QATrainer.Options Class

Dropout rate after activation functions in FFN layers. Should be within [0, 1).

Betas for Adam optimizer.

Epsilon for Adam optimizer.

Answer Index Start Column Name

Dropout rate for attention weights. Should be within [0, 1).

Number of samples to use for mini-batch training.

The clipping threshold of gradients. Should be within [0, +Inf). 0 means not to clip norm.

Context Column Name

Dropout rate for general situations. Should be within [0, 1).

Whether to use dynamic dropout.

Whether to apply layer normalization before each encoder block.

The final learning rate for polynomial decay scheduler.

Whether to freeze encoder parameters.

Whether to freeze transfer module parameters.

The label column name.

Whether to train layer norm parameters.

Learning rate for the first N epochs; all epochs >N using LR_N. Note: this may be interpreted differently depending on the scheduler.

How often to log the loss.

Stop training when reaching this number of epochs.

Dropout rate in the masked language model pooler layers. Should be within [0, 1).

Answer Column Name for the predicted answers

The Prediction column name.

Question Column Name

The Score column name.

The first sentence column.

The second sentence column.

The start learning rate for polynomial decay scheduler.

Task type, which is related to the model head.

Number of top predicted answers in question answering task.

Answer Column Name for the training data

The validation set used while training to improve model quality.

Proportion of warmup steps for polynomial decay scheduler.

Coefficiency of weight decay. Should be within [0, +Inf).