Closed rems75 closed 4 years ago
Update on this. Seems like my setup currently doesn't support RAPL, so the data does not contain ''cpu_time_seconds'' or "rapl_estimated_attributable_power_draw" which then crashes the statement generation.
I see, yes, we assume RAPL is available. In your use case would you like only GPU measurements if RAPL is not available?
Closing due to inactivity, but if having GPU-only recording is a requested feature, please open a new issue!
Hi there,
Cool library! Got this issue when running:
generate-carbon-impact-statement tracker "USA"
The script returns this error: `loading region bounding boxes for computing carbon emissions region, this may take a moment... 454/454... rate=556.09 Hz, eta=0:00:00, total=0:00:00, wall=15:15 UTCC Done! Traceback (most recent call last):
File "/home/retachet/.conda/envs/homer/bin/generate-carbon-impact-statement", line 7, in
exec(compile(f.read(), file, 'exec'))
File "/home/retachet/ExperimentImpactTracker/experiment-impact-tracker/scripts/generate-carbon-impact-statement", line 235, in
sys.exit(main(sys.argv[1:]))
File "/home/retachet/ExperimentImpactTracker/experiment-impact-tracker/scripts/generate-carbon-impact-statement", line 177, in main extracted_info = gather_additional_info(info, log_dir)
File "/home/retachet/ExperimentImpactTracker/experiment-impact-tracker/experiment_impact_tracker/utils.py", line 102, in gather_additional_info cpu_seconds = _get_cpu_hours_from_per_process_data(json_array)
File "/home/retachet/ExperimentImpactTracker/experiment-impact-tracker/experiment_impact_tracker/utils.py", line 94, in _get_cpu_hours_from_per_process_data cpu_point = datapoint["cpu_time_seconds"]
KeyError: 'cpu_time_seconds'`