|
- Ollama is making entry into the LLM world so simple that even school . . .
Ollama doesn't hide the configuration, it provides a nice dockerfile-like config file that can be easily distributed to your user This philosophy is much more powerful (it still needs maturing, tho)
- How does Ollama handle not having enough Vram? : r ollama - Reddit
I have been running phi3:3 8b on my GTX 1650 4GB and it's been great I was just wondering if I were to use a more complex model, let's say Llama3:7b, how will Ollama handle having only 4GB of VRAM available? Will it revert back to CPU usage and use my system memory (RAM) Or will it use both my system memory and GPU memory?
- Request for Stop command for Ollama Server : r ollama - Reddit
Ok so ollama doesn't Have a stop or exit command We have to manually kill the process And this is not very useful especially because the server respawns immediately So there should be a stop command as well Edit: yes I know and use these commands But these are all system commands which vary from OS to OS I am talking about a single command
- r ollama - Reddit
How good is Ollama on Windows? I have a 4070Ti 16GB card, Ryzen 5 5600X, 32GB RAM I want to run Stable Diffusion (already installed and working), Ollama with some 7B models, maybe a little heavier if possible, and Open WebUI
- How to add web search to ollama model : r ollama - Reddit
[Ollama WIP Project Demo] Stop paying for CoPilot Chat GPT, ollama + open models are powerful for daily
- r ollama on Reddit: Does anyone know how to change where your models . . .
OLLAMA_ORIGINS A comma separated list of allowed origins OLLAMA_MODELS The path to the models directory (default is "~ ollama models") OLLAMA_KEEP_ALIVE The duration that models stay loaded in memory (default is "5m") If you installed ollama the automatic way as in readme: open the systemd file
- Whats your go-to UI as of May 2024? : r LocalLLaMA - Reddit
Open-WebUI (former ollama-webui) is alright, and provides a lot of things out of the box, like using PDF or Word documents as a context, however I like it less and less because since ollama-webui it accumulated some bloat and the container size is ~2Gb, with quite rapid release cycle hence watchtower has to download ~2Gb every second night to
- Training a model with my own data : r LocalLLaMA - Reddit
I'm using ollama to run my models I want to use the mistral model, but create a lora to act as an assistant that primarily references data I've supplied during training This data will include things like test procedures, diagnostics help, and general process flows for what to do in different scenarios
|
|
|