Skip to content

P
Paddle

Crayon鑫 / Paddle
与 Fork 源项目一致

Fork自 PaddlePaddle / Paddle

通知 1
Star 1
Fork 0
P
Paddle
提交 9f6aaa3c 编写于 作者: S sandyhouse
浏览文件
操作

reformat code, test=develop

上级 b1a23b82
隐藏空白更改
内联 并排
Showing with 79 addition and 52 deletion
paddle/fluid/framework/pipeline_trainer.cc
浏览文件 @ 9f6aaa3c
...@@ -254,7 +254,7 @@ void PipelineTrainer::Finalize() { ...@@ -254,7 +254,7 @@ void PipelineTrainer::Finalize() {
const LoDTensor& minibatch_tensor = minibatch_ptr->Get<LoDTensor>(); const LoDTensor& minibatch_tensor = minibatch_ptr->Get<LoDTensor>();
TensorCopy(*static_cast<const Tensor*>(&minibatch_tensor), places_[0], TensorCopy(*static_cast<const Tensor*>(&minibatch_tensor), places_[0],
static_cast<Tensor*>(root_tensor)); static_cast<Tensor*>(root_tensor));
VLOG(4) << "Copy persitable var " << var->Name() << " to root scope"; VLOG(3) << "Copy persitable var " << var->Name() << " to root scope";
} }
} }
} }
......
paddle/fluid/framework/section_worker.cc
浏览文件 @ 9f6aaa3c
...@@ -154,6 +154,7 @@ void SectionWorker::TrainFiles() { ...@@ -154,6 +154,7 @@ void SectionWorker::TrainFiles() {
} }
} }
dev_ctx_->Wait(); dev_ctx_->Wait();
VLOG(0) << "real_microbatch_num for thread 0 " << real_microbatch_num; VLOG(0) << "real_microbatch_num for thread 0 " << real_microbatch_num;
// backward pass // backward pass
for (int i = 0; i < real_microbatch_num; ++i) { for (int i = 0; i < real_microbatch_num; ++i) {
...@@ -391,11 +392,13 @@ void SectionWorker::TrainFilesWithProfiler() { ...@@ -391,11 +392,13 @@ void SectionWorker::TrainFilesWithProfiler() {
op_total_time[op_idx] += time; op_total_time[op_idx] += time;
{ {
std::unique_lock<std::mutex> lk(cout_mutex); std::unique_lock<std::mutex> lk(cout_mutex);
std::cout << std::fixed; std::cout << std::fixed;
std::cout.precision(0); std::cout.precision(0);
std::cout << "::FWD:B[" << batch_id_ << "]:SEC[" << thread_id_ << "]:SCOPE[" << i std::cout << "::FWD:B[" << batch_id_ << "]:SEC[" << thread_id_
<< "]:OP[" << op->Type() << "]:START[" << start.tv_sec * 1e6 + start.tv_usec << "]:SCOPE[" << i << "]:OP[" << op->Type()
<< "]:END[" << end.tv_sec * 1e6 + end.tv_usec << "]" << std::endl; << "]:START[" << start.tv_sec * 1e6 + start.tv_usec
<< "]:END[" << end.tv_sec * 1e6 + end.tv_usec << "]"
<< std::endl;
} }
} }
op_idx++; op_idx++;
...@@ -403,11 +406,13 @@ void SectionWorker::TrainFilesWithProfiler() { ...@@ -403,11 +406,13 @@ void SectionWorker::TrainFilesWithProfiler() {
gettimeofday(&micro_end, NULL); gettimeofday(&micro_end, NULL);
{ {
std::unique_lock<std::mutex> lk(cout_mutex); std::unique_lock<std::mutex> lk(cout_mutex);
std::cout << std::fixed; std::cout << std::fixed;
std::cout.precision(0); std::cout.precision(0);
std::cout << "!!FWD:B[" << batch_id_ << "]:SEC[" << thread_id_ std::cout << "!!FWD:B[" << batch_id_ << "]:SEC[" << thread_id_
<< "]:START[" << micro_start.tv_sec * 1e6 + micro_start.tv_usec << "]:START["
<< "]:END[" << micro_end.tv_sec * 1e6 + micro_end.tv_usec << "]" << std::endl; << micro_start.tv_sec * 1e6 + micro_start.tv_usec
<< "]:END[" << micro_end.tv_sec * 1e6 + micro_end.tv_usec
<< "]" << std::endl;
} }
} catch (platform::EOFException&) { } catch (platform::EOFException&) {
