### Deploying FasterLivePortrait API (Python)

Source: https://github.com/warmshao/fasterliveportrait/blob/master/README_ZH.md

This command starts the API server for FasterLivePortrait, allowing programmatic access to its functionalities. Further details on API usage can be found in the provided tutorial.

```Python
python api.py
```

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

### Launching WebUI on Windows (Batch)

Source: https://github.com/warmshao/fasterliveportrait/blob/master/README_ZH.md

This batch script starts the Gradio web user interface for FasterLivePortrait on Windows, making the application accessible via a web browser at the default port.

```Batch
webui.bat
```

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

### Installing Onnxruntime for CPU Inference

Source: https://github.com/warmshao/fasterliveportrait/blob/master/README.md

This command installs the standard ONNX Runtime package for CPU inference. While simple to install, CPU inference is explicitly noted as extremely slow and not recommended for practical use.

```shell
pip install onnxruntime
```

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

### Downloading JoyVASA and Chinese-Hubert-Base Models (Shell)

Source: https://github.com/warmshao/fasterliveportrait/blob/master/README_ZH.md

These commands download the necessary models for JoyVASA support, including `chinese-hubert-base` and `JoyVASA` itself, enabling audio-driven video or image generation.

```Shell
huggingface-cli download TencentGameMate/chinese-hubert-base --local-dir .\checkpoints\chinese-hubert-base
```

```Shell
huggingface-cli download jdh-algo/JoyVASA --local-dir ./checkpoints/JoyVASA
```

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

### Launching WebUI with MediaPipe (Python)

Source: https://github.com/warmshao/fasterliveportrait/blob/master/README_ZH.md

These commands launch the Gradio web UI, enabling the use of the MediaPipe model instead of `insight_face`. Users can choose between TensorRT or ONNX Runtime for inference, providing flexibility for different hardware setups.

```Python
python webui.py --mode trt --mp
```

```Python
python webui.py --mode onnx --mp
```

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

### Installing espeak-ng on Linux (Shell)

Source: https://github.com/warmshao/fasterliveportrait/blob/master/README_ZH.md

This command installs the `espeak-ng` text-to-speech engine on Linux systems, which is a prerequisite for using the Kokoro-82M model for text-driven features.

```Shell
apt-get -qq -y install espeak-ng > /dev/null 2>&1
```

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

### Verifying FasterLivePortrait API Container Startup

Source: https://github.com/warmshao/fasterliveportrait/blob/master/assets/docs/API.md

This output, typically viewed via `docker logs $container_id`, indicates that the FasterLivePortrait API application has successfully started. It confirms that Uvicorn is running and listening on the specified port, `9871`.

```Shell
INFO:     Application startup complete.
INFO:     Uvicorn running on http://0.0.0.0:9871 (Press CTRL+C to quit)
```

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

### Installing espeak-ng on Linux

Source: https://github.com/warmshao/fasterliveportrait/blob/master/README.md

This command installs the `espeak-ng` text-to-speech engine on Linux systems. It is a prerequisite for utilizing the text-driven generation features, particularly with models like Kokoro-82M.

```Shell
apt-get -qq -y install espeak-ng > /dev/null 2>&1
```

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

### Launching Gradio WebUI with ONNX Runtime (Python)

Source: https://github.com/warmshao/fasterliveportrait/blob/master/README_ZH.md

This command starts the Gradio web user interface for FasterLivePortrait, configured to use the ONNX Runtime for model inference. It enables a web-based interaction with the live portrait generation features.

```Python
python webui.py --mode onnx
```

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

### Installing FFmpeg on MacOS (Shell)

Source: https://github.com/warmshao/fasterliveportrait/blob/master/README_ZH.md

This command installs FFmpeg on macOS using Homebrew, which is a necessary dependency for video processing within the FasterLivePortrait project.

```Shell
brew install ffmpeg
```

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

### Launching Web UI Demo on Windows

Source: https://github.com/warmshao/fasterliveportrait/blob/master/README.md

This snippet explains how to launch the FasterLivePortrait web user interface on Windows. Users can double-click the `webui.bat` file to start the web server, which will then be accessible in a browser at `http://localhost:9870/`.

```Batch
webui.bat
```

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

### Testing WebUI on MacOS (Python)

Source: https://github.com/warmshao/fasterliveportrait/blob/master/README_ZH.md

This command launches the Gradio web UI on macOS, configured to use ONNX Runtime for inference, allowing users to test the application's functionality after setup.

```Python
python webui.py --mode onnx
```

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

### Installing Compiled Onnxruntime GPU Wheel

Source: https://github.com/warmshao/fasterliveportrait/blob/master/README.md

This command installs the compiled onnxruntime-gpu Python wheel package from the local build directory. This step makes the custom-built GPU-enabled ONNX Runtime available for use in Python projects.

```shell
pip install build/Linux/Release/dist/onnxruntime_gpu-1.17.0-cp310-cp310-linux_x86_64.whl
```

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

### Launching Real-time Camera Demo on Windows

Source: https://github.com/warmshao/fasterliveportrait/blob/master/README.md

This snippet provides instructions for running the real-time camera demo on Windows. The demo can be started by double-clicking `camera.bat`, or by running `camera.bat assets/examples/source/s9.jpg` from the command line to specify a target image. Users can press 'q' to stop the operation.

```Batch
camera.bat
```

```Batch
camera.bat assets\examples\source\s9.jpg
```

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

### Building FasterLivePortrait API Docker Image

Source: https://github.com/warmshao/fasterliveportrait/blob/master/assets/docs/API.md

