nvidia-docker
llama.cpp
nvidia-docker | llama.cpp | |
---|---|---|
53 | 775 | |
16,998 | 57,463 | |
- | - | |
0.0 | 10.0 | |
5 months ago | 2 days ago | |
Makefile | C++ | |
Apache License 2.0 | MIT License |
Stars - the number of stars that a project has on GitHub. Growth - month over month growth in stars.
Activity is a relative number indicating how actively a project is being developed. Recent commits have higher weight than older ones.
For example, an activity of 9.0 indicates that a project is amongst the top 10% of the most actively developed projects that we are tracking.
nvidia-docker
-
What are the best AI tools you've ACTUALLY used?
Nvidia Docker on GitHub
-
Plex setup through Docker + Nvidia card, but hardware acceleration stops working after some time
Here's where I found discussion regarding this https://github.com/NVIDIA/nvidia-docker/issues/1671
-
Seeking Guidance on Leveraging Local Models and Optimizing GPU Utilization in containerized packages
I found the Faq, looks like Windows isn't supported which might indicate why I had this problem earlier. I might need to dual boot my machine if it won't work with WSL which I don't see mentioned in either page. WSL Cuda instructions found this I'll give it a try.
-
Which GPU for HW transcoding in PMS: Intel Arc or Nvidia?
Arc has linux kernel support from 6.0, been using an A770 with tdarr for a few months. Super solid and no issues like the nvidia docker toolkit just losing the GPU. The workaround doesn't hold for long - https://github.com/NVIDIA/nvidia-docker/issues/1730 It is one of the reasons I went with Intel over waiting for a low end 40 series. The other was I that basically stole it for $199. So far Plex is the only thing that doesn't work with Arc and all the HW transcoding falls on the iGPU. Knowing how they prioritize things nobody wants, Arc support and AV1 transcoding will be added when 16th gen Intel CPUs are released.
- [D] Would a Tesla M40 provide cheap inference acceleration for self-hosted LLMs?
-
Help! Accelerated-GPU with Cuda and CuPy
I use ubuntu with wsl2 under windows 11. I want to learn how to get GPU-accelerated computing working. My ASUS laptop has a NVIDIA graphics card as well as a built in Intel. Please ask me any needed questions and then tell me the steps I need to get this working? SOFTWARE DEVELOPER Thanks for sharing that information. Can you please confirm if you have already installed NVIDIA drivers on your Windows 11 system? If not, you'll need to install the appropriate NVIDIA drivers for your GPU in order to enable GPU acceleration. Additionally, what is your end goal is with GPU accelerated computing? yes I have the NVIDIA control panel installed. SOFTWARE DEVELOPER Great. To enable GPU acceleration in WSL2, you'll need to follow these steps: First, confirm you are running Windows 11 build 20145 or later, and have WSL2 installed and configured to use Ubuntu. Download the NVIDIA Linux driver for your GPU from the official NVIDIA Linux download page https://www.nvidia.com/Download/index.aspx. Install the latest NVIDIA driver, including the NVIDIA Docker plugin. You can follow instructions on the NVIDIA Docker plugin GitHub page (https://github.com/NVIDIA/nvidia-docker). Install the CUDA toolkit and cuDNN library. Once you've installed the NVIDIA driver successfully, you can download the CUDA toolkit and cuDNN library (if applicable) from the official NVIDIA CUDA download page https://developer.nvidia.com/cuda-downloads. Install the NVIDIA Container Toolkit. NVIDIA Container Toolkit is a nifty package that provides a set of tools for GPU-accelerated container deployments. You can find installation instructions and a getting started guide on the NVIDIA Container Toolkit GitHub page https://github.com/NVIDIA/nvidia-docker. Test the installation by running a TensorFlow or PyTorch script that uses the GPU for accelerated computing. Check if the GPU is recognized and being used in the output of the script. Once you've successfully followed these steps, you should have GPU-accelerated computing environments in WSL2. Hope that helps! Let me know if you have any other questions. yes to question #1. On #2, do I download the Game Ready driver or the Studio Driver. I don't play games as I'm a software devevloper but want accelerated graphics for development or data science purposes. SOFTWARE DEVELOPER Great to hear that you have WSL2 set up on your system. Regarding your question, you should download the Studio Driver version, as it's optimized for professional workloads, including data science and software development. The Game Ready version is optimized for gaming and might not always provide the best performance for non-gaming workloads. Once you have downloaded the Studio Driver for your GPU, follow the instructions provided by the driver installer to complete the installation process. Let me know if you have any further questions! Would you clarify #3? Is the NVIDIA Docker plugin included with the NVIDIA driver install? Or is it a separate install. I have Docker for Windows installed and working in WSL2. SOFTWARE DEVELOPER Sure, I'd be happy to clarify. The NVIDIA Docker plugin is an additional package that needs to be installed separately from the NVIDIA drivers. The plugin provides a convenient way for creating and running GPU-accelerated Docker containers in your system. To install the NVIDIA Docker plugin, follow the instructions on the NVIDIA Docker plugin GitHub page (https://github.com/NVIDIA/nvidia-docker). Note that the plugin is only compatible with Docker CE (Community Edition) or Docker EE (Enterprise Edition) on Linux-based systems, and does not work on Docker for Windows. Once you've installed the plugin, you can use it to run GPU-accelerated containers with Docker. It's worth noting that the NVIDIA Docker plugin requires access to your host system's NVIDIA GPU, so you'll need to ensure that your GPU has been correctly installed and recognized by your system. Hope that answers your question. Let me know if you have any additional questions!
