| exp_name: pretrained | |
| checkpoints_dir: ./checkpoints/${exp_name}/${dataset.dataset_name} | |
| preprocess: | |
| padding: true | |
| use_kinematic: true | |
| model: | |
| motion_encoder: vit_base_patch16_224_in21k | |
| text_encoder: distilbert-base-uncased | |
| train: | |
| batch_size: 256 | |
| epoch: 50 | |
| optimizer: | |
| motion_lr: 1.0e-05 | |
| text_lr: 1.0e-05 | |
| head_lr: 1.0e-05 | |
| train_motion_encoder: true | |
| train_text_encoder: true | |
| motion_encoder_pretrained: true | |
| patch_size: 16 | |
| seed: 42 | |
| eval: | |
| eval_train: false | |
| use_best_model: true | |
| dataset: | |
| dataset_name: KIT-ML | |
| data_root: ./data/KIT-ML | |
| motion_dir: ${dataset.data_root}/new_joints | |
| text_dir: ${dataset.data_root}/texts | |
| joints_num: 21 | |
| max_motion_length: 224 | |
| times: 1 | |
| fps: 12.5 | |