Skip to content

S
Serving

PaddlePaddle / Serving
1 年多 前同步成功

通知 186
Star 833
Fork 253
S
Serving
提交 02f04262 编写于 作者: M MRXLT
浏览文件
操作

support int32

上级 c8c7f4d0
隐藏空白更改
内联 并排
Showing with 168 addition and 36 deletion
core/general-client/include/general_model.h
浏览文件 @ 02f04262
......@@ -49,6 +49,8 @@ class ModelRes {
res._int64_value_map.end());
_float_value_map.insert(res._float_value_map.begin(),
res._float_value_map.end());
_int32_value_map.insert(res._int32_value_map.begin(),
res._int32_value_map.end());
_shape_map.insert(res._shape_map.begin(), res._shape_map.end());
_lod_map.insert(res._lod_map.begin(), res._lod_map.end());
}
......@@ -60,6 +62,9 @@ class ModelRes {
_float_value_map.insert(
std::make_move_iterator(std::begin(res._float_value_map)),
std::make_move_iterator(std::end(res._float_value_map)));
_int32_value_map.insert(
std::make_move_iterator(std::begin(res._int32_value_map)),
std::make_move_iterator(std::end(res._int32_value_map)));
_shape_map.insert(std::make_move_iterator(std::begin(res._shape_map)),
std::make_move_iterator(std::end(res._shape_map)));
_lod_map.insert(std::make_move_iterator(std::begin(res._lod_map)),
......@@ -78,6 +83,12 @@ class ModelRes {
std::vector<float>&& get_float_by_name_with_rv(const std::string& name) {
return std::move(_float_value_map[name]);
}
const std::vector<int32_t>& get_int32_by_name(const std::string& name) {
return _int32_value_map[name];
}
std::vector<int32_t>&& get_int32_by_name_with_rv(const std::string& name) {
return std::move(_int32_value_map[name]);
}
const std::vector<int>& get_shape_by_name(const std::string& name) {
return _shape_map[name];
}
......@@ -103,6 +114,9 @@ class ModelRes {
_float_value_map.insert(
std::make_move_iterator(std::begin(res._float_value_map)),
std::make_move_iterator(std::end(res._float_value_map)));
_int32_value_map.insert(
std::make_move_iterator(std::begin(res._int32_value_map)),
std::make_move_iterator(std::end(res._int32_value_map)));
_shape_map.insert(std::make_move_iterator(std::begin(res._shape_map)),
std::make_move_iterator(std::end(res._shape_map)));
_lod_map.insert(std::make_move_iterator(std::begin(res._lod_map)),
......@@ -115,6 +129,7 @@ class ModelRes {
std::string _engine_name;
std::map<std::string, std::vector<int64_t>> _int64_value_map;
std::map<std::string, std::vector<float>> _float_value_map;
std::map<std::string, std::vector<int32_t>> _int32_value_map;
std::map<std::string, std::vector<int>> _shape_map;
std::map<std::string, std::vector<int>> _lod_map;
};
......@@ -145,6 +160,14 @@ class PredictorRes {
const std::string& name) {
return std::move(_models[model_idx].get_float_by_name_with_rv(name));
}
const std::vector<int32_t>& get_int32_by_name(const int model_idx,
const std::string& name) {
return _models[model_idx].get_int32_by_name(name);
}
std::vector<int32_t>&& get_int32_by_name_with_rv(const int model_idx,
const std::string& name) {
return std::move(_models[model_idx].get_int32_by_name_with_rv(name));
}
const std::vector<int>& get_shape_by_name(const int model_idx,
const std::string& name) {
return _models[model_idx].get_shape_by_name(name);
......
