Closed snakers4 closed 2 years ago
Same happens here https://github.com/1ytic/warp-rna Probably I should have created a ticket in this repo https://github.com/1ytic/warp-rnnt
I tried these hacks, the error changed, it started complaining that
which g++
exits with error, manually it produces empty output
No CUDA 11 errors though, but maybe these are related
@1ytic Am I doing something wrong?
I checked this version pip3 install torch==1.9.0+cu111 -f https://download.pytorch.org/whl/cu111/torch_stable.html
and warp_rnnt
package compiled successfully. I also have the local cuda install with the same version 11.1.
I see. I tried using these images from scratch:
docker pull pytorch/pytorch:1.10.0-cuda11.3-cudnn8-devel
docker pull pytorch/pytorch:1.9.0-cuda11.1-cudnn8-devel
And it compiled successfully. Looks like the problem is with our images, we most likely borked something up inside of them. Many thanks.
Hi,
I have the following library / image versions:
pytorch/pytorch:1.9.0-cuda11.1-cudnn8-devel
When I try
pip install warp_rnnt
I have the following output:Error message
I used to have similar errors with
warp-ctc
, and they were resolved with hacks like this:Is there anything I can do about this without downgrading or changing the current environment?