jfia0107/BiSLTM_civil_comments

Trained using https://github.com/jfia0107/PyTorch-Lightning-NLP-lab and these parameters:
seed: 42
freeze_backbone: False
max_epochs: 5
monitor_metric: auc
monitor_mode: max
loss_fn.: torch.nn.BCEWithLogitsLoss
loss_fn.reduction: mean
optimizer: torch.optim.AdamW
optimizer.scheduler: none
optimizer.head_lr: 0.0003
input_size: 300
optimizer.weight_decay: 0.01
hidden_size: 128
num_layers: 2
dropout: 0.3
batch size: 128
max_len: 256
max_vocab_size: 30000