The NGC containers v18.06 for popular deep learning frameworks are now available at https://ngc.nvidia.com
These containers have support for:
- Python 2 and/or Python 3
- CUDA 9
- cuBLAS v9.0.333
- cuDNN 7.1.4
- NCCL 2.2.13
- Caffe2 0.8.1
- NVCaffe 0.17.0
- DIGITS 6.1.1
- Microsoft Cognitive Toolkit 2.5
- Inference Server 0.3
- MXNet 1.2.0
- PyTorch 0.4.0
- TensorFlow 1.8.0
- TensorRT 4.0.1
- Theano 1.0.1
- Torch7
- Ubuntu 16.04 with May 2018 updates
For the optimized framework release notes, see http://docs.nvidia.com/deeplearning/dgx/.
For the DIGITS release notes, see NVIDIA Deep Learning DIGITS Documentation.
For the SDK release notes, see https://docs.nvidia.com/deeplearning/sdk/index.html.
To view the support matrix across all optimized containers, see: https://docs.nvidia.com/deeplearning/dgx/support-matrix/index.html.