Vitis AI

https://github.com/xilinx/vitis-ai
Admin
Vitis AI is an integrated development environment for accelerating AI inference on AMD adaptable...
Tokens:360,993
Snippets:3,397
Trust Score:9.3
License:Apache-2.0
Update:1 month ago
Show doc for...
Tokens:
Raw
### Setup Environment

Source: https://github.com/xilinx/vitis-ai/blob/master/model_zoo/model-list/tf_superpoint_3.5/model_info.md

Install the required dependencies and configure paths for experiments and data.

```bash
cd code 
pip install -e .
```

--------------------------------

### Setup Host Cross-Compiler

Source: https://github.com/xilinx/vitis-ai/blob/master/docs/_sources/docs/quickstart/vek280.rst.txt

Install the cross-compilation environment on the host machine.

```Bash
[Host] $ cd Vitis-AI/board_setup/vek280
[Host] $ sudo chmod u+r+x host_cross_compiler_setup.sh
[Host] $ ./host_cross_compiler_setup.sh
```

--------------------------------

### Execute V70 Setup Script

Source: https://github.com/xilinx/vitis-ai/blob/master/docs/_sources/docs/quickstart/v70.rst.txt

Run the setup script on the host machine to install XRT, XRM, and the V70 platform. A manual reboot of the host is required after execution.

```Bash
[Host] $ cd <Vitis-AI install path>/Vitis-AI/board_setup/v70
[Host] $ source ./install.sh
```

--------------------------------

### Start WSL Distribution

Source: https://github.com/xilinx/vitis-ai/blob/master/docs/_sources/docs/quickstart/vek280.rst.txt

Launch a specific installed WSL distribution.

```Bash
[Powershell] > wsl -d Ubuntu-20.04
```

--------------------------------

### Install Vitis AI Runtime

Source: https://github.com/xilinx/vitis-ai/blob/master/docs/_sources/docs/workflow-system-integration.rst.txt

Execute the setup script on the target board to install the Vitis AI Runtime.

```Bash
cd ~/[TARGET]
bash target_vart_setup.sh
```

--------------------------------

### Install vai_q_pytorch Component

Source: https://github.com/xilinx/vitis-ai/blob/master/src/vai_quantizer/vai_q_pytorch/README.md

Build and install the main component, choosing between standard installation or development mode for frequent code changes.

```bash
cd ./pytorch_binding; python setup.py install
```

```bash
cd ./pytorch_binding; python setup.py develop
```

--------------------------------

### Build and Install from Source

Source: https://github.com/xilinx/vitis-ai/blob/master/src/vai_quantizer/vai_q_tensorflow2.x/tensorflow_model_optimization/python/core/quantization/keras/vitis/quantize_strategy/bfp/README.md

Execute the build script and install the generated wheel packages.

```bash
$ sh build.sh
$ pip install pkgs/*.whl
```

--------------------------------

### Compile Example

Source: https://github.com/xilinx/vitis-ai/blob/master/docs/_sources/docs/quickstart/v70.rst.txt

Set execution permissions for the build script and compile the example.

```Bash
[Docker] $ sudo chmod u+r+x build.sh
[Docker] $ bash -x build.sh
```

--------------------------------

### Navigate to Example Directory

Source: https://github.com/xilinx/vitis-ai/blob/master/docs/_sources/docs/quickstart/v70.rst.txt

Change the current working directory to the ResNet50 example folder.

```Bash
[Docker] $ cd /workspace/examples/vai_runtime/resnet50
```

--------------------------------

### Execute SDK Installation Script

Source: https://github.com/xilinx/vitis-ai/blob/master/src/vai_runtime/quick_start_for_embedded.md

Run the downloaded script to install the cross-compiler environment.

```bash
$ ./sdk-2023.1.0.0.sh
```

--------------------------------

### Verify Installation

Source: https://github.com/xilinx/vitis-ai/blob/master/src/vai_quantizer/vai_q_pytorch/README.md

Confirm the installation by importing the library without errors.

```bash
python -c "import pytorch_nndct"
```

--------------------------------

### Run Example

Source: https://github.com/xilinx/vitis-ai/blob/master/docs/_sources/docs/quickstart/v70.rst.txt

Execute the compiled ResNet50 example using the specified model file.

```Bash
[Docker] $ ./resnet50 /usr/share/vitis_ai_library/models/resnet50/resnet50.xmodel
```

--------------------------------

### Installation Rules

Source: https://github.com/xilinx/vitis-ai/blob/master/src/vai_runtime/vart/buffer-object/CMakeLists.txt

Defines installation paths for the library, headers, and CMake export files.

```cmake
install(
  TARGETS ${COMPONENT_NAME}
  COMPONENT dpu
  EXPORT ${COMPONENT_NAME}-targets
  ARCHIVE DESTINATION lib
  RUNTIME DESTINATION bin
  LIBRARY DESTINATION lib)
install(
  FILES include/xir/buffer_object.hpp
        include/xir/device_memory.hpp
        # include/xir/buffer_object_manager.hpp
        # include/xir/buffer_object_manager_store.hpp
        # include/xir/device_scheduler.hpp
  COMPONENT dpu
  DESTINATION include/xir)

install(
  EXPORT ${COMPONENT_NAME}-targets
  COMPONENT dpu
  NAMESPACE ${PROJECT_NAME}::
  DESTINATION share/cmake/${PROJECT_NAME})
```

--------------------------------

### Compile Example Application

Source: https://github.com/xilinx/vitis-ai/blob/master/src/vai_runtime/plugin-samples/samples/plugin-smfc/README.md

Commands to navigate to the example directory and build the ResNet50 application.

```sh
# Compile the example
cd /workspace/tools/Vitis-AI-Runtime/VART/vart/softmax-runner/test/resnet_v1_50_softmax
./build.sh
```

--------------------------------

### Setup Environment and Download Model

Source: https://github.com/xilinx/vitis-ai/blob/master/docs/_sources/docs/quickstart/v70.rst.txt

Initialize the board setup script and download the ResNet18 model weights.

```Bash
[Docker] $ source /workspace/board_setup/v70/setup.sh DPUCV2DX8G_v70
[Docker] $ cd resnet18/model
[Docker] $ wget https://download.pytorch.org/models/resnet18-5c106cde.pth -O resnet18.pth
```

--------------------------------

### Run Petalinux setup script

Source: https://github.com/xilinx/vitis-ai/blob/master/third_party/tvm/docs/running_on_zynq.md

Executes the setup script to install TVM and Vitis AI flow dependencies on the target device.

```sh
bash petalinux_setup.sh
```

--------------------------------

### Run Docker and Build Example

Source: https://github.com/xilinx/vitis-ai/blob/master/docs/docs/quickstart/vek280.html

Launch the Docker container and compile the resnet50_pt example application.

```bash
[Host]   $ cd ../..
[Host]   $ ./docker_run.sh xilinx/vitis-ai-pytorch-cpu:latest
```

```bash
[Docker] $ conda activate vitis-ai-pytorch
```

```bash
[Docker] $ cd examples/vai_runtime/resnet50_pt
[Docker] $ sudo chmod u+r+x build.sh
[Docker] $ bash –x build.sh
```

--------------------------------

### Run Demo

Source: https://github.com/xilinx/vitis-ai/blob/master/model_zoo/model-list/tf_superpoint_3.5/model_info.md

Execute the demonstration script.

```bash
bash run_demo.sh
```

--------------------------------

### Start Vitis AI Docker Container

Source: https://github.com/xilinx/vitis-ai/blob/master/docs/_sources/docs/install/install.rst.txt

Commands to navigate to the installation directory and launch a pre-built Vitis AI Docker container.

```bash
cd <Vitis-AI install path>/Vitis-AI
./docker_run.sh xilinx/vitis-ai-<pytorch|tensorflow2|tensorflow>-<cpu|rocm>:latest
```

--------------------------------

### Verify Docker Installation

Source: https://github.com/xilinx/vitis-ai/blob/master/docs/_sources/docs/install/install.rst.txt

Commands to test the Docker installation and check the installed version.

```bash
docker run hello-world
```

```bash
docker --version
```

--------------------------------

### Configure Python Installation Path

Source: https://github.com/xilinx/vitis-ai/blob/master/src/vai_runtime/vart/trace/vaitrace/CMakeLists.txt

