fairseq/examples/roberta/config/finetuning/sts_b.yaml

# @package _group_

common:
  fp16: true
  fp16_init_scale: 4
  threshold_loss_scale: 1
  fp16_scale_window: 128
  log_format: json
  log_interval: 200

task:
  _name: sentence_prediction
  data: ???
  init_token: 0
  separator_token: 2
  num_classes: 1
  max_positions: 512

checkpoint:
  restore_file: ???
  reset_optimizer: true
  reset_dataloader: true
  reset_meters: true
  no_epoch_checkpoints: true

distributed_training:
  find_unused_parameters: true
  distributed_world_size: 1

criterion:
  _name: sentence_prediction
  regression_target: true

dataset:
  batch_size: 16
  required_batch_size_multiple: 1
  max_tokens: 4400

optimizer:
  _name: adam
  weight_decay: 0.1
  adam_betas: (0.9,0.98)
  adam_eps: 1e-06

lr_scheduler:
  _name: polynomial_decay
  warmup_updates: 214

optimization:
  clip_norm: 0.0
  lr: [2e-05]
  max_update: 3598
  max_epoch: 10

model:
  _name: roberta
  dropout: 0.1
  attention_dropout: 0.1