/melodyt5

MelodyT5: A Unified Score-to-Score Transformer for Symbolic Music Processing [ISMIR 2024]

Primary LanguagePythonMIT LicenseMIT

MelodyT5: A Unified Score-to-Score Transformer for Symbolic Music Processing [ISMIR 2024]

This repository contains the code for the MelodyT5 model as described in the paper MelodyT5: A Unified Score-to-Score Transformer for Symbolic Music Processing.

Model Description

MelodyT5 is an unified framework for symbolic music processing, using an encoder-decoder architecture to handle multiple melody-centric tasks, such as generation, harmonization, and segmentation, by treating them as score-to-score transformations. Pre-trained on MelodyHub, a large dataset of melodies in ABC notation, it demonstrates the effectiveness of multi-task transfer learning in symbolic music processing.

We provide the weights of MelodyT5 on Hugging Face, which are based on pre-training with over one million task instances encompassing seven melody-centric tasks.

ABC Notation

ABC notation is an ASCII-based plain text musical notation system that is commonly used for transcribing traditional music and sharing sheet music online. It provides a simple and concise way to represent musical elements such as notes, rhythms, chords, and more.

For those looking to interact with ABC notation in various ways, there are several tools available:

  1. Online ABC Player: This web-based tool allows you to input ABC notation and hear the corresponding audio playback. By pasting your ABC code into the player, you can instantly listen to the tune as it would sound when played.

  2. ABC Sheet Music Editor - EasyABC: EasyABC is a user-friendly software application designed for creating, editing, and formatting ABC notation. Its graphical interface enables you to input your ABC code, preview the sheet music, and make adjustments as necessary.

To learn more about ABC notaton, please see ABC Examples and ABC Strandard.

Installation

To set up the MelodyT5 environment and install the necessary dependencies, follow these steps:

  1. Create and Activate Conda Environment

    conda create --name melodyt5 python=3.7.9
conda activate melodyt5

  2. Install Dependencies

    pip install -r requirements.txt

  3. Install Pytorch

    pip install torch==1.13.1+cu116 torchvision==0.14.1+cu116 torchaudio==0.13.1 --extra-index-url https://download.pytorch.org/whl/cu116

  4. Create Random Weights (Optional)

    If SHARE_WEIGHTS = True in config.py, run random_model.py to generate random weights for initializing the patch-level encoder and decoder with shared weights.

  5. Download Pre-trained MelodyT5 Weights (Optional)

    For those interested in starting with pre-trained models, MelodyT5 weights are available on Hugging Face.

Usage

  • config.py: Configuration settings for training and inference.
  • inference.py: Perform inference tasks (e.g., generation and harmonization) using trained models.
  • prompt.txt: Text file containing the prompt to specify a task and its input for the model.
  • random_model.py: Use this script to initialize the model with random weights when SHARE_WEIGHTS = TRUE in config.py (default setting).
  • train.py: Script for training the MelodyT5 model.
  • utils.py: Utility functions supporting model operations and data processing.

Setting Up Inference Parameters

Before running the inference script, you can configure the following parameters in config.py or directly via command-line arguments:

  • -num_tunes: Number of independently computed returned tunes (default: 3)
  • -max_patch: Maximum length in patches of each tune (default: 128)
  • -top_p: Cumulative probability threshold for nucleus sampling (default: 0.8)
  • -top_k: Number of highest probability tokens to keep for sampling (default: 8)
  • -temperature: Temperature of the sampling operation (default: 2.6)
  • -seed: Seed for random state (default: None)
  • -show_control_code: Whether to show control codes (default: True)

These parameters control how the model generates melodies based on the input provided in prompt.txt.

Running Inference

To perform inference tasks using MelodyT5, follow these steps:

  1. Prepare Your Prompt

    • Edit prompt.txt to specify the task and input for the model.

  2. Execute Inference

    • Run the following command to execute the inference script: 
      python inference.py -num_tunes 3 -max_patch 128 -top_p 0.8 -top_k 8 -temperature 2.6 -seed <seed_value> -show_control_code True
      Replace <seed_value> with your chosen seed value or leave it as None for a random seed.
    • The script will generate melodies based on the prompt specified in prompt.txt using the configured parameters and save the results in the output_tunes folder.

How to Use

Follow these steps to utilize MelodyT5 for symbolic music processing:

  1. Prepare XML Data

    Ensure your data is in single-track XML format, which can be .xml, .musicxml, or .mxl files.

  2. Convert to Standard Format

    • Place the folder containing your XML data in the xml2abc directory.

    • Navigate to the xml2abc directory.

    • Run the following command to convert your XML files into standard ABC notation format:

      python batch_xml2abc.py path_to_xml_directory

    • Replace path_to_xml_directory with the directory containing your XML files.

  3. Design Input-Output Pairs

    After conversion, design input-output pairs based on your specific task requirements. You can refer to the data formats used in MelodyHub for guidance.

  4. Configure Your Model

    Adjust model hyperparameters, training parameters, and file paths in the config.py file as you see fit.

  5. Train the Model

    Run the train.py script to train MelodyT5. Use the following command, adjusting for your specific setup:

    python -m torch.distributed.launch --nproc_per_node=8 --use_env train.py

    This command utilizes distributed training across multiple GPUs (modify --nproc_per_node as needed).

  6. Run Inference

    To perform inference tasks such as melody generation or harmonization, execute inference.py. The script reads prompts from prompt.txt to specify the task and input for the model. Customize prompts in prompt.txt to define different tasks and inputs for MelodyT5. Refer to the examples below for guidance on setting up prompts.

    %%input
%%variation
L:1/8
M:6/8
K:D
|: AFD DFA | Add B2 A | ABA F3 | GFG EFG | AFD DFA | Add B2 A | ABA F2 E | FDD D3 :: fdd ede |
fdd d2 g | fdd def | gfg e2 g | fed B2 A | AdF A3 | ABA F2 E | FDD D3 :|
%%output
E:8
L:1/8
M:6/8
K:D
|: B |

    Ensure the input (encoder input) is complete, while the output (decoder input) is optional. If you need the model to continue a given output, use %%input and %%output to mark the beginning of each section. Additionally, the output must not contain incomplete bars.

Inference Examples

Below are the MelodyT5 results on seven MelodyHub tasks, using random samples from the validation set. Three independent outputs were generated without cherry-picking. Each X:0 output corresponds to the original input for that task and is not generated by the model, while X:1, X:2, and X:3 are generated outputs.

To view the musical scores and listen to the tunes, you can use online ABC notation platforms such as Online ABC Player or the ABC Sheet Music Editor - EasyABC. Simply copy and paste the ABC notation into these tools to see the sheet music and hear the audio playback.

  1. Cataloging
%%input
%%cataloging
T:
O:
Y:
L:1/4
Q:1/4=120
M:3/4
K:C
A/0 A/0 B/0 (c/0 B/0) A/0 A/0 ^G/0 A/0 (e/0 f/0) (=g/0 f/0) e/0 | (f e) d | e2 e | (d c) B |
(c B) A | ^G2 e | f2 e | d2 c | B2 A | ^G2 c | (d e) d | (e f) g | e2 e | (e c) c | (d B) B |
(c A) A | ^G2 (A/B/) | (c d) c | c2 B | c2 (c/d/) | (e f) e | e2 d | e2 (e/f/) | g2 g | (g e) e |
f2 e | d2 e | (e d) e | (c d) e | (d e) (c/d/) | B2 c | B2 A | ^G2 A | A2 c | B2 A | ^G2 A |
A2 x |]

%%output
X:0
T:HET GODEN BANKET. Ter Huwfeest van Muzika en Poezy
O:Netherlands
Y:vocal

X:1
T:WILHELLTOF GERSCHANDS
O:Netherlands
Y:vocal


X:2
T:MIEN IS LOVEN, VRIENDAAT OP EEN
O:Netherlands
Y:vocal


X:3
T:Het lied en tijd over
O:Netherlands
Y:vocal
  1. Generation
%%input
%%generation

%%output
X:0
S:2
B:8
E:5
B:8
L:1/8
M:2/2
K:G
|: A3 c BG G2 | BGBd g2 fg | eA A2 BGBd | egdc BG G2 | A3 c BG G2 | BGBd g3 a | bgag egfa |
gedc BG G2 :: bg g2 agef | g3 e dega | bg g2 aged | eaag a2 ga | bg g2 agef | g3 e dega |
bgag egfa | gedc BG G2 :|

X:1
S:2
B:5
E:5
B:9
L:1/8
M:2/2
K:Amin
cd | e3 c d2 cd | e3 d c2 Ac | B3 c A3 A | A4 :: G2 | c3 c c3 c | d3 d e2 fe | f2 f2 e2 d2 | 
g3 a g2 cd | e3 d c2 dc | e3 d c2 Ac | B3 c A3 A | A4 :|

X:2
S:3
B:1
E:0
B:17
E:0
E:5
B:17
L:1/8
M:3/4
K:G
D |:"G" G2 GABc |"G/B" d2 B3 G |"C" E3 G E2 |"G" D3 DEF |"G" G2 GABc |"G/B" d2 B3 G | 
"D" E2 G2 B2 |"D" A4 DD |"G" G2 GABc |"G/B" d2 B3 G |"C" E3 G E2 |"G" D3 DEF |"G" G2 GABc | 
"G/B" d2 B3 G |"D" E2 G2 F2 |1"G" G4 GD :|2"G" G4 DG |:"C" E G3 E2 |"G" D3 E D2 |"G" B3 A G2 | 
"G" B d3 B2 |"C" e c3 e2 |"G" d3 B G2 |"Am" E2 G2 B2 |"D" A4 DG |"C" E G3 E2 |"G" D3 E D2 | 
"G" B3 A G2 |"G" B d3 B2 |"C" e c3 e2 |"G" d3 B G2 |"D" E2 G2 F2 |1"G" G4 DG :|2"G" G4 z2 |]

X:3
S:3
B:9
E:5
B:9
E:5
E:5
B:9
L:1/8
M:4/4
K:D
|: A2 | d2 d2 c2 c2 | B2 A2 F2 A2 | B2 B2 c2 c2 | d2 d2 d2 c2 | d2 d2 c2 c2 | B2 A2 F2 A2 | 
B2 B2 c2 c2 | d6 ::[K:D] A2 | d2 d2 cB c2 | Bc BA G2 FG | A3 B AG FE | D2 DE FG AB | d2 d2 cB c2 | 
Bc BA G2 FG | A3 B AG FE | D6 :: A2 | F2 DF A2 FA | G2 EC E2 A2 | F2 DF A2 FA | G2 EC A,2 A2 | 
F2 DF A2 FA | G2 EC E2 A2 | F2 DF A2 AF | G2 EC D2 :|
  1. Harmonization
%%input
%%harmonization
L:1/4
M:4/4
K:B
 B, | F D/C/ B, F | G3/4A/8B/8 G !fermata!F F |
 G A B A | G/B/A/G/ !fermata!F D |
 G F E D | C2 !fermata!B, :| z | F2 !fermata!D2 |
 F2 !fermata!D2 | D D C C | D D C D |
 E D C2 | !fermata!B,2 B A | G F E D |
 C2 !fermata!B, |]

%%output
X:0
E:5
L:1/4
M:4/4
K:B
"B" B, |"F#/A#" F"B" D/C/"G#m" B,"D#m" F |"G#m7/B" G3/4A/8B/8"C#" G"F#" !fermata!F"B" F |
"E" G"A#dim/D#" A"B/D#" B"F#" A |"G#m7/B" G/B/"A#m/C#"A/G/"F#" !fermata!F"B" D |
"E" G"B/D#" F"C#m7" E"B" D |"F#sus4" C2"B" !fermata!B, :| z |"F#/A#" F2"B" !fermata!D2 |
"F#" F2"B" !fermata!D2 |"B" D"B/D#" D"F#" C"F#" C |"B/D#" D"B/D#" D"F#" C"B#dim/E" D |
"C#m" E"G#m" D"F#7/E" C2 |"B" !fermata!B,2"G#m" B"D#" A |"E" G"D#m/F#" F"Emaj7/G#" E"B" D |
"F#7" C2"B" !fermata!B, |]

