Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

abacus-aibox-991 fix the bug of add_float_mask_data's error place #59

Merged
merged 1 commit into from
Mar 18, 2024
Merged

abacus-aibox-991 fix the bug of add_float_mask_data's error place #59

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
14 changes: 11 additions & 3 deletions paddle/fluid/framework/fleet/box_wrapper.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -731,8 +731,11 @@ class FloatMaskMetricMsg : public MetricMsg {
"the predict data length should be consistent with "
"the label data length"));
auto cal = GetCalculator();
auto pre_var_place = GetVarPlace(exe_scope, pred_varname_);
auto label_var_place = GetVarPlace(exe_scope, label_varname_);
auto mask_var_place = GetVarPlace(exe_scope, mask_varname_);
cal->add_float_mask_data(
pred_data, label_data, mask_data, label_len, place);
pred_data, label_data, mask_data, label_len, pre_var_place, label_var_place, mask_var_place);
}

protected:
Expand Down Expand Up @@ -775,8 +778,11 @@ class ContinueMaskMetricMsg : public MetricMsg {
"the predict data length should be consistent with "
"the label data length"));
auto cal = GetCalculator();
auto pre_var_place = GetVarPlace(exe_scope, pred_varname_);
auto label_var_place = GetVarPlace(exe_scope, label_varname_);
auto mask_var_place = GetVarPlace(exe_scope, mask_varname_);
cal->add_continue_mask_data(
pred_data, label_data, mask_data, label_len, place);
pred_data, label_data, mask_data, label_len, pre_var_place, label_var_place, mask_var_place);
}

protected:
Expand Down Expand Up @@ -913,8 +919,10 @@ class NanInfMetricMsg : public MetricMsg {
"the predict data length should be consistent with "
"the label data length"));
auto cal = GetCalculator();
auto pre_var_place = GetVarPlace(exe_scope, pred_varname_);
auto label_var_place = GetVarPlace(exe_scope, label_varname_);
cal->add_nan_inf_data(
pred_data, label_data, label_len, place);
pred_data, label_data, label_len, pre_var_place, label_var_place);
}
};

Expand Down
128 changes: 67 additions & 61 deletions paddle/fluid/framework/fleet/metrics.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -194,31 +194,34 @@ void BasicAucCalculator::add_mask_data(const float* d_pred,
void BasicAucCalculator::add_float_mask_data(const float* d_pred,
const float* d_label,
const int64_t* d_mask, int batch_size,
const paddle::platform::Place& place) {
if (platform::is_gpu_place(place) || platform::is_xpu_place(place)) {
thread_local std::vector<float> h_pred;
thread_local std::vector<float> h_label;
thread_local std::vector<int64_t> h_mask;
const paddle::platform::Place& place_pred,
const paddle::platform::Place& place_label,
const paddle::platform::Place& place_mask) {
thread_local std::vector<float> h_pred;
thread_local std::vector<float> h_label;
thread_local std::vector<int64_t> h_mask;
const float* add_pred = d_pred;
const float* add_label = d_label;
const int64_t* add_mask = d_mask;
if (platform::is_gpu_place(place_pred) || platform::is_xpu_place(place_pred)) {
h_pred.resize(batch_size);
SyncCopyD2H(h_pred.data(), d_pred, batch_size, place_pred);
add_pred = h_pred.data();
}
if (platform::is_gpu_place(place_label) || platform::is_xpu_place(place_label)) {
h_label.resize(batch_size);
SyncCopyD2H(h_label.data(), d_label, batch_size, place_label);
add_label = h_label.data();
}
if (platform::is_gpu_place(place_mask) || platform::is_xpu_place(place_mask)) {
h_mask.resize(batch_size);

SyncCopyD2H(h_pred.data(), d_pred, batch_size, place);
SyncCopyD2H(h_label.data(), d_label, batch_size, place);
SyncCopyD2H(h_mask.data(), d_mask, batch_size, place);

std::lock_guard<std::mutex> lock(_table_mutex);
for (int i = 0; i < batch_size; ++i) {
if (h_mask[i]) {
add_unlock_data_with_float_label(h_pred[i], h_label[i]);
}
}
} else {
std::lock_guard<std::mutex> lock(_table_mutex);
for (int i = 0; i < batch_size; ++i) {
if (d_mask[i]) {
add_unlock_data_with_float_label(d_pred[i], d_label[i]);
}
SyncCopyD2H(h_mask.data(), d_mask, batch_size, place_mask);
add_mask = h_mask.data();
}
std::lock_guard<std::mutex> lock(_table_mutex);
for (int i = 0; i < batch_size; ++i) {
if (add_mask[i]) {
add_unlock_data_with_float_label(add_pred[i], add_label[i]);
}
}
}
Expand All @@ -229,31 +232,34 @@ void BasicAucCalculator::add_continue_mask_data(
const float* d_label,
const int64_t* d_mask,
int batch_size,
const paddle::platform::Place& place) {
if (platform::is_gpu_place(place) || platform::is_xpu_place(place)) {
thread_local std::vector<float> h_pred;
thread_local std::vector<float> h_label;
thread_local std::vector<int64_t> h_mask;
const paddle::platform::Place& place_pred,
const paddle::platform::Place& place_label,
const paddle::platform::Place& place_mask) {
thread_local std::vector<float> h_pred;
thread_local std::vector<float> h_label;
thread_local std::vector<int64_t> h_mask;
const float* add_pred = d_pred;
const float* add_label = d_label;
const int64_t* add_mask = d_mask;
if (platform::is_gpu_place(place_pred) || platform::is_xpu_place(place_pred)) {
h_pred.resize(batch_size);
SyncCopyD2H(h_pred.data(), d_pred, batch_size, place_pred);
add_pred = h_pred.data();
}
if (platform::is_gpu_place(place_label) || platform::is_xpu_place(place_label)) {
h_label.resize(batch_size);
SyncCopyD2H(h_label.data(), d_label, batch_size, place_label);
add_label = h_label.data();
}
if (platform::is_gpu_place(place_mask) || platform::is_xpu_place(place_mask)) {
h_mask.resize(batch_size);

