In actual environments, it is impossible to fix the hardware from the beginning, so it may be necessary to switch the recognition hardware during the process. I have checked that the current version of OnnxRuntime is different, and in order to resolve conflicts, they are all modularized. However, it does not meet the actual application requirements. It is impossible to configure multiple software sets to cope with different hardware recognition. Can we create a Telegram group to discuss how to optimize and unify the next version
In actual environments, it is impossible to fix the hardware from the beginning, so it may be necessary to switch the recognition hardware during the process. I have checked that the current version of OnnxRuntime is different, and in order to resolve conflicts, they are all modularized. However, it does not meet the actual application requirements. It is impossible to configure multiple software sets to cope with different hardware recognition. Can we create a Telegram group to discuss how to optimize and unify the next version