std::unique_lock<std::mutex> lk(thread_mutex); std::unique_lock<std::mutex> lk(thread_mutex);
...@@ -467,11 +472,13 @@ void SectionWorker::TrainFilesWithProfiler() { ...@@ -467,11 +472,13 @@ void SectionWorker::TrainFilesWithProfiler() {
op_total_time[op_idx] += time; op_total_time[op_idx] += time;
{ {
std::unique_lock<std::mutex> lk(cout_mutex); std::unique_lock<std::mutex> lk(cout_mutex);
std::cout << std::fixed; std::cout << std::fixed;
std::cout.precision(0); std::cout.precision(0);
std::cout << "::BWD:B[" << batch_id_ << "]:SEC[" << thread_id_ << "]:SCOPE[" << i std::cout << "::BWD:B[" << batch_id_ << "]:SEC[" << thread_id_
<< "]:OP[" << op->Type() << "]:START[" << start.tv_sec * 1e6 + start.tv_usec << "]:SCOPE[" << i << "]:OP[" << op->Type()
<< "]:END[" << end.tv_sec * 1e6 + end.tv_usec << "]" << std::endl; << "]:START[" << start.tv_sec * 1e6 + start.tv_usec
<< "]:END[" << end.tv_sec * 1e6 + end.tv_usec << "]"
<< std::endl;
} }
} }
op_idx++; op_idx++;
...@@ -479,11 +486,13 @@ void SectionWorker::TrainFilesWithProfiler() { ...@@ -479,11 +486,13 @@ void SectionWorker::TrainFilesWithProfiler() {
gettimeofday(&micro_end, NULL); gettimeofday(&micro_end, NULL);
{ {
std::unique_lock<std::mutex> lk(cout_mutex); std::unique_lock<std::mutex> lk(cout_mutex);
std::cout << std::fixed; std::cout << std::fixed;
std::cout.precision(0); std::cout.precision(0);
std::cout << "!!BWD:B[" << batch_id_ << "]:SEC[" << thread_id_ std::cout << "!!BWD:B[" << batch_id_ << "]:SEC[" << thread_id_
<< "]:START[" << micro_start.tv_sec * 1e6 + micro_start.tv_usec << "]:START["
<< "]:END[" << micro_end.tv_sec * 1e6 + micro_end.tv_usec << "]" << std::endl; << micro_start.tv_sec * 1e6 + micro_start.tv_usec
<< "]:END[" << micro_end.tv_sec * 1e6 + micro_end.tv_usec
<< "]" << std::endl;
} }
} }
dev_ctx_->Wait(); dev_ctx_->Wait();
...@@ -522,11 +531,13 @@ void SectionWorker::TrainFilesWithProfiler() { ...@@ -522,11 +531,13 @@ void SectionWorker::TrainFilesWithProfiler() {
op_total_time[op_idx] += time; op_total_time[op_idx] += time;
{ {
std::unique_lock<std::mutex> lk(cout_mutex); std::unique_lock<std::mutex> lk(cout_mutex);
std::cout << std::fixed; std::cout << std::fixed;
std::cout.precision(0); std::cout.precision(0);
std::cout << "::UPD:B[" << batch_id_ << "]:SEC[" << thread_id_ << "]:SCOPE[" << num_microbatches_ std::cout << "::UPD:B[" << batch_id_ << "]:SEC[" << thread_id_
<< "]:OP[" << op->Type() << "]:START[" << start.tv_sec * 1e6 + start.tv_usec << "]:SCOPE[" << num_microbatches_ << "]:OP["
<< "]:END[" << end.tv_sec * 1e6 + end.tv_usec << "]" << std::endl; << op->Type() << "]:START["
<< start.tv_sec * 1e6 + start.tv_usec << "]:END["
<< end.tv_sec * 1e6 + end.tv_usec << "]" << std::endl;
} }
} }
op_idx++; op_idx++;
...@@ -534,11 +545,13 @@ void SectionWorker::TrainFilesWithProfiler() { ...@@ -534,11 +545,13 @@ void SectionWorker::TrainFilesWithProfiler() {
gettimeofday(&micro_end, NULL); gettimeofday(&micro_end, NULL);
{ {
std::unique_lock<std::mutex> lk(cout_mutex); std::unique_lock<std::mutex> lk(cout_mutex);
std::cout << std::fixed; std::cout << std::fixed;
std::cout.precision(0); std::cout.precision(0);
std::cout << "!!UPD:B[" << batch_id_ << "]:SEC[" << thread_id_ std::cout << "!!UPD:B[" << batch_id_ << "]:SEC[" << thread_id_
