aws-samples / aws-deepcomposer-samples

AWS DeepComposer

Welcome to AWS DeepComposer. This repository includes:

re:Invent 2019 workshop labs

Training a custom GAN model

Training an autoregressive convolutional neural network(AR-CNN) model

Training a Transformer-XL model

License

Licensing varies between subdirectories. Check each subdirectory for its corresponding license.

aws-deepcomposer-samples's Issues

np.float32 required?

data = np.asarray(data, dtype=np.float32) # {-1, 1}

Is it possible to use np.int8 or something else here? The range of values is only -1/1.

Thank you very very much :)

Hey guys,

I simply wanted to profusely thank you for GGA-MG (AR-CNN) code/repo, and especially for your absolutely fantastic and detailed Jupiter Notebook.

I was able to fix it up and make it work properly and as you have described in your paper so I was extremely pleased and satisfied with the results and my experience :) Thank you again :)

I wanted to invite you to my repo on my GitHub because I would like to submit my try with Amazon Deep Composer for people to enjoy and listen to, so I hope you do not mind links :)

https://github.com/asigalov61/Amazon-Deep-Composer

And here is also my SoundCloud account where I show-off all of my AI-assisted/AI-co-composed Music as Music AI is my specialty :)

https://soundcloud.com/aleksandr-sigalov-61/

Most sincerely,

Alex

[HELP] Deep Composer for compositions longer than 8 bars

Hey guys,

This is a question to those who created Deep Composer Music AI implementation/model...

Can you please help to make Deep Composer work with MIDIs longer than 8 bars? Because right now Deep Composer is really impractical IMHO compared to MuseNet or AIVA so if you serious about supporting your work, it will have to support more than 8 bars eventually.

Deep Composer is an amazing implementation of Music AI and it would be a shame if this great idea would be useless.

I tried to adjust model (hyper)parameters, inference parameters and MIDI pre-process parameters but it did not allow me to do so for some reason. So if you can help in any way, I would be very grateful 👍

Here is the link to the completed version of Google Colab Notebook for Amazon Deep Composer that I made :) I love it and so will you :)

https://github.com/asigalov61/Amazon-Deep-Composer/blob/master/Amazon-Deep-Composer.ipynb

Thank you

need an update in Lab 2/utils/midi_utils.py

Hi guys,

Thanks a lot for this amazing work!
I noticed that there's a new release of pianoroll on Nov.4. In this new release, there's change to the argument name in the function pypianoroll.Multitrack(), specifically, "beat_resolution" has been changed to "resolution".
So correspondingly, the code Lab 2/utils/midi_utils.py line 39 should be updated.
Could you please make the change? Thank you.

requirements.sh is not working correctly

FileNotFoundError: [Errno 2] No such file or directory: './requirements.sh': './requirements.sh'

Link given is not working

The link provided is not working,also deepcomposer is not showing up on aws console

`RuntimeError: No ffmpeg exe could be found` occured

I also faced #8 issue and solved it by executing the command below

!bash ./requirements.sh

But RuntimeError: No ffmpeg exe could be found occured.

---------------------------------------------------------------------------
RuntimeError                              Traceback (most recent call last)
<ipython-input-2-fb7c3305f190> in <module>()
      4 from PIL import Image
      5 import logging
----> 6 import pypianoroll
      7 import scipy.stats
      8 import pickle

~/anaconda3/envs/python3/lib/python3.6/site-packages/pypianoroll/__init__.py in <module>()
     16 
     17 import pypianoroll.metrics
---> 18 from pypianoroll.multitrack import Multitrack
     19 from pypianoroll.track import Track
     20 from pypianoroll.utilities import (

~/anaconda3/envs/python3/lib/python3.6/site-packages/pypianoroll/multitrack.py in <module>()
     11 from six import string_types
     12 
---> 13 from pypianoroll.track import Track
     14 from pypianoroll.visualization import plot_multitrack
     15 

~/anaconda3/envs/python3/lib/python3.6/site-packages/pypianoroll/track.py in <module>()
      7 from six import string_types
      8 
----> 9 from pypianoroll.visualization import plot_track
     10 
     11 

