Ecosyste.ms: Awesome
An open API service indexing awesome lists of open source software.
https://github.com/gpustack/gpustack
Manage GPU clusters for running LLMs
https://github.com/gpustack/gpustack
Last synced: 6 days ago
JSON representation
Manage GPU clusters for running LLMs
- Host: GitHub
- URL: https://github.com/gpustack/gpustack
- Owner: gpustack
- License: apache-2.0
- Created: 2024-05-11T03:41:58.000Z (8 months ago)
- Default Branch: main
- Last Pushed: 2024-10-24T14:18:39.000Z (3 months ago)
- Last Synced: 2024-10-24T19:50:55.713Z (3 months ago)
- Language: Python
- Homepage: https://gpustack.ai
- Size: 49.5 MB
- Stars: 439
- Watchers: 5
- Forks: 38
- Open Issues: 59
-
Metadata Files:
- Readme: README.md
- Contributing: docs/contributing.md
- License: LICENSE
- Code of conduct: docs/code-of-conduct.md
Awesome Lists containing this project
- Awesome-LLM - GPUStack - An open-source GPU cluster manager for running LLMs (LLM Deployment)
- StarryDivineSky - gpustack/gpustack
- awesome-llmops - GPUStack - source GPU cluster manager for running and managing LLMs |  | (LLMOps / Observability)
- awesome-repositories - gpustack/gpustack - Manage GPU clusters for running AI models (Python)
README
GPUStack is an open-source GPU cluster manager for running AI models.
### Key Features
- **Broad Hardware Compatibility:** Run with different brands of GPUs in Apple Macs, Windows PCs, and Linux servers.
- **Broad Model Support:** From LLMs and diffusion models to audio, embedding, and reranker models.
- **Scales with Your GPU Inventory:** Easily add more GPUs or nodes to scale up your operations.
- **Distributed Inference**: Supports both single-node multi-GPU and multi-node inference and serving.
- **Multiple Inference Backends**: Supports llama-box (llama.cpp & stable-diffusion.cpp), vox-box and vLLM as the inference backends.
- **Lightweight Python Package:** Minimal dependencies and operational overhead.
- **OpenAI-compatible APIs:** Serve APIs that are compatible with OpenAI standards.
- **User and API key management:** Simplified management of users and API keys.
- **GPU metrics monitoring:** Monitor GPU performance and utilization in real-time.
- **Token usage and rate metrics:** Track token usage and manage rate limits effectively.
## Installation
### Linux or macOS
GPUStack provides a script to install it as a service on systemd or launchd based systems with default port 80. To install GPUStack using this method, just run:
```bash
curl -sfL https://get.gpustack.ai | sh -s -
```
### Windows
Run PowerShell as administrator (**avoid** using PowerShell ISE), then run the following command to install GPUStack:
```powershell
Invoke-Expression (Invoke-WebRequest -Uri "https://get.gpustack.ai" -UseBasicParsing).Content
```
### Other Installation Methods
For manual installation, docker installation or detailed configuration options, please refer to the [Installation Documentation](https://docs.gpustack.ai/latest/installation/installation-script/).
## Getting Started
1. Run and chat with the **llama3.2** model:
```bash
gpustack chat llama3.2 "tell me a joke."
```
2. Run and generate an image with the **stable-diffusion-v3-5-large-turbo** model:
> ### 💡 Tip
>
> This command downloads the model (~12GB) from Hugging Face. The download time depends on your network speed. Ensure you have enough disk space and VRAM (12GB) to run the model. If you encounter issues, you can skip this step and move to the next one.
```bash
gpustack draw hf.co/gpustack/stable-diffusion-v3-5-large-turbo-GGUF:stable-diffusion-v3-5-large-turbo-Q4_0.gguf \
"A minion holding a sign that says 'GPUStack'. The background is filled with futuristic elements like neon lights, circuit boards, and holographic displays. The minion is wearing a tech-themed outfit, possibly with LED lights or digital patterns. The sign itself has a sleek, modern design with glowing edges. The overall atmosphere is high-tech and vibrant, with a mix of dark and neon colors." \
--sample-steps 5 --show
```
Once the command completes, the generated image will appear in the default viewer. You can experiment with the prompt and CLI options to customize the output.
3. Open `http://myserver` in the browser to access the GPUStack UI. Log in to GPUStack with username `admin` and the default password. You can run the following command to get the password for the default setup:
**Linux or macOS**
```bash
cat /var/lib/gpustack/initial_admin_password
```
**Windows**
```powershell
Get-Content -Path "$env:APPDATA\gpustack\initial_admin_password" -Raw
```
4. Click `Playground` in the navigation menu. Now you can chat with the LLM in the UI playground.
5. Click `API Keys` in the navigation menu, then click the `New API Key` button.
6. Fill in the `Name` and click the `Save` button.
7. Copy the generated API key and save it somewhere safe. Please note that you can only see it once on creation.
8. Now you can use the API key to access the OpenAI-compatible API. For example, use curl as the following:
```bash
export GPUSTACK_API_KEY=myapikey
curl http://myserver/v1-openai/chat/completions \
-H "Content-Type: application/json" \
-H "Authorization: Bearer $GPUSTACK_API_KEY" \
-d '{
"model": "llama3.2",
"messages": [
{
"role": "system",
"content": "You are a helpful assistant."
