search

Thinklab-SJTU / Bench2DriveZoo

BEVFormer, UniAD, VAD in Closed-Loop CARLA Evaluation with World Model RL Expert Think2Drive
Other
148 stars 9 forks source link

Estimated training time of UniAD and VAD on Bench2Drive-Base #51

Open kemaloksuz opened 2 days ago

kemaloksuz commented 2 days ago

Thanks for the great work. I think you used 8xA100 to train UniAD (tiny and base) and VAD (only base) by using 1 image/GPU. Please correct me if there is a mistake in this information. Also, how long should it take to train each of these models using this setting?

jayyoung0802 commented 16 hours ago

Hi @kemaloksuz , batch_size_per_gpu=1. As you train, you can see the time remaining in real time. And the details about training, you can refer to #5.