Downloads · 30 days
0
zhongyangggggggggg/chinartsong
chinartsong is a machine learning model from zhongyangggggggggg. Use it for the machine learning task on the model card, and read the license before you ship it in a product.
Reference implementation accompanying our paper submitted to Multimedia Systems. Given a natural-language description, the system composes a Chinese art song end to end: it infers the musical parameters, generates a m…
Downloads · 30 days
0
Access
Public
Updated Jul 30, 2026
Repo size
174 KB
Likes
0
Public
Click a slice to open those files.
.zip174 KB · 98%
From the Hugging Face model README
Reference implementation accompanying our paper submitted to Multimedia Systems. Given a natural-language description, the system composes a Chinese art song end to end: it infers the musical parameters, generates a melody with a fine-tuned NotaGen model, writes singable lyrics aligned to that melody, and exports MIDI and MusicXML.
pip install -r requirements.txtpip install torch==2.1.0 --index-url https://download.pytorch.org/whl/cu121abc2midi (from the abcMIDI package), installed at the system
level (macOS: brew install abcmidi; Debian/Ubuntu: apt install abcmidi).LLM access is configured through environment variables (see config.py). By default
the system targets an OpenAI-compatible aggregation gateway that exposes the five
backends evaluated in the paper (Claude Opus 4.6, GPT-5.2, DeepSeek-V3.2, GLM-4.6,
and MiniMax-M2.5):
export LLM_BASE_URL="https://api.ephone.ai/v1" # or your own OpenAI-compatible endpoint
export LLM_API_KEY="" # set your key if the endpoint requires one
export LLM_MODEL="claude-opus-4-6" # switch backend by changing the model id
The NotaGen melody weights can be downloaded from
Hugging Face.
Place the weight file at melody/models/weights_chinese_artsong.pth, or point the
NOTAGEN_WEIGHTS_PATH environment variable to its location.
The model can be fine-tuned by following the procedure described in the paper.
abc2midi tool as described above.export NOTAGEN_WEIGHTS_PATH="/absolute/path/to/weights_chinese_artsong.pth"
python main.py "我想要一首有希望的、充满诗意的歌。"
# Omit the description to enter it interactively:
python main.py
# Alternatively, use the convenience launcher with the settings in run.sh:
./run.sh "我想要一首有诗意的、充满希望的歌曲。"
output/.config.py Global configuration (LLM access, paths, generation limits)
main.py Command-line entry point
pipeline.py Four-stage composition pipeline
modules/ Emotion analysis, lyric generation, NotaGen interface, conversion
melody/ NotaGen melody model: data, pretrain, finetune, RL, inference, generate.py
requirements.txt Python dependencies
run.sh Convenience launcher