def test_line_has_time_elapsed():
"""Test that TrainingRunInfo can parse time elapsed values from console output."""
info = TrainingRunInfo()
assert info.line_has_time_elapsed('\t\tmax_steps:\t3.0e3') is False
assert (
info.line_has_time_elapsed(
'INFO:mlagents.trainers: 3DBall_00: 3DBallLearning: Step: 1000. Time Elapsed: 12.398 s Mean Reward: 1.259. Std of Reward: 0.799. Training.'
)
is True
)
def test_parse_max_steps(training_output):
"""Test that TrainingRunInfo can parse a max_steps value from console output."""
info = TrainingRunInfo()
info.update_from_training_output('Final Mean Reward: 1.763')
assert info.max_steps == 0
for line in training_output:
info.update_from_training_output(line)
assert info.max_steps == 3000
def test_parse_steps(training_output):
"""Test that TrainingRunInfo can parse a steps value from console output."""
info = TrainingRunInfo()
for line in training_output:
info.update_from_training_output(line)
assert info.step == 3000
info.update_from_training_output(
'INFO:mlagents.trainers: 3DBall_00: 3DBallLearning: Step: 1000. Time Elapsed: 12.398 s Mean Reward: 1.259. Std of Reward: 0.799. Training.'
)
assert info.step == 1000
def test_parse_exported_brains(training_output):
"""Test that TrainingRunInfo can parse exported brain values from console output."""
info = TrainingRunInfo()
for line in training_output:
info.update_from_training_output(line)
assert info.exported_brains == [Path('./models/3DBall_00/3DBallLearning.nn')]
info.update_from_training_output(
'INFO:mlagents.trainers:Exported ./models/3DBall_00/3DBallHardLearning.nn file'
)
assert info.exported_brains == [
Path('./models/3DBall_00/3DBallLearning.nn'),
Path('./models/3DBall_00/3DBallHardLearning.nn'),
]
def test_time_remaining():
"""Test that TrainingRunInfo can calculate steps_remaining and time_remaining as well as handle edge cases.
Edge cases include:
- all values have yet to be found
- max_steps has not been found
- steps matches max_steps value
- steps value exceeds max_steps value
"""
info = TrainingRunInfo()
assert info.time_remaining == 0
info.update_from_training_output(
'INFO:mlagents.trainers: 3DBall_00: 3DBallLearning: Step: 1000. Time Elapsed: 12.398 s Mean Reward: 1.259. Std of Reward: 0.799. Training.'
)
assert info.time_remaining == 0
info.max_steps = 3000
info.update_from_training_output(
'INFO:mlagents.trainers: 3DBall_00: 3DBallLearning: Step: 1000. Time Elapsed: 12.398 s Mean Reward: 1.259. Std of Reward: 0.799. Training.'
)
assert info.time_remaining == 24.796
info.update_from_training_output(
'INFO:mlagents.trainers: 3DBall_00: 3DBallLearning: Step: 3000. Time Elapsed: 30.652 s Mean Reward: 1.763. Std of Reward: 1.054. Training.'
)
assert info.time_remaining == 0.0
info.update_from_training_output(
'INFO:mlagents.trainers: 3DBall_00: 3DBallLearning: Step: 4000. Time Elapsed: 30.652 s Mean Reward: 1.763. Std of Reward: 1.054. Training.'
)
assert info.time_remaining == 0.0