| { | |
| "modelname": "openai/whisper-base", | |
| "gen_args": { | |
| "num_beams": 3 | |
| }, | |
| "fallback": [1,7,5], | |
| "corpora": { | |
| "train": { | |
| "thchs": { | |
| "languages": "cmn", | |
| "limit": [ | |
| 1000 | |
| ] | |
| } | |
| }, | |
| "dev": { | |
| "thchs": { | |
| "languages": "cmn", | |
| "limit": [ | |
| 200 | |
| ] | |
| } | |
| }, | |
| "thchs": { | |
| "asc": { | |
| "languages": "cmn", | |
| "limit": [ | |
| 0 | |
| ] | |
| } | |
| } | |
| }, | |
| "hyperparams": { | |
| "output_dir": "../models/lowhipa-base-thchs30", | |
| "overwrite_output_dir": "True", | |
| "peft": "True", | |
| "per_device_train_batch_size": 16, | |
| "gradient_accumulation_steps": 1, | |
| "learning_rate": 0.001, | |
| "warmup_ratio": 0.1, | |
| "max_steps": 630, | |
| "gradient_checkpointing": "True", | |
| "fp16": "True", | |
| "eval_strategy": "steps", | |
| "per_device_eval_batch_size": 8, | |
| "predict_with_generate": "True", | |
| "generation_max_length": 225, | |
| "save_steps": 126, | |
| "eval_steps": 126, | |
| "logging_steps": 63, | |
| "report_to": [ | |
| "tensorboard" | |
| ], | |
| "load_best_model_at_end": "True", | |
| "metric_for_best_model": "per", | |
| "greater_is_better": "False", | |
| "hub_private_repo": "True", | |
| "push_to_hub": "True", | |
| "torch_empty_cache_steps": 1 | |
| } | |
| } |