hydra_config_1737996741.0516832.yaml

experiment:
  seed: 42
  name: debug
  group: debug
  dry_run: true
  offline_run: false
  evaluate_segmentation: false
  evaluate_babyslm: false
  blimp_tasks: null
  resume_checkpoint_path: null
  resume_run_id: null
dataset:
  name: phonemetransformers/CHILDES
  subconfig: English
  text_column: phonemized_utterance
  is_phonemes: true
  max_age: null
  remove_child_utterances: true
  valid_size: 10000
tokenizer:
  name: phonemetransformers/CHILDES-English-phoneme-tokenizer
data_preprocessing:
  max_input_length: 64
  join_utts: dynamic
  remove_word_boundaries: true
  subsample: null
  subsample_type: examples
model:
  name: gpt2_lm
  model_kwargs:
    n_layer: 2
    n_head: 4
    n_embd: 128
    n_positions: 256
    n_inner: 512
trainer:
  batch_size: 32
  lr: 0.001
  num_warmup_steps: 10
  max_training_steps: 100
  logging_steps: 1
  save_steps: 50
  eval_steps: 50