BentoML

Hi <@U03UY1JB269>
thanks for your attention

In addition, I would like to draw your attention to the fact that you send the request to remote runner by `headers`.
and with request with such a length of headers - is invalid!

Please find the file with example attached.

hi <@U03UY1JB269>

We find that the problem happens when we use the input type as `ndarray`. The Bento service sends the data to the remote runner using `headers`. However, if we use the input type as a `pandas DataFrame`, the data sent from Bento to the remote runner is in the body of the request, and it works fine even with large datasets.

Please find the `notebook` file attached, containing a full simulation of the problem.