Magenta Docker Compose

This is a docker-compose implementation of Magenta Based on xychelsea's repo.
Tested with MacOS Ventura 13.5.1(M1), Docker version 4.22.1

How to Run

Clone this repo

git clone git@github.com:mj-life-is-once/magenta-docker.git

Navigate to the project root directory
Run docker-compose

docker-compose up -d

Run generation commands in docker. There are several options to run the command.

With docker exec

sudo docker exec -it magenta /bin/bash

# Run example command
melody_rnn_generate \
  --config=lookback_rnn \
  --bundle_file=/home/anaconda/magenta/models/lookback_rnn.mag \
  --output_dir=/home/anaconda/workspace/generated/lookback_rnn\
  --num_outputs=10 \
  --num_steps=128 \
  --primer_melody="[60]"

User Portainer

Open localhost:9000
Run bash in magenta container
Run the example command

  melody_rnn_generate \
    --config=lookback_rnn \
    --bundle_file=/home/anaconda/magenta/models/lookback_rnn.mag \
    --output_dir=/home/anaconda/workspace/generated/lookback_rnn\
    --num_outputs=10 \
    --num_steps=128 \
    --primer_melody="[60]"

Alternatively, you can directly run command through docker CLI

docker exec melody_rnn_generate \
    --config=lookback_rnn \
    --bundle_file=/home/anaconda/magenta/models/lookback_rnn.mag \
    --output_dir=/home/anaconda/workspace/generated/lookback_rnn\
    --num_outputs=10 \
    --num_steps=128 \
    --primer_melody="[60]"

List of example commands

1. melodyRNNGenerate

melody_rnn_generate \
 --config=lookback_rnn \
 --bundle_file=/home/anaconda/magenta/models/lookback_rnn.mag \
 --output_dir=/home/anaconda/workspace/generated/lookback_rnn\
 --num_outputs=10 \
 --num_steps=128 \
 --primer_melody="[60]"

2. polyphonyRNN

polyphony_rnn_generate \
--bundle_file=/home/anaconda/workspace/models/polyphony_rnn.mag \
--output_dir=/home/anaconda/workspace/generated/polyphony_rnn \
--num_outputs=1 \
--num_steps=128 \
--primer_melody="[60, -2, -2, -2, 60, -2, -2, -2, "\
"67, -2, -2, -2, 67, -2, -2, -2, 69, -2, -2, -2, "\
"69, -2, -2, -2, 67, -2, -2, -2, -2, -2, -2, -2]" \
--condition_on_primer=false \
--inject_primer_during_generation=true

Deployment in the GCP server

Find more detailed instruction in this README.md file and my blog post