Instructions to use NathanFradet/Maestro-TSD-bpe20k with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use NathanFradet/Maestro-TSD-bpe20k with Transformers:

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

pipe = pipeline("text-generation", model="NathanFradet/Maestro-TSD-bpe20k")

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("NathanFradet/Maestro-TSD-bpe20k")
model = AutoModelForCausalLM.from_pretrained("NathanFradet/Maestro-TSD-bpe20k")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use NathanFradet/Maestro-TSD-bpe20k with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "NathanFradet/Maestro-TSD-bpe20k"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "NathanFradet/Maestro-TSD-bpe20k",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/NathanFradet/Maestro-TSD-bpe20k

SGLang

How to use NathanFradet/Maestro-TSD-bpe20k 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 "NathanFradet/Maestro-TSD-bpe20k" \
    --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": "NathanFradet/Maestro-TSD-bpe20k",
		"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 "NathanFradet/Maestro-TSD-bpe20k" \
        --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": "NathanFradet/Maestro-TSD-bpe20k",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use NathanFradet/Maestro-TSD-bpe20k with Docker Model Runner:
```
docker model run hf.co/NathanFradet/Maestro-TSD-bpe20k
```

MidiTok version and code corrections

by asigalov61 - opened Dec 6, 2024

Discussion

asigalov61

Dec 6, 2024

Hey @Natooz

I am posting it here so that you can correct it...

Please update the recommended MidiTok version to the latest one because 2.7.1 does not work. The latest version (3.0.4) seems to work fine.

Also, please update the code snippet for both models. This is what worked for me:

import torch
from transformers import AutoModelForCausalLM
from miditok import TSD
from symusic import Score

torch.set_default_device("cuda")
model = AutoModelForCausalLM.from_pretrained("Natooz/Maestro-TSD-bpe20k", trust_remote_code=True, torch_dtype="auto")
tokenizer = TSD.from_pretrained("Natooz/Maestro-TSD-bpe20k")
input_midi = Score("Giant-Music-Transformer-Piano-Seed-1.mid")
input_tokens = tokenizer(input_midi)

generated_token_ids = model.generate(torch.LongTensor([input_tokens[0].ids]).cuda(), max_length=500)
generated_midi = tokenizer(generated_token_ids.tolist())
generated_midi.dump_midi("continued.mid")

Hope this is helpful :)

Alex

Upload images, audio, and videos by dragging in the text input, pasting, or clicking here.

Tap or paste here to upload images

· Sign up or log in to comment