CUDA out of memory error about m2i

[Karami-m]

The problem was solved by adding the following to the run.py file:

import tensorflow as tf  
gpus = tf.config.experimental.list_physical_devices('GPU')  
for gpu in gpus:  
    tf.config.experimental.set_memory_growth(gpu, True)

[larksq]

We used NVIDIA 3080 or sometimes the NVIDIA A10 for training. Usually changing the train_batch_size to a smaller value helps with the out-of-memory error problem. I am glad you found a solution that worked for you, hence your error might be related to a different version of TensorFlow.