moe-multilingual-translator / training_log.txt
arka7's picture
Upload Stage 1 model - Loss: 2.0218
780b318 verified
Training Completed Successfully!
Epoch: 1
Total Batches: 3743
Average Loss: 2.0218
Average Balance Loss: 0.0108
Expert Usage per Language:
en: [[0.20985517 0.16751863 0.31998625 0.30264 ]]
fr: [[0.24961634 0.21768875 0.26282057 0.26987436]]
hi: [[0.21246533 0.14122878 0.33271343 0.31359246]]
bn: [[0.24983221 0.22729187 0.25725418 0.26562175]]