added unsorted_segment_min to ivy functional API

[akshatvishu]

Close #17437

[zhumakhan]

Hi, thanks for your PR! Please use with_unsupported_dtypes decorator instead of validating data types inside the function.

[akshatvishu]

Hi, thanks for your PR! Please use with_unsupported_dtypes decorator instead of validating data types inside the function.

Hey @zhumakhan , thanks for the comment! Can you tell me how to use it for a particular argument ? Because num_segments and segment_ids can be only of int32/int64 regardless of the backend while the data can be float , int and at some backends unsigned_integers too is accepted as supported dtype.

[zhumakhan]

Yeah, you are right, decorators check all inputs. Maybe you can try something like this: https://github.com/unifyai/ivy/blob/master/ivy/functional/ivy/general.py#L3163. If you look at its backend implementations they used (e.g. in torch backend) ivy.utils.assertions.check_gather_input_valid(params, indices, axis, batch_dims) to check inputs

[akshatvishu]

Yeah, you are right, decorators check all inputs. Maybe you can try something like this: https://github.com/unifyai/ivy/blob/master/ivy/functional/ivy/general.py#L3163. If you look at its backend implementations they used (e.g. in torch backend) ivy.utils.assertions.check_gather_input_valid(params, indices, axis, batch_dims) to check inputs

Thank you for your insightful suggestion, @zhumakhan !! I found it very useful and have incorporated a similar approach into all backend implementations. I value your feedback, so please don't hesitate to share more suggestions or highlight any other areas where you think improvements could be made.😁

[zhumakhan]

LGTM!

[akshatvishu]

LGTM!

Thanks Zuma for the review!