Open vlad17 opened 7 years ago
Sure, I will update this on the next release :)
thanks! i've been seeing a related problem now:
2018-02-04 22:39:16.960722: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1093] Ignoring visible gpu device (device: 0, name: Tesla M60, pci bus id: 0000:00:1b.0, compute capability: 5.2) with Cuda compute capability 5.2. The minimum required Cuda capability is 7.0.
This stems from the same issue (on the dl/tensorflow/1.4.0/Dockerfile-py3.gpu.cuda9cudnn7_aws dockerfile)
ENV TF_CUDA_COMPUTE_CAPABILITIES=3.7,7.0
should perhaps be ENV TF_CUDA_COMPUTE_CAPABILITIES=3.7,5.2,7.0
?
Hi @vlad17, sorry for the late reply,
I've just labeled this issue as a feature request, we will add this in the next release.
The python 3 GPU docker file specifies
ENV TF_CUDA_COMPUTE_CAPABILITIES=3.7
, which is the compute capability for the K80s. AWS also has g3's M80 cards, which have compute capabilities5.2
. Could that line be changed toENV TF_CUDA_COMPUTE_CAPABILITIES=3.7,5.2
so that the TF that's built is optimized for all AWS GPU offerings?See nvidia for listing