dd7ff72
1
2
3
4
5
6
7
params: batch_size: 8 num_epochs: 50 learning_rate: 5e-5 weight_decay : 1e-4 gradient_accumulation_steps: 4 patience : 10