Magenta: Music and Art Generation with Machine Intelligence

Hi, I have finished training with the recommended hyperparameters after 4 days. But the generated sound with the midi file is very poor and lacks diversity. I trained only on the Acoustic Subset and used tensorflow 1.15.2. How should I reproduce the results in the GANsynth paper? @jesseengel

Lightweight MusicXML parser for Magenta with no dependencies. Handles uncompressed (.xml) and compressed (.mxl) files. Translates MusicXML file into NoteSequences similar to how MIDI is imported.

Because MusicXML differs from MIDI, different information is available to Magenta. For example, MusicXML does not contain MIDI CC data. However, MusicXML does contain other information (such as dynamic markings) which may be more useful in certain scenarios, such as using Magenta to generate sheet music (MusicXML files) in addition to MIDI files.

This parser currently only supports input and does not support output to MusicXML files.

Four public domain MusicXML files are included for unit testing.

I'm getting this error while trying to do the Data generation / preprocessing step of Score2Perf Music Transformer model on my computer. How can I fix it?

Hey all!

I'm on a Mac OSX attempting to build. I installed tensorflow in python 3. When I run the

bazel test //magenta:all

Command, all 6 tests fail locally. The fail log for each says that python can't import tensorflow. I'm assuming the tests are being run by my Mac's default python automatically (2.7.10, doesn't have tensorflow installed), is there a way to change that to python3?

Sorry if this is a dumb question!

Hello,

I'm running the last version MusicVAE from repository on ubuntu 18.04, cuda 10.1, tensorflow 2.2.0, configuration - hier-trio_16bar and have the error below (i tried for different batch sizes, even 1 and for different learning rates, but the problem is the same). Do you know how to fix it?

2020-06-09 21:01:27.365621: I tensorflow/core/common_runtime/bfc_allocator.cc:1010] Stats: Limit: 14684815360 InUse: 14684616704 MaxInUse: 14684815360 NumAllocs: 26588 MaxAllocSize: 181403648

2020-06-09 21:01:27.365991: W tensorflow/core/common_runtime/bfc_allocator.cc:439] **************************************************************************************************** 2020-06-09 21:01:27.366026: W tensorflow/core/framework/op_kernel.cc:1753] OP_REQUIRES failed at lstm_ops.cc:372 : Resource exhausted: OOM when allocating tensor with shape[256,1024] and type float on /job:localhost/replica:0/task:0/device:GPU:0 by allocator GPU_0_bfc 2020-06-09 21:01:34.147376: W tensorflow/core/kernels/data/cache_dataset_ops.cc:794] The calling iterator did not fully read the dataset being cached. In order to avoid unexpected truncation of the dataset, the partially cached contents of the dataset will be discarded. This can happen if you have an input pipeline similar to dataset.cache().take(k).repeat(). You should use dataset.take(k).cache().repeat() instead. Traceback (most recent call last): File "/home/burashnikova/env-tf22/lib/python3.6/site-packages/tensorflow/python/client/session.py", line 1365, in _do_call return fn(*args) File "/home/burashnikova/env-tf22/lib/python3.6/site-packages/tensorflow/python/client/session.py", line 1350, in _run_fn target_list, run_metadata) File "/home/burashnikova/env-tf22/lib/python3.6/site-packages/tensorflow/python/client/session.py", line 1443, in _call_tf_sessionrun run_metadata) tensorflow.python.framework.errors_impl.ResourceExhaustedError: 2 root error(s) found. (0) Resource exhausted: OOM when allocating tensor with shape[256,1114] and type float on /job:localhost/replica:0/task:0/device:GPU:0 by allocator GPU_0_bfc [[{{node swap_in_core_decoder_1/core_decoder_0/decoder/while/BasicDecoderStep/decoder/multi_rnn_cell/cell_0/lstm_cell/LSTMBlockCell_13_0}}]] Hint: If you want to see a list of allocated tensors when OOM happens, add report_tensor_allocations_upon_oom to RunOptions for current allocation info.

     [[add/_2901]]

Hint: If you want to see a list of allocated tensors when OOM happens, add report_tensor_allocations_upon_oom to RunOptions for current allocation info.