-
Plex Transcode (VC1 (HW) 1080p H264 (HW) 1080p) on Pixel 7 Pro
Im trying to determine how to troubleshoot & resolve the HW transcoding, but based on my testing Im assuming its some change to the NVIDIA toolkit https://github.com/NVIDIA/nvidia-docker
- jelyfin with nvidia acceleration stopped working
-
Dockerize CUDA-Accelerated Applications
NVIDIA Container Toolkit
-
Setting up a new unraid server with vgpu and plex docker transcodes
So I am in the initial planning stages of setting up a new unraid server. Looking at picking up an SC846 24bay 4u chassis. I've got a Gigabyte Aorus mb with an AMD 5950x, 32gb of ddr4 (adding more as needed) and an nvidia 3070ti. I plan on getting an LSI 8i for the drives and leaves room for expansion server plans. My goal is to have plex setup via docker and utilize the gpu transcoding to offload the cpu work. I also want to setup vms or a vm server to essentially also have a "gaming server" mainly for me and the kids. This means down the road I would be adding another GPU to split up with other users. Im trying to allow for a max of 4 people while also still allowing plex to transcode as needed. Now I know there's other ways to do this but I dont feel like splitting this up into multiple systems unless I have to. So really just trying to see if this might be possible. My worry is that in order to make the gpu available to the plex docker I have to setup an nvidia container. https://github.com/NVIDIA/nvidia-docker
llama.cpp
-
Ask HN: Affordable hardware for running local large language models?
Yes, Metal seems to allow a maximum of 1/2 of the RAM for one process, and 3/4 of the RAM allocated to the GPU overall. There’s a kernel hack to fix it, but that comes with the usual system integrity caveats. https://github.com/ggerganov/llama.cpp/discussions/2182
- Xmake: A modern C/C++ build tool
-
Better and Faster Large Language Models via Multi-Token Prediction
For anyone interested in exploring this, llama.cpp has an example implementation here:
https://github.com/ggerganov/llama.cpp/tree/master/examples/...
- Llama.cpp Bfloat16 Support
-
Fine-tune your first large language model (LLM) with LoRA, llama.cpp, and KitOps in 5 easy steps
Getting started with LLMs can be intimidating. In this tutorial we will show you how to fine-tune a large language model using LoRA, facilitated by tools like llama.cpp and KitOps.
- GGML Flash Attention support merged into llama.cpp
-
Phi-3 Weights Released
well https://github.com/ggerganov/llama.cpp/issues/6849
- Lossless Acceleration of LLM via Adaptive N-Gram Parallel Decoding
- Llama.cpp Working on Support for Llama3
-
Embeddings are a good starting point for the AI curious app developer
Have just done this recently for local chat with pdf feature in https://recurse.chat. (It's a macOS app that has built-in llama.cpp server and local vector database)
Running an embedding server locally is pretty straightforward:
- Get llama.cpp release binary: https://github.com/ggerganov/llama.cpp/releases
What are some alternatives?
Pytorch - Tensors and Dynamic neural networks in Python with strong GPU acceleration
ollama - Get up and running with Llama 3, Mistral, Gemma, and other large language models.
nvidia-container-runtime - NVIDIA container runtime
gpt4all - gpt4all: run open-source LLMs anywhere
Entware - Ultimate repo for embedded devices
text-generation-webui - A Gradio web UI for Large Language Models. Supports transformers, GPTQ, AWQ, EXL2, llama.cpp (GGUF), Llama models.
container-images
GPTQ-for-LLaMa - 4 bits quantization of LLaMA using GPTQ
Whisparr
ggml - Tensor library for machine learning
docker-to-linux - Make bootable Linux disk image abusing Docker
alpaca.cpp - Locally run an Instruction-Tuned Chat-Style LLM