When I try to create an empty tensorflow model, I notice that the VRAM consumption of my GPU increases from less than 1GB to 6.5GB.
This is problematic as it limits the capacity of the models I want to train later.
import tensorflow as tf
model = tf.keras.Sequential()
Hi @CapeItalist, Generally the empty model will not consume that much memory. I have verified in colab, Before defining the model the GPU ram utilization was at zero
After defining the empty model the GPU ram utilization went to 0.1.
The high consumption of ram in your case may be due to some other processes running in your system. Thank You.
