Closed neubig closed 1 year ago
The wmt20_metrics dataset was outputting systems based on the order provided by Python dictionaries, which was not stable across machines. This PR fixes this issue, sorting by the system name before outputting the data.
The wmt20_metrics dataset was outputting systems based on the order provided by Python dictionaries, which was not stable across machines. This PR fixes this issue, sorting by the system name before outputting the data.