Instructions to use bugrayildirim/story-to-music-analyzer with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use bugrayildirim/story-to-music-analyzer with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("text-generation", model="bugrayildirim/story-to-music-analyzer")

# Load model directly
from transformers import AutoTokenizer, AutoModelForSeq2SeqLM

tokenizer = AutoTokenizer.from_pretrained("bugrayildirim/story-to-music-analyzer")
model = AutoModelForSeq2SeqLM.from_pretrained("bugrayildirim/story-to-music-analyzer")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use bugrayildirim/story-to-music-analyzer with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "bugrayildirim/story-to-music-analyzer"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "bugrayildirim/story-to-music-analyzer",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/bugrayildirim/story-to-music-analyzer

SGLang

How to use bugrayildirim/story-to-music-analyzer with SGLang:

Install from pip and serve model

# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
    --model-path "bugrayildirim/story-to-music-analyzer" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "bugrayildirim/story-to-music-analyzer",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker images

docker run --gpus all \
    --shm-size 32g \
    -p 30000:30000 \
    -v ~/.cache/huggingface:/root/.cache/huggingface \
    --env "HF_TOKEN=<secret>" \
    --ipc=host \
    lmsysorg/sglang:latest \
    python3 -m sglang.launch_server \
        --model-path "bugrayildirim/story-to-music-analyzer" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "bugrayildirim/story-to-music-analyzer",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use bugrayildirim/story-to-music-analyzer with Docker Model Runner:
```
docker model run hf.co/bugrayildirim/story-to-music-analyzer
```

Story-to-Music Analyzer (Türkçe Metin → Müzik Metadata)

google/mt5-small'un Türkçe metinleri Suno ve Udio için müzik parametrelerine çeviren fine-tuned versiyonu. Story-to-Music MCP projesinin "Analyzer" bileşeni.

Görev

Türkçe bir metin (hikâye, şarkı sözü, konsept) verildiğinde, model şu JSON yapısını üretir:

{
  "emotion": "özlem",
  "energy": 4,
  "bpm": 80,
  "key": "A minor",
  "instruments": ["ney", "saz"],
  "vocal_style": "erkek, kısık, dramatik",
  "suno_style_prompt": "melancholic Turkish ballad with ney flute and saz, slow tempo"
}

Kullanım

from transformers import AutoTokenizer, AutoModelForSeq2SeqLM

tokenizer = AutoTokenizer.from_pretrained("bugrayildirim/story-to-music-analyzer")
model = AutoModelForSeq2SeqLM.from_pretrained("bugrayildirim/story-to-music-analyzer")

text = "Bir zamanlar büyük bir aşk yaşamıştım. Sonra o aşk gitti."
prompt = f"Şu metni analiz et ve müzik promptu üret: {text}"

inputs = tokenizer(prompt, return_tensors="pt", max_length=512, truncation=True)
outputs = model.generate(**inputs, max_new_tokens=192, num_beams=4, early_stopping=True)
print(tokenizer.decode(outputs[0], skip_special_tokens=True))

Eğitim

Base model: google/mt5-small (556M parametre)
Veri: 854 Türkçe şarkı sözü + Ollama (qwen2.5:7b) tarafından üretilen metadata etiketleri (distill_data.py pipeline'ı, GitHub)
Eğitim: Kaggle T4 GPU, 17 epoch, eval_loss 0.54'e kadar düştü
Önemli: fp16=False (mT5'in fp16'da NaN sorunu var), gradient checkpointing aktif

Detaylar: training/story_to_music_train.ipynb

Performans

JSON validity ve alan kapsamı (eval set 86 örnek):

Geçerli JSON: 100%
Tüm alanların kapsamı: 100%

Limitations

Sadece Türkçe metin destekler
CPU'da inference ~5 saniye (T4 GPU'da ~500ms)
suno_style_prompt alanı bazen yarı-Türkçe yarı-İngilizce çıktı verebilir (Story-to-Music MCP server bunu deterministik fallback ile düzeltir)

Lisans

MIT. Eğitim verisi olarak kullanılan şarkı sözlerinin telif hakları orijinal sahiplerine aittir; bu model stil/duygu örüntülerini öğrenmek için eğitildi, şarkı sözü ezberi yapmaz.

Atıf

@misc{story-to-music-analyzer,
  author = {Buğra Yıldırım},
  title = {Story-to-Music Analyzer: Turkish text to music metadata},
  year = {2026},
  publisher = {Hugging Face},
  howpublished = {\url{https://huggingface.co/bugrayildirim/story-to-music-analyzer}},
}

Downloads last month: 47

Safetensors

Model size

0.6B params

Tensor type

F32

Model tree for bugrayildirim/story-to-music-analyzer

Base model

google/mt5-small

Finetuned

(683)

this model

bugrayildirim
/

story-to-music-analyzer