TI Deep Learning Product User Guide
|
TIDL is packaged with SDK and already installed during SDK installation. It can be found inside ${PSDKRA_PATH}/tidl_j7_08_00_01_00
. Last few characters can be different depending on the version and device. User shall perform following steps to export required variables and install tools for ease of use.
Export the path of TIDL root directory as
TIDL_INSTALL_PATH
. All examples in this guide will use paths and filenames relative to this directory.Linux users:
user@ubuntu-pc$ export TIDL_INSTALL_PATH=${PSDKRA_PATH}/tidl_j7_08_00_00_00
Windows users:
C:\> set TIDL_INSTALL_PATH=${PSDKRA_PATH}/tidl_j7_08_00_00_00
If you want to visualize the network after importing, it is required to build tidlModelGraphviz
tool after installation. The visualization tool generates .svg
files indicating the network flow. The details of the network visualization tool can be found here. Please note that this tool is only supported for Linux environment and not available on windows.
To build
tidlModelGraphviz
tool, installgraphviz-dev
package and export the variableTIDL_GRAPHVIZ_PATH
user\@ubuntu-pc\$ sudo apt install graphviz-dev user\@ubuntu-pc\$ export TIDL_GRAPHVIZ_PATH=/usrBuild the
tidlModelGraphviz
tool by runningmake
inTIDL_INSTALL_PATH
.user\@ubuntu-pc\$ cd ${TIDL_INSTALL_PATH} user\@ubuntu-pc\$ TARGET_PLATFORM=PC make gv
All the required software componets by TIDL-RT are packaged with SDK except below external components
Software component | Version | Purpose |
---|---|---|
Google Protobuf | 3.11.3 | Only for model compilation on X86 |
Google FlatBuffers | 1.12.0 | Only for model compilation on X86 |
Open CV | 4.1.0 | Only for TIDL product Test bench both on X86 and cortex-A on TI device |
TIDL Build Instructions describes how to install these components to build TIDL-RT.
Note: TIDL-RT package comes with pre-built binaries and libraries, so you need to download these components only if you need to re-build the TIDL-RT software.
Note: Open CV dependency arise because some of the examples in test application of TIDL product make use of functions from open cv to perform image processing operations. For an application to use TIDL product, it doesn't need to depend on openCV as long as the application can provide the required tensor input to TIDL-RT