X:1
E:5
L:1/4
M:4/4
K:B
"B" B, |"F#/A#" F"B" D/C/"G#m" B,"B/D#" F |"G#m7/B" G3/4A/8B/8"C#" G"F#" !fermata!F"B" F |
 G"F#/A#" A"B/D#" B"F#" A |"G#m7/B" G/B/"A#m/C#"A/G/"F#" !fermata!F"B" D |
 G"B/D#" F"C#m" E"D#m7b5/C#" D"A" |"F#7" C2"B" !fermata!B, :|"F#" z |"F#/A#" F2"B" !fermata!D2 |
"B" F2"B" !fermata!D2 |"B" D"B/D#" D"F#" C"F#" C |"B/D#" D"B/D#" D"F#" C"B#dim/E" D |
 E"G#m" D"F#7/E" C"G#m"2 |"G#m" !fermata!B,2"G#m" B"A#" A |"G#m" G"D#m/F#" F"Emaj7/G#" E"B" D |
"F#7" C2"B" !fermata!B, |]

X:2
E:5
L:1/4
M:4/4
K:B
"B" B, |"F#/A#" F"B" D/C/"G#m" B,"D#m" F |"G#m7/B" G3/4A/8B/8"C#" G"F#" !fermata!F"B" F |
 G"F#" A"B/D#" B"F#" A"B" |"G#m/B" G/B/"A#dim/C#"A/G/"D#" !fermata!F"G#m" D |
"E" G"E/G#" F"A#m7b5/G#" E"B/F#" D |"F#7" C2"B" !fermata!B, :|"B" z |"D#m" F2"G#m" !fermata!D2 |
"D#m" F2"G#m" !fermata!D2 | D"G#m" D"F#/A#" C"C#m" C | D"B/D#" D"B/D#" C"B" D"F#sus4" |
"C#m" E"G#m" D"F#" C2 | !fermata!B,2"B" B"F#" A |"E" G"F#/A#" F"C#m" E"D#m" D |
"F#7" C2"B" !fermata!B, |]

X:3
E:5
L:1/4
M:4/4
K:B
"B" B, |"F#/A#" F"B" D/C/"G#m" B,"B/D#" F |"G#m7/B" G3/4A/8B/8"C#" G"F#" !fermata!F"B" F |
 G"F#" A"B/D#" B"F#" A"B" |"G#m/B" G/B/"A#dim/C#"A/G/"D#" !fermata!F"B" D |
 G"E/G#" F"C#m7" E"B" D"F#" |"F#sus4" C2"B" !fermata!B, :| z |"F#/A#" F2"B" !fermata!D2 |
"F#" F2"B" !fermata!D2 | D"B" D"B/D#" C"B/C#" C | D"B/D#" D"B/D#" C"B" D |
 E"C#m" D"F#sus4" C"D#7"2 |"G#m" !fermata!B,2"G#m" B"A#" A |"E" G"D#m/F#" F"E/G#" E"B" D |
"F#7" C2"B" !fermata!B, |]
  1. Melodization
%%input
%%melodization
L:1/8
M:6/8
K:G
|: z |"G" z6 | z6 |"Am" z6 |"C" z3"D7" z3 |"G" z6 | z6 |"Am" z3"D7" z3 |"G" z4 z :: z |"C" z6 |
z6 |"Bm" z6 |"Em" z6 |"C" z3"D7" z3 |"Em" z3"Am" z3 |"D7" z6 |"G" z4 :|

%%output
X:0
E:5
L:1/8
M:6/8
K:G
|: B/A/ |"G" GDE G2 A | Bgf gdB |"Am" ABc BGA |"C" BcA"D7" BGE |"G" GDE G2 A | Bgf gdB |
"Am" ABc"D7" BcA |"G" BGG G2 :: B/d/ |"C" e2 e e2 e | egf edB |"Bm" d2 d d2 d |"Em" dge dBG |
"C" c2 d"D7" e2 f |"Em" gdB"Am" A2 d |"D7" BGA BcA |"G" BGG G :|

