search

HaozheQi / P2B

P2B: Point-to-Box Network for 3D Object Tracking in Point Clouds
189 stars 36 forks source link

training time and machines #4

Closed WangZhouTao closed 4 years ago

WangZhouTao commented 4 years ago

Hello, thanks for your good job. I noticed that you mentioned the running speed on a single NVIDIA 1080Ti GPU. Can you share the training time and what machines are used?

HaozheQi commented 4 years ago

Hi, @WangZhouTao, actually, we don't count the training time strictly, it takes about 10 hours to converge the model with an Amax sever.

WangZhouTao commented 4 years ago

Hi, thanks for your reply. In fact, I only have a 1080ti GPU. Can you tell me the type and number of GPUs in your Amax sever? In this way, I can assess whether I can retrain this framework. looking forward to your reply.

HaozheQi commented 4 years ago

One GPU is OK for the training and you only need to change the settings in 'train_tracking.py' like that, parser.add_argument('--ngpu', type=int, default=1, help='# GPUs') os.environ["CUDA_VISIBLE_DEVICES"] = '0'

WangZhouTao commented 4 years ago

Thanks for your reply.