So I decided to move over to Ubuntu 20.04 from Windows 10 since it's a better environment that's widely used in industry. I did encounter some problems when I cloned my clean Windows code from Github I immediately encountered issues when trying to run it. For context, my code utilizes Tensorflow 2.2.0 to segment images. I'm training from scratch with a dataset. The problem immediately occurs when I ran my train.py program. Initially there was an error which was fixed by inserting the following two lines immediately after importing Tensorflow (after verifying that Tensorflow could see and access my GPU on Ubuntu)
gpus = tf.config.experimental.list_physical_devices('GPU')
tf.config.experimental.set_memory_growth(gpus[0], True)
Following this, I get OOM errors despite being able to run the same code and dataset on my Windows OS. I was able to run the program once when I initially cloned the repo but the day after, I got these OOM errors and have been encountering them since. The code still works on Windows. The following log file shows what I got when I ran in Ubuntu 20.04 on the same machine I house my Windows on (I am dual booting). I installed TF GPU support exactly as they outlined with the versions they mentioned. The only part that worries me is that TF doesn't have Ubuntu 20.04 support, which I speculate may be causing this issue but as a new Ubuntu user, I can't be entirely sure. The output of the Terminal and the train.py program are linked below.
https://drive.google.com/drive/folders/1GRkqCwwdnoPWzsPklq2NIS82P1bFfRr1?usp=sharing
Relevant specs:
GPU - NVIDIA 2070 Super
CPU - Ryzen 3600
RAM - 32 GB
Tensorflow - 2.2.0
NVIDIA Driver - 451.x
CUDA - 10.1
cuDNN - 7.6.5
Does anyone have any insight into what could be causing this issue?