X:1
E:5
L:1/8
M:6/8
K:G
|: d |"G" GBG GBG | BGG G2 B |"Am" cec ABc |"C" ecA"D7" A2 c |"G" BGG BGG | BGB Bcd |"Am" edc"D7" BcA | 
"G" BGG G2 :: d |"G" gfg GBd | gfg bag |"Bm" afd Adf |"Bm" afd def |"C" gfg"G" Bcd | 
"Em" gdB"Am" cde |"D7" dcB AGF |"G" BGG G2 :|

X:2
E:5
L:1/8
M:6/8
K:G
|: d/c/ |"G" BAB G2 E | D2 D DEG |"Am" ABA AGE |"C" cBc"D7" Adc |"G" BAB G2 E | D2 D DEG | 
"Am" ABA"D7" AGA |"G" BGG G2 :: B/c/ |"G" d2 d dBG | Bdd d2 B |"Am" c2 c cAA |"Em" B2 B B2 d | 
"C" e2 e"D7" dBA |"Em" B2 d"Am" dBA |"D7" GAB AGA |"G" BGG G2 :|

X:3
E:5
L:1/8
M:6/8
K:G
|: d/c/ |"G" BGG DGG | BGB dcB |"Am" cAA EAA |"C" cBc"D7" edc |"G" BGG DGG | BGB dcB | 
"Am" cBc"D7" Adc |"G" BGG G2 :: g/a/ |"G" bgg dgg | bgb bag |"Bm" aff dff |"Bm" afa agf | 
"C" egg"G" dgg |"Am" cgg"G" B2 B |"D7" cBc Adc |"G" BGG G2 :|
  1. Segmentation
%%input
%%segmentation
L:1/4
M:4/4
K:Eb
"Cm" c"Cm" c"Cm/Eb" g"Cm" g/a/ |"Bb/D" b"Eb" g"Ab" e"Ddim/F" f/g/ |"Bb7/F" a2"Eb" g2 |
"F7/A" f"Bbsus4" f"Cm" e"Bb/D" f |"Eb" g"Bbsus4" f"Cm" e"Bb/D" d |"Cm7/Eb" c2"Bb" B2 |
"Cm" e"Csus2" d"Cm" e/f/"Cm/Eb" g |"Fm" f/e/"G" d"Ab" c2 |
"G" d"Cm/Eb" c/d/"Cm" e"Gsus4" d |"C" c4 |]

%%output
X:0
E:9
L:1/4
M:4/4
K:Eb
"Cm" c"Cm" c"Cm/Eb" g"Cm" g/a/ |"Bb/D" b"Eb" g"Ab" e"Ddim/F" f/g/ |"Bb7/F" a2"Eb" !breath!g2 |
"F7/A" f"Bbsus4" f"Cm" e"Bb/D" f |"Eb" g"Bbsus4" f"Cm" e"Bb/D" d |"Cm7/Eb" c2"Bb" !breath!B2 |
"Cm" e"Csus2" d"Cm" e/f/"Cm/Eb" g |"Fm" f/e/"G" d"Ab" !breath!c2 |
"G" d"Cm/Eb" c/d/"Cm" e"Gsus4" d |"C" !breath!c4 |]

X:1
E:9
L:1/4
M:4/4
K:Eb
"Cm" c"Cm" c"Cm/Eb" g"Cm" g/a/ |"Bb/D" b"Eb" g"Ab" e"Ddim/F" f/g/ |"Bb7/F" a2"Eb" !breath!g2 |
"F7/A" f"Bbsus4" f"Cm" e"Bb/D" f |"Eb" g"Bbsus4" f"Cm" e"Bb/D" d |"Cm7/Eb" c2"Bb" !breath!B2 |
"Cm" e"Csus2" d"Cm" e/f/"Cm/Eb" g |"Fm" f/e/"G" d"Ab" !breath!c2 |
"G" d"Cm/Eb" c/d/"Cm" e"Gsus4" d |"C" !breath!c4 |]

