Magenta is an awesome project. Thank you so much for opening it up! I'm posting in 'tensorflow' because 'tensorflow-magenta' isn't open yet and I don't have enough reputation to christen it. Perhaps someone could tag this appropiately?
PROBLEM DESCRIPTION : I built tensorflow in ubuntu, installed dependencies, and I'm able to successfully built a target with:
bazel build //magenta:music_py_pb3
I can see the output in bazel-genfiles/magenta/protobuf, but running the python script doesn't actually output anything.
Am I missing something? Should I be looking out for a MIDI stream?
I am now trying to build the other targets, but:
bazel build //magenta:midi_io_test
...fails because there's no /tensorflow/core (I'm thinking at this point that you guys built it to tensorflow lives within the magenta repo as git submodule, so I've gone into it and hit ./configure then built:
bazel build -c opt //tensorflow/tools/pip_package:build_pip_package
...from within the /magenta/tensorflow directory, which built fine (after installing swig)...but the build still fails with:
tensorflow/core/framework/step_stats.proto: File not found.
tensorflow/core/framework/device_attributes.proto: File not found.
tensorflow/core/framework/graph.proto: File not found.
(etc.)
Since I've already done a pip install of tensorflow and I am able to use it in python2.7 from python console without a problem, wouldn't it make more sense to use existing installation? And if not, why does it fail even though I've built the submoduled tensorflow? I realise that after calling bazel to build pip package of this, I ended up with 'core' in /magenta/tensorflow/tensorflow/core, and the WORKSPACE in /magenta expects
tf_workspace("tensorflow/", "@tf"
...but that doesn't seem to be a path problem, because merging the contents of tensorflow/tensorflow to tensorflow/ still produces the same error.
Sorry to open this tag, for such an AWESOME project, with such a menial question (I promise I have better ones coming up), but I'm new to bazel and tensorflow (though not new to music and neural networks and AI in general).
UPDATE:
Discussion of this issue (I know it wasn't meant to be for that, sorry, and thanks for updating the post with magenta tag) is being carried here:
https://groups.google.com/a/tensorflow.org/forum/#!topic/magenta-discuss/HC_squh5TbA
and Carlos Cardona is updating a gist (which might eventually make a good n00b readme?) here:
https://gist.github.com/cgcardona/69015852f193d7f879996863c87b6769
on progress we're making trying to figure it out.
