.info("[DAG Executor] Stop")
def _get_next_data_id(self):
with self._id_lock:
if self._id_counter >= self._reset_max_id:
self._id_counter -= self._reset_max_id
self._id_counter += 1
return self._id_counter - 1
data_id = self._id_generator.next()
cond_v = threading.Condition()
with self._cv_for_cv_pool:
self._cv_pool[data_id] = cond_v
self._fetch_buffer[data_id] = None
return data_id, cond_v
def _set_in_channel(self, in_channel):
if not isinstance(in_channel, (ThreadChannel, ProcessChannel)):
raise TypeError(
self._log('in_channel must be Channel type, but get {}'.format(
type(in_channel))))
_LOGGER.critical("[DAG Executor] Failed to set in_channel: "
"in_channel must be Channel type, but get {}".
format(type(in_channel)))
os._exit(-1)
in_channel.add_producer(self.name)
self._in_channel = in_channel
def _set_out_channel(self, out_channel):
if not isinstance(out_channel, (ThreadChannel, ProcessChannel)):
raise TypeError(
self._log('out_channel must be Channel type, but get {}'.format(
type(out_channel))))
_LOGGER.critical("[DAG Executor] Failed to set out_channel: "
"must be Channel type, but get {}".format(
type(out_channel)))
os._exit(-1)
out_channel.add_consumer(self.name)
self._out_channel = out_channel
......@@ -121,7 +138,7 @@ class DAGExecutor(object):
try:
channeldata_dict = self._out_channel.front(self.name)
except ChannelStopError:
_LOGGER.debug(self._log("stop."))
_LOGGER.info("[DAG Executor] Stop.")
with self._cv_for_cv_pool:
for data_id, cv in self._cv_pool.items():
closed_errror_data = ChannelData(
......@@ -129,46 +146,61 @@ class DAGExecutor(object):
error_info="dag closed.",
data_id=data_id)
with cv:
self._fetch_buffer = closed_errror_data
self._fetch_buffer[data_id] = closed_errror_data
cv.notify_all()
break
if len(channeldata_dict) != 1:
_LOGGER.error("out_channel cannot have multiple input ops")
_LOGGER.critical(
"[DAG Executor] Failed to fetch result: out_channel "
"cannot have multiple input ops")
os._exit(-1)
(_, channeldata), = channeldata_dict.items()
if not isinstance(channeldata, ChannelData):
raise TypeError(
self._log('data must be ChannelData type, but get {}'.
format(type(channeldata))))
_LOGGER.critical(
'[DAG Executor] Failed to fetch result: data in out_channel" \
" must be ChannelData type, but get {}'
.format(type(channeldata)))
os._exit(-1)
data_id = channeldata.id
_LOGGER.debug("recive thread fetch data: {}".format(data_id))
_LOGGER.debug("(logid={}) [recive thread] Fetched data".format(
data_id))
with self._cv_for_cv_pool:
cv = self._cv_pool[data_id]
with cv:
self._fetch_buffer = channeldata
cv.notify_all()
cond_v = self._cv_pool[data_id]
with cond_v:
self._fetch_buffer[data_id] = channeldata
cond_v.notify_all()
def _get_channeldata_from_fetch_buffer(self, data_id, cond_v):
ready_data = None
def _get_channeldata_from_fetch_buffer(self, data_id):
resp = None
cv = threading.Condition()
with cond_v:
with self._cv_for_cv_pool:
self._cv_pool[data_id] = cv
with cv:
cv.wait()
_LOGGER.debug("resp func get lock (data_id: {})".format(data_id))
resp = copy.deepcopy(self._fetch_buffer)
if self._fetch_buffer[data_id] is not None:
# The requested data is already ready
ready_data = self._fetch_buffer[data_id]
self._cv_pool.pop(data_id)
self._fetch_buffer.pop(data_id)
if ready_data is None:
# Wait for data ready
cond_v.wait()
with self._cv_for_cv_pool:
ready_data = self._fetch_buffer[data_id]
self._cv_pool.pop(data_id)
return resp
self._fetch_buffer.pop(data_id)
_LOGGER.debug("(logid={}) [resp thread] Got data".format(data_id))
return ready_data
def _pack_channeldata(self, rpc_request, data_id):
_LOGGER.debug(self._log('start inferce'))
dictdata = None
try:
dictdata = self._unpack_rpc_func(rpc_request)
except Exception as e:
_LOGGER.error(
"(logid={}) Failed to parse RPC request package: {}"
.format(data_id, e),
exc_info=True)
return ChannelData(
ecode=ChannelDataEcode.RPC_PACKAGE_ERROR.value,
error_info="rpc package error: {}".format(e),
......@@ -181,58 +213,86 @@ class DAGExecutor(object):
if key == self._client_profile_key:
profile_value = rpc_request.value[idx]
break
client_need_profile = (profile_value == self._client_profile_value)
_LOGGER.debug("(logid={}) Need profile in client: {}".format(
data_id, client_need_profile))
return ChannelData(
datatype=ChannelDataType.DICT.value,
dictdata=dictdata,
data_id=data_id,
client_need_profile=(
profile_value == self._client_profile_value))
client_need_profile=client_need_profile)
def call(self, rpc_request):
data_id = self._get_next_data_id()
if self._tracer is not None:
trace_buffer = self._tracer.data_buffer()
data_id, cond_v = self._get_next_data_id()
_LOGGER.info("(logid={}) Succ generate id".format(data_id))
start_call, end_call = None, None
if not self._is_thread_op:
self._profiler.record("call_{}#DAG-{}_0".format(data_id, data_id))
start_call = self._profiler.record("call_{}#DAG-{}_0".format(
data_id, data_id))
else:
self._profiler.record("call_{}#DAG_0".format(data_id))
start_call = self._profiler.record("call_{}#DAG_0".format(data_id))
_LOGGER.debug("(logid={}) Parsing RPC request package".format(data_id))
self._profiler.record("prepack_{}#{}_0".format(data_id, self.name))
req_channeldata = self._pack_channeldata(rpc_request, data_id)
self._profiler.record("prepack_{}#{}_1".format(data_id, self.name))
resp_channeldata = None
for i in range(self._retry):
_LOGGER.debug(self._log('push data'))
#self._profiler.record("push_{}#{}_0".format(data_id, self.name))
_LOGGER.debug("(logid={}) Pushing data into Graph engine".format(
data_id))
try:
self._in_channel.push(req_channeldata, self.name)
except ChannelStopError:
_LOGGER.debug(self._log("stop."))
_LOGGER.debug("[DAG Executor] Stop")
with self._cv_for_cv_pool:
self._cv_pool.pop(data_id)
return self._pack_for_rpc_resp(
ChannelData(
ecode=ChannelDataEcode.CLOSED_ERROR.value,
error_info="dag closed.",
data_id=data_id))
#self._profiler.record("push_{}#{}_1".format(data_id, self.name))
_LOGGER.debug(self._log('wait for infer'))
#self._profiler.record("fetch_{}#{}_0".format(data_id, self.name))
resp_channeldata = self._get_channeldata_from_fetch_buffer(data_id)
#self._profiler.record("fetch_{}#{}_1".format(data_id, self.name))
_LOGGER.debug("(logid={}) Wait for Graph engine...".format(data_id))
resp_channeldata = self._get_channeldata_from_fetch_buffer(data_id,
cond_v)
if resp_channeldata.ecode == ChannelDataEcode.OK.value:
_LOGGER.info("(logid={}) Succ predict".format(data_id))
break
else:
_LOGGER.error("(logid={}) Failed to predict: {}"
.format(data_id, resp_channeldata.error_info))
if resp_channeldata.ecode != ChannelDataEcode.TIMEOUT.value:
break
if i + 1 < self._retry:
_LOGGER.warn("retry({}): {}".format(
i + 1, resp_channeldata.error_info))
_LOGGER.warning("(logid={}) DAGExecutor retry({}/{})".format(
data_id, i + 1, self._retry))
_LOGGER.debug("(logid={}) Packing RPC response package".format(data_id))
self._profiler.record("postpack_{}#{}_0".format(data_id, self.name))
rpc_resp = self._pack_for_rpc_resp(resp_channeldata)
self._profiler.record("postpack_{}#{}_1".format(data_id, self.name))
if not self._is_thread_op:
self._profiler.record("call_{}#DAG-{}_1".format(data_id, data_id))
end_call = self._profiler.record("call_{}#DAG-{}_1".format(data_id,
data_id))
else:
self._profiler.record("call_{}#DAG_1".format(data_id))
#self._profiler.print_profile()
end_call = self._profiler.record("call_{}#DAG_1".format(data_id))
if self._tracer is not None:
trace_buffer.put({
"name": "DAG",
"id": data_id,
"succ": resp_channeldata.ecode == ChannelDataEcode.OK.value,
"actions": {
"call_{}".format(data_id): end_call - start_call,
},
})
profile_str = self._profiler.gen_profile_str()
if self._server_use_profile:
......@@ -250,25 +310,33 @@ class DAGExecutor(object):
return rpc_resp
def _pack_for_rpc_resp(self, channeldata):
_LOGGER.debug(self._log('get channeldata'))
try:
return self._pack_rpc_func(channeldata)
def _log(self, info_str):
return "[{}] {}".format(self.name, info_str)
except Exception as e:
_LOGGER.error(
"(logid={}) Failed to pack RPC response package: {}"
.format(channeldata.id, e),
exc_info=True)
resp = pipeline_service_pb2.Response()
resp.ecode = ChannelDataEcode.RPC_PACKAGE_ERROR.value
resp.error_info = "rpc package error: {}".format(e)
return resp
class DAG(object):
def __init__(self, request_name, response_op, use_profile, is_thread_op,
client_type, channel_size, show_info):
client_type, channel_size, build_dag_each_worker, tracer):
self._request_name = request_name
self._response_op = response_op
self._use_profile = use_profile
self._is_thread_op = is_thread_op
self._channel_size = channel_size
self._client_type = client_type
self._show_info = show_info
self._build_dag_each_worker = build_dag_each_worker
self._tracer = tracer
if not self._is_thread_op:
self._manager = multiprocessing.Manager()
self._manager = PipelineProcSyncManager()
_LOGGER.info("[DAG] Succ init")
def get_use_ops(self, response_op):
unique_names = set()
......@@ -288,8 +356,10 @@ class DAG(object):
used_ops.add(pred_op)
# check the name of op is globally unique
if pred_op.name in unique_names:
raise Exception("the name of Op must be unique: {}".