core/general-client/src/general_model.cpp
浏览文件 @ 02f04262
......@@ -207,17 +207,28 @@ int PredictorClient::batch_predict(
for (auto &name : int_feed_name) {
int idx = _feed_name_to_idx[name];
Tensor *tensor = tensor_vec[idx];
VLOG(2) << "prepare int feed " << name << " shape size "
<< int_shape[vec_idx].size();
if (_type[idx] == 0) {
VLOG(2) << "prepare int64 feed " << name << " shape size "
<< int_shape[vec_idx].size();
VLOG(3) << "feed var name " << name << " index " << vec_idx
<< "first data " << int_feed[vec_idx][0];
for (uint32_t j = 0; j < int_feed[vec_idx].size(); ++j) {
tensor->add_int64_data(int_feed[vec_idx][j]);
}
} else if (_type[idx] == 2) {
VLOG(2) << "prepare int32 feed " << name << " shape size "
<< int_shape[vec_idx].size();
VLOG(3) << "feed var name " << name << " index " << vec_idx
<< "first data " << int32_t(int_feed[vec_idx][0]);
for (uint32_t j = 0; j < int_feed[vec_idx].size(); ++j) {
tensor->add_int_data(int32_t(int_feed[vec_idx][j]));
}
}
for (uint32_t j = 0; j < int_shape[vec_idx].size(); ++j) {
tensor->add_shape(int_shape[vec_idx][j]);
}
tensor->set_elem_type(0);
VLOG(3) << "feed var name " << name << " index " << vec_idx
<< "first data " << int_feed[vec_idx][0];
for (uint32_t j = 0; j < int_feed[vec_idx].size(); ++j) {
tensor->add_int64_data(int_feed[vec_idx][j]);
}
tensor->set_elem_type(_type[idx]);
vec_idx++;
}
......@@ -284,7 +295,7 @@ int PredictorClient::batch_predict(
for (auto &name : fetch_name) {
// int idx = _fetch_name_to_idx[name];
if (_fetch_name_to_type[name] == 0) {
VLOG(2) << "ferch var " << name << "type int";
VLOG(2) << "fetch var " << name << " type int64";
model._int64_value_map[name].resize(
output.insts(0).tensor_array(idx).int64_data_size());
int size = output.insts(0).tensor_array(idx).int64_data_size();
......@@ -292,8 +303,8 @@ int PredictorClient::batch_predict(
model._int64_value_map[name][i] =
output.insts(0).tensor_array(idx).int64_data(i);
}
} else {
VLOG(2) << "fetch var " << name << "type float";
} else if (_fetch_name_to_type[name] == 1) {
VLOG(2) << "fetch var " << name << " type float";
model._float_value_map[name].resize(
output.insts(0).tensor_array(idx).float_data_size());
int size = output.insts(0).tensor_array(idx).float_data_size();
......@@ -301,7 +312,10 @@ int PredictorClient::batch_predict(
model._float_value_map[name][i] =
output.insts(0).tensor_array(idx).float_data(i);
}
} else if (_fetch_name_to_type[name] == 2) {
VLOG(2) << "fetch var " << name << " type int32";
}
idx += 1;
}
predict_res_batch.add_model_res(std::move(model));
......@@ -448,12 +462,19 @@ int PredictorClient::numpy_predict(
for (auto &name : int_feed_name) {
int idx = _feed_name_to_idx[name];
Tensor *tensor = tensor_vec[idx];
VLOG(2) << "prepare int feed " << name << " shape size "
<< int_shape[vec_idx].size();
for (uint32_t j = 0; j < int_shape[vec_idx].size(); ++j) {
tensor->add_shape(int_shape[vec_idx][j]);
}
tensor->set_elem_type(0);
tensor->set_elem_type(_type[idx]);
if (_type[idx] == 0) {
VLOG(2) << "prepare int feed " << name << " shape size "
<< int_shape[vec_idx].size();
} else {
VLOG(2) << "prepare int32 feed " << name << " shape size "
<< int_shape[vec_idx].size();
}
const int int_shape_size = int_shape[vec_idx].size();
switch (int_shape_size) {
......@@ -463,7 +484,11 @@ int PredictorClient::numpy_predict(
for (ssize_t j = 0; j < int_array.shape(1); j++) {
for (ssize_t k = 0; k < int_array.shape(2); k++) {
for (ssize_t l = 0; k < int_array.shape(3); l++) {
tensor->add_int64_data(int_array(i, j, k, l));
if (_type[idx] == 0) {
tensor->add_int64_data(int_array(i, j, k, l));
} else {
tensor->add_int_data(int_array(i, j, k, l));
}
}
}
}
......@@ -475,7 +500,11 @@ int PredictorClient::numpy_predict(
for (ssize_t i = 0; i < int_array.shape(0); i++) {
for (ssize_t j = 0; j < int_array.shape(1); j++) {
for (ssize_t k = 0; k < int_array.shape(2); k++) {
tensor->add_int64_data(int_array(i, j, k));
if (_type[idx] == 0) {
