limits in the demo.
I'm doing a complete test seeing everything I could do. but it has a text output limit. and there are issues that cannot be proven. for example when trying to generate a report. It tries to generate it long and fails to generate it completely. this limits testing to the model. Could you extend the output text a little more?
Not clear what you are asking. Models generate until they think they should stop in general. Are you sure that's not simply what's happening? What kind of length are you expecting
It has a text output limit that is indicated by a sign in the chat. that I reached the limit of text to generate per message. For example, if I ask you to generate a long story with a few words. This does not finish generating and the following message appears.
(Reached maximum output tokens for DBRX Playground). The same thing happens if I ask you for a detailed report on programming codes. When you have to write a long output message it stops and doesn't finish it.
That's a safety valve on the playground, not the model in this repository itself.
That's why I was asking if they could extend it to perform more complex tests. in this case the output text. I see that the model itself is quite optimized compared to other models. That's why I try to test the demo in a more complex way. If it might be possible to extend the output text.
Thanks @Codigoabierto for raising this, we are unable to raise the output text limit unfortunately, as the HF space is really meant for demo purposes. The model is not available on other providers as well (e.g. https://labs.perplexity.ai/), so you can also try them there. Or, if you are an enterprise customer, it's also available on Databricks through our Model Serving API.