| all_vis_tokens: false | |
| append_eos_token: true | |
| batch_size_test: 64 | |
| batch_size_train: 8 | |
| dataset_name: audiocaps | |
| end_layer_idx: 31 | |
| freqm_p: 24 | |
| image_res: 224 | |
| injected_hidden_states: 6 | |
| lm_loss_weight: 0.1 | |
| melbins: 128 | |
| modality: audio | |
| noise: false | |
| norm_mean: -4.2677393 | |
| norm_std: 4.5689974 | |
| num_tries: 8 | |
| num_workers: 4 | |
| optimizer: {lr: 2e-05, opt: adamW, prompt_lr: 1e-05, weight_decay: 0.02} | |
| pretrained_model: /gpfswork/rech/dyf/ugz83ue/.cache/torch/hub/checkpoints/audioset_10_10_0.4593.pth | |
| prompt_len: 10 | |
| prompt_tuning: true | |
| replace_added_tokens: true | |
| schedular: {cooldown_epochs: 0, decay_rate: 1, epochs: 30, lr: 2e-05, min_lr: 1e-06, | |
| sched: cosine, scheduler_groups: 0, warmup_epochs: 4, warmup_lr: 1e-05} | |
| shift_labels: false | |
| skip_norm: false | |
| start_layer_idx: 19 | |
| target_length: 1024 | |
| test_split: audiocaps_caption_test | |
| timem_p: 96 | |
| train_split: audiocaps_caption_train | |
| unfreeze_text_layer_norm: false | |
| unfreeze_vision_layer_norm: false | |
| use_cache: false | |
| use_vis_prefix: true | |
| val_split: audiocaps_caption_val | |
| vision_model_name: ast | |
| warm_up: true | |