I've been using local LLMs for privacy reasons and tried Ollama and LMStudio
The models when imported into Ollama will be split into tens of files and their names changed into sha256-babble. This makes it impossible to find out which version was imported after some time has passed, and locks it to Ollama (can't be transferred or decrypted). I used the CLI to check and it said dolphin-mistral:latest
Yeah, latest but when was it last updated? How tf do I know what latest is supposed to mean and what are the parameters? I checked on other forums and apparently Ollama has a reputation for misnaming models.
Also, you can't change system prompts without going through a convoluted process of cloning the model and AFAIK you can't anything other than the prompt and temperature anyway. Not sampling, response length... Why would anyone use this over LMStudio? I mean the latter isn't perfect but at least they try
I don't know if it's my fault honestly I wasted so much time poking around the files and now it seems like everything was deliberately obfuscated.