Natural language processing

https://github.com/huggingface/transformers/blob/546dc24e0883e5e9f5eb06ec8060e3e6ccc5f6d7/src/transformers/models/gpt2/modeling_gpt2.py#L698

Assertions can't be relied upon for control flow because they can be disabled, as per the following:

$ python --help
usage: python [option] ... [-c cmd | -m mod | file | -] [arg] ...
...
-O     : remove assert and __debug__-dependent statem

Describe the bug
A harmless but distracting warning is raised when prefetch is passed as argument to the Flow.

Reproduce

from jina import Flow, Document
f = Flow(prefetch=10).add()
with f:
     f.index(inputs=Document())

Is your feature request related to a problem? Please describe.
I typically used compressed datasets (e.g. gzipped) to save disk space. This works fine with AllenNLP during training because I can write my dataset reader to load the compressed data. However, the predict command opens the file and reads lines for the Predictor. This fails when it tries to load data from my compressed files.

Maybe add something like this as a pre or post processing step?

It might make sense to download the emoji list and store it as part of the build, so people do not need to load the emojis module, ...

import emoji
from emoji import unicode_codes
import re


EMOJI_UNICODE = unicode_codes.EMOJI_UNICODE['en']
emojis = sorted(EMOJI_UNICODE.values(), key=len, reverse=True)
print (emojis