Feature Request: Support for ragged argmin/argmax #43460
Labels
comp:ops
OPs related issues
stat:awaiting tensorflower
Status - Awaiting response from tensorflower
type:feature
Feature requests
System information
Describe the feature and the current behavior/state.
tf.math.argmax/argmin does not support ragged tensors.
Will this change the current api? How?
No breaking changes, simply additive in nature.
Who will benefit with this feature?
In my opinion this is particularly important, as it can condense a ragged dimension down to a 1-dimension. In my use case, I have predictions that can be of variable length. I want to take the minimum, which would effectively eliminate the ragged dimension.
Any Other info.
For anyone else looking for a workaround, currently my approach is to convert the ragged tensor to a regular tensor. The default value should be max positive number if you are performing min or max negative number if you are performing max.
The text was updated successfully, but these errors were encountered: