fix_bug simplify.py

[Vibsteamer]

expected behavior:
when "labeled":true in dpgen simplify, 02.fp will soft-link "labeled data", and the soft-linked "task dir" will also be created, for format consistency.

it is expected to be data.000 and task.000.000000,
being respectively guaranteed by funcs data_system_fmt and fp_task_fmt

bug:
the typo_bug used data_system_fmt for the "task dir" instead of fp_task_fmt,
then gives task.000 instead of task.000.000000,

which makes _check_empty_iter (who checks glob.glog("task.000.*")) in generator/run.py sentence this iter empty,
then 00.train of the next iter is always skipped.

consequence:
this make the "simplify_labeled" process never starts correctly,

no iter0 model presents and randomly-picked data in iter0 are never trained,
then iter1 gives error that can't finding the graph file from iter0 when trying copying them due to the train-skip.

BTW
thought "simplify_labeled" valuable in some complex or big-data scenarios but seems not loved by users yet.
pity : (

Signed-off-by: Wanrun Jiang [email protected]

[njzjz]

dpgen/dpgen/generator/run.py

Lines 118 to 121 in 355f8ed

	def _check_empty_iter(iter_index, max_v = 0) :
	fp_path = os.path.join(make_iter_name(iter_index), fp_name)
	# check the number of collected data
	sys_data = glob.glob(os.path.join(fp_path, "data.*"))

_check_empty_iter only checks data.*, so it should not be a problem.

[Vibsteamer]

dpgen/dpgen/generator/run.py

Lines 118 to 121 in 355f8ed

	def _check_empty_iter(iter_index, max_v = 0) :
	fp_path = os.path.join(make_iter_name(iter_index), fp_name)
	# check the number of collected data
	sys_data = glob.glob(os.path.join(fp_path, "data.*"))

_check_empty_iter only checks data.*, so it should not be a problem.

OK, I will do the upgrade. Thanks.