Working in text-geration-webui, with a little surprise...

#8
by mwettach - opened

For the model to load in text-generation-webui, I had to remove these 2 lines from quantize_config.json:
"model_name_or_path": null,
"model_file_base_name": null
I probably need to update auto-gptq.

In a first test I asked the model:
Write a vba function count_vowels(text) that counts the vowels in a string. It should recognize and count both uppercase and lowercase vowels.

The response was a VBA macro which upon first sight seems to do the requested count, but I would have needed to rewrite it as a function.

As a surprise, I got back an additional text with a return question:

  My question would be how can I use this code?

Isn't this cheeky? :-)

After coming up with the following prompt, the response was completely as expected.

I want to write a VBA FUNCTION count_vowels(text) that counts the vowels in a string and returns the number of vowels as an integer value. It should recognize and count both uppercase and lowercase vowels. The function would look like this:

We discussed this in the other thread - yes you need to update AutoGPTQ. And yes this model works fine in text-generation-webui, as long as the AutoGPTQ loader is used.

Sounds good re the prompting!

TheBloke changed discussion status to closed
TheBloke changed discussion status to open

Sign up or log in to comment