_LOGGER.critical("Failed to get used Ops: the"
" name of Op must be unique: {}".
format(pred_op.name))
os._exit(-1)
unique_names.add(pred_op.name)
return used_ops, succ_ops_of_use_op
......@@ -301,10 +371,13 @@ class DAG(object):
else:
channel = ProcessChannel(
self._manager, name=name_gen.next(), maxsize=self._channel_size)
_LOGGER.debug("[DAG] Generate channel: {}".format(channel.name))
return channel
def _gen_virtual_op(self, name_gen):
return VirtualOp(name=name_gen.next())
vir_op = VirtualOp(name=name_gen.next())
_LOGGER.debug("[DAG] Generate virtual_op: {}".format(vir_op.name))
return vir_op
def _topo_sort(self, used_ops, response_op, out_degree_ops):
out_degree_num = {
......@@ -318,7 +391,9 @@ class DAG(object):
if len(op.get_input_ops()) == 0:
zero_indegree_num += 1
if zero_indegree_num != 1:
raise Exception("DAG contains multiple input Ops")
_LOGGER.critical("Failed to topo sort: DAG contains "
"multiple RequestOps")
os._exit(-1)
last_op = response_op.get_input_ops()[0]
ques[que_idx].put(last_op)
......@@ -342,37 +417,47 @@ class DAG(object):
break
que_idx = (que_idx + 1) % 2
if sorted_op_num < len(used_ops):
raise Exception("not legal DAG")
_LOGGER.critical("Failed to topo sort: not legal DAG")
os._exit(-1)
return dag_views, last_op
def _build_dag(self, response_op):
if response_op is None:
raise Exception("response_op has not been set.")
_LOGGER.critical("Failed to build DAG: ResponseOp"
" has not been set.")
os._exit(-1)
used_ops, out_degree_ops = self.get_use_ops(response_op)
if self._show_info:
if not self._build_dag_each_worker:
_LOGGER.info("================= USED OP =================")
for op in used_ops:
if op.name != self._request_name:
_LOGGER.info(op.name)
_LOGGER.info("-------------------------------------------")
if len(used_ops) <= 1:
raise Exception(
"Besides RequestOp and ResponseOp, there should be at least one Op in DAG."
)
_LOGGER.critical(
"Failed to build DAG: besides RequestOp and ResponseOp, "
"there should be at least one Op in DAG.")
os._exit(-1)
if self._build_dag_each_worker:
_LOGGER.info("Because `build_dag_each_worker` mode is used, "
"Auto-batching is set to the default config: "
"batch_size=1, auto_batching_timeout=None")
for op in used_ops:
op.use_default_auto_batching_config()
dag_views, last_op = self._topo_sort(used_ops, response_op,
out_degree_ops)
dag_views = list(reversed(dag_views))
if self._show_info:
_LOGGER.info("================== DAG ====================")
if not self._build_dag_each_worker:
_LOGGER.debug("================== DAG ====================")
for idx, view in enumerate(dag_views):
_LOGGER.info("(VIEW {})".format(idx))
_LOGGER.debug("(VIEW {})".format(idx))
for op in view:
_LOGGER.info(" [{}]".format(op.name))
_LOGGER.debug(" [{}]".format(op.name))
for out_op in out_degree_ops[op.name]:
_LOGGER.info(" - {}".format(out_op.name))
_LOGGER.info("-------------------------------------------")
_LOGGER.debug(" - {}".format(out_op.name))
_LOGGER.debug("-------------------------------------------")
# create channels and virtual ops
virtual_op_name_gen = NameGenerator("vir")
......@@ -414,7 +499,6 @@ class DAG(object):
continue
channel = self._gen_channel(channel_name_gen)
channels.append(channel)
_LOGGER.debug("{} => {}".format(channel.name, op.name))
op.add_input_channel(channel)
pred_ops = pred_op_of_next_view_op[op.name]
if v_idx == 0:
......@@ -422,8 +506,6 @@ class DAG(object):
else:
# if pred_op is virtual op, it will use ancestors as producers to channel
for pred_op in pred_ops:
_LOGGER.debug("{} => {}".format(pred_op.name,
channel.name))
pred_op.add_output_channel(channel)
processed_op.add(op.name)
# find same input op to combine channel
......@@ -439,8 +521,6 @@ class DAG(object):
same_flag = False
break
if same_flag:
_LOGGER.debug("{} => {}".format(channel.name,
other_op.name))
other_op.add_input_channel(channel)
processed_op.add(other_op.name)
output_channel = self._gen_channel(channel_name_gen)
......@@ -458,14 +538,19 @@ class DAG(object):
actual_ops.append(op)
for c in channels:
_LOGGER.debug(c.debug())
_LOGGER.debug("Channel({}):\n\t- producers: {}\n\t- consumers: {}"
.format(c.name, c.get_producers(), c.get_consumers()))
return (actual_ops, channels, input_channel, output_channel, pack_func,
unpack_func)
def get_channels(self):
return self._channels
def build(self):
(actual_ops, channels, input_channel, output_channel, pack_func,
unpack_func) = self._build_dag(self._response_op)
_LOGGER.info("[DAG] Succ build DAG")
self._actual_ops = actual_ops
self._channels = channels
......@@ -474,18 +559,24 @@ class DAG(object):
self._pack_func = pack_func
self._unpack_func = unpack_func
if self._tracer is not None:
self._tracer.set_channels(self._channels)
return self._input_channel, self._output_channel, self._pack_func, self._unpack_func
def start(self):
self._threads_or_proces = []
for op in self._actual_ops:
op.use_profiler(self._use_profile)
op.set_tracer(self._tracer)
if self._is_thread_op:
self._threads_or_proces.extend(
op.start_with_thread(self._client_type))
else:
self._threads_or_proces.extend(
op.start_with_process(self._client_type))
_LOGGER.info("[DAG] start")
# not join yet
return self._threads_or_proces
......
python/pipeline/logger.py 0 → 100644
浏览文件 @ f3953aa5
# Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
import logging
import logging.config
import os
class SectionLevelFilter(object):
def __init__(self, levels):
self._levels = levels
def filter(self, logRecord):
return logRecord.levelno in self._levels
log_dir = "PipelineServingLogs"
if not os.path.exists(log_dir):
os.makedirs(log_dir)
logger_config = {
"version": 1,
"formatters": {
"normal_fmt": {
"format":
"%(levelname)s %(asctime)s [%(filename)s:%(lineno)d] %(message)s",
},
"tracer_fmt": {
"format": "%(asctime)s %(message)s",
},
},
"handlers": {
"f_pipeline.log": {
"class": "logging.FileHandler",
"level": "INFO",
"formatter": "normal_fmt",
"filename": os.path.join(log_dir, "pipeline.log"),
},
"f_pipeline.log.wf": {
"class": "logging.FileHandler",
"level": "WARNING",
"formatter": "normal_fmt",
"filename": os.path.join(log_dir, "pipeline.log.wf"),
},
"f_tracer.log": {
"class": "logging.FileHandler",
"level": "INFO",
"formatter": "tracer_fmt",
"filename": os.path.join(log_dir, "pipeline.tracer"),
},
},
"loggers": {
# propagate = True
".".join(__name__.split(".")[:-1] + ["profiler"]): {
"level": "INFO",
"handlers": ["f_tracer.log"],
},
},
"root": {
"level": "DEBUG",
"handlers": ["f_pipeline.log", "f_pipeline.log.wf"],
},
}
logging.config.dictConfig(logger_config)
