search

quangnhat185 / darknet_for_colab

Darknet repository for training YOLOv4 on Google Colab with Custom Dataset
MIT License
31 stars 57 forks source link

Make failed saying *nvcc fatal : Unsupported gpu architecture 'compute_30'* #2

Closed nagasanjay closed 1 year ago

nagasanjay commented 3 years ago

Support for 'compute_30' is removed from CUDA version 10.2. So it is unable to compile for this architecture.

divyadreddy commented 3 years ago

Even I'm getting this error, can someone please fix this

Saisrivathsan15 commented 3 years ago

https://stackoverflow.com/questions/64774548/unsupported-gpu-architecture-compute-30-on-a-cuda-5-capable-gpu

It worked when compute_30 was replaced with compute_50 in Makefile as given in the above link.

nagasanjay commented 3 years ago

remove the line with compute_30 in the makefile. It works

https://stackoverflow.com/questions/64774548/unsupported-gpu-architecture-compute-30-on-a-cuda-5-capable-gpu

It worked when compute_30 was replaced with compute_50 in Makefile as given in the above link.

fyuan12 commented 3 years ago

Changing the line ARCH= -gencode arch=compute_30,code=sm_30 \ to ARCH= -gencode arch=compute_50,code=sm_50 \ worked for me.

jcads commented 3 years ago

Changing the line ARCH= -gencode arch=compute_30,code=sm_30 \ to ARCH= -gencode arch=compute_50,code=sm_50 \ worked for me.

Can I ask where exactly is this line located? I searched it in CMakeLists.txt and it's not there.

nagasanjay commented 3 years ago

Changing the line ARCH= -gencode arch=compute_30,code=sm_30 \ to ARCH= -gencode arch=compute_50,code=sm_50 \ worked for me.

Can I ask where exactly is this line located? I searched it in CMakeLists.txt and it's not there.

It is in the "Makefile"

quangnhat185 commented 1 year ago

This happens due to the upgrade CUDA on Colab. MAKEFILE has been updated to fix this issue