| # visopt_audio_audiocaps_caption_reptok_6clst_l19_31_nocache_noeostk_ptl10_ptlr1e5fixed_ast_bs8_longt_lessmask_opt_2_7b | |
| config=./configs/audio/ePALM_audio_caption.yaml | |
| data_dir=data/audiocaps | |
| output_dir=logs/epalm/ePALM_audio_caption | |
| accelerate launch --multi_gpu --mixed_precision=fp16 --num_processes=4 --num_machines=1 accelerate_training/audio_caption.py \ | |
| --config $config \ | |
| --output_dir $output_dir \ | |
| --data_dir $data_dir \ | |
| --save_best \ | |
| --text_model facebook/opt-2.7b \ | |
| --low_cpu | |
| # --evaluate --checkpoint $WORK/logs/epalm/ePALM_audio_caption/checkpoint_best.pth | |