Sets the installation directory for Python files based on the build environment and installation preferences.

```cmake
if(BUILD_PYTHON)
  if(MSVC)
    set(VAITRACE_PY_INSTALL_DIR ${PYTHON_SITE_PACKAGES_USER})
  else(MSVC)
    if(INSTALL_HOME)
      set(VAITRACE_PY_INSTALL_DIR ${PYTHON_SITE_PACKAGES_USER})
    else(INSTALL_HOME)
      set(VAITRACE_PY_INSTALL_DIR ${PYTHON_SITE_PACKAGES})
    endif(INSTALL_HOME)
  endif(MSVC)
  install(
    DIRECTORY python/vaitrace_py
    DESTINATION ${VAITRACE_PY_INSTALL_DIR}
    FILES_MATCHING
    PATTERN "*.py")
endif()
```

--------------------------------

### Download and Extract OnBoard Example Recipes

Source: https://github.com/xilinx/vitis-ai/blob/master/examples/OnBoard/README.md

Commands to download and unpack the necessary model and image files for OnBoard examples.

```bash
wget http://www.xilinx.com/bin/public/openDownload?filename=onboard_example_recipes.tar.gz -O onboard_example_recipes.tar.gz
tar -xvf onboard_example_recipes.tar.gz ./
```

--------------------------------

### Install tf2onnx

Source: https://github.com/xilinx/vitis-ai/blob/master/src/vai_quantizer/tensorflow-onnx/tutorials/efficientnet-edge.ipynb

Installs the tf2onnx conversion tool.

```bash
!pip install -U tf2onnx
```

--------------------------------

### Start TensorBoard Server

Source: https://github.com/xilinx/vitis-ai/blob/master/examples/OnBoard/README.md

Launch the TensorBoard server to visualize logs generated by the OnBoard tool.

```bash
tensorboard --logdir=<your_log_dir> --bind_all
```

--------------------------------

### Install xir_util

Source: https://github.com/xilinx/vitis-ai/blob/master/src/vai_runtime/xir/tools/CMakeLists.txt

Specifies the installation destination for the xir_util binary.

```cmake
install(TARGETS xir_util DESTINATION bin)
```

--------------------------------

### Compile Sample Application

Source: https://github.com/xilinx/vitis-ai/blob/master/docs/docs/quickstart/v70.html

Navigate to the classification sample directory and execute the build script.

```bash
[Docker] $ cd /workspace/examples/vai_library/samples/classification
[Docker] $ ./build.sh
```

--------------------------------

### Install XNNC Wheel

Source: https://github.com/xilinx/vitis-ai/blob/master/src/vai_quantizer/xnnc4xir/README.md

Install the generated wheel file.

```bash
pip3 install xnnc-3.5.0-py3-none-any.whl
```

--------------------------------

### Download Runtime Examples

Source: https://github.com/xilinx/vitis-ai/blob/master/docs/_sources/docs/quickstart/v70.rst.txt

Download the Vitis AI runtime image and video examples package.

```Bash
[Docker] $ cd ~
[Docker] $ wget https://www.xilinx.com/bin/public/openDownload?filename=vitis_ai_runtime_r3.5.0_image_video.tar.gz -O vitis_ai_runtime_r3.5.0_image_video.tar.gz
```

--------------------------------

### Install dependencies

Source: https://github.com/xilinx/vitis-ai/blob/master/src/vai_quantizer/tensorflow-onnx/tutorials/keras-resnet50.ipynb

Installs the necessary TensorFlow and tf2onnx packages.

```python
!pip install tensorflow tf2onnx
```

--------------------------------

### vart.RunnerExt Usage Example

Source: https://github.com/xilinx/vitis-ai/blob/master/docs/doxygen/api/namespace/namespacerunnerext__example.html

Demonstrates the standard workflow for initializing a runner, retrieving input/output buffers, and executing an asynchronous task.

```APIDOC
## vart.RunnerExt Example

### Description
This example demonstrates how to instantiate a runner using `vart.RunnerExt.create_runner`, retrieve input and output tensor buffers, and perform an asynchronous execution of the graph.

### Code Example
```python
# create runner
runner = vart.RunnerExt.create_runner(subgraph, "run")

# get input and output tensor buffers
input_tensor_buffers = runner.get_inputs()
output_tensor_buffers = runner.get_outputs()

# run graph runner
v = runner.execute_async(input_tensor_buffers, output_tensor_buffers)
runner.wait(v)

# process output
output_data = np.asarray(output_tensor_buffers[0])
```
```

--------------------------------

### Verify Installation

Source: https://github.com/xilinx/vitis-ai/blob/master/src/vai_optimizer/pytorch_binding/pytorch_nndct/pruning/README.md

