I was working on writing a custom training script for my keras GAN model training and when I start training or when I try generating images with my model, I get spammed with this message along with the intended output.
2024-01-23 05:31:31.705379: E ./tensorflow/compiler/xla/stream_executor/stream_executor_internal.h:124] SetPriority unimplemented for this stream.
almost 60 lines of this above message gets displayed when I use my model to train or evaluate or just generate. images I only encounter this problem when using using TPU. When using GPU I dont encounter this message.
Can anyone explain me this issue, why this arises and how to prevent it?
I am having the same issue. Did you find any solution?
The message “SetPriority unimplemented for this stream” you’re seeing is a common informational warning when using TensorFlow with TPUs, and it’s different from using GPUs. This is due to some features or methods in TensorFlow being implemented for GPUs but not for TPUs. If this message isn’t causing any actual failure in your model’s training or evaluation, it’s likely harmless. To reduce log clutter, you can adjust TensorFlow’s logging level to ERROR:
import tensorflow as tf
This will suppress non-error messages. Keep in mind that TensorFlow is regularly updated, so staying current with the latest version can also help resolve such issues.