20B Models Not Working about misc-scripts HOT 5 OPEN

Which model did you try, which quantization method, and which API endpoint?

from misc-scripts.

Which model did you try, which quantization method, and which API endpoint?

I was trying TheBloke/Emerhyst-20B-GPTQ with this one. I set the quantization to GPTQ on the cell. I used the kobold api endpoint. I copied the kobold url and pasted it on SillyTavern using the Kobold Classic API. I get this same error message with every 20b model I attempt to use though.

from misc-scripts.

I can confirm that it doesn't work and stalls indefinitely (I don't receive any errors though). I can't test much right now, can you try increasing/decreasing gpu memory utilization and see if that does anything?

from misc-scripts.

I tried to increase gpu memory utilization, and I managed to get four messages before I was met with same error again or a different error that says "Internal server error." It said it in the cell as well.

Decreasing gpu memory utilization only tells me that there isn't enough memory and that I need to increase it. Another thing I tried was adjusting the response length by lowering it even more (such as 150 and then 100) but still I don't get any messages. It just loads indefinitely with no output.

from misc-scripts.

Adding this because I don't know if it means anything but once I stopped the cell, I received this:

from misc-scripts.