Skip to content
This repository has been archived by the owner on Nov 3, 2023. It is now read-only.

Fixes train_model worldlogging for multitask with mutators. #4414

Merged
merged 2 commits into from Mar 17, 2022
Merged

Fixes train_model worldlogging for multitask with mutators. #4414

Show file tree
Hide file tree
Changes from 1 commit
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
5ef3987
Fixes train_model worldlogging for multitask with mutators.
kauterry Mar 11, 2022
2ad8b18
Fix bug in train_model when evaltask doesn't match task.
kauterry Mar 16, 2022
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Jump to
Jump to file
Failed to load files.
Diff view
Diff view
9 changes: 5 additions & 4 deletions parlai/scripts/train_model.py
View file
Open in desktop
Expand Up @@ -619,7 +619,7 @@ def validate(self):
return True
return False

def _run_single_eval(self, opt, valid_world, max_exs, datatype, is_multitask):
def _run_single_eval(self, opt, valid_world, max_exs, datatype, is_multitask, task):

# run evaluation on a single world
valid_world.reset()
Expand All @@ -629,7 +629,7 @@ def _run_single_eval(self, opt, valid_world, max_exs, datatype, is_multitask):
# set up world logger for the "test" fold
if opt['world_logs'] and datatype == 'test':
task_opt['world_logs'] = get_task_world_logs(
valid_world.getID(), opt['world_logs'], is_multitask
task, opt['world_logs'], is_multitask
)
world_logger = WorldLogger(task_opt)

Expand Down Expand Up @@ -691,9 +691,10 @@ def _run_eval(

max_exs_per_worker = max_exs / (len(valid_worlds) * num_workers())
is_multitask = len(valid_worlds) > 1
for v_world in valid_worlds:
for index, v_world in enumerate(valid_worlds):
task = opt['task'].split(',')[index]
task_report = self._run_single_eval(
opt, v_world, max_exs_per_worker, datatype, is_multitask
opt, v_world, max_exs_per_worker, datatype, is_multitask, task
)
reports.append(task_report)

Expand Down
27 changes: 27 additions & 0 deletions tests/test_train_model.py
View file
Open in desktop
Expand Up @@ -271,6 +271,33 @@ def test_save_multiple_world_logs(self):
json_lines = f.readlines()
assert len(json_lines) == 5

def test_save_multiple_world_logs_mutator(self):
"""
Test that we can save multiple world_logs from train model on multiple tasks
with mutators present.
"""
with testing_utils.tempdir() as tmpdir:
log_report = os.path.join(tmpdir, 'world_logs.jsonl')
multitask = 'integration_tests:mutators=flatten,integration_tests:ReverseTeacher:mutator=reverse'
Copy link
Contributor

@jxmsML jxmsML Mar 15, 2022 •
edited

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I wonder if we can test this with integration_tests:mutators=flatten,integration_tests:mutator=reverse to highlight that you wanted to test if world logs works for exact same teacher except for mutators

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I was just trying to convey that the two tasks can be anything, it's the number of tasks that matter. So it works for any two tasks.

valid, test = testing_utils.train_model(
{
'task': multitask,
'validation_max_exs': 10,
'model': 'repeat_label',
'short_final_eval': True,
'num_epochs': 1.0,
'world_logs': log_report,
}
)

for task in multitask.split(','):
task_log_report = get_task_world_logs(
task, log_report, is_multitask=True
)
with PathManager.open(task_log_report) as f:
json_lines = f.readlines()
assert len(json_lines) == 5


@register_agent("fake_report")
class FakeReportAgent(Agent):
Expand Down