Efficient allreduce is not supported for 2 IndexedSlices

Asked Jul 02 '19 at 00:22

Active Jul 02 '19 at 16:06

Viewed 932 times

I am trying to run a Subclassed Keras Model on multiple GPUs. The code is running as expected, however, the following "warning" crops up during the execution of the code:

"Efficient allreduce is not supported for 2 IndexedSlices"

What does this mean? I followed the Multi-GPU tutorial on Tensorflow 2.0 Beta guide. I am also using the Dataset API for my input pipeline.

edited Jul 02 '19 at 16:06

asked Jul 02 '19 at 00:22

Damayanti Sengupta

1

Did you discover this in the end? :) – Roelant Jan 20 '20 at 13:10
anyone from Tensorflow team can answer this? – thinkdeep Jan 08 '21 at 23:04
This might be a relevant thread: https://github.com/tensorflow/tensorflow/issues/41898 – S. P Mar 05 '22 at 05:12

Efficient allreduce is not supported for 2 IndexedSlices

0 Answers0

Linked