<< "]:START[" << micro_start.tv_sec * 1e6 + micro_start.tv_usec << "]:START["
<< "]:END[" << micro_end.tv_sec * 1e6 + micro_end.tv_usec << "]" << std::endl; << micro_start.tv_sec * 1e6 + micro_start.tv_usec << "]:END["
<< micro_end.tv_sec * 1e6 + micro_end.tv_usec << "]"
<< std::endl;
} }
dev_ctx_->Wait(); dev_ctx_->Wait();
batch_timer.Pause(); batch_timer.Pause();
...@@ -582,7 +595,8 @@ void SectionWorker::TrainFilesWithProfiler() { ...@@ -582,7 +595,8 @@ void SectionWorker::TrainFilesWithProfiler() {
lk.unlock(); lk.unlock();
VLOG(0) << "============timeline============"; VLOG(0) << "============timeline============";
for (size_t i = 0; i < ops_.size(); ++i) { for (size_t i = 0; i < ops_.size(); ++i) {
VLOG(0) << "op: " << op_name[i] << ", max_time: " << op_max_time[i] VLOG(0) << "op: " << op_name[i]
<< ", max_time: " << op_max_time[i]
<< ", min_time: " << op_min_time[i] << ", min_time: " << op_min_time[i]
<< ", mean_time: " << op_total_time[i] / op_count[i]; << ", mean_time: " << op_total_time[i] / op_count[i];
} }
...@@ -632,11 +646,13 @@ void SectionWorker::TrainFilesWithProfiler() { ...@@ -632,11 +646,13 @@ void SectionWorker::TrainFilesWithProfiler() {
op_total_time[op_idx] += time; op_total_time[op_idx] += time;
{ {
std::unique_lock<std::mutex> lk(cout_mutex); std::unique_lock<std::mutex> lk(cout_mutex);
std::cout << std::fixed; std::cout << std::fixed;
std::cout.precision(0); std::cout.precision(0);
std::cout << "::FWD:B[" << local_batch_id_ << "]:SEC[" << thread_id_ << "]:SCOPE[" << i std::cout << "::FWD:B[" << local_batch_id_ << "]:SEC["
<< "]:OP[" << op->Type() << "]:START[" << start.tv_sec * 1e6 + start.tv_usec << thread_id_ << "]:SCOPE[" << i << "]:OP["
<< "]:END[" << end.tv_sec * 1e6 + end.tv_usec << "]" << std::endl; << op->Type() << "]:START["
<< start.tv_sec * 1e6 + start.tv_usec << "]:END["
<< end.tv_sec * 1e6 + end.tv_usec << "]" << std::endl;
} }
} }
op_idx++; op_idx++;
...@@ -644,11 +660,13 @@ void SectionWorker::TrainFilesWithProfiler() { ...@@ -644,11 +660,13 @@ void SectionWorker::TrainFilesWithProfiler() {
gettimeofday(&micro_end, NULL); gettimeofday(&micro_end, NULL);
{ {
std::unique_lock<std::mutex> lk(cout_mutex); std::unique_lock<std::mutex> lk(cout_mutex);
std::cout << std::fixed; std::cout << std::fixed;
std::cout.precision(0); std::cout.precision(0);
std::cout << "!!FWD:B[" << batch_id_ << "]:SEC[" << thread_id_ std::cout << "!!FWD:B[" << batch_id_ << "]:SEC[" << thread_id_
<< "]:START[" << micro_start.tv_sec * 1e6 + micro_start.tv_usec << "]:START["
<< "]:END[" << micro_end.tv_sec * 1e6 + micro_end.tv_usec << "]" << std::endl; << micro_start.tv_sec * 1e6 + micro_start.tv_usec
<< "]:END[" << micro_end.tv_sec * 1e6 + micro_end.tv_usec
<< "]" << std::endl;
} }
} }
dev_ctx_->Wait(); dev_ctx_->Wait();
...@@ -685,11 +703,13 @@ void SectionWorker::TrainFilesWithProfiler() { ...@@ -685,11 +703,13 @@ void SectionWorker::TrainFilesWithProfiler() {
op_total_time[op_idx] += time; op_total_time[op_idx] += time;
{ {
std::unique_lock<std::mutex> lk(cout_mutex); std::unique_lock<std::mutex> lk(cout_mutex);
std::cout << std::fixed; std::cout << std::fixed;
