Model not loading - nothing happens

[AlphaCapella]

Prerequisites

Please answer the following questions for yourself before submitting an issue.

• [X ] I am running the latest code. Development is very rapid so there are no tagged versions as of now.
• [X ] I carefully followed the README.md.
• [X ] I searched using keywords relevant to my issue to make sure that I am creating a new issue that is not already open (or closed).
• [ X] I reviewed the Discussions, and have a new bug or useful enhancement to share.

Expected Behavior

Trying to get llama.cpp to run on windows 10

Current Behavior

Generally everything seems to work, but the model does not load. After it tries to load the model, it just exits after a few seconds and that is it. No error, message, nothing. RAM gets loaded for a few seconds, same with CPU and GPU.
I have tried every binary available, always with the same outcome. Also the two models I have tried are: ggml-vicuna-7b-1.1 and ggml-vicuna-13b-1.1

If I try for example ".\main.exe --help" the output is correct. So it seems to work in principle, also my GPU gets detected if the binary is with clblast.
Below I attached the console input and output. I am really out of words. Hours of googling did not seem to yield anything, because there is apparently nobody with that problem.

Environment and Context

• 32 GB of RAM
• AMD Ryzen 9 5950x
• Nvidia RTX 3090

[SlyEcho]

The file format has changed for Q4_0, Q4_1 and Q8_0. The files on that repo are too old.

What you're seeing is the program silently crashing. True, that's not something desirable, maybe an issue on its own.

Q5_0 and Q5_1 models should still work as before.

It was changed here: #1508

[AlphaCapella]

Interesting, I was using the model "ggml-vic7b-uncensored-q4_0.bin", which has been updated for the new quantization. Apparently not correctly, because after what you said I tried the q5 version and it works like a charm. Happenend with the 7b and 13b model.

Anyways, thank you for your help.

[SlyEcho]

Yeah, it was updated, but then the format changed again :)

[VioletFlare]

I just checked out an older version that dates back to september 11 so that i can use older GGML models :)