tensor->add_int64_data(int_array(i, j, k));
} else {
tensor->add_int_data(int_array(i, j, k));
}
}
}
}
......@@ -485,7 +514,11 @@ int PredictorClient::numpy_predict(
auto int_array = int_feed[vec_idx].unchecked<2>();
for (ssize_t i = 0; i < int_array.shape(0); i++) {
for (ssize_t j = 0; j < int_array.shape(1); j++) {
tensor->add_int64_data(int_array(i, j));
if (_type[idx] == 0) {
tensor->add_int64_data(int_array(i, j));
} else {
tensor->add_int_data(int_array(i, j));
}
}
}
break;
......@@ -493,7 +526,11 @@ int PredictorClient::numpy_predict(
case 1: {
auto int_array = int_feed[vec_idx].unchecked<1>();
for (ssize_t i = 0; i < int_array.shape(0); i++) {
tensor->add_int64_data(int_array(i));
if (_type[idx] == 0) {
tensor->add_int64_data(int_array(i));
} else {
tensor->add_int_data(int_array(i));
}
}
break;
}
......@@ -563,7 +600,7 @@ int PredictorClient::numpy_predict(
for (auto &name : fetch_name) {
// int idx = _fetch_name_to_idx[name];
if (_fetch_name_to_type[name] == 0) {
VLOG(2) << "ferch var " << name << "type int";
VLOG(2) << "ferch var " << name << "type int64";
model._int64_value_map[name].resize(
output.insts(0).tensor_array(idx).int64_data_size());
int size = output.insts(0).tensor_array(idx).int64_data_size();
......@@ -571,7 +608,7 @@ int PredictorClient::numpy_predict(
model._int64_value_map[name][i] =
output.insts(0).tensor_array(idx).int64_data(i);
}
} else {
} else if (_fetch_name_to_type[name] == 1) {
VLOG(2) << "fetch var " << name << "type float";
model._float_value_map[name].resize(
output.insts(0).tensor_array(idx).float_data_size());
......@@ -580,6 +617,15 @@ int PredictorClient::numpy_predict(
model._float_value_map[name][i] =
output.insts(0).tensor_array(idx).float_data(i);
}
} else if (_fetch_name_to_type[name] == 2) {
VLOG(2) << "fetch var " << name << "type int32";
model._int32_value_map[name].resize(
output.insts(0).tensor_array(idx).int_data_size());
int size = output.insts(0).tensor_array(idx).int_data_size();
for (int i = 0; i < size; ++i) {
model._int64_value_map[name][i] =
output.insts(0).tensor_array(idx).int_data(i);
}
}
idx += 1;
}
......@@ -613,7 +659,6 @@ int PredictorClient::numpy_predict(
_api.thrd_clear();
return 0;
}
} // namespace general_model
} // namespace paddle_serving
} // namespace baidu
core/general-server/op/general_reader_op.cpp
浏览文件 @ 02f04262
......@@ -126,9 +126,12 @@ int GeneralReaderOp::inference() {
if (elem_type[i] == 0) { // int64
elem_size[i] = sizeof(int64_t);
lod_tensor.dtype = paddle::PaddleDType::INT64;
} else {
} else if (elem_type[i] == 1) {
elem_size[i] = sizeof(float);
lod_tensor.dtype = paddle::PaddleDType::FLOAT32;
} else if (elem_type[i] == 2) {
elem_size[i] = sizeof(int32_t);
lod_tensor.dtype = paddle::PaddleDType::INT32;
}
if (model_config->_is_lod_feed[i]) {
......@@ -159,8 +162,10 @@ int GeneralReaderOp::inference() {
int data_len = 0;
if (tensor.int64_data_size() > 0) {
data_len = tensor.int64_data_size();
} else {
} else if (tensor.float_data_size() > 0) {
data_len = tensor.float_data_size();
} else if (tensor.int_data_size() > 0) {
data_len = tensor.int_data_size();
}
VLOG(2) << "tensor size for var[" << i << "]: " << data_len;
tensor_size += data_len;
......@@ -210,7 +215,7 @@ int GeneralReaderOp::inference() {
offset += capacity[i];
}
}
} else {
} else if (elem_type[i] == 1) {
float *dst_ptr = static_cast<float *>(out->at(i).data.data());
int offset = 0;
for (int j = 0; j < batch_size; ++j) {
......@@ -224,6 +229,20 @@ int GeneralReaderOp::inference() {
offset += capacity[i];
}
}
} else if (elem_type[i] == 2) {
int32_t *dst_ptr = static_cast<int32_t *>(out->at(i).data.data());
int offset = 0;
for (int j = 0; j < batch_size; ++j) {
int elem_num = req->insts(j).tensor_array(i).int_data_size();
for (int k = 0; k < elem_num; ++k) {
dst_ptr[offset + k] = req->insts(j).tensor_array(i).int_data(k);
}
if (out->at(i).lod.size() == 1) {
offset = out->at(i).lod[0][j + 1];
} else {
offset += capacity[i];
}
}
}
}
......
