Instructions to use Serveurperso/ACE-Step-1.5-GGUF with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use Serveurperso/ACE-Step-1.5-GGUF with llama-cpp-python:

# !pip install llama-cpp-python

from llama_cpp import Llama

llm = Llama.from_pretrained(
	repo_id="Serveurperso/ACE-Step-1.5-GGUF",
	filename="Qwen3-Embedding-0.6B-BF16.gguf",
)

output = llm(
	"Once upon a time,",
	max_tokens=512,
	echo=True
)
print(output)

Notebooks
Google Colab
Kaggle
Local Apps

llama.cpp

How to use Serveurperso/ACE-Step-1.5-GGUF with llama.cpp:

Install from brew

brew install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama-server -hf Serveurperso/ACE-Step-1.5-GGUF:Q4_K_M
# Run inference directly in the terminal:
llama-cli -hf Serveurperso/ACE-Step-1.5-GGUF:Q4_K_M

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama-server -hf Serveurperso/ACE-Step-1.5-GGUF:Q4_K_M
# Run inference directly in the terminal:
llama-cli -hf Serveurperso/ACE-Step-1.5-GGUF:Q4_K_M

Use pre-built binary

# Download pre-built binary from:
# https://github.com/ggerganov/llama.cpp/releases
# Start a local OpenAI-compatible server with a web UI:
./llama-server -hf Serveurperso/ACE-Step-1.5-GGUF:Q4_K_M
# Run inference directly in the terminal:
./llama-cli -hf Serveurperso/ACE-Step-1.5-GGUF:Q4_K_M

Build from source code

git clone https://github.com/ggerganov/llama.cpp.git
cd llama.cpp
cmake -B build
cmake --build build -j --target llama-server llama-cli
# Start a local OpenAI-compatible server with a web UI:
./build/bin/llama-server -hf Serveurperso/ACE-Step-1.5-GGUF:Q4_K_M
# Run inference directly in the terminal:
./build/bin/llama-cli -hf Serveurperso/ACE-Step-1.5-GGUF:Q4_K_M

Use Docker

docker model run hf.co/Serveurperso/ACE-Step-1.5-GGUF:Q4_K_M

LM Studio
Jan
Ollama
How to use Serveurperso/ACE-Step-1.5-GGUF with Ollama:
```
ollama run hf.co/Serveurperso/ACE-Step-1.5-GGUF:Q4_K_M
```

Unsloth Studio new

How to use Serveurperso/ACE-Step-1.5-GGUF with Unsloth Studio:

Install Unsloth Studio (macOS, Linux, WSL)

curl -fsSL https://unsloth.ai/install.sh | sh
# Run unsloth studio
unsloth studio -H 0.0.0.0 -p 8888
# Then open http://localhost:8888 in your browser
# Search for Serveurperso/ACE-Step-1.5-GGUF to start chatting

Install Unsloth Studio (Windows)

irm https://unsloth.ai/install.ps1 | iex
# Run unsloth studio
unsloth studio -H 0.0.0.0 -p 8888
# Then open http://localhost:8888 in your browser
# Search for Serveurperso/ACE-Step-1.5-GGUF to start chatting

Using HuggingFace Spaces for Unsloth

# No setup required
# Open https://huggingface.co/spaces/unsloth/studio in your browser
# Search for Serveurperso/ACE-Step-1.5-GGUF to start chatting

Docker Model Runner
How to use Serveurperso/ACE-Step-1.5-GGUF with Docker Model Runner:
```
docker model run hf.co/Serveurperso/ACE-Step-1.5-GGUF:Q4_K_M
```

Lemonade

How to use Serveurperso/ACE-Step-1.5-GGUF with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull Serveurperso/ACE-Step-1.5-GGUF:Q4_K_M

Run and chat with the model

lemonade run user.ACE-Step-1.5-GGUF-Q4_K_M

List all available models

lemonade list

Workflow?

by nored2222 - opened 17 days ago

Discussion

nored2222

17 days ago

Does anyone have a working workflow? I'm especially looking for covering.

Serveurperso

Owner 10 days ago

Start with LLM 4B and XL-Turbo and a fairly basic, mainstream vocal track. Open your own music (with lyrics, it really helps with model conditioning). Use "understand" to approximate the prompt. Paste the actual lyrics (found online, with the tags), check "Src audio" and "Timbre ref" on your song. Switch task to cover-nofsq mode, set the "cover strength" between 0.3 and 0.6, and generate the music (directly Synthezize button). aYou will get the clone that the model is capable of generating. The closer the copy, the more the lyrics and style can be modified. The "cover" mode (without NOFSQ) will diverge even further from your original music. The model is extremely good at pop/electronic music (4 to the floor). All styles can be done, but it requires some familiarity with the model.

Serveurperso

Owner 10 days ago

My screenshot system doesn't show the menus, but the only thing I did was use "LM understand" at the beginning to initialize the prompt and choose the task type (cover-nofsq / cover). The model is large to explore. You have to send it lots of different music tracks and test lots of different prompts plus parameters... It's limitless. On video I've just "forked" a track.
You should consider generating many of them with a random seed (-1) or doing batches (cherry-picking).

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