(1) Resource exhausted: OOM when allocating tensor with shape[256,1114] and type float on /job:localhost/replica:0/task:0/device:GPU:0 by allocator GPU_0_bfc [[{{node swap_in_core_decoder_1/core_decoder_0/decoder/while/BasicDecoderStep/decoder/multi_rnn_cell/cell_0/lstm_cell/LSTMBlockCell_13_0}}]] Hint: If you want to see a list of allocated tensors when OOM happens, add report_tensor_allocations_upon_oom to RunOptions for current allocation info.

0 successful operations. 0 derived errors ignored.

During handling of the above exception, another exception occurred:

Traceback (most recent call last): File "music_vae_train.py", line 340, in console_entry_point() File "music_vae_train.py", line 336, in console_entry_point tf.app.run(main) File "/home/burashnikova/env-tf22/lib/python3.6/site-packages/tensorflow/python/platform/app.py", line 40, in run _run(main=main, argv=argv, flags_parser=_parse_flags_tolerate_undef) File "/home/burashnikova/env-tf22/lib/python3.6/site-packages/absl/app.py", line 299, in run _run_main(main, args) File "/home/burashnikova/env-tf22/lib/python3.6/site-packages/absl/app.py", line 250, in _run_main sys.exit(main(argv)) File "music_vae_train.py", line 331, in main run(configs.CONFIG_MAP) File "music_vae_train.py", line 312, in run task=FLAGS.task) File "music_vae_train.py", line 211, in train is_chief=is_chief) File "/home/burashnikova/env-tf22/lib/python3.6/site-packages/tf_slim/training/training.py", line 551, in train loss = session.run(train_op, run_metadata=run_metadata) File "/home/burashnikova/env-tf22/lib/python3.6/site-packages/tensorflow/python/training/monitored_session.py", line 778, in run run_metadata=run_metadata) File "/home/burashnikova/env-tf22/lib/python3.6/site-packages/tensorflow/python/training/monitored_session.py", line 1283, in run run_metadata=run_metadata) File "/home/burashnikova/env-tf22/lib/python3.6/site-packages/tensorflow/python/training/monitored_session.py", line 1384, in run raise six.reraise(*original_exc_info) File "/home/burashnikova/env-tf22/lib/python3.6/site-packages/six.py", line 703, in reraise raise value File "/home/burashnikova/env-tf22/lib/python3.6/site-packages/tensorflow/python/training/monitored_session.py", line 1369, in run return self._sess.run(*args, **kwargs) File "/home/burashnikova/env-tf22/lib/python3.6/site-packages/tensorflow/python/training/monitored_session.py", line 1442, in run run_metadata=run_metadata) File "/home/burashnikova/env-tf22/lib/python3.6/site-packages/tensorflow/python/training/monitored_session.py", line 1200, in run return self._sess.run(*args, **kwargs) File "/home/burashnikova/env-tf22/lib/python3.6/site-packages/tensorflow/python/client/session.py", line 958, in run run_metadata_ptr) File "/home/burashnikova/env-tf22/lib/python3.6/site-packages/tensorflow/python/client/session.py", line 1181, in _run feed_dict_tensor, options, run_metadata) File "/home/burashnikova/env-tf22/lib/python3.6/site-packages/tensorflow/python/client/session.py", line 1359, in _do_run run_metadata) File "/home/burashnikova/env-tf22/lib/python3.6/site-packages/tensorflow/python/client/session.py", line 1384, in _do_call raise type(e)(node_def, op, message) tensorflow.python.framework.errors_impl.ResourceExhaustedError: 2 root error(s) found. (0) Resource exhausted: OOM when allocating tensor with shape[256,1114] and type float on /job:localhost/replica:0/task:0/device:GPU:0 by allocator GPU_0_bfc [[{{node swap_in_core_decoder_1/core_decoder_0/decoder/while/BasicDecoderStep/decoder/multi_rnn_cell/cell_0/lstm_cell/LSTMBlockCell_13_0}}]] Hint: If you want to see a list of allocated tensors when OOM happens, add report_tensor_allocations_upon_oom to RunOptions for current allocation info.

     [[add/_2901]]

Hint: If you want to see a list of allocated tensors when OOM happens, add report_tensor_allocations_upon_oom to RunOptions for current allocation info.

