Olivia Lee

Ok, now the problem is both llama and mistral are failing the compile static cache tests on A10 because of the same error: ``` ============================= FAILURES SHORT STACK ============================= ________________...

> Ok, now the problem is both llama and mistral are failing the compile static cache tests on A10 because of the same error: > > ``` > ============================= FAILURES...

I merged from current main and again did slow tests on my dev and aws a10 machine, I believe this PR is good to merge now @ArthurZucker @ydshieh

Please merge this when available @ArthurZucker , for I don't have write access to the library

according to the official documents, it could only be saved as h5 or tf format, how it could be saved as pb format

I even tried to convert h5 to pb but when I use model.save() to save the file as "model.h5" , and later try to load using load_model("model.h5"), some error occurs....

After all pb file is more important for deployment and acceleration