X:2
E:9
L:1/4
M:4/4
K:Eb
"Cm" c"Cm" c"Cm/Eb" g"Cm" g/a/ |"Bb/D" b"Eb" g"Ab" e"Ddim/F" f/g/ |"Bb7/F" a2"Eb" !breath!g2 |
"F7/A" f"Bbsus4" f"Cm" e"Bb/D" f |"Eb" g"Bbsus4" f"Cm" e"Bb/D" d |"Cm7/Eb" c2"Bb" !breath!B2 |
"Cm" e"Csus2" d"Cm" e/f/"Cm/Eb" g |"Fm" f/e/"G" d"Ab" !breath!c2 |
"G" d"Cm/Eb" c/d/"Cm" e"Gsus4" d |"C" !breath!c4 |]

X:3
E:9
L:1/4
M:4/4
K:Eb
"Cm" c"Cm" c"Cm/Eb" g"Cm" g/a/ |"Bb/D" b"Eb" g"Ab" e"Ddim/F" f/g/ |"Bb7/F" a2"Eb" !breath!g2 |
"F7/A" f"Bbsus4" f"Cm" e"Bb/D" f |"Eb" g"Bbsus4" f"Cm" e"Bb/D" d |"Cm7/Eb" c2"Bb" !breath!B2 |
"Cm" e"Csus2" d"Cm" e/f/"Cm/Eb" g |"Fm" f/e/"G" d"Ab" !breath!c2 |
"G" d"Cm/Eb" c/d/"Cm" e"Gsus4" d |"C" !breath!c4 |]
  1. Transcription
%%input
%%transcription
L:1/8
M:3/4
K:A
EG A2 A2 | BA G2 A2 | Bc/d/ e2 A2 | BA GF E2 | F2 G2 A2 | Bc d2 e2 |
fd c/4B/4c/4B/4c/4B/4c/4B/4 A2 | A2 A4 | EG A2 A2 | BA G2 A2 | Bc/d/ e2 A2 | BA GF E2 |
F2 G2 A2 | Bc d2 e2 | fd c/4B/4c/4B/4c/4B/4c/4B/4 A2 | A2 A4 | cd e2 ef | =g2 f3 e | dc d2 dB |
AB G>F E2 | F2 G2 A2 | Bc d2 e2 | fd c/4B/4c/4B/4c/4B/4c/4B/4 A2 | A2 A4- | A2 cd e2 |
ef =g2 f2- | fe dc d2 | dB AB G>F | E2 F2 G2 | A2 Bc d2 | e2 fd c/4B/4c/4B/4c/4B/4c/4B/4 |
A2 A2 A2- | A4 z2 |]

%%output
X:0
E:3
L:1/8
M:3/4
K:A
EG | A2 A2 BA | G2 A2 Bc/d/ | e2 A2 BA | GF E2 F2 | G2 A2 Bc | d2 e2 fd | TB2 A2 A2 | A4 :: cd |
e2 ef =g2 | f3 edc | d2 dB AB | G>F E2 F2 | G2 A2 Bc | d2 e2 fd | TB2 A2 A2 | A6 :|

X:1
E:3
L:1/8
M:3/4
K:A
EG |"A" A2 A2 BA |"E" G2 A2 Bc/d/ |"A" e2 A2 BA |"E" GF E2 F2 |"E" G2 A2 Bc |"D" d2 e2 fd | 
"E" TB2 A2 A2 |"A" A4 :| cd |"A" e2 ef =g2 |"D" f3 e dc |"D" d2 dB AB |"E" G>F E2 F2 | 
"E" G2 A2 Bc |"D" d2 e2 fd |"E" TB2 A2 A2 |"A" A6 cd |"A" e2 ef =g2 |"D" f3 e dc |"D" d2 dB AB | 
"E" G>F E2 F2 |"E" G2 A2 Bc |"D" d2 e2 fd |"E" TB2 A2 A2 |"A" A4 |]