(1) Resource exhausted: OOM when allocating tensor with shape[256,1114] and type float on /job:localhost/replica:0/task:0/device:GPU:0 by allocator GPU_0_bfc [[{{node swap_in_core_decoder_1/core_decoder_0/decoder/while/BasicDecoderStep/decoder/multi_rnn_cell/cell_0/lstm_cell/LSTMBlockCell_13_0}}]] Hint: If you want to see a list of allocated tensors when OOM happens, add report_tensor_allocations_upon_oom to RunOptions for current allocation info.

0 successful operations. 0 derived errors ignored.

Hello, I'm doing manipulations on midi files using Magenta. I convert the midi file to NoteSequence using

import magenta.music as mm
mm.midi_file_to_sequence_proto(midi)

The problem is that if I have unused midi channels (lets say 5 and 6) and used midi channels (lets say 7 and 8), the conversion put the used channels in place of empty ones, so I loose the instrument true setup, and when converting it back to midi and playing it with midi player, the instruments are mixed up. Any suggestion how to avoid this wrong conversion (only happens when there are some unused midi channels in the midi file). Thanks!

Hi,

I'm trying to generate a basic melody using the code shown in the tutorial, and I've got 2 questions. Note I'm a beginner, hence my questions may be really simple.

1. When I type in Python: BUNDLE_PATH=<E:\Storage\Documents\FEI_AI\magenta\BUNDLES> It returns Syntax Error,pointing at '<'

So I used: BUNDLE_PATH='E:\Storage\Documents\FEI_AI\magenta\BUNDLES' It proceeds without error. But is it correct? Same goes with the next line involving CONFIG.

2. After typing: melody_rnn_generate \ I typed: --config=${CONFIG} It returns Invalid Syntax, and points at '$'.

Need advice, thank you!

Hi, by two days the GANSynth's Colab Demo give me an error on Environment Setup:

Load model from /content/gansynth/acoustic_only/stage_00012/./model.ckpt-11000000
---------------------------------------------------------------------------
KeyError                                  Traceback (most recent call last)
<ipython-input-1-bad1649e76e0> in <module>()
     65 tf.reset_default_graph()
     66 flags = lib_flags.Flags({'batch_size_schedule': [BATCH_SIZE]})
---> 67 model = lib_model.Model.load_from_path(CKPT_DIR, flags)
     68 
     69 # Helper functions

4 frames
/usr/local/lib/python3.6/dist-packages/magenta/models/gansynth/lib/model.py in load_from_path(cls, path, flags)
    175     batch_size = flags.get('eval_batch_size',
    176                            train_util.get_batch_size(stage_id, **flags))
--> 177     model = cls(stage_id, batch_size, flags)
    178     model.saver.restore(model.sess, ckpt)
    179     return model

/usr/local/lib/python3.6/dist-packages/magenta/models/gansynth/lib/model.py in __init__(self, stage_id, batch_size, config)
    191     """
    192     data_helper = data_helpers.registry[config['data_type']](config)
--> 193     real_images, real_one_hot_labels = data_helper.provide_data(batch_size)
    194 
    195     # gen_one_hot_labels = real_one_hot_labels

/usr/local/lib/python3.6/dist-packages/magenta/models/gansynth/lib/data_helpers.py in provide_data(self, batch_size)
     64     with tf.name_scope('inputs'):
     65       with tf.device('/cpu:0'):
---> 66         dataset = self.dataset.provide_dataset()
     67         dataset = dataset.shuffle(buffer_size=1000)
     68         dataset = dataset.map(self._map_fn, num_parallel_calls=4)

/usr/local/lib/python3.6/dist-packages/magenta/models/gansynth/lib/datasets.py in provide_dataset(self)
    116       return wave, one_hot_label, label, example['instrument']['source']
    117 
--> 118     dataset = self._get_dataset_from_tfds()
    119     dataset = dataset.map(
    120         _parse_nsynth, num_parallel_calls=tf.data.experimental.AUTOTUNE)

/usr/local/lib/python3.6/dist-packages/magenta/models/gansynth/lib/datasets.py in _get_dataset_from_tfds(self)
     65       dataset = tfds.load(
     66           'nsynth/gansynth_subset:2.3.*',
---> 67           data_dir=self._config['tfds_data_dir'],
     68           split=tfds.Split.TRAIN,
     69           download=False)

KeyError: 'tfds_data_dir'

How can I resolve?

