RuntimeError: rnnt_loss status 1

Question

RuntimeError: rnnt_loss status 1

CLC2019 opened this issue 3 years ago · comments

CLC2019 commented 3 years ago

Hello,the error occurs when I run my project.Then I go to run 'python -m warp_rnnt.test',but the same error occurs which is showed as follows. Could you give me some ideas to solve it?The pytorch version is 1.6.0.The CUDA version is 10.1.The gcc version is 7.5.0.

Ivan Sorokin · Answer 1 · Sun Aug 08 2021 04:59:43 GMT+0800 (China Standard Time)

Has it worked successfully before? Maybe you changed the gpu card? Try to reinstall the cuda driver.

CLC2019 · Answer 2 · Sat Sep 04 2021 17:27:40 GMT+0800 (China Standard Time)

@1ytic I am really sorry for replying late.It's my first test on my server.During this time,I tested it in other servers without errors.And I also reinstalled the cuda driver with the same error on my sersver,The pytorch version is 1.6.0.The CUDA version is 10.0.The cudatoolkit version is 10.1.Could you give me some advice?Thanks a lot.

Ivan Sorokin · Answer 3 · Mon Sep 06 2021 15:19:53 GMT+0800 (China Standard Time)

Try to recompile the warp-rnnt package.

CLC2019 · Answer 4 · Mon Sep 06 2021 16:42:28 GMT+0800 (China Standard Time)

@1ytic Thanks a lot.The error is solved.

jiay7 · Answer 5 · Fri Mar 18 2022 17:15:24 GMT+0800 (China Standard Time)

@1ytic Thanks a lot.The error is solved.

Hi, I got the same problem. How did you solve it? python=3.8 torch verison=1.10.2 cudatoolkit=10.2.89 CUDA version=10.2