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

fix test_tipc/train logic #2879

Merged
merged 7 commits into from
Aug 2, 2022
Merged

fix test_tipc/train logic #2879

Changes from all commits
Commits
Show all changes
7 commits
Select commit Hold shift + click to select a range
7ddc870
fix test_tipc/train logic
veyron95 Jul 26, 2022
88db0de
recover some code
veyron95 Jul 26, 2022
c8cd725
polish code
veyron95 Jul 26, 2022
0571b55
update train.py
veyron95 Jul 26, 2022
c169143
polish code
veyron95 Jul 27, 2022
10e362e
Merge branch 'develop' into fix_xlnet_for_ips_count
FrostML Aug 2, 2022
fbb721c
Merge branch 'develop' into fix_xlnet_for_ips_count
FrostML Aug 2, 2022
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
44 changes: 26 additions & 18 deletions tests/test_tipc/train.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -116,17 +116,31 @@ def do_train(args):
optimizer.step()
optimizer.clear_grad()

train_batch_cost = time.time() - batch_start
reader_cost_avg.record(train_reader_cost)
batch_cost_avg.record(train_batch_cost)
batch_ips_avg.record(train_batch_cost, sample_per_cards)

if args.profiler_options is not None:
profiler.add_profiler_step(args.profiler_options)

if args.max_steps and step_id == args.max_steps:
if args.save_model and rank == 0:
model_dir = args.save_model
if not os.path.exists(model_dir):
os.makedirs(model_dir)
paddle.save(model.state_dict(),
os.path.join(model_dir, "model.pdparams"))
paddle.save(optimizer.state_dict(),
os.path.join(model_dir, "model.pdopt"))
return

if args.lr_scheduler is not None and not args.scheduler_update_by_epoch:
lr.step()

if step_id % args.logging_steps == 0:
total_avg_loss = loss.numpy()

train_batch_cost = time.time() - batch_start
reader_cost_avg.record(train_reader_cost)
batch_cost_avg.record(train_batch_cost)
batch_ips_avg.record(train_batch_cost, sample_per_cards)

benchmark_model.logger(
args,
step_id=step_id,
Expand All @@ -141,22 +155,16 @@ def do_train(args):
reader_cost_avg.reset()
batch_cost_avg.reset()
batch_ips_avg.reset()
else:
train_batch_cost = time.time() - batch_start
reader_cost_avg.record(train_reader_cost)
batch_cost_avg.record(train_batch_cost)
batch_ips_avg.record(train_batch_cost, sample_per_cards)

batch_start = time.time()

if args.max_steps and step_id == args.max_steps:
if args.save_model and rank == 0:
model_dir = args.save_model
if not os.path.exists(model_dir):
os.makedirs(model_dir)
paddle.save(model.state_dict(),
os.path.join(model_dir, "model.pdparams"))
paddle.save(optimizer.state_dict(),
os.path.join(model_dir, "model.pdopt"))
return
batch_id += 1
step_id += 1
if args.lr_scheduler is not None and not args.scheduler_update_by_epoch:
lr.step()
batch_start = time.time()

if args.lr_scheduler is not None and args.scheduler_update_by_epoch:
lr.step()
Expand Down