Closed Rastapar closed 4 months ago
Correct use of LLM
A sentence can have many phrases. List in the transcript per phrase, not sentence.
When talking about some error, go to the phrase mentioned in the chat. Create a function that just highlights that wrong phrase and scrolls automatically to it.
The scroll feature would take more time than expected so I will leave it for later The LLM is quantized but even so it takes more than 4GB of GPU VRAM
Highlight the user sentences when selected in the dropdown.