Error "Unexpected CUDA error: out of memory" when running train.py

Question

Error "Unexpected CUDA error: out of memory" when running train.py

Zhang-Xiaoxue opened this issue 3 years ago · comments

Lisnol commented 3 years ago

When I run the code train.py, the following error code is displayed:

the nvidia-smi results are:

My tensorflow version: 1.15.0. Carla version: 0.9.11

thoithoi58 · Answer 1 · Thu Jul 01 2021 20:06:33 GMT+0800 (China Standard Time)

This repo only work with 0.9.5, so give it a try

Lisnol · Answer 2 · Thu Jul 01 2021 22:19:03 GMT+0800 (China Standard Time)

This repo only work with 0.9.5, so give it a try

@thoithoi58 I tried 0.9.5. The carla 0.9.5 requires that the Python version is 3.5. But when I use python=3.5, there are many errors appearing. For example, it cannot import dataclasses.

Could you help me?

thoithoi58 · Answer 3 · Fri Jul 02 2021 00:12:59 GMT+0800 (China Standard Time)

This repo only work with 0.9.5, so give it a try

@thoithoi58 I tried 0.9.5. The carla 0.9.5 requires that the Python version is 3.5. But when I use python=3.5, there are many errors appearing. For example, it cannot import dataclasses.

Could you help me?

did you use the pretrained carla? try to make it manually, i use the windows pretrained carla which support python 3.7, you can try it