Description
DLProf Release for 21.08, available in the NVIDIA TensorFlow 1.x, TensorFlow 2.x, and PyTorch NGC containers, and as a Python Wheel on the NVIDIA PY Index.
Note:
NVTX-Plugins are being removed from the Deep Learning Frameworks TensorFLow 1.x & TensorFLow 2.x containers starting with the 21.09 release. They can still be manually installed from https://github.com/NVIDIA/nvtx-plugins.
Driver Requirements
- Requires DLProf SQLite database generated by DLProf v1.2 or later.
- Ensure that you have access and are logged into NGC. For step-by-step instructions, see the NGC Getting Started Guide.
- Install Docker and nvidia-docker. For DGX users, see Preparing to use NVIDIA Containers. For users other than DGX, see nvidia-docker installation documentation.
New Features
The key features of DLProf v1.4.0 / r21.08 are:
- Released in the TensorFlow 1.x 21.08, TensorFlow 2.x 21.08 and PyTorch 21.08 NGC container.
- Multi-GPU calculation and reporting enhancement.
- Added parameters and output shapes into the iteration report for GEMM and CONV operations.
Known Issues
- This software is only supported for TensorFlow 1.15, TensorFlow 2.34, PyTorch 1.9, and TensorRT 7.2.
- nvtx-plugins is being removed from the DLFW TensorFlow 1 & TensorFlow 2 containers starting with the 21.09 release.
Resolved Issues
- None.