Verify that the pytorch_nndct package is correctly installed by importing it.

```shell
$ python -c "import pytorch_nndct"
```

--------------------------------

### Prepare VART C++ Profiling

Source: https://github.com/xilinx/vitis-ai/blob/master/examples/vai_profiler/examples.md

Commands to navigate to the sample directory.

```bash
# cd path-to-vart-resnet50-samples
```

--------------------------------

### Build XIR from source

Source: https://github.com/xilinx/vitis-ai/blob/master/src/vai_runtime/xir/README.md

Standard command to build XIR and install it to the local user directory.

```bash
./cmake.sh
```

--------------------------------

### Install pytorch_nndct

Source: https://github.com/xilinx/vitis-ai/blob/master/src/vai_optimizer/pytorch_binding/pytorch_nndct/pruning/README.md

Install the required dependencies and the pytorch_nndct package from source.

```shell
$ cd nndct && pip install -r requirements.txt
$ cd pytorch_binding && python setup.py install
```

--------------------------------

### Install Python requirements

Source: https://github.com/xilinx/vitis-ai/blob/master/src/vai_quantizer/vai_q_onnx/example/resnet_ptq_cpu/README.md

Install the necessary dependencies for the quantization workflow.

```bash
python -m pip install -r requirements.txt
```

--------------------------------

### Download Cross-Compiler SDK

Source: https://github.com/xilinx/vitis-ai/blob/master/src/vai_runtime/quick_start_for_embedded.md

Commands to create the SDK directory and download the installation script.

```bash
$ mkdir petalinux_sdk_2023.1
$ cd petalinux_sdk_2023.1
$ wget https://www.xilinx.com/bin/public/openDownload?filename=sdk-2023.1.0.0.sh -O sdk-2023.1.0.0.sh
$ chmod +x sdk-2023.1.0.0.sh
```

--------------------------------

### Install tf2onnx

Source: https://github.com/xilinx/vitis-ai/blob/master/src/vai_quantizer/tensorflow-onnx/tutorials/efficientdet.ipynb

Installs the latest version of the tf2onnx converter from GitHub.

```bash
# via saved-model
!pip install git+https://github.com/onnx/tensorflow-onnx
```

--------------------------------

### Default Ubuntu sources.list entry

Source: https://github.com/xilinx/vitis-ai/blob/master/docs/_sources/docs/install/China_Ubuntu_servers.rst.txt

Example of the default Ubuntu package server configuration pointing to US-based mirrors.

```text
     deb http://us.archive.ubuntu.com/ubuntu/ focal universe
```

--------------------------------

### Install TensorFlow on aarch64

Source: https://github.com/xilinx/vitis-ai/blob/master/third_party/tflite/README.md

Install the full TensorFlow package on the aarch64 host.

```bash
pip install tensorflow=2.3.0
```

--------------------------------

### Setup and Preprocessing Functions

Source: https://github.com/xilinx/vitis-ai/blob/master/src/vai_quantizer/tensorflow-onnx/tutorials/efficientnet-edge.ipynb

