Closed hasan-alj88 closed 1 year ago
Thanks for bringing this to our attention. I think after updating the Nvidia drivers and CUDA, something broke in our TensorFlow environments. We will look into a fix for this issue.
@hasan-alj88 the issue should be fixed now on the base Conda environment and on a newly created environment called py3.11-tf2.12
.
You can test that on the cli:
# base environment
conda activate
python3 -c "import tensorflow as tf; print(tf.config.list_physical_devices('GPU'))"
# py3.11-tf2.12 environment
conda activate py3.11-tf2.12
python3 -c "import tensorflow as tf; print(tf.config.list_physical_devices('GPU'))"
Or you can run the following cell in a notebook:
import tensorflow as tf
print(tf.config.list_physical_devices('GPU'))
The output should be a list of available GPUs, similar to the following:
[PhysicalDevice(name='/physical_device:GPU:0', device_type='GPU'), PhysicalDevice(name='/physical_device:GPU:1', device_type='GPU')]
Please test and confirm.
the new conda does reginise GPU you may mark this issue resolved.
I am using tensorflow v2 to train my AI model. However, I cant connect to [conda env:tf2] kernel since it always disconnect every time. Note I tried to use [conda env:tf2-rapids-py39], However it does not use GPU which should be available in the instance I have selected thus the training is slow due to relaying on CPU only.
Kindly advice.