Fix tests for new tensorboard >= 2.6

[awaelchli]

Unblocks master branch for failing tests.
Tensorboard recently released 2.6.0, which breaks one of our tests.

==================================== ERRORS ====================================
______________ ERROR collecting tests/loggers/test_tensorboard.py ______________
tests/loggers/test_tensorboard.py:23: in <module>
    from tensorboard.backend.event_processing.event_accumulator import EventAccumulator
/usr/local/lib/python3.7/dist-packages/tensorboard/backend/event_processing/event_accumulator.py:21: in <module>
    from tensorboard.backend.event_processing import event_file_loader
/usr/local/lib/python3.7/dist-packages/tensorboard/backend/event_processing/event_file_loader.py:21: in <module>
    from tensorboard import dataclass_compat
/usr/local/lib/python3.7/dist-packages/tensorboard/dataclass_compat.py:35: in <module>
    from tensorboard.plugins.mesh import metadata as mesh_metadata
/usr/local/lib/python3.7/dist-packages/tensorboard/plugins/mesh/metadata.py:20: in <module>
    from tensorboard.plugins.mesh import plugin_data_pb2
/usr/local/lib/python3.7/dist-packages/tensorboard/plugins/mesh/plugin_data_pb2.py:23: in <module>
    serialized_pb=_b('\n*tensorboard/plugins/mesh/plugin_data.proto\x12\x10tensorboard.mesh\"\xea\x01\n\x0eMeshPluginData\x12\x0f\n\x07version\x18\x01 \x01(\x05\x12\x0c\n\x04name\x18\x02 \x01(\t\x12\x42\n\x0c\x63ontent_type\x18\x03 \x01(\x0e\x32,.tensorboard.mesh.MeshPluginData.ContentType\x12\x13\n\x0bjson_config\x18\x05 \x01(\t\x12\r\n\x05shape\x18\x06 \x03(\x05\x12\x12\n\ncomponents\x18\x07 \x01(\r\"=\n\x0b\x43ontentType\x12\r\n\tUNDEFINED\x10\x00\x12\n\n\x06VERTEX\x10\x01\x12\x08\n\x04\x46\x41\x43\x45\x10\x02\x12\t\n\x05\x43OLOR\x10\x03\x62\x06proto3')
/usr/local/lib/python3.7/dist-packages/google/protobuf/descriptor.py:983: in __new__
    return _message.default_pool.AddSerializedFile(serialized_pb)
E   TypeError: Couldn't build proto file into descriptor pool!
E   Invalid proto descriptor for file "tensorboard/plugins/mesh/plugin_data.proto":
E     tensorboard.mesh.MeshPluginData.version: "tensorboard.mesh.MeshPluginData.version" is already defined in file "tensorboardX/proto/plugin_mesh.proto".
E     tensorboard.mesh.MeshPluginData.name: "tensorboard.mesh.MeshPluginData.name" is already defined in file "tensorboardX/proto/plugin_mesh.proto".
E     tensorboard.mesh.MeshPluginData.content_type: "tensorboard.mesh.MeshPluginData.content_type" is already defined in file "tensorboardX/proto/plugin_mesh.proto".
E     tensorboard.mesh.MeshPluginData.json_config: "tensorboard.mesh.MeshPluginData.json_config" is already defined in file "tensorboardX/proto/plugin_mesh.proto".
E     tensorboard.mesh.MeshPluginData.shape: "tensorboard.mesh.MeshPluginData.shape" is already defined in file "tensorboardX/proto/plugin_mesh.proto".
E     tensorboard.mesh.MeshPluginData.UNDEFINED: "tensorboard.mesh.MeshPluginData.UNDEFINED" is already defined in file "tensorboardX/proto/plugin_mesh.proto".
E     tensorboard.mesh.MeshPluginData.UNDEFINED: Note that enum values use C++ scoping rules, meaning that enum values are siblings of their type, not children of it.  Therefore, "UNDEFINED" must be unique within "tensorboard.mesh.MeshPluginData", not just within "ContentType".
E     tensorboard.mesh.MeshPluginData.VERTEX: "tensorboard.mesh.MeshPluginData.VERTEX" is already defined in file "tensorboardX/proto/plugin_mesh.proto".
E     tensorboard.mesh.MeshPluginData.VERTEX: Note that enum values use C++ scoping rules, meaning that enum values are siblings of their type, not children of it.  Therefore, "VERTEX" must be unique within "tensorboard.mesh.MeshPluginData", not just within "ContentType".
E     tensorboard.mesh.MeshPluginData.FACE: "tensorboard.mesh.MeshPluginData.FACE" is already defined in file "tensorboardX/proto/plugin_mesh.proto".
E     tensorboard.mesh.MeshPluginData.FACE: Note that enum values use C++ scoping rules, meaning that enum values are siblings of their type, not children of it.  Therefore, "FACE" must be unique within "tensorboard.mesh.MeshPluginData", not just within "ContentType".
E     tensorboard.mesh.MeshPluginData.COLOR: "tensorboard.mesh.MeshPluginData.COLOR" is already defined in file "tensorboardX/proto/plugin_mesh.proto".
E     tensorboard.mesh.MeshPluginData.COLOR: Note that enum values use C++ scoping rules, meaning that enum values are siblings of their type, not children of it.  Therefore, "COLOR" must be unique within "tensorboard.mesh.MeshPluginData", not just within "ContentType".
E     tensorboard.mesh.MeshPluginData.ContentType: "tensorboard.mesh.MeshPluginData.ContentType" is already defined in file "tensorboardX/proto/plugin_mesh.proto".
E     tensorboard.mesh.MeshPluginData: "tensorboard.mesh.MeshPluginData" is already defined in file "tensorboardX/proto/plugin_mesh.proto".
E     tensorboard.mesh.MeshPluginData.content_type: "tensorboard.mesh.MeshPluginData.ContentType" seems to be defined in "tensorboardX/proto/plugin_mesh.proto", which is not imported by "tensorboard/plugins/mesh/plugin_data.proto".  To use it here, please add the necessary import.

Before submitting

• Was this discussed/approved via a GitHub issue? (not for typos and docs)
• Did you read the contributor guideline, Pull Request section?
• Did you make sure your PR does only one thing, instead of bundling different changes together?
• Did you make sure to update the documentation with your changes? (if necessary)
• Did you write any new necessary tests? (not for typos and docs)
• Did you verify new and existing tests pass locally with your changes?
• Did you update the CHANGELOG? (not for typos, docs, test updates, or internal minor changes/refactorings)

PR review

Anyone in the community is free to review the PR once the tests have passed.
Before you start reviewing make sure you have read Review guidelines. In short, see the following bullet-list:

• Is this pull request ready for review? (if not, please submit in draft mode)
• Check that all items from Before submitting are resolved
• Make sure the title is self-explanatory and the description concisely explains the PR
• Add labels and milestones (and optionally projects) to the PR so it can be classified

Did you have fun?

I made sure I had fun coding 🙃

[codecov]

Codecov Report

Merging #8789 (16b2d32) into master (87093a3) will not change coverage.
The diff coverage is n/a.

@@          Coverage Diff           @@
##           master   #8789   +/-   ##
======================================
  Coverage      89%     89%           
======================================
  Files         169     169           
  Lines       14068   14068           
======================================
  Hits        12460   12460           
  Misses       1608    1608           

[carmocca]

Should we open a GH issue in https://github.com/tensorflow/tensorboard?

[awaelchli]

Yes although I'm still missing some information. The issue is that on the CPU installation it works fine. There must be another dependency (of tensorboard?) that plays a role here and which is different in the GPU container.