Release 18.05.02

This patch release allows building the ArmNN driver 18.05 on Android P, still using HAL 1.0.

Note to consumers: when using the setup script provided (setup.sh) be aware that it will download the latest version available of both Arm NN and the Compute Library.
If you are using an earlier version of the driver you will have to manually check out the correct version of the dependency libraries, as backward compatibility is not guaranteed.

Release 18.05

This release of Arm NN integrates the latest Compute Library and adds additional NNAPI support.

New Features:

• In general, the amount of RAM needed for a loaded network has been reduced by 20-30% compared to release 18.03.
• The MUL operation now supports operands of different shapes.
• The CONV2D operation supports more shapes and padding options when used with QUANT8_ASYMM tensors.
• The latest 8-bit quantized operations from Compute Library have been integrated. In testing, 8-bit quantized mobilenets models are 3x faster compared to release 18.03.

Other changes:

• This release of ArmNN requires at least release 18.05 of the Compute Library.
• Fixed an issue where pooling operations with different pooling width and height would produce the wrong output.
• The CONCAT layer is now supported for all concatenation dimensions for rank 4 tensors
• Overall efficiency has been improved by reducing the number of reshapes required for intermediate tensors

Known issues:

• CONCAT is not supported for rank 1 and rank 2 tensors. This restriction will be lifted in a future release.

Release 18.03

v18.03

Release 18.03

Release 18.02

Release 18.02