diff --git a/docs/tidl_osr_debug.md b/docs/tidl_osr_debug.md index 7ff6a9c..133e017 100644 --- a/docs/tidl_osr_debug.md +++ b/docs/tidl_osr_debug.md @@ -46,7 +46,8 @@ As an example for ONNX out of box example script user can run in ARM only mode a ``` For user's custom model they can refer [here](../examples/osrt_python/README.md#example-apis-for-tidl-offload-or-delegation-in-osrts) to enable ARM only mode * User can set debug_level = 1 or 2 to enable verbose debug log during model compilation and during model inference -* If model infernece works fine in ARM only mode but model compilation fails with C7x-MMA offload, then try dispatching some of the layers (less commonly used layer type) to ARM by using “deny_list” option. +* If model inference works fine in ARM only mode but model compilation fails with C7x-MMA offload, then try dispatching some of the layers (less commonly used layer type) to ARM by using “deny_list” option. +* Additionally, it is recommended to disable default onnx graph optimizations (i.e. set session option for graph_optimization_level to onnxruntime.GraphOptimizationLevel.ORT_DISABLE_ALL) # Steps to Debug Error Scenarios for target(EVM/device) execution diff --git a/docs/version_compatibility_table.md b/docs/version_compatibility_table.md index f152250..5e40fdd 100644 --- a/docs/version_compatibility_table.md +++ b/docs/version_compatibility_table.md @@ -5,6 +5,7 @@ |EdgeAI TIDL Tools TAG | AM62 | AM62A | AM68A/J721S2 (TDA4AL, TDA4VL) | AM68PA/J721E (TDA4VM)| AM69A/J784S4(TDA4AP, TDA4VP,TDA4AH, TDA4VH)| | ---------------------------- |:--------------:|:---------------:|:--------------:|:--------------:|:-------------:| +| 09_01_07_00 | 09_01_00_08 | Processor SDK LINUX : 09.01.00.07 | Processor SDK LINUX 09.01.00.06
Processor SDK RTOS 09.01.00.06 | Processor SDK LINUX 09.01.00.06
Processor SDK RTOS 09.01.00.06 | Processor SDK LINUX 09.01.00.06
Processor SDK RTOS 09.01.00.06 | | 09_01_06_00 | 09_01_00_08 | Processor SDK LINUX : 09.01.00.07 | Processor SDK LINUX 09.01.00.06
Processor SDK RTOS 09.01.00.06 | Processor SDK LINUX 09.01.00.06
Processor SDK RTOS 09.01.00.06 | Processor SDK LINUX 09.01.00.06
Processor SDK RTOS 09.01.00.06 | | 09_01_04_00 | 09_01_00_08 | Processor SDK LINUX : 09.01.00.07 | Processor SDK LINUX 09.01.00.06
Processor SDK RTOS 09.01.00.06 | Processor SDK LINUX 09.01.00.06
Processor SDK RTOS 09.01.00.06 | Processor SDK LINUX 09.01.00.06
Processor SDK RTOS 09.01.00.06 | | 09_01_03_00 | 09_01_00_08 | Processor SDK LINUX : 09.01.00.07 | Processor SDK LINUX 09.01.00.06
Processor SDK RTOS 09.01.00.06 | Processor SDK LINUX 09.01.00.06
Processor SDK RTOS 09.01.00.06 | Processor SDK LINUX 09.01.00.06
Processor SDK RTOS 09.01.00.06 | diff --git a/examples/osrt_cpp/README.md b/examples/osrt_cpp/README.md index 9a2b9ee..b9dbb2d 100644 --- a/examples/osrt_cpp/README.md +++ b/examples/osrt_cpp/README.md @@ -10,16 +10,16 @@ ## Introduction - - CPP APIs os the DL runtime offered by solutions only supports the model inference. So the user is expeted to run the [Python Examples](../../README.md#python-exampe) on PC to generate the model artifacts. + - CPP APIs for DL runtimes only support model inference. So the user is expeted to run the [Python Examples](../../README.md#python-exampe) on PC to generate the model artifacts. - CPP API require yaml file reading lib. So the user is expected to install libyaml-cpp-dev by running command "sudo apt-get install libyaml-cpp-dev" -> Note : We are plannign to clean-up and unify the user inetrface for CPP examples by next release. We are also planning to add more CPP exmaples. + ## Setup - Prepare the Environment for the Model compilation by follwoing the setup section [here](../../README.md#setup) ## Build - - Build the CPP examples using cmake from repository base directory. Create a build folder for your generated build files. + - Build the CPP examples using cmake from repository base directory. Create a build folder for your generated build files ``` mkdir build && cd build @@ -96,12 +96,16 @@ - -v : verbose (set to 1) ## Validation on Target -- Build and run steps remains same for PC emaultionn and target. Copy the below folders from PC to the EVM where this repo is cloned before ruunning the examples - +- Build and run steps remains same for PC emulation and target. Copy the below folders from PC to the EVM where this repo is cloned before ruunning the examples ``` ./model-artifacts ./models ``` +- For ONNX runtime export the following on device, prior to execution: + ``` + export TIDL_RT_ONNX_VARDIM=1 + ``` + ## Running pre-compiled model from modelzoo - To run precomiled model from model zoo run the follwoing commands( as an example: cl-0000_tflitert_mlperf_mobilenet_v1_1.0_224_tflite) - Fetch the tar link from model zoo and wget the file @@ -118,9 +122,10 @@ cd ../ ./bin/Release/tfl_main -z "cl-0000_tflitert_mlperf_mobilenet_v1_1.0_224_tflite/" -v 1 -i "test_data/airshow.jpg" -l "test_data/labels.txt" -a 1 -d 1 ``` -- to run on target , copy the below folders from PC to the EVM where this repo is cloned before ruunning the examples +- To run on target , copy the below folders from PC to the EVM where this repo is cloned before running the examples ``` ./model-artifacts ./models ``` + diff --git a/setup.sh b/setup.sh index f92ffe3..6da410b 100755 --- a/setup.sh +++ b/setup.sh @@ -196,7 +196,7 @@ cp_osrt_lib() SCRIPTDIR=`pwd` -REL=09_01_06_00 +REL=09_01_07_00 skip_cpp_deps=0 skip_arm_gcc_download=0 skip_x86_python_install=0 @@ -240,13 +240,20 @@ shift # past argument done set -- "${POSITIONAL[@]}" # restore positional parameters -#Check if tools are built for -if [ $TIDL_TOOLS_TYPE == GPU ];then - tidl_gpu_tools=1 +#Check if CPU or GPU tools +if [ -z "$TIDL_TOOLS_TYPE" ];then + echo "Defaulting to CPU tools" + tidl_gpu_tools=0 else - tidl_gpu_tools=0 + echo "TIDL_TOOLS_TYPE set to :$TIDL_TOOLS_TYPE" + if [ $TIDL_TOOLS_TYPE == GPU ];then + tidl_gpu_tools=1 + else + tidl_gpu_tools=0 + fi fi + version_match=`python3 -c 'import sys;r=0 if sys.version_info >= (3,6) else 1;print(r)'` if [ $version_match -ne 0 ]; then echo 'python version must be >= 3.6'