I have a video transcription with timestamps, and I want to ask llama2 questions and as an answer to get an index of transcription line or list of lines (prefereably subsequent) so that playback timestamp could be retrieved - the goal is to seek to a video timestamp based on questions.
I’ve managed to run Llama-2-7b-instruct-hf (half precision on RTX 4090) in chat mode and wonder if there is a way to pass transcript data in a way to reliable retrieve referenced transcript lines (or line indeces).