I'm trying to follow the instructions for arbitrary style transfer here, and it says that I need to download a pretrained model from this link. However, what I get is a tar.gz file and whenever I extract it, I only get the following files:

$ tar xvzf arbitrary_style_transfer.tar.gz
$ ls arbitrary_style_transfer
model.ckpt-data-00000-of-000001 model.ckpt.index model.ckpt.meta

Which one here should I use?

Hi, I'm following the instructions on https://github.com/tensorflow/magenta for setting up a development environment.

I've installed Bazel aok. I've installed Anaconda aok. I've created a tensorflow env and installed it.

But when running bazel test //magenta/... I get

ERROR: E:/music/magenta/magenta/magenta/models/arbitrary_image_stylization/BUILD:48:1: PythonZipper magenta/models/arbitrary_image_stylization/arbitrary_image_stylization_with_weights.zip failed (Exit -1)

INFO: Elapsed time: 1.734s, Critical Path: 0.15s INFO: 0 processes. FAILED: Build did NOT complete successfully //magenta/common:beam_search_test NO STATUS //magenta/common:concurrency_test NO STATUS //magenta/common:nade_test NO STATUS etc etc

Any clues?

By following the tutorial, when converting midi files to note sequences, it encounters following error

ERROR:tensorflow:Midi decoding error <type 'exceptions.TypeError'>: Bad header in MIDI file

when running

bazel run //magenta/scripts:convert_midi_dir_to_note_sequences -- \
--midi_dir=$MIDI_DIRECTORY \
--output_file=$SEQUENCES_TFRECORD \
--recursive

No tfrecord file generated.

I've tried many different midi files from midiworld website, the same error remains. So I guess it might not be the problem with midi files.

Environment: Bazel 0.30 TensorFlow 0.9 under virtual environment

Greetings!

I'm starting recently to learn about magenta, in particular I'm following a book by packt and I'm quite excited. I am, however, very confused about the status of magenta.

Let me elaborate on the confusion:

The README on this repository says that it's inactive, and I should check the website for active stuff. The website (python section) says that magenta is active, and links to github.com/tensorflow/magenta which redirects to this repository, which says that it's inactive. Getting vague Liar's Paradox vibes.

This repository seems to be the upstream for whatever I get with pip install magenta, and it is still getting commits - which seems to imply that magenta as a python module is alive, as the website says. However in issue 2003 you say that magenta is developed, just not this repository.

So what does this all mean?

Is magenta as a python module alive or dead? It currently does not work on my main computer, is this something that is likely to change at some point? Does it make sense to create issues like I did?

If the python module is still being developed but this repository is inactive, which is the active one?

If the python module is not being developed but something is taking its place (please don't say javascript ❤️ ), what is the new project that i can use to train an AI with my music and have it make new stuff from it? And is there a book that I could buy to learn to do it?

Thanks for all your work!

I'm trying to convert midi to a note sequence, but it is not working. I've installed magenta 2.1.4 and tensorflow 2.11.0

convert_dir_to_note_sequences \
--input_dir= /Users/My_name/downloads/classic \
--output_file=/Users/My_name/downloads/classic_out/noteseq/notesequences.tfrecord \
--recursive

I try to run the code and I get this error:

File "/Users/My_name/magenta-env/lib/python3.8/site-packages/resampy/interpn.py", line 73, in <module>
    @guvectorize(
TypeError: guvectorize() missing 1 required positional argument: 'signature'

Magenta has the following problematic dependencies:

• numpy == 1.21.6

This version of numpy gives runtime errors for reasons I'm not completely sure of, such as:

RuntimeError: module compiled against API version 0xf but this version of numpy is 0xe

This runtime problem disappears when numpy is upgraded, but naturally others arise.

• numba == 0.49.1

This is the worst offender as it depends on an old version of llvmlite, which in turn depends on an old version of llvm (8.x) that does not support the newer Apple targets. This must be upgraded to a much newer, possibly the newest release, as even llvm 9.x is problematic. llvm 11.x must be supported in order to run on this platform.

• tensorflow == 2.9.1

This is tensorflow-macos on Apple silicon, but I'm not sure whether you can add dependency alternatives to include it. A minor problem anyway, as it can be overridden by editing setup.py and installing locally.

numba and numpy need upgrading though, until then there will be no way to run Magenta on new macs.