python/pipeline/operator.py
浏览文件 @ f3953aa5
......@@ -12,7 +12,8 @@
# See the License for the specific language governing permissions and
# limitations under the License.
# pylint: disable=doc-string-missing
from time import time as _time
import time
import threading
import multiprocessing
from paddle_serving_client import MultiLangClient, Client
......@@ -21,16 +22,24 @@ import logging
import func_timeout
import os
import sys
import collections
import numpy as np
from numpy import *
if sys.version_info.major == 2:
import Queue
elif sys.version_info.major == 3:
import queue as Queue
else:
raise Exception("Error Python version")
from .proto import pipeline_service_pb2
from .channel import (ThreadChannel, ProcessChannel, ChannelDataEcode,
ChannelData, ChannelDataType, ChannelStopError)
ChannelData, ChannelDataType, ChannelStopError,
ChannelTimeoutError)
from .util import NameGenerator
from .profiler import TimeProfiler
from .profiler import UnsafeTimeProfiler as TimeProfiler
_LOGGER = logging.getLogger()
_LOGGER = logging.getLogger(__name__)
_op_name_gen = NameGenerator("Op")
......@@ -43,7 +52,9 @@ class Op(object):
client_config=None,
concurrency=1,
timeout=-1,
retry=1):
retry=1,
batch_size=1,
auto_batching_timeout=None):
if name is None:
name = _op_name_gen.next()
self.name = name # to identify the type of OP, it must be globally unique
......@@ -57,41 +68,82 @@ class Op(object):
self._client_config = client_config
self._fetch_names = fetch_list
self._timeout = timeout
if timeout > 0:
self._timeout = timeout / 1000.0
else:
self._timeout = -1
self._retry = max(1, retry)
self._input = None
self._outputs = []
self._batch_size = batch_size
self._auto_batching_timeout = auto_batching_timeout
if self._auto_batching_timeout is not None:
if self._auto_batching_timeout <= 0 or self._batch_size == 1:
_LOGGER.warning(
self._log(
"Because auto_batching_timeout <= 0 or batch_size == 1,"
" set auto_batching_timeout to None."))
self._auto_batching_timeout = None
else:
self._auto_batching_timeout = self._auto_batching_timeout / 1000.0
if not isinstance(self, RequestOp) and not isinstance(self, ResponseOp):
_LOGGER.info(
self._log("\n\tinput_ops: {},"
"\n\tserver_endpoints: {}"
"\n\tfetch_list: {}"
"\n\tclient_config: {}"
"\n\tconcurrency: {},"
"\n\ttimeout(s): {},"
"\n\tretry: {},"
"\n\tbatch_size: {},"
"\n\tauto_batching_timeout(s): {}".format(
", ".join([op.name for op in input_ops
]), self._server_endpoints,
self._fetch_names, self._client_config,
self.concurrency, self._timeout, self._retry,
self._batch_size, self._auto_batching_timeout)))
self._server_use_profile = False
self._tracer = None
# only for multithread
# only for thread op
self._for_init_op_lock = threading.Lock()
self._for_close_op_lock = threading.Lock()
self._succ_init_op = False
self._succ_close_op = False
def use_default_auto_batching_config(self):
if self._batch_size != 1:
_LOGGER.warning("Op({}) reset batch_size=1 (original: {})"
.format(self.name, self._batch_size))
self._batch_size = 1
if self._auto_batching_timeout != None:
_LOGGER.warning(
"Op({}) reset auto_batching_timeout=None (original: {})"
.format(self.name, self._auto_batching_timeout))
self._auto_batching_timeout = None
def use_profiler(self, use_profile):
self._server_use_profile = use_profile
def _profiler_record(self, string):
if self._profiler is None:
return
self._profiler.record(string)
def set_tracer(self, tracer):
self._tracer = tracer
def init_client(self, client_type, client_config, server_endpoints,
fetch_names):
if self.with_serving == False:
_LOGGER.debug("{} no client".format(self.name))
_LOGGER.info("Op({}) has no client (and it also do not "
"run the process function)".format(self.name))
return None
_LOGGER.debug("{} client_config: {}".format(self.name, client_config))
_LOGGER.debug("{} fetch_names: {}".format(self.name, fetch_names))
if client_type == 'brpc':
client = Client()
client.load_client_config(client_config)
elif client_type == 'grpc':
client = MultiLangClient()
else:
raise ValueError("unknow client type: {}".format(client_type))
raise ValueError("Failed to init client: unknow client "
"type {}".format(client_type))
client.connect(server_endpoints)
self._fetch_names = fetch_names
return client
......@@ -105,16 +157,19 @@ class Op(object):
self._input_ops = []
for op in ops:
if not isinstance(op, Op):
raise TypeError(
self._log('input op must be Op type, not {}'.format(
type(op))))
_LOGGER.critical(
self._log("Failed to set input_ops: input op "
"must be Op type, not {}".format(type(op))))
os._exit(-1)
self._input_ops.append(op)
def add_input_channel(self, channel):
if not isinstance(channel, (ThreadChannel, ProcessChannel)):
raise TypeError(
self._log('input channel must be Channel type, not {}'.format(
_LOGGER.critical(
self._log("Failed to set input_channel: input "
"channel must be Channel type, not {}".format(
type(channel))))
os._exit(-1)
channel.add_consumer(self.name)
self._input = channel
......@@ -126,9 +181,10 @@ class Op(object):
def add_output_channel(self, channel):
if not isinstance(channel, (ThreadChannel, ProcessChannel)):
raise TypeError(
self._log('output channel must be Channel type, not {}'.format(
type(channel))))
_LOGGER.critical(
self._log("Failed to add output_channel: output channel "
"must be Channel type, not {}".format(type(channel))))
os._exit(-1)
channel.add_producer(self.name)
self._outputs.append(channel)
......@@ -141,21 +197,28 @@ class Op(object):
def preprocess(self, input_dicts):
# multiple previous Op
if len(input_dicts) != 1:
raise NotImplementedError(
'this Op has multiple previous inputs. Please override this func.'
)
_LOGGER.critical(
self._log(
"Failed to run preprocess: this Op has multiple previous "
"inputs. Please override this func."))
os._exit(-1)
(_, input_dict), = input_dicts.items()
return input_dict
def process(self, feed_dict):
err, err_info = ChannelData.check_npdata(feed_dict)
def process(self, feed_batch, typical_logid):
err, err_info = ChannelData.check_batch_npdata(feed_batch)
if err != 0:
raise NotImplementedError(
"{} Please override preprocess func.".format(err_info))
_LOGGER.critical(
self._log("Failed to run process: {}. Please override "
"preprocess func.".format(err_info)))
os._exit(-1)
call_result = self.client.predict(
feed=feed_dict, fetch=self._fetch_names)
_LOGGER.debug(self._log("get call_result"))
feed=feed_batch, fetch=self._fetch_names, log_id=typical_logid)
if isinstance(self.client, MultiLangClient):
if call_result is None or call_result["serving_status_code"] != 0:
return None
call_result.pop("serving_status_code")
return call_result
def postprocess(self, input_dict, fetch_dict):
......@@ -184,43 +247,51 @@ class Op(object):
data,
channels,
name=None,
profile_str=None,
client_need_profile=False,
profile_set=None):
if name is None:
name = self.name
self._add_profile_into_channeldata(data, client_need_profile,
profile_set)
for channel in channels:
channel.push(data, name)
def _add_profile_into_channeldata(self, data, client_need_profile,
profile_set):
profile_str = self._profiler.gen_profile_str()
if self._server_use_profile:
sys.stderr.write(profile_str)
# add profile into channeldata
if client_need_profile and profile_set is not None:
if profile_str is not None:
profile_set.add(profile_str)
data.add_profile(profile_set)
for channel in channels:
channel.push(data, name)
def start_with_process(self, client_type):
trace_buffer = None
if self._tracer is not None:
trace_buffer = self._tracer.data_buffer()
proces = []
for concurrency_idx in range(self.concurrency):
p = multiprocessing.Process(
target=self._run,
args=(concurrency_idx, self._get_input_channel(),
self._get_output_channels(), client_type, False))
self._get_output_channels(), client_type, False,
trace_buffer))
p.daemon = True
p.start()
proces.append(p)
return proces
def start_with_thread(self, client_type):
trace_buffer = None
if self._tracer is not None:
trace_buffer = self._tracer.data_buffer()
threads = []
for concurrency_idx in range(self.concurrency):
t = threading.Thread(
target=self._run,
args=(concurrency_idx, self._get_input_channel(),
self._get_output_channels(), client_type, True))
self._get_output_channels(), client_type, True,
trace_buffer))
