PyTorch is a robust, open source machine learning framework that supports basic mathematical and tensor operations. You can run PyTorch to achieve various types of deep learning. CPU optimization is implemented by using multithreading, vectorization, and the oneAPI Deep Neural Network (oneDNN) library, which accelerates model training and inference without affecting model accuracy. The Zen Deep Neural Network (ZenDNN) library is a neural network acceleration library that is optimized for the AMD CPU architecture to improve deep learning inference performance. The PyTorch-AMD image is optimized for AMD CPUs and contains PyTorch and the ZenDNN library. The PyTorch-AMD image also uses TCMalloc instead of the default ptmalloc (pthreads malloc) to provide ready-to-use high-performance PyTorch images for deep learning research and practice.
Images
|
Image |
Image address |
|
pytorch-amd |
ac2-registry.cn-hangzhou.cr.aliyuncs.com/ac2/pytorch-amd:1.13.1-3.2304-zendnn4.1 |
Image content
pytorch-amd:1.13.1-3.2304-zendnn4.1
-
BaseOS: Alinux 3.2304
-
Python: 3.10.13
-
certifi: 2023.7.22
-
charset-normalizer: 3.1.0
-
idna: 3.4
-
libcomps: 0.1.19
-
numpy: 1.24.2
-
olefile: 0.46
-
Pillow: 9.4.0
-
pip: 23.3.1
-
PySocks: 1.7.1
-
requests: 2.31.0
-
setuptools: 65.5.1
-
six: 1.16.0
-
torch: 1.13.1
-
torchvision: 0.14.1+cpu
-
typing_extensions: 4.5.0
-
urllib3: 1.26.18
Operational requirements
The PyTorch-AMD image is integrated with ZenDNN optimization for the AMD CPU architecture. We recommended that you run the PyTorch-AMD image on AMD CPUs.
Important features
The pytorch-amd:1.13.1-3.2304-zendnn4.1 image is integrated with PyTorch 1.13.1. The image contains ZenDNN 4.1 and uses TCMalloc instead of the default ptmalloc. For more information about ZenDNN 4.1, see ZenDNN User Guide.
Updates
2024.01: Released the PyTorch-AMD image.