Skip to content

TensorFlow device configuration fails after importing tensorflow_quantum #911

@aapo-kossi

Description

@aapo-kossi

TensorFlow device configuration fails after importing tensorflow_quantum

With the current version of tfq, importing the package before setting TensorFlow device configuration (e.g., enabling memory growth) results in a runtime error:

RuntimeError: Physical devices cannot be modified after being initialized

Expected Behavior
I would expect that importing tensorflow_quantum should not need to trigger device initialization yet, since I have not at this point created any tensors on any device.

Reproducer

import tensorflow as tf
import tensorflow_quantum as tfq

gpus = tf.config.list_physical_devices('GPU')
tf.config.experimental.set_memory_growth(gpus[0], True)  # RuntimeError here

Workaround
Set device configuration before importing tensorflow_quantum. This is fine for me, but since this is unexpected behaviour to me I am documenting it here.

Environment

  • TensorFlow version: 2.15.0.post1
  • TensorFlow Quantum version: 0.7.3
  • Platform: Linux

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions