- model_config.model_name=openai/whisper-base
- model_config.attn_implementation=flash_attention_2
- data_config.name=mitermix/audiosnippets
- data_config.audio_key=mp3
- data_config.caption_key=caption
- trainer_config.bf16=true
- trainer_config.dataloader_num_workers=16
- trainer_config.per_device_train_batch_size=128
- trainer_config.max_steps=50000
- ++trainer_config.push_to_hub=true
- ++trainer_config.hub_model_id=seastar105/whisper-base-emo-speech-caption
- ++trainer_config.save_total_limit=5
- ++trainer_config.save_steps=5000
- ++trainer_config.dataloader_prefetch_factor=4
- ++trainer_config.warmup_steps=5000