# When a process exits, it attempts to terminate
# all of its daemonic child processes.
t.daemon = True
t.start()
threads.append(t)
return threads
......@@ -228,103 +299,209 @@ class Op(object):
def init_op(self):
pass
def _run_preprocess(self, parsed_data, data_id, log_func):
def _run_preprocess(self, parsed_data_dict, op_info_prefix):
_LOGGER.debug("{} Running preprocess".format(op_info_prefix))
preped_data_dict = collections.OrderedDict()
err_channeldata_dict = collections.OrderedDict()
for data_id, parsed_data in parsed_data_dict.items():
preped_data, error_channeldata = None, None
try:
preped_data = self.preprocess(parsed_data)
except NotImplementedError as e:
# preprocess function not implemented
error_info = log_func(e)
_LOGGER.error(error_info)
error_channeldata = ChannelData(
ecode=ChannelDataEcode.NOT_IMPLEMENTED.value,
error_info=error_info,
data_id=data_id)
except TypeError as e:
# Error type in channeldata.datatype
error_info = log_func(e)
_LOGGER.error(error_info)
error_info = "(logid={}) {} Failed to preprocess: {}".format(
data_id, op_info_prefix, e)
_LOGGER.error(error_info, exc_info=True)
error_channeldata = ChannelData(
ecode=ChannelDataEcode.TYPE_ERROR.value,
error_info=error_info,
data_id=data_id)
except Exception as e:
error_info = log_func(e)
_LOGGER.error(error_info)
error_info = "(logid={}) {} Failed to preprocess: {}".format(
data_id, op_info_prefix, e)
_LOGGER.error(error_info, exc_info=True)
error_channeldata = ChannelData(
ecode=ChannelDataEcode.UNKNOW.value,
error_info=error_info,
data_id=data_id)
return preped_data, error_channeldata
def _run_process(self, preped_data, data_id, log_func):
midped_data, error_channeldata = None, None
if error_channeldata is not None:
err_channeldata_dict[data_id] = error_channeldata
else:
preped_data_dict[data_id] = preped_data
_LOGGER.debug("{} Succ preprocess".format(op_info_prefix))
return preped_data_dict, err_channeldata_dict
def _run_process(self, preped_data_dict, op_info_prefix):
_LOGGER.debug("{} Running process".format(op_info_prefix))
midped_data_dict = collections.OrderedDict()
err_channeldata_dict = collections.OrderedDict()
if self.with_serving:
data_ids = preped_data_dict.keys()
typical_logid = data_ids[0]
if len(data_ids) != 1:
for data_id in data_ids:
_LOGGER.info(
"(logid={}) {} During access to PaddleServingService,"
" we selected logid={} (from batch: {}) as a "
"representative for logging.".format(
data_id, op_info_prefix, typical_logid, data_ids))
# combine samples to batch
one_input = preped_data_dict[data_ids[0]]
feed_batch = []
input_offset = None
if isinstance(one_input, dict):
# sample input
feed_batch = [preped_data_dict[data_id] for data_id in data_ids]
input_offset = list(range(len(data_ids) + 1))
elif isinstance(one_input, list):
# batch input
input_offset = [0]
for data_id in data_ids:
batch_input = preped_data_dict[data_id]
offset = input_offset[-1] + len(batch_input)
feed_batch += batch_input
input_offset.append(offset)
else:
_LOGGER.critical(
"{} Failed to process: expect input type is dict(sample"
" input) or list(batch input), but get {}".format(
op_info_prefix, type(one_input)))
os._exit(-1)
midped_batch = None
ecode = ChannelDataEcode.OK.value
if self._timeout <= 0:
try:
midped_data = self.process(preped_data)
midped_batch = self.process(feed_batch, typical_logid)
except Exception as e:
ecode = ChannelDataEcode.UNKNOW.value
error_info = log_func(e)
_LOGGER.error(error_info)
error_info = "(logid={}) {} Failed to process(batch: {}): {}".format(
typical_logid, op_info_prefix, data_ids, e)
_LOGGER.error(error_info, exc_info=True)
else:
for i in range(self._retry):
try:
midped_data = func_timeout.func_timeout(
self._timeout, self.process, args=(preped_data, ))
midped_batch = func_timeout.func_timeout(
self._timeout,
self.process,
args=(feed_batch, typical_logid))
except func_timeout.FunctionTimedOut as e:
if i + 1 >= self._retry:
ecode = ChannelDataEcode.TIMEOUT.value
error_info = log_func(e)
error_info = "(logid={}) {} Failed to process(batch: {}): " \
"exceeded retry count.".format(
typical_logid, op_info_prefix, data_ids)
_LOGGER.error(error_info)
else:
_LOGGER.warn(
log_func("timeout, retry({})".format(i + 1)))
_LOGGER.warning(
"(logid={}) {} Failed to process(batch: {}): timeout,"
" and retrying({}/{})...".format(
typical_logid, op_info_prefix, data_ids, i +
1, self._retry))
except Exception as e:
ecode = ChannelDataEcode.UNKNOW.value
error_info = log_func(e)
_LOGGER.error(error_info)
error_info = "(logid={}) {} Failed to process(batch: {}): {}".format(
typical_logid, op_info_prefix, data_ids, e)
_LOGGER.error(error_info, exc_info=True)
break
else:
break
if ecode != ChannelDataEcode.OK.value:
error_channeldata = ChannelData(
for data_id in data_ids:
err_channeldata_dict[data_id] = ChannelData(
ecode=ecode, error_info=error_info, data_id=data_id)
elif midped_data is None:
elif midped_batch is None:
# op client return None
error_channeldata = ChannelData(
error_info = "(logid={}) {} Failed to predict, please check if " \
"PaddleServingService is working properly.".format(
typical_logid, op_info_prefix)
_LOGGER.error(error_info)
for data_id in data_ids:
err_channeldata_dict[data_id] = ChannelData(
ecode=ChannelDataEcode.CLIENT_ERROR.value,
error_info=log_func(
"predict failed. pls check the server side."),
error_info=error_info,
data_id=data_id)
else:
midped_data = preped_data
return midped_data, error_channeldata
def _run_postprocess(self, input_dict, midped_data, data_id, log_func):
output_data, error_channeldata = None, None
# transform np format to dict format
var_names = midped_batch.keys()
lod_var_names = set()
lod_offset_names = set()
for name in var_names:
lod_offset_name = "{}.lod".format(name)
if lod_offset_name in var_names:
_LOGGER.debug("(logid={}) {} {} is LodTensor".format(
typical_logid, op_info_prefix, name))
lod_var_names.add(name)
lod_offset_names.add(lod_offset_name)
for idx, data_id in enumerate(data_ids):
midped_data_dict[data_id] = {}
for name, value in midped_batch.items():
if name in lod_offset_names:
continue
if name in lod_var_names:
# lodtensor
lod_offset_name = "{}.lod".format(name)
lod_offset = midped_batch[lod_offset_name]
for idx, data_id in enumerate(data_ids):
data_offset_left = input_offset[idx]
data_offset_right = input_offset[idx + 1]
lod_offset_left = lod_offset[data_offset_left]
lod_offset_right = lod_offset[data_offset_right]
midped_data_dict[data_id][name] = value[
lod_offset_left:lod_offset_right]
midped_data_dict[data_id][lod_offset_name] = \
lod_offset[data_offset_left:data_offset_right + 1] - lod_offset[data_offset_left]
else:
# normal tensor
for idx, data_id in enumerate(data_ids):
left = input_offset[idx]
right = input_offset[idx + 1]
midped_data_dict[data_id][name] = value[left:right]
else:
midped_data_dict = preped_data_dict
_LOGGER.debug("{} Succ process".format(op_info_prefix))
return midped_data_dict, err_channeldata_dict
def _run_postprocess(self, parsed_data_dict, midped_data_dict,
op_info_prefix):
_LOGGER.debug("{} Running postprocess".format(op_info_prefix))
postped_data_dict = collections.OrderedDict()
err_channeldata_dict = collections.OrderedDict()
for data_id, midped_data in midped_data_dict.items():
postped_data, err_channeldata = None, None
try:
postped_data = self.postprocess(input_dict, midped_data)
postped_data = self.postprocess(parsed_data_dict[data_id],
midped_data)
except Exception as e:
error_info = log_func(e)
_LOGGER.error(error_info)
error_channeldata = ChannelData(
error_info = "(logid={}) {} Failed to postprocess: {}".format(
data_id, op_info_prefix, e)
_LOGGER.error(error_info, exc_info=True)
err_channeldata = ChannelData(
ecode=ChannelDataEcode.UNKNOW.value,
error_info=error_info,
data_id=data_id)
return output_data, error_channeldata
if err_channeldata is not None:
err_channeldata_dict[data_id] = err_channeldata
continue
else:
if not isinstance(postped_data, dict):
error_info = log_func("output of postprocess funticon must be " \
"dict type, but get {}".format(type(postped_data)))
error_info = "(logid={}) {} Failed to postprocess: " \
"output of postprocess funticon must be " \
"dict type, but get {}".format(
data_id, op_info_prefix,
type(postped_data))
_LOGGER.error(error_info)
error_channeldata = ChannelData(
err_channeldata = ChannelData(
ecode=ChannelDataEcode.UNKNOW.value,
error_info=error_info,
data_id=data_id)
return output_data, error_channeldata
err_channeldata_dict[data_id] = err_channeldata
continue
output_data = None