std::cout.precision(0); std::cout.precision(0);
std::cout << "::BWD:B[" << local_batch_id_ << "]:SEC[" << thread_id_ << "]:SCOPE[" << i std::cout << "::BWD:B[" << local_batch_id_ << "]:SEC["
<< "]:OP[" << op->Type() << "]:START[" << start.tv_sec * 1e6 + start.tv_usec << thread_id_ << "]:SCOPE[" << i << "]:OP["
<< "]:END[" << end.tv_sec * 1e6 + end.tv_usec << "]" << std::endl; << op->Type() << "]:START["
<< start.tv_sec * 1e6 + start.tv_usec << "]:END["
<< end.tv_sec * 1e6 + end.tv_usec << "]" << std::endl;
} }
} }
op_idx++; op_idx++;
...@@ -697,11 +717,13 @@ void SectionWorker::TrainFilesWithProfiler() { ...@@ -697,11 +717,13 @@ void SectionWorker::TrainFilesWithProfiler() {
gettimeofday(&micro_end, NULL); gettimeofday(&micro_end, NULL);
{ {
std::unique_lock<std::mutex> lk(cout_mutex); std::unique_lock<std::mutex> lk(cout_mutex);
std::cout << std::fixed; std::cout << std::fixed;
std::cout.precision(0); std::cout.precision(0);
std::cout << "!!BWD:B[" << batch_id_ << "]:SEC[" << thread_id_ std::cout << "!!BWD:B[" << batch_id_ << "]:SEC[" << thread_id_
<< "]:START[" << micro_start.tv_sec * 1e6 + micro_start.tv_usec << "]:START["
<< "]:END[" << micro_end.tv_sec * 1e6 + micro_end.tv_usec << "]" << std::endl; << micro_start.tv_sec * 1e6 + micro_start.tv_usec
<< "]:END[" << micro_end.tv_sec * 1e6 + micro_end.tv_usec
<< "]" << std::endl;
} }
} }
dev_ctx_->Wait(); dev_ctx_->Wait();
...@@ -738,11 +760,13 @@ void SectionWorker::TrainFilesWithProfiler() { ...@@ -738,11 +760,13 @@ void SectionWorker::TrainFilesWithProfiler() {
op_total_time[op_idx] += time; op_total_time[op_idx] += time;
{ {
std::unique_lock<std::mutex> lk(cout_mutex); std::unique_lock<std::mutex> lk(cout_mutex);
std::cout << std::fixed; std::cout << std::fixed;
std::cout.precision(0); std::cout.precision(0);
std::cout << "::UPD:B[" << batch_id_ << "]:SEC[" << thread_id_ << "]:SCOPE[" << num_microbatches_ std::cout << "::UPD:B[" << batch_id_ << "]:SEC[" << thread_id_
<< "]:OP[" << op->Type() << "]:START[" << start.tv_sec * 1e6 + start.tv_usec << "]:SCOPE[" << num_microbatches_ << "]:OP["
<< "]:END[" << end.tv_sec * 1e6 + end.tv_usec << "]" << std::endl; << op->Type() << "]:START["
<< start.tv_sec * 1e6 + start.tv_usec << "]:END["
<< end.tv_sec * 1e6 + end.tv_usec << "]" << std::endl;
} }
} }
op_idx++; op_idx++;
...@@ -750,11 +774,13 @@ void SectionWorker::TrainFilesWithProfiler() { ...@@ -750,11 +774,13 @@ void SectionWorker::TrainFilesWithProfiler() {
gettimeofday(&micro_end, NULL); gettimeofday(&micro_end, NULL);
{ {
std::unique_lock<std::mutex> lk(cout_mutex); std::unique_lock<std::mutex> lk(cout_mutex);
std::cout << std::fixed; std::cout << std::fixed;
std::cout.precision(0); std::cout.precision(0);
std::cout << "!!UPD:B[" << batch_id_ << "]:SEC[" << thread_id_ std::cout << "!!UPD:B[" << batch_id_ << "]:SEC[" << thread_id_
<< "]:START[" << micro_start.tv_sec * 1e6 + micro_start.tv_usec << "]:START["
<< "]:END[" << micro_end.tv_sec * 1e6 + micro_end.tv_usec << "]" << std::endl; << micro_start.tv_sec * 1e6 + micro_start.tv_usec << "]:END["
<< micro_end.tv_sec * 1e6 + micro_end.tv_usec << "]"
<< std::endl;
} }
dev_ctx_->Wait(); dev_ctx_->Wait();
if (local_completed) { if (local_completed) {
......