core/general-server/op/general_response_op.cpp
浏览文件 @ 02f04262
......@@ -157,6 +157,27 @@ int GeneralResponseOp::inference() {
}
VLOG(2) << "fetch var [" << model_config->_fetch_name[idx] << "] ready";
var_idx++;
} else if (in->at(idx).dtype == paddle::PaddleDType::INT32) {
VLOG(2) << "Prepare float var [" << model_config->_fetch_name[idx]
<< "].";
int32_t *data_ptr = static_cast<int32_t *>(in->at(idx).data.data());
if (model_config->_is_lod_fetch[idx]) {
FetchInst *fetch_p = output->mutable_insts(0);
for (int j = 0; j < in->at(idx).lod[0].size(); ++j) {
fetch_p->mutable_tensor_array(var_idx)->add_lod(
in->at(idx).lod[0][j]);
}
for (int j = 0; j < cap; ++j) {
fetch_p->mutable_tensor_array(var_idx)->add_int_data(data_ptr[j]);
}
} else {
FetchInst *fetch_p = output->mutable_insts(0);
for (int j = 0; j < cap; ++j) {
fetch_p->mutable_tensor_array(var_idx)->add_int_data(data_ptr[j]);
}
}
VLOG(2) << "fetch var [" << model_config->_fetch_name[idx] << "] ready";
var_idx++;
}
}
}
......
python/paddle_serving_client/__init__.py
浏览文件 @ 02f04262
......@@ -28,8 +28,11 @@ sys.path.append(
os.path.join(os.path.abspath(os.path.dirname(__file__)), 'proto'))
from .proto import multi_lang_general_model_service_pb2_grpc
int_type = 0
float_type = 1
int64_type = 0
float32_type = 1
int32_type = 2
int_type = set([int64_type, int32_type])
float_type = set([float32_type])
class _NOPProfiler(object):
......@@ -279,7 +282,7 @@ class Client(object):
raise ValueError("Wrong feed name: {}.".format(key))
#if not isinstance(feed_i[key], np.ndarray):
self.shape_check(feed_i, key)
if self.feed_types_[key] == int_type:
if self.feed_types_[key] in int_type:
if i == 0:
int_feed_names.append(key)
if isinstance(feed_i[key], np.ndarray):
......@@ -292,7 +295,7 @@ class Client(object):
else:
int_slot.append(feed_i[key])
self.all_numpy_input = False
elif self.feed_types_[key] == float_type:
elif self.feed_types_[key] in float_type:
if i == 0:
float_feed_names.append(key)
if isinstance(feed_i[key], np.ndarray):
......@@ -339,7 +342,7 @@ class Client(object):
result_map = {}
# result map needs to be a numpy array
for i, name in enumerate(fetch_names):
if self.fetch_names_to_type_[name] == int_type:
if self.fetch_names_to_type_[name] == int64_type:
# result_map[name] will be py::array(numpy array)
result_map[name] = result_batch_handle.get_int64_by_name(
mi, name)
......@@ -348,7 +351,7 @@ class Client(object):
if name in self.lod_tensor_set:
result_map["{}.lod".format(
name)] = result_batch_handle.get_lod(mi, name)
elif self.fetch_names_to_type_[name] == float_type:
elif self.fetch_names_to_type_[name] == float32_type:
result_map[name] = result_batch_handle.get_float_by_name(
mi, name)
shape = result_batch_handle.get_shape(mi, name)
......@@ -356,6 +359,16 @@ class Client(object):
if name in self.lod_tensor_set:
result_map["{}.lod".format(
name)] = result_batch_handle.get_lod(mi, name)