err, _ = ChannelData.check_npdata(postped_data)
if err == 0:
output_data = ChannelData(
......@@ -336,144 +513,245 @@ class Op(object):
ChannelDataType.DICT.value,
dictdata=postped_data,
data_id=data_id)
return output_data, error_channeldata
postped_data_dict[data_id] = output_data
_LOGGER.debug("{} Succ postprocess".format(op_info_prefix))
return postped_data_dict, err_channeldata_dict
def _run(self, concurrency_idx, input_channel, output_channels, client_type,
is_thread_op):
def get_log_func(op_info_prefix):
def log_func(info_str):
return "{} {}".format(op_info_prefix, info_str)
def _auto_batching_generator(self, input_channel, op_name, batch_size,
timeout, op_info_prefix):
while True:
batch = []
while len(batch) == 0:
endtime = None
if timeout is not None:
endtime = _time() + timeout
for idx in range(batch_size):
try:
channeldata_dict = None
if timeout is not None:
remaining = endtime - _time()
if remaining <= 0.0:
_LOGGER.debug("{} Failed to generate batch: "
"timeout".format(op_info_prefix))
break
channeldata_dict = input_channel.front(op_name,
timeout)
else:
channeldata_dict = input_channel.front(op_name)
batch.append(channeldata_dict)
except ChannelTimeoutError:
_LOGGER.debug("{} Failed to generate batch: "
"timeout".format(op_info_prefix))
break
_LOGGER.debug("{} Got actual batch_size: {}".format(op_info_prefix,
len(batch)))
yield batch
def _parse_channeldata_batch(self, batch, output_channels):
parsed_data_dict = collections.OrderedDict()
need_profile_dict = {}
profile_dict = {}
for channeldata_dict in batch:
(data_id, error_channeldata, parsed_data,
client_need_profile, profile_set) = \
self._parse_channeldata(channeldata_dict)
if error_channeldata is None:
parsed_data_dict[data_id] = parsed_data
need_profile_dict[data_id] = client_need_profile
profile_dict[data_id] = profile_set
else:
# error data in predecessor Op
# (error_channeldata with profile info)
self._push_to_output_channels(error_channeldata,
output_channels)
return log_func
return parsed_data_dict, need_profile_dict, profile_dict
def _run(self, concurrency_idx, input_channel, output_channels, client_type,
is_thread_op, trace_buffer):
op_info_prefix = "[{}|{}]".format(self.name, concurrency_idx)
log = get_log_func(op_info_prefix)
tid = threading.current_thread().ident
# init op
self.concurrency_idx = concurrency_idx
profiler = None
try:
if is_thread_op:
with self._for_init_op_lock:
if not self._succ_init_op:
# init profiler
self._profiler = TimeProfiler()
self._profiler.enable(True)
# init client
self.client = self.init_client(
client_type, self._client_config,
self._server_endpoints, self._fetch_names)
# user defined
self.init_op()
self._succ_init_op = True
self._succ_close_op = False
else:
# init profiler
self._profiler = TimeProfiler()
self._profiler.enable(True)
# init client
self.client = self.init_client(client_type, self._client_config,
self._server_endpoints,
self._fetch_names)
# user defined
self.init_op()
profiler = self._initialize(is_thread_op, client_type,
concurrency_idx)
except Exception as e:
_LOGGER.error(log(e))
_LOGGER.critical(
"{} Failed to init op: {}".format(op_info_prefix, e),
exc_info=True)
os._exit(-1)
_LOGGER.info("{} Succ init".format(op_info_prefix))
batch_generator = self._auto_batching_generator(
input_channel=input_channel,
op_name=self.name,
batch_size=self._batch_size,
timeout=self._auto_batching_timeout,
op_info_prefix=op_info_prefix)
start, end = None, None
trace_que = collections.deque()
while True:
#self._profiler_record("get#{}_0".format(op_info_prefix))
start = int(round(_time() * 1000000))
try:
channeldata_dict = input_channel.front(self.name)
channeldata_dict_batch = next(batch_generator)
except ChannelStopError:
_LOGGER.debug(log("stop."))
if is_thread_op:
with self._for_close_op_lock:
if not self._succ_close_op:
self._profiler = None
self.client = None
self._succ_init_op = False
self._succ_close_op = True
_LOGGER.debug("{} Stop.".format(op_info_prefix))
self._finalize(is_thread_op)
break
#self._profiler_record("get#{}_1".format(op_info_prefix))
_LOGGER.debug(log("input_data: {}".format(channeldata_dict)))
end = int(round(_time() * 1000000))
in_time = end - start
(data_id, error_channeldata, parsed_data, client_need_profile,
profile_set) = self._parse_channeldata(channeldata_dict)
# error data in predecessor Op
if error_channeldata is not None:
# parse channeldata batch
try:
# error_channeldata with profile info
self._push_to_output_channels(error_channeldata,
output_channels)
parsed_data_dict, need_profile_dict, profile_dict \
= self._parse_channeldata_batch(
channeldata_dict_batch, output_channels)
except ChannelStopError:
_LOGGER.debug(log("stop."))
_LOGGER.debug("{} Stop.".format(op_info_prefix))
self._finalize(is_thread_op)
break
if len(parsed_data_dict) == 0:
# data in the whole batch is all error data
continue
# preprecess
self._profiler_record("prep#{}_0".format(op_info_prefix))
preped_data, error_channeldata = self._run_preprocess(parsed_data,
data_id, log)
self._profiler_record("prep#{}_1".format(op_info_prefix))
if error_channeldata is not None:
start = profiler.record("prep#{}_0".format(op_info_prefix))
preped_data_dict, err_channeldata_dict \
= self._run_preprocess(parsed_data_dict, op_info_prefix)
end = profiler.record("prep#{}_1".format(op_info_prefix))
prep_time = end - start
try:
for data_id, err_channeldata in err_channeldata_dict.items():
self._push_to_output_channels(
error_channeldata,
output_channels,
client_need_profile=client_need_profile,
profile_set=profile_set)
data=err_channeldata,
channels=output_channels,
client_need_profile=need_profile_dict[data_id],
profile_set=profile_dict[data_id])
except ChannelStopError:
_LOGGER.debug(log("stop."))
_LOGGER.debug("{} Stop.".format(op_info_prefix))
self._finalize(is_thread_op)
break
if len(preped_data_dict) == 0:
continue
# process
self._profiler_record("midp#{}_0".format(op_info_prefix))
midped_data, error_channeldata = self._run_process(preped_data,
data_id, log)
self._profiler_record("midp#{}_1".format(op_info_prefix))
if error_channeldata is not None:
start = profiler.record("midp#{}_0".format(op_info_prefix))
midped_data_dict, err_channeldata_dict \
= self._run_process(preped_data_dict, op_info_prefix)
end = profiler.record("midp#{}_1".format(op_info_prefix))
midp_time = end - start
try:
for data_id, err_channeldata in err_channeldata_dict.items():
self._push_to_output_channels(
error_channeldata,
output_channels,
client_need_profile=client_need_profile,
profile_set=profile_set)
data=err_channeldata,
channels=output_channels,
client_need_profile=need_profile_dict[data_id],
profile_set=profile_dict[data_id])
except ChannelStopError:
_LOGGER.debug(log("stop."))
_LOGGER.debug("{} Stop.".format(op_info_prefix))
self._finalize(is_thread_op)
break
if len(midped_data_dict) == 0:
continue
# postprocess
self._profiler_record("postp#{}_0".format(op_info_prefix))
output_data, error_channeldata = self._run_postprocess(
parsed_data, midped_data, data_id, log)
self._profiler_record("postp#{}_1".format(op_info_prefix))
if error_channeldata is not None:
start = profiler.record("postp#{}_0".format(op_info_prefix))
postped_data_dict, err_channeldata_dict \
= self._run_postprocess(
parsed_data_dict, midped_data_dict, op_info_prefix)
end = profiler.record("postp#{}_1".format(op_info_prefix))
postp_time = end - start
try:
for data_id, err_channeldata in err_channeldata_dict.items():
self._push_to_output_channels(
error_channeldata,
output_channels,
client_need_profile=client_need_profile,
profile_set=profile_set)
data=err_channeldata,
channels=output_channels,
client_need_profile=need_profile_dict[data_id],
profile_set=profile_dict[data_id])
except ChannelStopError:
_LOGGER.debug(log("stop."))
_LOGGER.debug("{} Stop.".format(op_info_prefix))
self._finalize(is_thread_op)
break
if len(postped_data_dict) == 0:
continue
# push data to channel (if run succ)
#self._profiler_record("push#{}_0".format(op_info_prefix))
start = int(round(_time() * 1000000))
try:
profile_str = profiler.gen_profile_str()
for data_id, postped_data in postped_data_dict.items():
if self._server_use_profile:
sys.stderr.write(profile_str)
self._push_to_output_channels(
output_data,
output_channels,
client_need_profile=client_need_profile,
profile_set=profile_set)
data=postped_data,
channels=output_channels,
profile_str=profile_str,
client_need_profile=need_profile_dict[data_id],
profile_set=profile_dict[data_id])
except ChannelStopError:
_LOGGER.debug(log("stop."))