},
{
"role": "user",
"content": "Hello!"
}
],
"stream": true
}'
```
## Supported Platforms
- [x] macOS
- [x] Linux
- [x] Windows
## Supported Accelerators
- [x] Apple Metal (M-series chips)
- [x] NVIDIA CUDA ([Compute Capability](https://developer.nvidia.com/cuda-gpus) 6.0 and above)
- [x] Ascend CANN
- [x] Moore Threads MUSA
We plan to support the following accelerators in future releases.
- [ ] AMD ROCm
- [ ] Intel oneAPI
- [ ] Qualcomm AI Engine
## Supported Models
GPUStack uses [llama-box](https://github.com/gpustack/llama-box) (bundled [llama.cpp](https://github.com/ggerganov/llama.cpp) and [stable-diffusion.cpp](https://github.com/leejet/stable-diffusion.cpp) server), [vLLM](https://github.com/vllm-project/vllm) and [vox-box](https://github.com/gpustack/vox-box) as the backends and supports a wide range of models. Models from the following sources are supported:
1. [Hugging Face](https://huggingface.co/)
2. [ModelScope](https://modelscope.cn/)
3. [Ollama Library](https://ollama.com/library)
4. Local File Path
### Example Models:
| **Category** | **Models** |
| -------------------------------- | -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| **Large Language Models(LLMs)** | [Qwen](https://huggingface.co/models?search=Qwen/Qwen), [LLaMA](https://huggingface.co/meta-llama), [Mistral](https://huggingface.co/mistralai), [Deepseek](https://huggingface.co/models?search=deepseek-ai/deepseek), [Phi](https://huggingface.co/models?search=microsoft/phi), [Yi](https://huggingface.co/models?search=01-ai/Yi) |
| **Vision Language Models(VLMs)** | [Llama3.2-Vision](https://huggingface.co/models?pipeline_tag=image-text-to-text&search=llama3.2), [Pixtral](https://huggingface.co/models?search=pixtral) , [Qwen2-VL](https://huggingface.co/models?search=Qwen/Qwen2-VL), [LLaVA](https://huggingface.co/models?search=llava), [InternVL2](https://huggingface.co/models?search=internvl2) |
| **Diffusion Models** | [Stable Diffusion](https://huggingface.co/models?search=gpustack/stable-diffusion), [FLUX](https://huggingface.co/models?search=gpustack/flux) |
| **Rerankers** | [GTE](https://huggingface.co/gpustack/gte-multilingual-reranker-base-GGUF), [BCE](https://huggingface.co/gpustack/bce-reranker-base_v1-GGUF), [BGE](https://huggingface.co/gpustack/bge-reranker-v2-m3-GGUF), [Jina](https://huggingface.co/models?search=gpustack/jina) |
| **Audio Models** | [Whisper](https://huggingface.co/models?search=Systran/faster) (speech-to-text), [CosyVoice](https://huggingface.co/models?search=FunAudioLLM/CosyVoice) (text-to-speech) |
For full list of supported models, please refer to the supported models section in the [inference backends](https://docs.gpustack.ai/latest/user-guide/inference-backends/) documentation.
## OpenAI-Compatible APIs
GPUStack serves the following OpenAI compatible APIs under the `/v1-openai` path:
- [x] [List Models](https://platform.openai.com/docs/api-reference/models/list)
- [x] [Create Completion](https://platform.openai.com/docs/api-reference/completions/create)
- [x] [Create Chat Completion](https://platform.openai.com/docs/api-reference/chat/create)
- [x] [Create Embeddings](https://platform.openai.com/docs/api-reference/embeddings/create)
- [x] [Create Image](https://platform.openai.com/docs/api-reference/images/create)
- [x] [Create Image Edit](https://platform.openai.com/docs/api-reference/images/createEdit)
- [x] [Create Speech](https://platform.openai.com/docs/api-reference/audio/createSpeech)
- [x] [Create Transcription](https://platform.openai.com/docs/api-reference/audio/createTranscription)
For example, you can use the official [OpenAI Python API library](https://github.com/openai/openai-python) to consume the APIs:
```python
from openai import OpenAI
client = OpenAI(base_url="http://myserver/v1-openai", api_key="myapikey")
completion = client.chat.completions.create(
model="llama3.2",
messages=[
{"role": "system", "content": "You are a helpful assistant."},
{"role": "user", "content": "Hello!"}
]
)
print(completion.choices[0].message)
```
GPUStack users can generate their own API keys in the UI.
## Documentation
Please see the [official docs site](https://docs.gpustack.ai) for complete documentation.
## Build
1. Install Python (version 3.10 to 3.12).
2. Run `make build`.
You can find the built wheel package in `dist` directory.
## Contributing
Please read the [Contributing Guide](./docs/contributing.md) if you're interested in contributing to GPUStack.
## License
Copyright (c) 2024 The GPUStack authors
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at [LICENSE](./LICENSE) file for details.
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
