Deep Learning Profiler 20.09 Release Notes

Description

DLProf release for 20.09, available in the NVIDIA TensorFlow 1.x, TensorFlow 2.x, and PyTorch NGC containers.

Driver Requirements

Release 20.09 is based on NVIDIA CUDA 11.0.3, which requires NVIDIA Driver release 450 or later. However, if you are running on Tesla (for example, T4 or any other Tesla board), you may use NVIDIA driver release 418.xx or 440.30. The CUDA driver's compatibility package only supports particular drivers. For a complete list of supported drivers, see the CUDA Application Compatibility topic. For more information, see CUDA Compatibility and Upgrades.

New Features

The key features of DLProf v0.15.0 / r20.09 are:

• Released in the TensorFlow 1.x 20.09, TensorFlow 2.x 20.09 and PyTorch 20.09 NGC container
• Latest DLProf build is based on TensorFlow 1.15.2, TensorBoard 1.15.0, PyTorch 1.6.0, and Nsight Systems 2020.3.2.
• A stand-alone version of DLProf is available as a PIP wheel file.
• Expert Systems is now able to auto-detect and recommend an ideal iterations range
• PyTorch features:
  • Expert Systems detects the use of slow debugging API calls and recommends that these calls not be used.
  • DLProf recognizes the profiler.start and profiler.stop parameters in PyTorch scripts to set specific profile ranges.

Known Issues

• This software is only accessible in the NGC TensorFlow and PyTorch containers and as a separate PIP wheel in the nvidia-pyindex.
• This software is only supported for TensorFlow 1.15, and PyTorch 1.6 and TensorBoard 1.15.
• Partial simple mode profiling is supported for TensorFlow 2.

Resolved Issues

• None