X:2
E:3
L:1/8
M:3/4
K:A
|: EG |"A" A2 A2 BA |"E" G2 A2 Bc/d/ |"A" e2 A2 BA |"E" GF E2 F2 |"G" G2 A2 Bc |"D" d2 e2 fd | 
"E" TB2 A2 A2 |"A" A4 :| cd |"A" e2 ef =g2 |"D" f3 e dc |"D" d2 dB AB |"E" G>F E2 F2 | 
"G" G2 A2 Bc |"D" d2 e2 fd |"E" TB2 A2 A2 |"A" A6 cd |"A" e2 ef =g2 |"D" f3 e dc |"D" d2 dB AB | 
"E" G>F E2 F2 |"G" G2 A2 Bc |"D" d2 e2 fd |"E" TB2 A2 A2 |"A" A6 ||

X:3
E:4
L:1/8
M:3/4
K:A
EG | A2 A2 BA | G2 A2 Bc/d/ | e2 A2 BA | GF E2 F2 | G2 A2 Bc | d2 e2 fd | TB2 A2 A2 | A4 :| cd | 
e2 ef =g2 | f3 e dc | d2 dB AB | G>F E2 F2 | G2 A2 Bc | d2 e2 fd | TB2 A2 A2 | A6 || cd | e2 ef =g2 | 
f3 e dc | d2 dB AB | G>F E2 F2 | G2 A2 Bc | d2 e2 fd | TB2 A2 A2 | A6 ||
  1. Variation
%%input
%%variation
L:1/8
M:6/8
K:D
|: AFD DFA | Add B2 A | ABA F3 | GFG EFG | AFD DFA | Add B2 A | ABA F2 E | FDD D3 :: fdd ede |
fdd d2 g | fdd def | gfg e2 g | fed B2 A | AdF A3 | ABA F2 E | FDD D3 :|

%%output
X:0
E:8
L:1/8
M:6/8
K:D
|: B | AFD DFA | Add B2 A | ABA F2 E | FEE E2 B | AFD DF/G/A | Add B2 A | ABA F2 E | FDD D2 :: e |
fdd dcd | fdd d2 e | f^ef d=ef | g2 f efg | ff/e/d B2 d | Add F2 G | ABA F2 E | FDD D2 :|

X:1
E:8
L:1/8
M:6/8
K:D
|: B | AFD DFA | ded B2 A | ABA F2 D | GFG E2 B | AFD DF/G/A | df/e/d B2 A | ABA F2 E | EDD D2 :: 
e | fdd e^de | fdd d2 e | f2 f def | g2 f e2 g | fed B2 d | A2 d F3 | ABA F2 E | EDD D2 :|

X:2
E:8
L:1/8
M:6/8
K:D
|: B | AFD DFA | BdB BAF | ABA F2 D | FEE E2 B | AFD DFA | BdB BAF | ABA F2 E |1 FDD D2 :|2 
FDD D2 e |: fdd dcd | fdd d2 e | fef def | gfg eag | fed B2 d | A2 d F2 G | ABA F2 E |1 
FDD D2 e :|2 FDD D2 ||

X:3
E:5
L:1/8
M:6/8
K:D
|: (d/B/) | AFD DFA | B2 d F2 A | AFD DEF | GFG EFG | AFD DFA | B2 d F2 A | Bdd F2 E | FDD D2 :: 
fed dB/c/d | efe efg | fed daa | agf eag | fed B2 d | A2 d F2 A | Bdd F2 E | FDD D2 :|

BibTeX Citation

@misc{wu2024melodyt5unifiedscoretoscoretransformer,
      title={MelodyT5: A Unified Score-to-Score Transformer for Symbolic Music Processing}, 
      author={Shangda Wu and Yashan Wang and Xiaobing Li and Feng Yu and Maosong Sun},
      year={2024},
      eprint={2407.02277},
      archivePrefix={arXiv},
      primaryClass={cs.SD},
      url={https://arxiv.org/abs/2407.02277}, 
}