ollama-webui VS llama.cpp

I'm a podman beginner, trying to install ollama-webui(1) using Podman on M2 MBA.

I started up Podman Desktop, and did a terminal command "docker run -d -p 3000:8080 --add-host=host.docker.internal:host-gateway -v ollama-webui:/app/backend/data --name ollama-webui --restart always ghcr.io/ollama-webui/ollama-webui:main" based on Github's instructions, but it gave a error message something about "host".

Do you know what is the problem and how do I overcome this?

If I run the above command using Docker Desktop, it runs and installs Ollama-WebUI just fine.

(1) https://github.com/ollama-webui/ollama-webui ("Installing with Docker")

Thank you.

I set up Ollama in a docker container (really easy) and I use the Ollama web-ui here, that's very much like ChatGPT, and I have that in a container as well.

The primary use case here seems to be that it might be possible to use this tool to spend <$20/mo for the same feature set as ChatGPT+. It does not currently make any effort to support locally-hosted open source models, which is what I would have assumed from its name.

If you're interested in a fully Libre LLM stack, I've had fun lately with ollama [0] and ollama-webui [1]. It was pretty trivial to take ollama-webui's docker-compose file and set up a locally-running chat server with Mistral 7B. Trying out different models and prompts was likewise very easy to get started with.

Mistral isn't anything like as good as GPT-4, but it's Apache licensed and fully local, which meets my definition of Libre. I'll continue to use both while the FOSS stacks catch up, but it's fun to keep up with the progress on the open source stuff as tooling develops.

[0] https://github.com/jmorganca/ollama

[1] https://github.com/ollama-webui/ollama-webui

if you can compile stuff, then looking at llama.cpp (what ollama uses) is also interesting: https://github.com/ggerganov/llama.cpp

the server is here: https://github.com/ggerganov/llama.cpp/tree/master/examples/...

And you can search for any GGUF on huggingface

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

For anyone interested in exploring this, llama.cpp has an example implementation here:

https://github.com/ggerganov/llama.cpp/tree/master/examples/...

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.

well https://github.com/ggerganov/llama.cpp/issues/6849