_LOGGER.debug("{} Stop.".format(op_info_prefix))
self._finalize(is_thread_op)
break
end = int(round(_time() * 1000000))
out_time = end - start
if trace_buffer is not None:
trace_que.append({
"name": self.name,
"actions": {
"in": in_time,
"prep": prep_time,
"midp": midp_time,
"postp": postp_time,
"out": out_time,
}
})
while trace_que:
info = trace_que[0]
try:
trace_buffer.put_nowait(info)
trace_que.popleft()
except Queue.Full:
break
#self._profiler_record("push#{}_1".format(op_info_prefix))
def _initialize(self, is_thread_op, client_type, concurrency_idx):
if is_thread_op:
with self._for_init_op_lock:
if not self._succ_init_op:
# for the threaded version of Op, each thread cannot get its concurrency_idx
self.concurrency_idx = None
# init client
self.client = self.init_client(
client_type, self._client_config,
self._server_endpoints, self._fetch_names)
# user defined
self.init_op()
self._succ_init_op = True
self._succ_close_op = False
else:
self.concurrency_idx = concurrency_idx
# init client
self.client = self.init_client(client_type, self._client_config,
self._server_endpoints,
self._fetch_names)
# user defined
self.init_op()
# use a separate TimeProfiler per thread or process
profiler = TimeProfiler()
profiler.enable(True)
return profiler
def _finalize(self, is_thread_op):
if is_thread_op:
with self._for_close_op_lock:
if not self._succ_close_op:
self._profiler = None
self.client = None
self._succ_init_op = False
self._succ_close_op = True
def _log(self, info):
return "{} {}".format(self.name, info)
......@@ -483,13 +761,13 @@ class RequestOp(Op):
""" RequestOp do not run preprocess, process, postprocess. """
def __init__(self):
# PipelineService.name = "@G"
super(RequestOp, self).__init__(name="@G", input_ops=[])
# PipelineService.name = "@DAGExecutor"
super(RequestOp, self).__init__(name="@DAGExecutor", input_ops=[])
# init op
try:
self.init_op()
except Exception as e:
_LOGGER.error(e)
_LOGGER.critical("Op(Request) Failed to init: {}".format(e))
os._exit(-1)
def unpack_request_package(self, request):
......@@ -508,12 +786,14 @@ class ResponseOp(Op):
""" ResponseOp do not run preprocess, process, postprocess. """
def __init__(self, input_ops):
super(ResponseOp, self).__init__(name="@R", input_ops=input_ops)
super(ResponseOp, self).__init__(
name="@DAGExecutor", input_ops=input_ops)
# init op
try:
self.init_op()
except Exception as e:
_LOGGER.error(e)
_LOGGER.critical("Op(ResponseOp) Failed to init: {}".format(
e, exc_info=True))
os._exit(-1)
def pack_response_package(self, channeldata):
......@@ -524,7 +804,7 @@ class ResponseOp(Op):
feed = channeldata.parse()
# ndarray to string:
# https://stackoverflow.com/questions/30167538/convert-a-numpy-ndarray-to-stringor-bytes-and-convert-it-back-to-numpy-ndarray
np.set_printoptions(threshold=np.nan)
np.set_printoptions(threshold=sys.maxsize)
for name, var in feed.items():
resp.value.append(var.__repr__())
resp.key.append(name)
......@@ -536,14 +816,19 @@ class ResponseOp(Op):
resp.error_info = self._log(
"fetch var type must be str({}).".format(
type(var)))
_LOGGER.error("(logid={}) Failed to pack RPC "
"response package: {}".format(
channeldata.id, resp.error_info))
break
resp.value.append(var)
resp.key.append(name)
else:
resp.ecode = ChannelDataEcode.TYPE_ERROR.value
resp.error_info = self._log(
"Error type({}) in datatype.".format(channeldata.datatype))
_LOGGER.error(resp.error_info)
"error type({}) in datatype.".format(channeldata.datatype))
_LOGGER.error("(logid={}) Failed to pack RPC response"
" package: {}".format(channeldata.id,
resp.error_info))
else:
resp.error_info = channeldata.error_info
return resp
......@@ -561,6 +846,7 @@ class VirtualOp(Op):
self._virtual_pred_ops.append(op)
def _actual_pred_op_names(self, op):
# can use disjoint-set, but it's not necessary
if not isinstance(op, VirtualOp):
return [op.name]
names = []
......@@ -570,9 +856,11 @@ class VirtualOp(Op):
def add_output_channel(self, channel):
if not isinstance(channel, (ThreadChannel, ProcessChannel)):
raise TypeError(
self._log('output channel must be Channel type, not {}'.format(
_LOGGER.critical(
self._log("Failed to add output_channel: output_channel"
" must be Channel type, not {}".format(
type(channel))))
os._exit(-1)
for op in self._virtual_pred_ops:
for op_name in self._actual_pred_op_names(op):
channel.add_producer(op_name)
......@@ -580,27 +868,31 @@ class VirtualOp(Op):
def _run(self, concurrency_idx, input_channel, output_channels, client_type,
is_thread_op):
def get_log_func(op_info_prefix):
def log_func(info_str):
return "{} {}".format(op_info_prefix, info_str)
return log_func
op_info_prefix = "[{}|{}]".format(self.name, concurrency_idx)
log = get_log_func(op_info_prefix)
tid = threading.current_thread().ident
batch_generator = self._auto_batching_generator(
input_channel=input_channel,
op_name=self.name,
batch_size=1,
timeout=None,
log_func=log)
while True:
try:
channeldata_dict = input_channel.front(self.name)
channeldata_dict_batch = next(batch_generator)
except ChannelStopError:
_LOGGER.debug(log("stop."))
_LOGGER.debug("{} Stop.".format(op_info_prefix))
self._finalize(is_thread_op)
break
try:
for channeldata_dict in channeldata_dict_batch:
for name, data in channeldata_dict.items():
self._push_to_output_channels(
data, channels=output_channels, name=name)
except ChannelStopError:
_LOGGER.debug(log("stop."))
_LOGGER.debug("{} Stop.".format(op_info_prefix))
self._finalize(is_thread_op)
break
python/pipeline/pipeline_client.py
浏览文件 @ f3953aa5
......@@ -18,10 +18,11 @@ import numpy as np
from numpy import *
import logging
import functools
from .channel import ChannelDataEcode
from .proto import pipeline_service_pb2
from .proto import pipeline_service_pb2_grpc
_LOGGER = logging.getLogger()
_LOGGER = logging.getLogger(__name__)
class PipelineClient(object):
......@@ -59,7 +60,11 @@ class PipelineClient(object):
def _unpack_response_package(self, resp, fetch):
if resp.ecode != 0:
return {"ecode": resp.ecode, "error_info": resp.error_info}
return {
"ecode": resp.ecode,
"ecode_desc": ChannelDataEcode(resp.ecode),
"error_info": resp.error_info,
}
fetch_map = {"ecode": resp.ecode}
for idx, key in enumerate(resp.key):
if key == self._profile_key:
......
python/pipeline/pipeline_server.py
浏览文件 @ f3953aa5
......@@ -15,6 +15,7 @@
from concurrent import futures
import grpc
import logging
import json
import socket
import contextlib
from contextlib import closing
......@@ -25,24 +26,21 @@ from .proto import pipeline_service_pb2_grpc
from .operator import ResponseOp
from .dag import DAGExecutor
_LOGGER = logging.getLogger()
_LOGGER = logging.getLogger(__name__)
class PipelineService(pipeline_service_pb2_grpc.PipelineServiceServicer):
def __init__(self, response_op, dag_config, show_info):
super(PipelineService, self).__init__()
class PipelineServicer(pipeline_service_pb2_grpc.PipelineServiceServicer):
def __init__(self, response_op, dag_conf, worker_idx=-1):
super(PipelineServicer, self).__init__()
# init dag executor
self._dag_executor = DAGExecutor(
response_op, dag_config, show_info=show_info)
self._dag_executor = DAGExecutor(response_op, dag_conf, worker_idx)
self._dag_executor.start()
_LOGGER.info("[PipelineServicer] succ init")
def inference(self, request, context):
resp = self._dag_executor.call(request)
return resp
def __del__(self):
self._dag_executor.stop()
@contextlib.contextmanager
def _reserve_port(port):
......@@ -66,9 +64,11 @@ class PipelineServer(object):
def set_response_op(self, response_op):
if not isinstance(response_op, ResponseOp):
raise Exception("response_op must be ResponseOp type.")
raise Exception("Failed to set response_op: response_op "
"must be ResponseOp type.")
if len(response_op.get_input_ops()) != 1:
raise Exception("response_op can only have one previous op.")
raise Exception("Failed to set response_op: response_op "
"can only have one previous op.")
