networking
networking copied to clipboard
Cannot build due to "Unable to find package for @org_tensorflow//tensorflow/core..."
After trying ./configure and bazel build -c opt //tensorflow_networking/mpi:all
I get the error
ERROR: Skipping '//tensorflow_networking/mpi:all': while parsing '//tensorflow_networking/mpi:all': error loading package 'tensorflow_networking/mpi': Unable to find package for @org_tensorflow//tensorflow/core:platform/default/build_config.bzl: The repository '@org_tensorflow' could not be resolved.
WARNING: Target pattern parsing failed.
ERROR: while parsing '//tensorflow_networking/mpi:all': error loading package 'tensorflow_networking/mpi': Unable to find package for @org_tensorflow//tensorflow/core:platform/default/build_config.bzl: The repository '@org_tensorflow' could not be resolved.
INFO: Elapsed time: 9.969s
INFO: 0 processes.
FAILED: Build did NOT complete successfully (1 packages loaded)
Bazel is completely obscure to me so no idea what this is meant to say.
I built TF 2.1 using Bazel right before this. Do I need to place this repo files inside the TF tree? Do I need to tell it about TF?
@Flamefire MPI plugin needs some renovation which unfortunately is still WIP. In other words, it does not build now and we apologise for that.
I will update this issue when the MPI plugin is ready for use.
This is really sad. TF without MPI on clusters will use the very slow Ethernet connections
This is really sad. TF without MPI on clusters will use the very slow Ethernet connections
You could build TF 1.x with built-in MPI support. For TF 2.x, you could try the built-in NCCL support with MultiWorkerMirroredStrategy, which gives decent performance using RDMA without MPI.
Also ping @jbedorf who is the author of the MPI plugin.
Getting the same error... Following for an update