elif self.fetch_names_to_type_[name] == int32_type:
# result_map[name] will be py::array(numpy array)
result_map[name] = result_batch_handle.get_int32_by_name(
mi, name)
shape = result_batch_handle.get_shape(mi, name)
result_map[name].shape = shape
if name in self.lod_tensor_set:
result_map["{}.lod".format(
name)] = result_batch_handle.get_lod(mi, name)
multi_result_map.append(result_map)
ret = None
if len(model_engine_names) == 1:
......@@ -454,6 +467,8 @@ class MultiLangClient(object):
data = np.array(var, dtype="int64")
elif v_type == 1: # float32
data = np.array(var, dtype="float32")
elif v_type == 2: #int32
data = np.array(var, dtype="int32")
else:
raise Exception("error type.")
else:
......@@ -472,6 +487,11 @@ class MultiLangClient(object):
tensor.float_data.extend(var.reshape(-1).tolist())
else:
tensor.float_data.extend(self._flatten_list(var))
elif v_type == 2: #int32
if isinstance(car, np.array):
tensor.int_data.extend(var.reshape(-1).tolist())
else:
tensor.int_data.extend(self._flatten_list(var))
else:
raise Exception("error type.")
if isinstance(var, np.ndarray):
......@@ -503,6 +523,9 @@ class MultiLangClient(object):
elif v_type == 1: # float32
result_map[name] = np.array(
list(var.float_data), dtype="float32")
elif v_type == 2: # int32
result_map[name] = np.array(
list(var.int_data), dtype="int32")
else:
raise Exception("error type.")
result_map[name].shape = list(var.shape)
......
python/paddle_serving_client/io/__init__.py
浏览文件 @ 02f04262
......@@ -48,16 +48,18 @@ def save_model(server_model_folder,
config = model_conf.GeneralModelConfig()
#int64 = 0; float32 = 1; int32 = 2;
for key in feed_var_dict:
feed_var = model_conf.FeedVar()
feed_var.alias_name = key
feed_var.name = feed_var_dict[key].name
feed_var.is_lod_tensor = feed_var_dict[key].lod_level >= 1
if feed_var_dict[key].dtype == core.VarDesc.VarType.INT32 or \
feed_var_dict[key].dtype == core.VarDesc.VarType.INT64:
if feed_var_dict[key].dtype == core.VarDesc.VarType.INT64:
feed_var.feed_type = 0
if feed_var_dict[key].dtype == core.VarDesc.VarType.FP32:
feed_var.feed_type = 1
if feed_var_dict[key].dtype == core.VarDesc.VarType.INT32:
feed_var.feed_type = 2
if feed_var.is_lod_tensor:
feed_var.shape.extend([-1])
else:
......@@ -73,13 +75,12 @@ def save_model(server_model_folder,
fetch_var.alias_name = key
fetch_var.name = fetch_var_dict[key].name
fetch_var.is_lod_tensor = fetch_var_dict[key].lod_level >= 1
if fetch_var_dict[key].dtype == core.VarDesc.VarType.INT32 or \
fetch_var_dict[key].dtype == core.VarDesc.VarType.INT64:
if fetch_var_dict[key].dtype == core.VarDesc.VarType.INT64:
fetch_var.fetch_type = 0
if fetch_var_dict[key].dtype == core.VarDesc.VarType.FP32:
fetch_var.fetch_type = 1
if fetch_var_dict[key].dtype == core.VarDesc.VarType.INT32:
fetch_var.fetch_type = 2
if fetch_var.is_lod_tensor:
fetch_var.shape.extend([-1])
else:
......
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