Closed Luna-yu closed 3 years ago
+1 Also experiencing this.
Edit: I resolved my issue by commenting out the old nvcc flags in CMakeLists.txt. Specifically, lines 35-38
Thank you very much, I solved the problem through this method
Thanks, @Chudbrochil. This worked for me as well trying to setup warp_ctc using cuda 11.2.
@Luna-yu: if @Chudbrochil solution worked for you, want to close the issue?
(yu_tf_py) root@gao-PowerEdge-T640:~/yu_ex/warp-ctc3333/pytorch_binding# python setup.py install
Traceback (most recent call last):
File "setup.py", line 6, in
This seems like a separate issue. We can continue discussing it on this thread, but you may want to close the issue as it seems like you have successfully run the make
command.
Have you tried uninstalling and re-installing pytorch?
What OS are you running? If you're running linux, you can run this command to double check what Cuda version you have installed:
/usr/local/cuda/bin/nvcc --version
This seems like a separate issue. We can continue discussing it on this thread, but you may want to close the issue as it seems like you have successfully run the
make
command.Have you tried uninstalling and re-installing pytorch?
What OS are you running? If you're running linux, you can run this command to double check what Cuda version you have installed:
/usr/local/cuda/bin/nvcc --version
thank you!
+1 Also experiencing this.
Edit: I resolved my issue by commenting out the old nvcc flags in CMakeLists.txt. Specifically, lines 35-38
Thank you so much! I have switched to nn.CTCLoss because I could not install warp-ctc but after switching to nn.CTCLoss my accuracy was decreasing...
+1 Also experiencing this. Edit: I resolved my issue by commenting out the old nvcc flags in CMakeLists.txt. Specifically, lines 35-38
Thank you so much! I have switched to nn.CTCLoss because I could not install warp-ctc but after switching to nn.CTCLoss my accuracy was decreasing...
Lastest version of Pytorch includes CTC in its module, but there are still many project using the early version. :(
(yu_tf_py) root@localhost:/home/yu/yu_ex/warp-ctc/build# make -- cuda found TRUE -- Building shared library with GPU support -- Configuring done -- Generating done -- Build files have been written to: /home/yu/yu_ex/warp-ctc/build [ 11%] Building NVCC (Device) object CMakeFiles/warpctc.dir/src/warpctc_generated_reduce.cu.o nvcc fatal : Unsupported gpu architecture 'compute_30' CMake Error at warpctc_generated_reduce.cu.o.cmake:219 (message): Error generating /home/yu/yu_ex/warp-ctc/build/CMakeFiles/warpctc.dir/src/./warpctc_generated_reduce.cu.o
CMakeFiles/warpctc.dir/build.make:70: recipe for target 'CMakeFiles/warpctc.dir/src/warpctc_generated_reduce.cu.o' failed make[2]: [CMakeFiles/warpctc.dir/src/warpctc_generated_reduce.cu.o] Error 1 CMakeFiles/Makefile2:104: recipe for target 'CMakeFiles/warpctc.dir/all' failed make[1]: [CMakeFiles/warpctc.dir/all] Error 2 Makefile:129: recipe for target 'all' failed make: *** [all] Error 2