Shut Down, no further development

polymuse-future

Making the music real In development phase, once completed repo will change name to polymuse

Features

Need to discuss .......

Table of Contents

• Overview
• Components
• Links
• Installing ..
• Training(Note model)
• Loading
• Player

Overview

This is BE project aiming to generate the musical patterns from the midi file that are the audibes to ears

Components

Will be added soon

Links

This to ...

Installing ...

This is pre complete installation, package may not run as expected

$ pip install polymuse

OR

$ pip install polymuse-future

install the polymuse-future recommended

Train

Only NOTE training available

Note Training

from polymuse import train

F = dataset_path # It should be absolute PATH(recomended) where midi file are

train.train_gpu(F, maxx = 5) #Only if GPU is available, It uses CuDNNLstm version which performs operation on GPU
train.train(F, maxx = 5) #if GPU version do not works 

@dataset_path : It should be absolute PATH(recomended) where midi file are @maxx : It is parameters that specifies maximum no of files used to training in case there are large no of files in dataset_path given

This snapshot will train the model on dataset given, 3 files will generated and stored in following dir strucure : .h5_models :...chorus
:....... stateless
:...........wlvv.h5
:...drum
:......stateless
:...........vyvh.h5
:...lead
:......stateless
:.......... vyvh.h5

Load Pretrain Models

Below code snapshot downloads the default model, and make above directory structure in current working directory

from polymuse import loader
loader.load(mname = 'default')

Load sample midis

Below code snapshot downloads the default midi and download in current directory

from polymuse import loader
loader.load_midi()

Testing midi : Single Song Trainig

Song : Candi Staton := Young Hearts Run Free

Note Player

Before using the player please train the models on dataset or load pre trained models

from polymuse import player
# Before this please make sure the h5_models are loaded locally

midi_file = "F:\\rushikesh\\project\\dataset\\lakh_dataset\\Kenny G" # directory where midi file will
midi_file = dutils.get_all_files(F)[0] # Midi file must be of atleast 3 tracks

player.play_3_track_no_time(midi_file, midi_fname = 'midi00')

The above will store midi file in current directory with file name midi00XXX

""" If you need to use plot option to see the model, need to install conda, pydot , pip graphviz, conda graphviz """