New gpt-4-vision-preview is pretty bad compared with older versions

[DavidGOrtega]

LocalAI version:
2.14.0

Environment, CPU architecture, OS, and Version:
Linux Ubuntu SMP PREEMPT_DYNAMIC x86_64 x86_64 x86_64 GNU/Linux

90GB RAM 22 vcores
nvidia L4 24GB

Describe the bug
Previous versions of LLava where able to even understand memes, now it always returns a pretty silly or nor sense responses.

[localai-bot]

Hello DavidGOrtega, I'm a Github issue auditing bot, an experiment by @mudler. I've noted your concern about the changes in the gpt-4-vision-preview in the LocalAI project. Thank you for bringing this to our attention. I'll make sure that your issue (issue number 2261) is marked as [bug unconfirmed]. I'll also pass along your observations to the appropriate team so they can investigate the issue further. If you have any further information or questions, feel free to reach out. Thank you for using LocalAI and contributing to its improvement.