SyncCopyD2H(h_pred.data(), d_pred, batch_size, place);
SyncCopyD2H(h_label.data(), d_label, batch_size, place);
SyncCopyD2H(h_mask.data(), d_mask, batch_size, place);

std::lock_guard<std::mutex> lock(_table_mutex);
for (int i = 0; i < batch_size; ++i) {
if (h_mask[i]) {
add_unlock_data_with_continue_label(h_pred[i], h_label[i]);
}
}
} else {
std::lock_guard<std::mutex> lock(_table_mutex);
for (int i = 0; i < batch_size; ++i) {
if (d_mask[i]) {
add_unlock_data_with_continue_label(d_pred[i], d_label[i]);
}
SyncCopyD2H(h_mask.data(), d_mask, batch_size, place_mask);
add_mask = h_mask.data();
}
std::lock_guard<std::mutex> lock(_table_mutex);
for (int i = 0; i < batch_size; ++i) {
if (add_mask[i]) {
add_unlock_data_with_continue_label(add_pred[i], add_label[i]);
}
}
}
Expand Down Expand Up @@ -438,26 +444,26 @@ void BasicAucCalculator::add_uid_data(const float* d_pred,
void BasicAucCalculator::add_nan_inf_data(const float* d_pred,
const int64_t* d_label,
int batch_size,
const paddle::platform::Place& place){
if (platform::is_gpu_place(place) || platform::is_xpu_place(place)) {
thread_local std::vector<float> h_pred;
thread_local std::vector<int64_t> h_label;
const paddle::platform::Place& place_pred,
const paddle::platform::Place& place_label){
thread_local std::vector<float> h_pred;
thread_local std::vector<int64_t> h_label;
const float* add_pred = d_pred;
const int64_t* add_label = d_label;
if (platform::is_gpu_place(place_pred) || platform::is_xpu_place(place_pred)) {
h_pred.resize(batch_size);
SyncCopyD2H(h_pred.data(), d_pred, batch_size, place_pred);
add_pred = h_pred.data();
}
if (platform::is_gpu_place(place_label) || platform::is_xpu_place(place_label)) {
h_label.resize(batch_size);
SyncCopyD2H(h_pred.data(), d_pred, batch_size, place);
SyncCopyD2H(h_label.data(), d_label, batch_size, place);

std::lock_guard<std::mutex> lock(_table_mutex);
for (int i = 0; i < batch_size; ++i) {
add_nan_inf_unlock_data(h_pred[i], h_label[i]);
}
} else {
std::lock_guard<std::mutex> lock(_table_mutex);
for (int i = 0; i < batch_size; ++i) {
add_nan_inf_unlock_data(d_pred[i], d_label[i]);
}
SyncCopyD2H(h_label.data(), d_label, batch_size, place_label);
add_label = h_label.data();
}
std::lock_guard<std::mutex> lock(_table_mutex);
for (int i = 0; i < batch_size; ++i) {
add_nan_inf_unlock_data(add_pred[i], add_label[i]);
}

}

void BasicAucCalculator::add_uid_unlock_data(double pred,
Expand Down
11 changes: 8 additions & 3 deletions paddle/fluid/framework/fleet/metrics.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -83,13 +83,17 @@ class BasicAucCalculator {
const float* d_label,
const int64_t* d_mask,
int batch_size,
const paddle::platform::Place& place);
const paddle::platform::Place& place_pred,
const paddle::platform::Place& place_label,
const paddle::platform::Place& place_mask);
// add continue data
void add_continue_mask_data(const float* d_pred,
const float* d_label,
const int64_t* d_mask,
int batch_size,
const paddle::platform::Place& place);
const paddle::platform::Place& place_pred,
const paddle::platform::Place& place_label,
const paddle::platform::Place& place_mask);
// add sample data
void add_sample_data(const float* d_pred,
const int64_t* d_label,
Expand All @@ -109,7 +113,8 @@ class BasicAucCalculator {
void add_nan_inf_data(const float* d_pred,
const int64_t* d_label,
int batch_size,
const paddle::platform::Place& place);
const paddle::platform::Place& place_pred,
const paddle::platform::Place& place_label);

void compute();
void computeContinueMsg();
Expand Down