Computes a tensor such that
\\(output_i = \max_j(data_j)\\) where `max` is over `j` such
that `segment_ids[j] == i`.

If the max is empty for a given segment ID `i`, `output[i] = 0`.

Caution: On CPU, values in `segment_ids` are always validated to be sorted,
and an error is thrown for indices that are not increasing. On GPU, this
does not throw an error for unsorted indices. On GPU, out-of-order indices

contrary way, so as to make the Rays fall more obliquely on the first
refracting Surface, the Image soon became an Inch or two shorter. And
therefore in trying this Experiment, I was as curious as I could be in
placing the Prism by the above-mention'd Rule exactly in such a Posture,
that the Refractions of the Rays at their Emergence out of the Prism
might be equal to that at their Incidence on it. This Prism had some

t-area-height solid transparent;\n    border-bottom: $carousel-indicator-hit-area-height solid transparent;\n    opacity: .5;\n    @include transition($carousel-indicator-transition);\n  }\n\n  .active {\n    opacity: 1;\n  }\n}\n\n\n// Optional captions\n//\n//\n\n.carousel-caption {\n  position: absolute;\n  right: (100% - $carousel-caption-width) / 2;\n  bottom: 20px;\n  left: (100% - $carousel-caption-width) / 2;\n  z-index: 10;\n  padding-top: 20px;\n  padding-bottom: 20px;\n  color: $carou...

t-area-height solid transparent;\n    border-bottom: $carousel-indicator-hit-area-height solid transparent;\n    opacity: .5;\n    @include transition($carousel-indicator-transition);\n  }\n\n  .active {\n    opacity: 1;\n  }\n}\n\n\n// Optional captions\n//\n//\n\n.carousel-caption {\n  position: absolute;\n  right: (100% - $carousel-caption-width) / 2;\n  bottom: 20px;\n  left: (100% - $carousel-caption-width) / 2;\n  z-index: 10;\n  padding-top: 20px;\n  padding-bottom: 20px;\n  color: $carou...