Thanks!

I get error messages for building wheels in various moduls (llvmlite, numba, python-rtmidi). I'm using Python 3.9, already tried to reinstall everything, installed other package versions. It just doesn't work. Can you help me please?

(I removed some not so important code to not exceed the max. char. length here)

Building wheels for collected packages: numba, python-rtmidi, llvmlite
  Building wheel for numba (setup.py) ... error
  error: subprocess-exited-with-error

  × python setup.py bdist_wheel did not run successfully.
  │ exit code: 1
  ╰─> [826 lines of output]
      TBB not found
      OpenMP disabled
      running bdist_wheel
      running build
      got version from file C:\Users\Dell\AppData\Local\Temp\pip-install-s_n9pbef\numba_9c8741e8713142ffb7ba6d62968ea471\numba/_version.py {'version': '0.49.1', 'full': 'd2cac8597ad2aa4074147d9c7595f5b5e9919901'}
      running build_py
      creating build
    
      running build_ext
      No module named 'numpy.distutils._msvccompiler' in numpy.distutils; trying from distutils
      building 'numba._dynfunc' extension
      error: Microsoft Visual C++ 14.0 or greater is required. Get it with "Microsoft C++ Build Tools": https://visualstudio.microsoft.com/visual-cpp-build-tools/
      [end of output]

  note: This error originates from a subprocess, and is likely not a problem with pip.
  ERROR: Failed building wheel for numba
  Running setup.py clean for numba
  Building wheel for python-rtmidi (setup.py) ... error
  error: subprocess-exited-with-error

  × python setup.py bdist_wheel did not run successfully.
  │ exit code: 1
  ╰─> [13 lines of output]
      running bdist_wheel
      running build
      running build_py
      creating build
      creating build\lib.win-amd64-cpython-39
      creating build\lib.win-amd64-cpython-39\rtmidi
      copying rtmidi\midiconstants.py -> build\lib.win-amd64-cpython-39\rtmidi
      copying rtmidi\midiutil.py -> build\lib.win-amd64-cpython-39\rtmidi
      copying rtmidi\release.py -> build\lib.win-amd64-cpython-39\rtmidi
      copying rtmidi\__init__.py -> build\lib.win-amd64-cpython-39\rtmidi
      running build_ext
      building 'rtmidi._rtmidi' extension
      error: Microsoft Visual C++ 14.0 or greater is required. Get it with "Microsoft C++ Build Tools": https://visualstudio.microsoft.com/visual-cpp-build-tools/
      [end of output]

  note: This error originates from a subprocess, and is likely not a problem with pip.
  ERROR: Failed building wheel for python-rtmidi
  Running setup.py clean for python-rtmidi
  Building wheel for llvmlite (setup.py) ... error
  error: subprocess-exited-with-error

  × python setup.py bdist_wheel did not run successfully.
  │ exit code: 1
  ╰─> [40 lines of output]
      running bdist_wheel
      C:\Users\Dell\anaconda3\python.exe C:\Users\Dell\AppData\Local\Temp\pip-install-s_n9pbef\llvmlite_4ce3f81337d447c18fde83ea427f20d5\ffi\build.py
      -- Selecting Windows SDK version  to target Windows 10.0.19044.
      CMake Error at CMakeLists.txt:3 (project):
        Failed to run MSBuild command:

          MSBuild.exe

        to get the value of VCTargetsPath:

          Das System kann die angegebene Datei nicht finden



      -- Configuring incomplete, errors occurred!
      See also "C:/Users/Dell/AppData/Local/Temp/tmp_9vh8mbn/CMakeFiles/CMakeOutput.log".
      CMake Error at CMakeLists.txt:3 (project):
        Generator

          Visual Studio 15 2017 Win64

        could not find any instance of Visual Studio.



