Open snova-zoltanc opened 1 week ago
It looks like as we scale the number of workers, the tokenization time does decrease linearly with number of workers like it should. Lets look into how much time is being wasted by these locks and try to minimize that
It looks like as we scale the number of workers, the tokenization time does decrease linearly with number of workers like it should. Lets look into how much time is being wasted by these locks and try to minimize that