Closed Yiir-1 closed 1 year ago
Thanks for the good work! I want to know how long does it takes for training the model using 8 GPUs. And which GPU do you use to train the model?
We use 8 A100 GPUs to train the models, and the training time for camera-based, LiDAR-based and multimodal baselines are approximately 20h, 12h, 25h. For camera-based CONet, LiDAR-based CONet and multimodal CONet, the training time are 34h, 20h, 36h. Note that the validation is enabled per epoch, so you can disable it for fast training.
Thanks for the good work! I want to know how long does it takes for training the model using 8 GPUs? And which GPU do you use to train the model?