Decomposer-Qwen3-8B/README.md

license, library_name, pipeline_tag, base_model, datasets, language, tags

license	library_name	pipeline_tag	base_model	datasets	language	tags
apache-2.0	transformers	text-generation	Qwen/Qwen3-8B	haiyewon/Strudel-Synth	en	music midi strudel symbolic-music music-decompilation code-generation qwen3

Decomposer

Decomposer decompiles symbolic music (MIDI) into executable Strudel programs. This model is the 8B model from the paper Decomposer: Learning to Decompile Symbolic Music to Programs, post-trained from Qwen3-8B.

• 📄 Paper: arXiv:2607.01849
• 🌐 Project page: yewon-kim.com/decomposer
• 🎹 Live demo: haiyewon/decomposer-demo
• 💻 Code: github.com/elianakim/Decomposer
• 🤗 Dataset: haiyewon/Strudel-Synth

Model Description

• Base model: Qwen/Qwen3-8B
• Model size: 8.2B
• Type: Causal Language Model
• Precision: bfloat16

Quick Start

Clone our GitHub code repo, run through the setup steps, and try:

git clone https://github.com/elianakim/Decomposer.git
cd Decomposer
uv sync

uv run generate_transformers.py \
    --model haiyewon/Decomposer-Qwen3-8B \
    --midi_path path/to/input.mid \
    --n_outputs 5

The repo and inference scripts provide a more complete usage guide.

Model Details

Input Format

MIDI is serialized as instrument-wise note events. Onsets are given in cycle coordinates instead of seconds:

[BPM=120 Meter=4]
[acoustic_grand_piano] C4@0.00 E4@0.25 G4@0.50 C5@1.00
[drum] bass_drum@0.00 closed_hi_hat@0.25 closed_hi_hat@0.50 snare@0.75

• Header [BPM=… Meter=…] — tempo/meter; if absent, madmom is used to estimate them.
• Events — one line per instrument, pitch@cycle_onset tokens; melodic pitches as note names (C4, F#3), drums as GM drum names (bass_drum, closed_hi_hat).

Training Data

Strudel-Synth is a synthetic corpus of 21,174 (MIDI, Strudel) pairs (20,152 train / 1,022 test) built by distilling Strudel programs from Claude-Opus-4.6 and rendering each to MIDI with the Strudel runtime. Its ~20K training split is divided into two disjoint ~10K halves, one per stage:

• SFT uses the first half as paired (MIDI, Strudel) examples.
• RL uses the second half (MIDI only), plus short-fragment (<30 s) LMD MIDI.

Inference Hyperparameters

Recommended settings matching the paper evaluation:

Setting	Value
temperature	1.0
max_new_tokens	4096
enable_thinking	false

Citation

If you find our model useful, please cite our research as

@article{kim2026decomposer,
  title   = {Decomposer: Learning to Decompile Symbolic Music to Programs},
  author  = {Kim, Yewon and Gandhi, Apurva and Chung, David and Neubig, Graham and Donahue, Chris},
  journal = {arXiv preprint arXiv:2607.01849},
  year    = {2026}
}