Update functional.py

[liyown]

Optimized for slow segmentation of data when the Dilliclet coefficient is small，According to this issue #230, but it might be difficult to replicate the results reported in the original paper.

[liyown]

In the code, if the number of samples for a category is 0, it will result in a process of randomly re-selecting the category. The code adds an exception handling mechanism to ensure that sample selection continues even if the number of samples for a category is 0. For example, when the sample number of a class is 0, a class with a non-zero sample number is randomly selected for sample selection.

[AgentDS]

Hi, Sorry for the late reply.
Could you please rename the new version of the function as client_inner_dirichlet_partition_faster, and keep the old version of the function with the old name? so we can keep both versions for this function.

[liyown]

Modified

[AgentDS]

Thank you very much!