Docker file instructions for building gpu dockers are not compatible with new tensorflow code.
rrkarim opened this issue · comments
Bug Report
If this is a bug report, please fill out the following form in full:
System information
- OS Platform and Distribution (e.g., Linux Ubuntu 16.04): 18.04
Describe the problem
Trying to build the Dockerfile.devel-gpu image I'm running into the problem. Compiler can't find CUBLASLT_MATRIX_LAYOUT_STRIDED_BATCH_OFFSET (and more) attribute in cublas headers. I think this is the breaking change: tensorflow/tensorflow@fb9dcd1#diff-ad24359a46ac20404fa3d12efb6fbc3eb0ee6095a9fa02b2bf7cb3425ea66be2.
I guess we need to support more recent cublas libraries in docker instructions.
Could you please try Dockerfile.gpu file and let us know if that works. Thank you!
Dockerfile.gpu doesn't build tf serving
@singhniraj08 sorry, no, when I said Dockerfile.gpu doesn't build I meant it doesnt run bazel build but the dockerfile is successful in building the image. Dockerfile.devel-gpu is not though and the problem is with cublas version.
@rrkarim, Thank you for the clarification.
@christisg, Dockerfile.devel-gpu image fails to build TF serving because of cublas version. Could you please take a look. Thank you!
This issue has been marked stale because it has no recent activity since 7 days. It will be closed if no further activity occurs. Thank you.
This issue was closed due to lack of activity after being marked stale for past 7 days.