python/paddle/fluid/optimizer.py
浏览文件 @ 9f6aaa3c
...@@ -1142,7 +1142,7 @@ class MomentumOptimizer(Optimizer): ...@@ -1142,7 +1142,7 @@ class MomentumOptimizer(Optimizer):
class DGCMomentumOptimizer(Optimizer): class DGCMomentumOptimizer(Optimizer):
""" """
:api_attr: Static Graph :api_attr: Static Graph
DGC (Deep Gradient Compression) Momentum Optimizer. Original paper is https://arxiv.org/abs/1712.01887 DGC (Deep Gradient Compression) Momentum Optimizer. Original paper is https://arxiv.org/abs/1712.01887
...@@ -3068,7 +3068,7 @@ Lamb = LambOptimizer ...@@ -3068,7 +3068,7 @@ Lamb = LambOptimizer
class ModelAverage(Optimizer): class ModelAverage(Optimizer):
""" """
:api_attr: Static Graph :api_attr: Static Graph
The ModelAverage optimizer accumulates specific continuous historical parameters The ModelAverage optimizer accumulates specific continuous historical parameters
during training. The accumulated historical range can be controlled by the passed during training. The accumulated historical range can be controlled by the passed
...@@ -3377,7 +3377,7 @@ class ModelAverage(Optimizer): ...@@ -3377,7 +3377,7 @@ class ModelAverage(Optimizer):
class ExponentialMovingAverage(object): class ExponentialMovingAverage(object):
""" """
:api_attr: Static Graph :api_attr: Static Graph
Compute the moving average of parameters with exponential decay. Compute the moving average of parameters with exponential decay.
Given a parameter :math:`\\theta`, its exponential moving average (EMA) Given a parameter :math:`\\theta`, its exponential moving average (EMA)
...@@ -3627,7 +3627,7 @@ class ExponentialMovingAverage(object): ...@@ -3627,7 +3627,7 @@ class ExponentialMovingAverage(object):
class PipelineOptimizer(object): class PipelineOptimizer(object):
""" """
:api_attr: Static Graph :api_attr: Static Graph
Pipeline Optimizer: Make a program to run as pipeline, that is splitting a Pipeline Optimizer: Make a program to run as pipeline, that is splitting a
program into multiple sections (sub-programs) and each section run on a program into multiple sections (sub-programs) and each section run on a
...@@ -4132,7 +4132,7 @@ class PipelineOptimizer(object): ...@@ -4132,7 +4132,7 @@ class PipelineOptimizer(object):
index=0, index=0,
type='fill_constant', type='fill_constant',
inputs={}, inputs={},
outputs={'Out':[grad_var]}, outputs={'Out': [grad_var]},
attrs={ attrs={
'shape': grad_var.shape, 'shape': grad_var.shape,
'dtype': grad_var.dtype, 'dtype': grad_var.dtype,
...@@ -4326,6 +4326,7 @@ class PipelineOptimizer(object): ...@@ -4326,6 +4326,7 @@ class PipelineOptimizer(object):
# attribute have not been set yet. Then check all ops have the # attribute have not been set yet. Then check all ops have the
# op_device attribute. # op_device attribute.
self._add_default_opdevice_attr(main_block) self._add_default_opdevice_attr(main_block)
device_specs = self._check_validation(main_block) device_specs = self._check_validation(main_block)
# Step3: add enqueue and dequeue ops between section boundaries # Step3: add enqueue and dequeue ops between section boundaries
...@@ -4388,7 +4389,7 @@ class PipelineOptimizer(object): ...@@ -4388,7 +4389,7 @@ class PipelineOptimizer(object):
class RecomputeOptimizer(Optimizer): class RecomputeOptimizer(Optimizer):
""" """
:api_attr: Static Graph :api_attr: Static Graph
Recompute Optimizer Wrapper Recompute Optimizer Wrapper
...@@ -4473,7 +4474,7 @@ class RecomputeOptimizer(Optimizer): ...@@ -4473,7 +4474,7 @@ class RecomputeOptimizer(Optimizer):
def load(self, stat_dict): def load(self, stat_dict):
""" """
:api_attr: Static Graph :api_attr: Static Graph
load function is not supported by Recompute Optimizer for now. load function is not supported by Recompute Optimizer for now.
:return: None :return: None
...@@ -4697,7 +4698,7 @@ class RecomputeOptimizer(Optimizer): ...@@ -4697,7 +4698,7 @@ class RecomputeOptimizer(Optimizer):
class LookaheadOptimizer(object): class LookaheadOptimizer(object):
""" """
:api_attr: Static Graph :api_attr: Static Graph
This implements the Lookahead optimizer of the This implements the Lookahead optimizer of the
paper : https://arxiv.org/abs/1907.08610. paper : https://arxiv.org/abs/1907.08610.
......
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