use better tensorboard distribution

[DeNeutoy]

This uses a better version of tensorboard which also contains the frontend visualisation stuff, so you don't have to use a different virtualenv to actually look at the logs.

I ran a quick test experiment, everything looks fine:

Fixes #342

[schmmd]

What's there not to like? These even gets rid of pillow...

[DeNeutoy]

It doesn't actually get rid of pillow in the sense that it still is installed, it just gets installed properly by this version of tensorboard, but I guess it's good that we don't have to explicitly depend on it

[nelson-liu]

maybe worth noting: After this change, I couldn't get tensorboardX (formerly called tensorboard-pytorch) working while using allennlp as a module. I'd get the following error:

Traceback (most recent call last):
  File "train_model.py", line 13, in <module>
    from tensorboardX import SummaryWriter
  File "/home/nfliu/miniconda3/envs/447/lib/python3.6/site-packages/tensorboardX/__init__.py", line 4, in <module>
    from .writer import FileWriter, SummaryWriter
  File "/home/nfliu/miniconda3/envs/447/lib/python3.6/site-packages/tensorboardX/writer.py", line 24, in <module>
    from .src import event_pb2
  File "/home/nfliu/miniconda3/envs/447/lib/python3.6/site-packages/tensorboardX/src/event_pb2.py", line 16, in <module>
    from tensorboardX.src import summary_pb2 as tensorboardX_dot_src_dot_summary__pb2
  File "/home/nfliu/miniconda3/envs/447/lib/python3.6/site-packages/tensorboardX/src/summary_pb2.py", line 16, in <module>
    from tensorboardX.src import tensor_pb2 as tensorboardX_dot_src_dot_tensor__pb2
  File "/home/nfliu/miniconda3/envs/447/lib/python3.6/site-packages/tensorboardX/src/tensor_pb2.py", line 17, in <module>
    from tensorboardX.src import tensor_shape_pb2 as tensorboardX_dot_src_dot_tensor__shape__pb2
  File "/home/nfliu/miniconda3/envs/447/lib/python3.6/site-packages/tensorboardX/src/tensor_shape_pb2.py", line 22, in <module>
    serialized_pb=_b('\n#tensorboardX/src/tensor_shape.proto\x12\x0btensorboard\"{\n\x10TensorShapeProto\x12.\n\x03\x64im\x18\x02 \x03(\x0b\x32!.tensorboard.TensorShapeProto.Dim\x12\x14\n\x0cunknown_rank\x18\x03 \x01(\x08\x1a!\n\x03\x44im\x12\x0c\n\x04size\x18\x01 \x01(\x03\x12\x0c\n\x04name\x18\x02 \x01(\tB2\n\x18org.tensorflow.frameworkB\x11TensorShapeProtosP\x01\xf8\x01\x01\x62\x06proto3')
  File "/home/nfliu/miniconda3/envs/447/lib/python3.6/site-packages/google/protobuf/descriptor.py", line 829, in __new__
    return _message.default_pool.AddSerializedFile(serialized_pb)
TypeError: Couldn't build proto file into descriptor pool!
Invalid proto descriptor for file "tensorboardX/src/tensor_shape.proto":
  tensorboard.TensorShapeProto.dim: "tensorboard.TensorShapeProto.dim" is already defined in file "tensorboard/src/tensor_shape.proto".
  tensorboard.TensorShapeProto.unknown_rank: "tensorboard.TensorShapeProto.unknown_rank" is already defined in file "tensorboard/src/tensor_shape.proto".
  tensorboard.TensorShapeProto.Dim.size: "tensorboard.TensorShapeProto.Dim.size" is already defined in file "tensorboard/src/tensor_shape.proto".
  tensorboard.TensorShapeProto.Dim.name: "tensorboard.TensorShapeProto.Dim.name" is already defined in file "tensorboard/src/tensor_shape.proto".
  tensorboard.TensorShapeProto.Dim: "tensorboard.TensorShapeProto.Dim" is already defined in file "tensorboard/src/tensor_shape.proto".
  tensorboard.TensorShapeProto: "tensorboard.TensorShapeProto" is already defined in file "tensorboard/src/tensor_shape.proto".
  tensorboard.TensorShapeProto.dim: "tensorboard.TensorShapeProto.Dim" seems to be defined in "tensorboard/src/tensor_shape.proto", which is not imported by "tensorboardX/src/tensor_shape.proto".  To use it here, please add the necessary import.

It's an easy enough fix --- just use the tensorboard package from DMLC instead of tensorboardX. Mostly putting this here since it might be helpful to others that get this error as well. Weirdly enough, they can co-exist on my OSX machine but not on my linux box.