Faulty boolean question in the "Structured data prediction using Vertex AI Platform" lab?
MrCsabaToth opened this issue · comments
The "Structured data prediction using Vertex AI Platform" lab (https://www.cloudskillsboost.google/course_sessions/2920302/labs/394190) has the following question:
Currently as far as Tensorflow is concerned, the only supported GPUs are that of NVIDIA.
True
False
The lab claims the True selection to be the correct one (a.k.a. TensorFlow is nVidia exclusive only), however this is false.
Here is what Bard says:
"False. TensorFlow currently supports GPUs from other vendors, including AMD and Intel. However, NVIDIA GPUs are still the most popular choice for TensorFlow, as they offer the best performance.
Here are the supported GPUs for TensorFlow:
- NVIDIA GPUs with CUDA® architectures 3.5, 5.0, 6.0, 7.0, 7.5, 8.0 and higher.
- AMD GPUs with ROCm® 3.9 and higher.
- Intel GPUs with oneAPI Deep Neural Network Library (oneDNN) 2.6 and higher.
TensorFlow does not currently support GPUs from the following vendors:
- ARM
- Qualcomm
- Samsung
It is possible that TensorFlow will support GPUs from other vendors in the future."
Here is how to use TensorFlow with an AMD ROCm GPU: https://rocm.docs.amd.com/en/latest/how_to/tensorflow_install/tensorflow_install.html
Maybe I interpret "supported" too loosely. Because Bard also notes that only nVidia is guaranteed to work, but as we see above it's possible with AMD and Intel as well.