| OUTPUT_DIR="${HOME}/checkpoints/cwm_cvpr_checkpoints/ablation_3frame_no_clumping_mr0.60/" |
| DATA_PATH="${HOME}/BBNet/bbnet/models/VideoMAE-main/video_file_lists/kinetics_400_train_list.txt" |
|
|
| python3 ~/BBNet/bbnet/models/VideoMAE-main/run_cwm_pretraining.py \ |
| --data_path ${DATA_PATH} \ |
| --mask_type rotated_table \ |
| --mask_ratio 0.60 \ |
| --mask_kwargs '{"tube_length": 1}' \ |
| --model vitbase_8x8patch_3frames_1tube \ |
| --context_frames 2 \ |
| --target_frames 1 \ |
| --temporal_units 'ms' \ |
| --sampling_rate 150 \ |
| --context_target_gap 150 150 \ |
| --batch_size 16 \ |
| --accum_iter 2 \ |
| --opt adamw \ |
| --opt_betas 0.9 0.95 \ |
| --warmup_epochs 40 \ |
| --save_ckpt_freq 50 \ |
| --epochs 800 \ |
| --no_normlize_target \ |
| --rescale_size 224 \ |
| --augmentation_type 'multiscale' \ |
| --augmentation_scales 1.0 0.875 0.75 0.66 \ |
| --log_dir ${OUTPUT_DIR} \ |
| --output_dir ${OUTPUT_DIR} \ |
| --print_freq 1 \ |
| --num_workers 16 \ |
| --use_xla \ |
| --min_lr 1e-5 |
|
|
