dtensorflow && tensorflow out of memory

[yy270]

Hello, I recently encountered an issue when using the dpgen program to call the deepmd program for model training on a GPU 4090. I set up two GPUs, each with around 24GB of memory, but during training, it indicates that the total memory is approximately 24GB and reports an out-of-memory error. I checked the GPU memory usage, and everything seemed fine.Python 3.8 version, DeepMD 2.2.11, TensorFlow 2.13.1 version