Imports necessary libraries and defines helper functions for image statistics, cropping, resizing, and normalization for EdgeTPU models.

```python
%matplotlib inline

import numpy as np
import math
import matplotlib.pyplot as plt
import onnxruntime as rt
import cv2
import json

labels = json.load(open("labels_map.txt", "r"))


def img_stats(a, name={}):
    return {
        "name": name,
        "size": a.shape,
        "mean": "{:.2f}".format(a.mean()),
        "std": "{:.2f}".format(a.std()),
        "max": a.max(),
        "min": a.min(),
        "median": "{:.2f}".format(np.median(a)),
    }


def center_crop(img, out_height, out_width):
    height, width, _ = img.shape
    left = int((width - out_width) / 2)
    right = int((width + out_width) / 2)
    top = int((height - out_height) / 2)
    bottom = int((height + out_height) / 2)
    img = img[top:bottom, left:right]
    return img


def resize_with_aspectratio(img, out_height, out_width, scale=87.5, inter_pol=cv2.INTER_LINEAR):
    height, width, _ = img.shape
    new_height = int(100. * out_height / scale)
    new_width = int(100. * out_width / scale)
    if height > width:
        w = new_width
        h = int(new_height * height / width)
    else:
        h = new_height
        w = int(new_width * width / height)
    img = cv2.resize(img, (w, h), interpolation=inter_pol)
    return img


def pre_process_mobilenet(img, dims):
    output_height, output_width, _ = dims
    img = resize_with_aspectratio(img, output_height, output_width, inter_pol=cv2.INTER_LINEAR)
    img = center_crop(img, output_height, output_width)
    img = np.asarray(img, dtype='float32')

    img /= 255.0
    img -= 0.5
    img *= 2
    return img

def pre_process_edgetpu(img, dims):
    output_height, output_width, _ = dims
    img = resize_with_aspectratio(img, output_height, output_width, inter_pol=cv2.INTER_LINEAR)
    img = center_crop(img, output_height, output_width)
    img = np.asarray(img, dtype='float32')
    img -= [127.0, 127.0, 127.0]
    img /= [128.0, 128.0, 128.0]
    return img


# read the image
fname = "panda.jpg"
img = cv2.imread(fname)
img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)

# pre-process the image like mobilenet and resize it to 300x300
img = pre_process_edgetpu(img, (300, 300, 3))
plt.axis('off')
plt.imshow(img)
plt.show()

# create a batch of 1 (that batch size is buned into the saved_model)
img_batch = np.expand_dims(img, axis=0)
```

--------------------------------

### Validate Installation

Source: https://github.com/xilinx/vitis-ai/blob/master/src/vai_quantizer/vai_q_tensorflow1.x/readme.md

Verify the installation by checking the version or accessing the help menu.

```shell
$ vai_q_tensorflow --version
$ vai_q_tensorflow --help
```

--------------------------------

### Install Test Targets

Source: https://github.com/xilinx/vitis-ai/blob/master/src/vai_runtime/vart/buffer-object/CMakeLists.txt

Configures the installation rules for the compiled test binaries.

```cmake
install(
  TARGETS mem_write mem_read mem_save dump_op_weights dump_op_weights_2
  COMPONENT dpu
  ARCHIVE DESTINATION lib
  RUNTIME DESTINATION bin
  LIBRARY DESTINATION lib)
```

--------------------------------

### Verify XNNC Installation

Source: https://github.com/xilinx/vitis-ai/blob/master/src/vai_quantizer/xnnc4xir/README.md

Check if the installation was successful by displaying help information.

```bash
xnnc-run --help
```

--------------------------------

### Install XNNC via pip

Source: https://github.com/xilinx/vitis-ai/blob/master/src/vai_quantizer/xnnc4xir/README.md

Command to install dependencies from the requirements file.

```bash
pip3 install -r requirements.txt
```

--------------------------------

### Build and Install Quantizer

Source: https://github.com/xilinx/vitis-ai/blob/master/src/vai_quantizer/vai_q_tensorflow1.x/readme.md

Build the wheel package for either GPU or CPU-only versions of the quantizer.

```shell
# GPU version
$ sh build.sh
# CPU-only version
$ sh build.sh CPU
```

--------------------------------

### Install dependencies

Source: https://github.com/xilinx/vitis-ai/blob/master/src/vai_quantizer/tensorflow-onnx/tutorials/ConvertingSSDMobilenetToONNX.ipynb

Installs the necessary tf2onnx and onnxruntime packages via pip.

```bash
!pip install --user -U tf2onnx
!pip install --user -U onnxruntime
```

--------------------------------

### Install TIMM dependency

Source: https://github.com/xilinx/vitis-ai/blob/master/src/vai_quantizer/vai_q_pytorch/example/timm/README.md

Install the required timm package via pip.

```bash
pip install timm
```

--------------------------------

### Run PyTorch Basic Usage Example

Source: https://github.com/xilinx/vitis-ai/blob/master/examples/OnBoard/README.md

Activate the PyTorch environment and execute the basic usage demonstration script.

```bash
conda activate vitis-ai-wego-torch
python demo_pytorch.py
```

--------------------------------

### Validate Installation

Source: https://github.com/xilinx/vitis-ai/blob/master/src/vai_optimizer/tensorflow/README.md

Verifies that the package is correctly installed by importing the runner class.

```shell
$ python -c "from tf_nndct import IterativePruningRunner"
```
Tokens:
Raw