In the single file mode, all the metrics are calculated serially. Thus the serial and parallel dictionaries are merged before calling compute. However, the dictionaries were being merged using:
serial_metric_dict.update(parallel_metric_dict)
which overwrites the heavy metrics originally present in serial_metric_dict. This patch fixes it by instead doing:
for metric_type, metric_list in parallel_metric_dict.items():
if metric_type in serial_metric_dict:
serial_metric_dict[metric_type].extend(metric_list)
else:
serial_metric_dict[metric_type] = metric_list
This PR fixes #76.
In the single file mode, all the metrics are calculated serially. Thus the serial and parallel dictionaries are merged before calling compute. However, the dictionaries were being merged using:
which overwrites the heavy metrics originally present in
serial_metric_dict
. This patch fixes it by instead doing: