| batch_size: 8 |
| ckpt_freq: 500 |
| data: |
| eval_data: '' |
| shuffle: true |
| train_data: /data/train.jsonl |
| do_ckpt: true |
| do_eval: false |
| duration_sec: 80.0 |
| eval_freq: 300 |
| first_codebook_weight_multiplier: 100.0 |
| full_finetuning: false |
| gradient_checkpointing: true |
| log_freq: 10 |
| lora: |
| enable: true |
| ft_embed: false |
| rank: 64 |
| scaling: 2.0 |
| max_norm: 1.0 |
| max_steps: 3000 |
| moshi_paths: |
| config_path: null |
| hf_repo_id: kyutai/moshiko-pytorch-bf16 |
| mimi_path: null |
| moshi_path: null |
| tokenizer_path: null |
| num_ckpt_keep: 3 |
| num_microbatches: 1 |
| optim: |
| lr: 2.0e-06 |
| pct_start: 0.05 |
| weight_decay: 0.1 |
| overwrite_run_dir: true |
| param_dtype: bfloat16 |
| run_dir: /checkpoints/run_001 |
| save_adapters: true |
| seed: 42 |
| text_padding_weight: 0.5 |
| wandb: |
| key: null |
| offline: false |
| project: null |
| run_name: null |
| world_size: 1 |
|
|
