Installation and Verification

Install TensorFlow 2.21.0 (GPU version).

python3 -m pip install "tensorflow[and-cuda]==2.21.0"

Configure the path of the CUDA runtime library installed using the NVIDIA pip package.

export LD_LIBRARY_PATH=/usr/lib64:\
/usr/local/lib/python3.11/site-packages/nvidia/cublas/lib:\
/usr/local/lib/python3.11/site-packages/nvidia/cuda_cupti/lib:\
/usr/local/lib/python3.11/site-packages/nvidia/cuda_nvrtc/lib:\
/usr/local/lib/python3.11/site-packages/nvidia/cuda_runtime/lib:\
/usr/local/lib/python3.11/site-packages/nvidia/cudnn/lib:\
/usr/local/lib/python3.11/site-packages/nvidia/cufft/lib:\
/usr/local/lib/python3.11/site-packages/nvidia/curand/lib:\
/usr/local/lib/python3.11/site-packages/nvidia/cusolver/lib:\
/usr/local/lib/python3.11/site-packages/nvidia/cusparse/lib:\
/usr/local/lib/python3.11/site-packages/nvidia/nccl/lib:\
/usr/local/lib/python3.11/site-packages/nvidia/nvjitlink/lib:${LD_LIBRARY_PATH:-}

Run the following commands to verify the TensorFlow version, GPU device visibility, and GPU tensor computation.

python3 - <<'PY'
import tensorflow as tf

print("tensorflow_version=" + tf.__version__)
print("physical_gpus=" + repr(tf.config.list_physical_devices("GPU")))
print("built_with_cuda=" + str(tf.test.is_built_with_cuda()))
assert tf.__version__ == "2.21.0"
assert tf.config.list_physical_devices("GPU") 
assert tf.test.is_built_with_cuda()  
with tf.device("/GPU:0"):     
    a = tf.constant([[1.0, 2.0], [3.0, 4.0]])     
    b = tf.constant([[1.0, 2.0], [3.0, 4.0]])     
    c = tf.matmul(a, b)  

print("tensorflow_gpu_matmul=" + str(c.numpy().tolist())) 
PY

The expected output is as follows.

tensorflow_version=2.21.0 
physical_gpus=[PhysicalDevice(name='/physical_device:GPU:0', device_type='GPU')] 
built_with_cuda=True 
tensorflow_gpu_matmul=[[7.0, 10.0], [15.0, 22.0]]

Parent topic: Installation Guide