Hi Mark, I was using this technique that you just mention in this video, which to me, is more of metaprompt with chain of thoughts. The images I sent into the LLM is more of a continuous images, extracted from PDF, and chain thoughts with past prompts and answers from the LLM is essential to me to recreate the information from the images. However, chaining images and prompts are very challenging, and the entire process is tedious and time consuming. The output has to be embed within a master json string template that I need the vision model to answer. Sometimes, a few images add up to only 1 json string output that I am looking for. Still I need to aggregate every answer from the vision model into a dictionary of json string and filter out those partial json string in order to get the final output that I am look for. I hope you understand what I am trying to achieve here. If you have a better approach to my task, please give me some advice, or another video of yours that can help to address this challenge of mine, thanks
@@lmek1624 do you have any models that you heard good things about? I've seen a few posted about the place, but not sure which one is best to start with
![BLACK BAG - Official Trailer [HD] - Only in Theaters March 14](http://i.ytimg.com/vi/Du0Xp8WX_7I/mqdefault.jpg)
Thank you.
How to pass two images for comparison in a single prompt to llama vision instruct? Eg: gemini supports multi image prompts
Hi Mark, I was using this technique that you just mention in this video, which to me, is more of metaprompt with chain of thoughts. The images I sent into the LLM is more of a continuous images, extracted from PDF, and chain thoughts with past prompts and answers from the LLM is essential to me to recreate the information from the images. However, chaining images and prompts are very challenging, and the entire process is tedious and time consuming. The output has to be embed within a master json string template that I need the vision model to answer. Sometimes, a few images add up to only 1 json string output that I am looking for. Still I need to aggregate every answer from the vision model into a dictionary of json string and filter out those partial json string in order to get the final output that I am look for. I hope you understand what I am trying to achieve here. If you have a better approach to my task, please give me some advice, or another video of yours that can help to address this challenge of mine, thanks
Hey Mark, right now tts is very very hot. Could you do a locally cpu text-to-speech tutorial?
Yea we were waiting for this but he ignores it lol
@@lmek1624 do you have any models that you heard good things about? I've seen a few posted about the place, but not sure which one is best to start with
Forget it, he doesnt ever reply zz
sometimes I do, I'm just very good/bad at procrastinating!