Closed LeningWang closed 2 weeks ago
We used the following script to visualize based on the provided model files, but the generated video is completely black.
Your ann_file
for testing is wrong.
We have updated the readme for the link to all pre-generated data files. Could you try again?
Sorry, I forgot to update some code.
It is a known issue that overflow would happen when xformers
is disabled. Could you pull the latest commit and try again?
Maybe if we don't use xformers, we won't need to pull the latest commit? In other words, we only download nuscenes_mmdet3d-12Hz/nuscenes_interp_12Hz_infos_val.pkl and run python tools/test.py resume_from_checkpoint=${RUN_LOG_DIR} task_id=${ANY} runner.validation_times=4 runner.pipeline_param.init_noise=rand_all ++dataset.data.val.ann_file=${ROOT}/../data/nuscenes_mmdet3d-12Hz/nuscenes_interp_12Hz_infos_val.pkl
I do not understand why you refuse to take updates.
Anyway, you are free to choose to use or not to use xformers with a6b0d8d. Without it, please always enable xformers.
Thanks, my problems are well solved. Thank you, nice person, and nice work.
I would like to directly obtain visualizations in MagicDrive-t using pretrained weights for rawbox_mv2.0t_0.4.3.yaml on OneDrive. Could you please provide specific execution scripts, preferably for generating frames at different time intervals? Thank you very much.