gpt2_train_cfg.yaml

data_config:
  path: https://raw.githubusercontent.com/karpathy/char-rnn/master/data/tinyshakespeare/input.txt
  block_size: 128
  train_split: 0.9
  truncate: 0.05
gpt_config:
  n_layer: 8
  n_head: 8
  n_embd: 512
trainer_config:
  max_epochs: 10
  batch_size: 216
  data_loader_workers: 4
  grad_norm_clip: 1.0
  snapshot_path: gpt_snapshot.pt
  save_every: 3
  use_amp: True
optimizer_config:
  weight_decay: 0.1
  learning_rate: 0.0003

hydra:
  run:
    dir: ./