model_name: Llama-3.2-1B
out_dir: /home/aiops/zhuty/litgpt/out/pretrain/llama3-1b-amv1-4epoch-2k
precision: bf16-mixed
initial_checkpoint_dir: checkpoints/meta-llama/Llama-3.2-1B
resume: auto
data:
  class_path: litgpt.data.TextFiles
  init_args:
    train_data_path: /home/aiops/zhuty/cont_data/bio/amv1/train
    val_data_path: /home/aiops/zhuty/cont_data/bio/amv1/test
    seed: 42
    num_workers: 10
    add_eos: true
train:
  save_interval: 2500
  save_interval_epochs: 1
  save_optimizer_state: false
  log_interval: 1
  global_batch_size: 256
  micro_batch_size: 1
  lr_warmup_fraction: 0.01
  epochs: 4
  max_seq_length: 2048
  max_norm: 1.0
  min_lr: 5.0e-06
eval:
  interval: 20
  max_iters: 100
  initial_validation: true
  final_validation: true
  evaluate_example: first
  num_generation_examples: 1
  calculate_exact_match: false
log:
  project: mathcont
optimizer:
  class_path: torch.optim.AdamW
  init_args:
    lr: 5.0e-05
    weight_decay: 0.1
    betas:
    - 0.9
    - 0.95
devices: auto
num_nodes: 1
tokenizer_dir: checkpoints/meta-llama/Llama-3.2-1B
logger_name: wandb
seed: 42
compiler: torch
executors:
- sdpa
- torchcompile
- torch
strategy: fsdp