search

georgeliu233 / OPGP

[ITSC'23] Code for 'Occupancy Prediction-Guided Neural Planner for Autonomous Driving'
60 stars 3 forks source link

what is the calculation time of OPGP #1

Closed teddyluo closed 1 year ago

teddyluo commented 1 year ago

Hi, I roughly read the whole paper and want to know two details: 1) on what kind of hardware was this method tested? 2）what is the computation time of OPGP (training? and inference?)?

Can you provide more details about OPGP? many thanks

georgeliu233 commented 1 year ago

Hi, I roughly read the whole paper and want to know two details:

  1. on what kind of hardware was this method tested? 2）what is the computation time of OPGP (training? and inference?)?

Can you provide more details about OPGP? many thanks

Hi Teddy, thank you for your interest of our work :) For your questions: For Q1: 1) We conducted open-loop testing for planning and testing for prediction on log-replayed random-splited WOMD (validation) data, and we did not conduct HIL (hardware in loop) test; 2) Currently we train and test for our result on the same GPU configs (on 4 A100 40G) For Q2: cumputational time is about 0.03s for training and about 0.15-0.2s for testing (planning+refinement) per sample;

I plan to provide more of the related information after this year's WOMD challenge