Simplest way(TensorFlow 2.2+)
import tensorflow as tfgpus = tf.config.experimental.list_physical_devices('GPU')for gpu in gpus: tf.config.experimental.set_memory_growth(gpu, True)
Or set environment variable
set
TF_FORCE_GPU_ALLOW_GROWTH
to
true
.
if TensorFlow 2.0 and 2.1
import tensorflow as tftf.config.gpu.set_per_process_memory_growth(True)
Source:
- https://www.tensorflow.org/guide/gpu#limiting_gpu_memory_growth
- https://stackoverflow.com/questions/34199233/how-to-prevent-tensorflow-from-allocating-the-totality-of-a-gpu-memory/34200194#34200194