Open trtd56 opened 3 years ago
BERT Based Multilingual Machine Comprehension in English and Hindi
の比較
パラメータ
Lower the case for input: False
Max (sequence, query, answer) length: 384, 64, 30
Train, predict batch sizes: 12,8
Learning rate, warmup proportion: 5e-5, 0.1
Number of epochs: 3.0
前回のコンペ