如何在所有其他请求中使用在第一个请求中创建的sklearn泡菜对象

Question

I have the following use case In my project- Python. We have a prediction system in place for every user request through API (We are using IIS, need to look at flask capabilities).

The size of a pickle file is huge, like 700MB, I don't want to load it for every request (since it is consuming more RAM for every request).

Can we load the pickle file into flask cache and use it for all the requests? (or) Can we share the data among all the requests, so that we don't need to load it again and again?

Please help with the sample implementation of sharing pickle data across all the requests

Answer 1

You can try using something like Flask-Cache , and do a function like this:

@cache.cached(timeout=None)
def get_pickle_object():
    return your_pickle_object

This way the pickle object will only be imported the first time this function is called and afterward it will be dumped from cache.

You can also call this once during app initialization so that it is already loaded for the first request.

Then every time you need the object, use get_pickle_object() instead of the object itself.