| model_name: micro-llama-300M-v2 | |
| out_dir: out/pretrain/micro-llama-v2 | |
| precision: bf16-mixed | |
| initial_checkpoint_dir: /root/litgpt/out_lightning_ai/step-00128000-converted | |
| resume: false | |
| data: | |
| class_path: litgpt.data.MicroLlama | |
| init_args: | |
| data_path: s3://microllama-v2 | |
| seed: 42 | |
| num_workers: 8 | |
| train: | |
| save_interval: 1000 | |
| log_interval: 10 | |
| global_batch_size: 32 | |
| micro_batch_size: 4 | |
| lr_warmup_steps: 2000 | |
| max_tokens: 3000000000000 | |
| max_seq_length: 2048 | |
| max_norm: 1.0 | |
| min_lr: 4.0e-05 | |
| eval: | |
| interval: 1000 | |
| max_iters: 100 | |
| initial_validation: false | |
| final_validation: true | |
| evaluate_example: first | |
| optimizer: | |
| class_path: torch.optim.AdamW | |
| init_args: | |
| lr: 0.0004 | |
| 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 | |