      -- Configuring incomplete, errors occurred!
      See also "C:/Users/Dell/AppData/Local/Temp/tmp5dfq2vh1/CMakeFiles/CMakeOutput.log".
      Trying generator 'Visual Studio 14 2015 Win64'
      Trying generator 'Visual Studio 15 2017 Win64'
      Traceback (most recent call last):
        File "C:\Users\Dell\AppData\Local\Temp\pip-install-s_n9pbef\llvmlite_4ce3f81337d447c18fde83ea427f20d5\ffi\build.py", line 192, in <module>
          main()
        File "C:\Users\Dell\AppData\Local\Temp\pip-install-s_n9pbef\llvmlite_4ce3f81337d447c18fde83ea427f20d5\ffi\build.py", line 180, in main
          main_win32()
        File "C:\Users\Dell\AppData\Local\Temp\pip-install-s_n9pbef\llvmlite_4ce3f81337d447c18fde83ea427f20d5\ffi\build.py", line 89, in main_win32
          generator = find_win32_generator()
        File "C:\Users\Dell\AppData\Local\Temp\pip-install-s_n9pbef\llvmlite_4ce3f81337d447c18fde83ea427f20d5\ffi\build.py", line 85, in find_win32_generator
          raise RuntimeError("No compatible cmake generator installed on this machine")
      RuntimeError: No compatible cmake generator installed on this machine
      error: command 'C:\\Users\\Dell\\anaconda3\\python.exe' failed with exit code 1
      [end of output]

  note: This error originates from a subprocess, and is likely not a problem with pip.
  ERROR: Failed building wheel for llvmlite
  Running setup.py clean for llvmlite
Failed to build numba python-rtmidi llvmlite
Installing collected packages: python-rtmidi, pygtrie, mido, llvmlite, keras, flatbuffers, tensorflow-estimator, sox, scipy, protobuf, promise, numba, keras-preprocessing, importlib_resources, imageio, etils, absl-py, tf-slim, tensorflow-probability, sk-video, resampy, mir-eval, matplotlib, googleapis-common-protos, dm-sonnet, tensorflow-metadata, librosa, tensorflow-datasets, tensorboard, note-seq, tensorflow, magenta
  Attempting uninstall: python-rtmidi
    Found existing installation: python-rtmidi 1.4.9
    Uninstalling python-rtmidi-1.4.9:
      Successfully uninstalled python-rtmidi-1.4.9
  Running setup.py install for python-rtmidi ... error
  error: subprocess-exited-with-error

  × Running setup.py install for python-rtmidi did not run successfully.
  │ exit code: 1
  ╰─> [15 lines of output]
      running install
      C:\Users\Dell\anaconda3\lib\site-packages\setuptools\command\install.py:34: SetuptoolsDeprecationWarning: setup.py install is deprecated. Use build and pip and other standards-based tools.
        warnings.warn(
      running build
      running build_py
      creating build
      creating build\lib.win-amd64-cpython-39
      creating build\lib.win-amd64-cpython-39\rtmidi
      copying rtmidi\midiconstants.py -> build\lib.win-amd64-cpython-39\rtmidi
      copying rtmidi\midiutil.py -> build\lib.win-amd64-cpython-39\rtmidi
      copying rtmidi\release.py -> build\lib.win-amd64-cpython-39\rtmidi
      copying rtmidi\__init__.py -> build\lib.win-amd64-cpython-39\rtmidi
      running build_ext
      building 'rtmidi._rtmidi' extension
      error: Microsoft Visual C++ 14.0 or greater is required. Get it with "Microsoft C++ Build Tools": https://visualstudio.microsoft.com/visual-cpp-build-tools/
      [end of output]

  note: This error originates from a subprocess, and is likely not a problem with pip.
  Rolling back uninstall of python-rtmidi
  Moving to c:\users\dell\anaconda3\lib\site-packages\python_rtmidi-1.4.9.dist-info\
   from C:\Users\Dell\anaconda3\Lib\site-packages\~ython_rtmidi-1.4.9.dist-info
  Moving to c:\users\dell\anaconda3\lib\site-packages\rtmidi\
   from C:\Users\Dell\anaconda3\Lib\site-packages\~tmidi
error: legacy-install-failure

× Encountered error while trying to install package.
╰─> python-rtmidi

note: This is an issue with the package mentioned above, not pip.
hint: See above for output from the failure.

I am using mac venutra.

Using the conda install I get this error during installation: After this I attempted the manual install which gave me this error:

I don't believe downgrading to python 3.6 would work, as in the documentation it is said that verisons >= 3.5 work (correct me if i am wrong!)

Cheers