~/anaconda3/envs/python3/lib/python3.6/site-packages/pypianoroll/visualization.py in <module>()
     15 
     16 try:
---> 17     from moviepy.editor import VideoClip
     18     from moviepy.video.io.bindings import mplfig_to_npimage
     19 

~/anaconda3/envs/python3/lib/python3.6/site-packages/moviepy/editor.py in <module>()
     34 
     35 # Clips
---> 36 from .video.io.VideoFileClip import VideoFileClip
     37 from .video.io.ImageSequenceClip import ImageSequenceClip
     38 from .video.io.downloader import download_webfile

~/anaconda3/envs/python3/lib/python3.6/site-packages/moviepy/video/io/VideoFileClip.py in <module>()
      1 import os
      2 
----> 3 from moviepy.audio.io.AudioFileClip import AudioFileClip
      4 from moviepy.Clip import Clip
      5 from moviepy.video.io.ffmpeg_reader import FFMPEG_VideoReader

~/anaconda3/envs/python3/lib/python3.6/site-packages/moviepy/audio/io/AudioFileClip.py in <module>()
      1 from __future__ import division
      2 
----> 3 from moviepy.audio.AudioClip import AudioClip
      4 from moviepy.audio.io.readers import FFMPEG_AudioReader
      5 

~/anaconda3/envs/python3/lib/python3.6/site-packages/moviepy/audio/AudioClip.py in <module>()
      5 from tqdm import tqdm
      6 
----> 7 from moviepy.audio.io.ffmpeg_audiowriter import ffmpeg_audiowrite
      8 from moviepy.Clip import Clip
      9 from moviepy.decorators import requires_duration

~/anaconda3/envs/python3/lib/python3.6/site-packages/moviepy/audio/io/ffmpeg_audiowriter.py in <module>()
      5 
      6 from moviepy.compat import DEVNULL
----> 7 from moviepy.config import get_setting
      8 from moviepy.decorators import requires_duration
      9 

~/anaconda3/envs/python3/lib/python3.6/site-packages/moviepy/config.py in <module>()
     34 if FFMPEG_BINARY=='ffmpeg-imageio':
     35     from imageio.plugins.ffmpeg import get_exe
---> 36     FFMPEG_BINARY = get_exe()
     37 
     38 elif FFMPEG_BINARY=='auto-detect':

~/anaconda3/envs/python3/lib/python3.6/site-packages/imageio/plugins/ffmpeg.py in get_exe()
     47     import imageio_ffmpeg
     48 
---> 49     return imageio_ffmpeg.get_ffmpeg_exe()
     50 
     51 

~/anaconda3/envs/python3/lib/python3.6/site-packages/imageio_ffmpeg/_utils.py in get_ffmpeg_exe()
     48     # Nothing was found
     49     raise RuntimeError(
---> 50         "No ffmpeg exe could be found. Install ffmpeg on your system, "
     51         "or set the IMAGEIO_FFMPEG_EXE environment variable."
     52     )

RuntimeError: No ffmpeg exe could be found. Install ffmpeg on your system, or set the IMAGEIO_FFMPEG_EXE environment variable.

I think the command such as !conda install ffmpeg -y should be included in requrements.sh

Make own dataset

Hi, thanks for this great work.

I am trying to make my own dataset, but I am getting an error when I try to load them into the model. how can i fix this? My code below.
Thanks.

import numpy as np
import os
from pypianoroll import Multitrack

dir = './my_dataset'

for midi in os.listdir(dir):
    midi_data = Multitrack(os.path.join(dir, midi))
    tracks = [track.pianoroll for track in midi_data.tracks]
    sample = np.stack(tracks, axis=-1)
    
print(sample.shape)
np.save('./dataset/new_train.npy',sample)

ModuleNotFoundError: No module named 'pypianoroll'

hi there I was trying to do Lab 2 – Train a custom GAN model and when I ran the IMPORTS cell I get this error and I don't know what about.