self._response_op = response_op
def _port_is_available(self, port):
......@@ -78,28 +78,26 @@ class PipelineServer(object):
return result != 0
def prepare_server(self, yml_file):
with open(yml_file) as f:
yml_config = yaml.load(f.read())
self._port = yml_config.get('port')
if self._port is None:
raise SystemExit("Please set *port* in [{}] yaml file.".format(
yml_file))
conf = ServerYamlConfChecker.load_server_yaml_conf(yml_file)
self._port = conf["port"]
if not self._port_is_available(self._port):
raise SystemExit("Prot {} is already used".format(self._port))
self._worker_num = yml_config.get('worker_num', 1)
self._build_dag_each_worker = yml_config.get('build_dag_each_worker',
False)
raise SystemExit("Failed to prepare_server: prot {} "
"is already used".format(self._port))
self._worker_num = conf["worker_num"]
self._build_dag_each_worker = conf["build_dag_each_worker"]
_LOGGER.info("============= PIPELINE SERVER =============")
_LOGGER.info("port: {}".format(self._port))
_LOGGER.info("worker_num: {}".format(self._worker_num))
servicer_info = "build_dag_each_worker: {}".format(
self._build_dag_each_worker)
_LOGGER.info("\n{}".format(
json.dumps(
conf, indent=4, separators=(',', ':'))))
if self._build_dag_each_worker is True:
servicer_info += " (Make sure that install grpcio whl with --no-binary flag)"
_LOGGER.info(servicer_info)
_LOGGER.warning(
"(Make sure that install grpcio whl with --no-binary flag: "
"pip install grpcio --no-binary grpcio)")
_LOGGER.info("-------------------------------------------")
self._dag_config = yml_config.get("dag", {})
self._conf = conf
def run_server(self):
if self._build_dag_each_worker:
......@@ -110,29 +108,167 @@ class PipelineServer(object):
show_info = (i == 0)
worker = multiprocessing.Process(
target=self._run_server_func,
args=(bind_address, self._response_op,
self._dag_config))
args=(bind_address, self._response_op, self._conf, i))
worker.start()
workers.append(worker)
for worker in workers:
worker.join()
else:
server = grpc.server(
futures.ThreadPoolExecutor(max_workers=self._worker_num))
futures.ThreadPoolExecutor(max_workers=self._worker_num),
options=[('grpc.max_send_message_length', 256 * 1024 * 1024),
('grpc.max_receive_message_length', 256 * 1024 * 1024)
])
pipeline_service_pb2_grpc.add_PipelineServiceServicer_to_server(
PipelineService(self._response_op, self._dag_config, True),
server)
PipelineServicer(self._response_op, self._conf), server)
server.add_insecure_port('[::]:{}'.format(self._port))
server.start()
server.wait_for_termination()
def _run_server_func(self, bind_address, response_op, dag_config):
options = (('grpc.so_reuseport', 1), )
def _run_server_func(self, bind_address, response_op, dag_conf, worker_idx):
options = [('grpc.so_reuseport', 1),
('grpc.max_send_message_length', 256 * 1024 * 1024),
('grpc.max_send_message_length', 256 * 1024 * 1024)]
server = grpc.server(
futures.ThreadPoolExecutor(
max_workers=1, ), options=options)
pipeline_service_pb2_grpc.add_PipelineServiceServicer_to_server(
PipelineService(response_op, dag_config, False), server)
PipelineServicer(response_op, dag_conf, worker_idx), server)
server.add_insecure_port(bind_address)
server.start()
server.wait_for_termination()
class ServerYamlConfChecker(object):
def __init__(self):
pass
@staticmethod
def load_server_yaml_conf(yml_file):
with open(yml_file) as f:
conf = yaml.load(f.read())
ServerYamlConfChecker.check_server_conf(conf)
ServerYamlConfChecker.check_dag_conf(conf["dag"])
ServerYamlConfChecker.check_tracer_conf(conf["dag"]["tracer"])
return conf
@staticmethod
def check_conf(conf, default_conf, conf_type, conf_qualification):
ServerYamlConfChecker.fill_with_default_conf(conf, default_conf)
ServerYamlConfChecker.check_conf_type(conf, conf_type)
ServerYamlConfChecker.check_conf_qualification(conf, conf_qualification)
@staticmethod
def check_server_conf(conf):
default_conf = {
"port": 9292,
"worker_num": 1,
"build_dag_each_worker": False,
"dag": {},
}
conf_type = {
"port": int,
"worker_num": int,
"build_dag_each_worker": bool,
}
conf_qualification = {
"port": [(">=", 1024), ("<=", 65535)],
"worker_num": (">=", 1),
}
ServerYamlConfChecker.check_conf(conf, default_conf, conf_type,
conf_qualification)
@staticmethod
def check_tracer_conf(conf):
default_conf = {"interval_s": -1, }
conf_type = {"interval_s": int, }
conf_qualification = {}
ServerYamlConfChecker.check_conf(conf, default_conf, conf_type,
conf_qualification)
@staticmethod
def check_dag_conf(conf):
default_conf = {
"retry": 1,
"client_type": "brpc",
"use_profile": False,
"channel_size": 0,
"is_thread_op": True,
"tracer": {},
}
conf_type = {
"retry": int,
"client_type": str,
"use_profile": bool,
"channel_size": int,
"is_thread_op": bool,
}
conf_qualification = {
"retry": (">=", 1),
"client_type": ("in", ["brpc", "grpc"]),
"channel_size": (">=", 0),
}
ServerYamlConfChecker.check_conf(conf, default_conf, conf_type,
conf_qualification)
@staticmethod
def fill_with_default_conf(conf, default_conf):
for key, val in default_conf.items():
if conf.get(key) is None:
_LOGGER.warning("[CONF] {} not set, use default: {}"
.format(key, val))
conf[key] = val
@staticmethod
def check_conf_type(conf, conf_type):
for key, val in conf_type.items():
if not isinstance(conf[key], val):
raise SystemExit("[CONF] {} must be {} type, but get {}."
.format(key, val, type(conf[key])))
@staticmethod
def check_conf_qualification(conf, conf_qualification):
for key, qualification in conf_qualification.items():
if not isinstance(qualification, list):
qualification = [qualification]
if not ServerYamlConfChecker.qualification_check(conf[key],
qualification):
raise SystemExit("[CONF] {} must be {}, but get {}."
.format(key, ", ".join([
"{} {}"
.format(q[0], q[1]) for q in qualification
]), conf[key]))
@staticmethod
def qualification_check(value, qualifications):
if not isinstance(qualifications, list):
qualifications = [qualifications]
ok = True
for q in qualifications:
operator, limit = q
if operator == "<":
ok = value < limit
elif operator == "==":
ok = value == limit
elif operator == ">":
ok = value > limit
elif operator == "<=":
ok = value <= limit
elif operator == ">=":
ok = value >= limit
elif operator == "in":
ok = value in limit
else:
raise SystemExit("unknow operator: {}".format(operator))
if ok == False:
break
return ok
python/pipeline/profiler.py
浏览文件 @ f3953aa5
......@@ -22,10 +22,165 @@ elif sys.version_info.major == 3:
import queue as Queue
else:
raise Exception("Error Python version")
from time import time as _time
import time
import threading
import multiprocessing
_LOGGER = logging.getLogger()
_LOGGER = logging.getLogger(__name__)
_LOGGER.propagate = False
class PerformanceTracer(object):
def __init__(self, is_thread_mode, interval_s, server_worker_num):
self._is_thread_mode = is_thread_mode
if is_thread_mode:
# Because the Channel in the thread mode cannot be
# accessed across processes, when using thread mode,
# the PerformanceTracer is also the thread mode.
# However, performance may be affected by GIL.
self._data_buffer = Queue.Queue()
else:
self._data_buffer = multiprocessing.Manager().Queue()
self._interval_s = interval_s
self._thrd = None
self._proc = None
self._channels = []
# The size of data in Channel will not exceed server_worker_num
self._server_worker_num = server_worker_num
def data_buffer(self):
return self._data_buffer
def start(self):
if self._is_thread_mode:
self._thrd = threading.Thread(
target=self._trace_func, args=(self._channels, ))
self._thrd.daemon = True
self._thrd.start()
else:
self._proc = multiprocessing.Process(
target=self._trace_func, args=(self._channels, ))
self._proc.daemon = True
self._proc.start()
def set_channels(self, channels):
self._channels = channels
def _trace_func(self, channels):
all_actions = ["in", "prep", "midp", "postp", "out"]
calcu_actions = ["prep", "midp", "postp"]
while True:
op_cost = {}
err_request = []
err_count = 0
_LOGGER.info("==================== TRACER ======================")
# op
while True:
try:
item = self._data_buffer.get_nowait()
name = item["name"]
actions = item["actions"]
if name == "DAG":
succ = item["succ"]
req_id = item["id"]
if not succ:
err_count += 1
err_request.append(req_id)
if name not in op_cost:
op_cost[name] = {}
for action, cost in actions.items():
if action not in op_cost[name]:
op_cost[name][action] = []
op_cost[name][action].append(cost)
except Queue.Empty:
break
if len(op_cost) != 0:
for name in op_cost:
tot_cost, calcu_cost = 0.0, 0.0
for action, costs in op_cost[name].items():
op_cost[name][action] = sum(costs) / (1e3 * len(costs))
tot_cost += op_cost[name][action]
if name != "DAG":
_LOGGER.info("Op({}):".format(name))
for action in all_actions:
if action in op_cost[name]:
_LOGGER.info("\t{}[{} ms]".format(
action, op_cost[name][action]))
for action in calcu_actions:
if action in op_cost[name]:
calcu_cost += op_cost[name][action]
_LOGGER.info("\tidle[{}]".format(1 - 1.0 * calcu_cost /
tot_cost))
if "DAG" in op_cost:
calls = op_cost["DAG"].values()
calls.sort()
tot = len(calls)
qps = 1.0 * tot / self._interval_s
ave_cost = sum(calls) / tot
latencys = [50, 60, 70, 80, 90, 95, 99]
_LOGGER.info("DAGExecutor:")
_LOGGER.info("\tQuery count[{}]".format(tot))
_LOGGER.info("\tQPS[{} q/s]".format(qps))
_LOGGER.info("\tSucc[{}]".format(1 - 1.0 * err_count / tot))
_LOGGER.info("\tError req[{}]".format(", ".join(
[str(x) for x in err_request])))
_LOGGER.info("\tLatency:")
_LOGGER.info("\t\tave[{} ms]".format(ave_cost))
for latency in latencys:
_LOGGER.info("\t\t.{}[{} ms]".format(latency, calls[int(
tot * latency / 100.0)]))
# channel
_LOGGER.info("Channel (server worker num[{}]):".format(
self._server_worker_num))
for channel in channels:
_LOGGER.info("\t{}(In: {}, Out: {}) size[{}/{}]".format(
channel.name,
channel.get_producers(),
channel.get_consumers(),
channel.size(), channel.get_maxsize()))
time.sleep(self._interval_s)
class UnsafeTimeProfiler(object):
""" thread unsafe profiler """
def __init__(self):
self.pid = os.getpid()
self.print_head = 'PROFILE\tpid:{}\t'.format(self.pid)
self.time_record = [self.print_head]
self._enable = False
def enable(self, enable):
self._enable = enable
def record(self, name):
if self._enable is False:
return
timestamp = int(round(_time() * 1000000))
self.time_record.append('{}:{} '.format(name, timestamp))
return timestamp
def print_profile(self):
if self._enable is False:
return
sys.stderr.write(self.gen_profile_str())
def gen_profile_str(self):
if self._enable is False:
return
self.time_record.append('\n')
profile_str = ''.join(self.time_record)
self.time_record = [self.print_head]
return profile_str
class TimeProfiler(object):
......@@ -42,12 +197,13 @@ class TimeProfiler(object):
def record(self, name_with_tag):
if self._enable is False:
return
timestamp = int(round(time.time() * 1000000))
timestamp = int(round(_time() * 1000000))
name_with_tag = name_with_tag.split("_")
tag = name_with_tag[-1]
name = '_'.join(name_with_tag[:-1])
with self._lock:
self._time_record.put((name, tag, timestamp))
return timestamp
def print_profile(self):
if self._enable is False:
......
python/pipeline/util.py
浏览文件 @ f3953aa5
......@@ -13,13 +13,101 @@
# limitations under the License.
import sys
import logging
import threading
import multiprocessing
import multiprocessing.managers
if sys.version_info.major == 2:
import Queue
from Queue import PriorityQueue
elif sys.version_info.major == 3:
import queue as Queue
from queue import PriorityQueue
else:
raise Exception("Error Python version")
_LOGGER = logging.getLogger(__name__)
class NameGenerator(object):
# use unsafe-id-generator
def __init__(self, prefix):
self._idx = -1
self._prefix = prefix
self._id_generator = UnsafeIdGenerator(1000000000000000000)
def next(self):
next_id = self._id_generator.next()
return "{}{}".format(self._prefix, next_id)
class UnsafeIdGenerator(object):
def __init__(self, max_id, base_counter=0, step=1):
self._base_counter = base_counter
self._counter = self._base_counter
self._step = step
self._max_id = max_id # for reset
def next(self):
if self._counter >= self._max_id:
self._counter = self._base_counter
_LOGGER.info("Reset Id: {}".format(self._counter))
next_id = self._counter
self._counter += self._step
return next_id
class ThreadIdGenerator(UnsafeIdGenerator):
def __init__(self, max_id, base_counter=0, step=1, lock=None):
# if you want to use your lock, you may need to use Reentrant-Lock
self._lock = lock
if self._lock is None:
self._lock = threading.Lock()
super(ThreadIdGenerator, self).__init__(max_id, base_counter, step)
def next(self):
self._idx += 1
return "{}{}".format(self._prefix, self._idx)
next_id = None
with self._lock:
if self._counter >= self._max_id:
self._counter = self._base_counter
_LOGGER.info("Reset Id: {}".format(self._counter))
next_id = self._counter
self._counter += self._step
return next_id
class ProcessIdGenerator(UnsafeIdGenerator):
def __init__(self, max_id, base_counter=0, step=1, lock=None):
# if you want to use your lock, you may need to use Reentrant-Lock
self._lock = lock
if self._lock is None:
self._lock = multiprocessing.Lock()
self._base_counter = base_counter
self._counter = multiprocessing.Manager().Value('i', 0)
self._step = step
self._max_id = max_id
def next(self):
next_id = None
with self._lock:
if self._counter.value >= self._max_id:
self._counter.value = self._base_counter
_LOGGER.info("Reset Id: {}".format(self._counter.value))
next_id = self._counter.value
self._counter.value += self._step
return next_id
def PipelineProcSyncManager():
"""
add PriorityQueue into SyncManager, see more:
https://stackoverflow.com/questions/25324560/strange-queue-priorityqueue-behaviour-with-multiprocessing-in-python-2-7-6?answertab=active#tab-top
"""
class PipelineManager(multiprocessing.managers.SyncManager):
pass
PipelineManager.register("PriorityQueue", PriorityQueue)
m = PipelineManager()
m.start()
return m
python/requirements.txt
浏览文件 @ f3953aa5
......@@ -3,6 +3,7 @@ wheel>=0.34.0, <0.35.0
setuptools>=44.1.0
opencv-python==4.2.0.32
google>=2.0.3
opencv-python==4.2.0.32
protobuf>=3.12.2
grpcio-tools>=1.28.1
grpcio>=1.28.1
......
python/setup.py.app.in
浏览文件 @ f3953aa5
......@@ -42,7 +42,7 @@ if '${PACK}' == 'ON':
REQUIRED_PACKAGES = [
'six >= 1.10.0', 'sentencepiece', 'opencv-python', 'pillow',
'six >= 1.10.0', 'sentencepiece', 'opencv-python<=4.2.0.32', 'pillow',
'shapely', 'pyclipper'
]
......
tools/serving_build.sh
浏览文件 @ f3953aa5
......@@ -776,7 +776,7 @@ function python_test_pipeline(){
# test: thread servicer & thread op
cat << EOF > config.yml
port: 18080
worker_num: 2
worker_num: 4
build_dag_each_worker: false
dag:
is_thread_op: true
......@@ -793,7 +793,7 @@ EOF
# test: thread servicer & process op
cat << EOF > config.yml
port: 18080
worker_num: 2
worker_num: 4
build_dag_each_worker: false
dag:
is_thread_op: false
......@@ -807,13 +807,13 @@ EOF
ps -ef | grep "pipeline_server" | grep -v grep | awk '{print $2}' | xargs kill
kill_process_by_port 18080
# test: process servicer & thread op
# test: process servicer & process op
cat << EOF > config.yml
port: 18080
worker_num: 2
build_dag_each_worker: true
worker_num: 4
build_dag_each_worker: false
dag:
is_thread_op: flase
is_thread_op: false
client_type: brpc
retry: 1
use_profile: false
......@@ -824,11 +824,13 @@ EOF
ps -ef | grep "pipeline_server" | grep -v grep | awk '{print $2}' | xargs kill
kill_process_by_port 18080
# test: process servicer & process op
# test: process servicer & thread op
pip uninstall grpcio -y
pip install grpcio --no-binary=grpcio
cat << EOF > config.yml
port: 18080
worker_num: 2
build_dag_each_worker: false
worker_num: 4
build_dag_each_worker: true
dag:
is_thread_op: false
client_type: brpc
......@@ -851,7 +853,7 @@ EOF
sleep 5
cat << EOF > config.yml
port: 18080
worker_num: 2
worker_num: 4
build_dag_each_worker: false
dag:
is_thread_op: false
......
Markdown is supported
0% 或 .
You are about to add 0 people to the discussion. Proceed with caution.
先完成此消息的编辑！
想要评论请 注册 或