GitHub - ebattenberg/crbm-drum-patterns: Python code to do Conditional Restricted Boltzmann machine training applied to drum pattern generation.

Branches Tags

Name		Name	Last commit message	Last commit date
Latest commit History 6 Commits
data		data
.gitignore		.gitignore
LICENSE		LICENSE
README		README
crbm.py		crbm.py
crbm_gen.py		crbm_gen.py
crbm_train.py		crbm_train.py
data_helper.py		data_helper.py
gnumpy.py		gnumpy.py
gnumpy_example.py		gnumpy_example.py
midi_tools.py		midi_tools.py
npmat.py		npmat.py

Repository files navigation

This package is meant to be used for training a Conditional Restricted Boltzmann machine
to generate drum patterns.  

Example training and generation scripts are: crbm_train.py and crbm_gen.py


Requirements:

    * CUDA
    * CUDAMat
    * python-midi 

Included:

    * Gnumpy: http://www.cs.toronto.edu/~tijmen/gnumpy.html
    * npmat: http://www.cs.toronto.edu/~ilya/npmat.py


Obtaining and Installing the above:

* CUDA can be downloaded from: https://developer.nvidia.com/cuda-downloads
    please note that a CUDA-capable Nvidia GPU is required
* CUDAMat
    checkout:
        hg clone https://code.google.com/p/cudamat/

    install:
        cd cudamat
        # for mac: add '-m 64' to nvcc lines in Makefile
        make
        export PYTHONPATH=$PYTHONPATH:`pwd`

    test:
        python test_cudamat.py
* python-midi
    git clone https://github.com/ebattenberg/python-midi.git
    cd python-midi
    python setup.py install