Hi,
Thank you for your impressive work on this project! I am very interested in your research and am currently trying to reproduce the results. I have a few questions regarding the implementation details:
-
As far as I know, Bench2Drive is based on CARLA, which typically requires Python 3.8. However, AutoVLA appears to depend on Python 3.9. Could you please share how you managed to resolve this version conflict? Did you use specific environment configurations or patches?
-
When training on Bench2Drive, did you follow a process similar to nuPlan by creating a Chain-of-Thought (CoT) dataset? If so, do you have a tentative timeline for when these datasets or the generation scripts will be made public?
-
Could you share the pre-trained weights and the processed dataset with me for academic purposes? I strictly commit to using them only for academic research. My email address is: zcuangzhao@hust.edu.cn.
Thank you for your time and help!
Hi,
Thank you for your impressive work on this project! I am very interested in your research and am currently trying to reproduce the results. I have a few questions regarding the implementation details:
As far as I know, Bench2Drive is based on CARLA, which typically requires Python 3.8. However, AutoVLA appears to depend on Python 3.9. Could you please share how you managed to resolve this version conflict? Did you use specific environment configurations or patches?
When training on Bench2Drive, did you follow a process similar to nuPlan by creating a Chain-of-Thought (CoT) dataset? If so, do you have a tentative timeline for when these datasets or the generation scripts will be made public?
Could you share the pre-trained weights and the processed dataset with me for academic purposes? I strictly commit to using them only for academic research. My email address is: zcuangzhao@hust.edu.cn.
Thank you for your time and help!