Hi, I am trying to reproduce the results of CE on LSMDC dataset. I follow the instructions to evaluate the provided pre-trained model with config on LSMDC but get the results somehow lower than expected:
Please be aware that all the numbers from the tables are reported over three seeded runs. The results you are obtaining are the exact numbers for the first seeded experiment. You can check the results of every seed in the log.
Hi, I am trying to reproduce the results of CE on LSMDC dataset. I follow the instructions to evaluate the provided pre-trained model with config on LSMDC but get the results somehow lower than expected:
I evaluate it with the following script:
How could I reproduce the reported results below:
Many thanks!