This command builds the Docker image for the FasterLivePortrait API. It tags the image with `shaoguo/faster_liveportrait_api:v1.0` and uses `DockerfileAPI` as the Dockerfile. Users should replace the tag with their desired image name.

```Shell
docker build -t shaoguo/faster_liveportrait_api:v1.0 -f DockerfileAPI .
```

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

### Running Camera Demo on Windows (Batch)

Source: https://github.com/warmshao/fasterliveportrait/blob/master/README_ZH.md

This batch script initiates the real-time camera demo on Windows. Users can optionally specify a source image path as an argument to change the target image for the demo. Press 'q' to stop the demo.

```Batch
camera.bat
```

```Batch
camera.bat assets/examples/source/s9.jpg
```

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

### Installing MacOS Python Dependencies (Python)

Source: https://github.com/warmshao/fasterliveportrait/blob/master/README_ZH.md

This command installs all Python packages listed in `requirements_macos.txt`, which are specific dependencies required for running FasterLivePortrait on macOS.

```Python
pip install -r requirements_macos.txt
```

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

### Installing FFmpeg on macOS

Source: https://github.com/warmshao/fasterliveportrait/blob/master/README.md

This snippet demonstrates how to install FFmpeg on macOS using Homebrew. FFmpeg is a crucial dependency for multimedia processing within the FasterLivePortrait project, and this command ensures its installation.

```Shell
brew install ffmpeg
```

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

### Setting up Python Virtual Environment on MacOS (Shell)

Source: https://github.com/warmshao/fasterliveportrait/blob/master/README_ZH.md

These commands create and activate a new Conda virtual environment named 'flip' with Python 3.10, providing an isolated environment for project dependencies on macOS.

```Shell
conda create -n flip python=3.10 && conda activate flip
```

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

### Downloading Kokoro-82M Model (Shell)

Source: https://github.com/warmshao/fasterliveportrait/blob/master/README_ZH.md

This command downloads the Kokoro-82M model from Hugging Face Hub and saves it to the local checkpoints directory, enabling text-driven video or image generation.

```Shell
huggingface-cli download hexgrad/Kokoro-82M --local-dir .\checkpoints\Kokoro-82M
```

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

### Configuring CMake for TensorRT Plugin

Source: https://github.com/warmshao/fasterliveportrait/blob/master/README.md

This command configures the CMake build system for the `grid-sample3d-trt-plugin`, pointing it to the root directory of the TensorRT installation. This is crucial for CMake to find the necessary TensorRT libraries and headers to build the plugin.

```shell
cmake .. -DTensorRT_ROOT=$TENSORRT_HOME
```

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

### Running Inference with Paste-Back Optimization

Source: https://github.com/warmshao/fasterliveportrait/blob/master/README.md

This example command demonstrates running the `run.py` script with specific source and driving videos, utilizing TensorRT inference and enabling the `paste_back` optimization. It also activates support for animal models, significantly improving processing speed.

```Python
python run.py --src_image assets/examples/source/s39.jpg --dri_video assets/examples/driving/d0.mp4 --cfg configs/trt_infer.yaml --paste_back --animal
```

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

### Launching Gradio WebUI with TensorRT

Source: https://github.com/warmshao/fasterliveportrait/blob/master/README.md

This command starts the Gradio web user interface, optimized to leverage TensorRT for accelerated model inference. TensorRT typically offers higher performance compared to ONNX Runtime for compatible hardware.

```Python
python webui.py --mode trt
```

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

### Downloading ONNX Models for MacOS (Shell)

Source: https://github.com/warmshao/fasterliveportrait/blob/master/README_ZH.md

This command downloads the necessary ONNX model files for FasterLivePortrait from Hugging Face Hub to the local checkpoints directory, enabling model inference on macOS.

```Shell
huggingface-cli download warmshao/FasterLivePortrait --local-dir ./checkpoints
```

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

### Installing Python Dependencies on macOS

Source: https://github.com/warmshao/fasterliveportrait/blob/master/README.md

This snippet shows how to install the necessary Python packages for FasterLivePortrait on macOS. It uses `pip` to install dependencies listed in the `requirements_macos.txt` file, ensuring all required libraries are available within the activated virtual environment.

```Shell
pip install -r requirements_macos.txt
```

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

### Testing FasterLivePortrait Pipeline with Onnxruntime

Source: https://github.com/warmshao/fasterliveportrait/blob/master/README.md

This Python command executes the FasterLivePortrait inference pipeline using ONNX Runtime. It specifies a source image, a driving video, and a configuration file for the ONNX inference process to test the setup.

```python
python run.py \
--src_image assets/examples/source/s10.jpg \
--dri_video assets/examples/driving/d14.mp4 \
--cfg configs/onnx_infer.yaml
```

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

### Launching Gradio WebUI with TensorRT (Python)

Source: https://github.com/warmshao/fasterliveportrait/blob/master/README_ZH.md

This command initiates the Gradio web user interface for FasterLivePortrait, optimized to utilize TensorRT for accelerated model inference. It provides a high-performance web-based interface for the application.

```Python
python webui.py --mode trt
```

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

### Running Locally with MediaPipe (Python)

Source: https://github.com/warmshao/fasterliveportrait/blob/master/README_ZH.md

This command executes the `run.py` script to process a source image with a driving video using the MediaPipe model, configured for TensorRT inference. This is for local, non-web UI operation.

```Python
python run.py --src_image assets/examples/source/s12.jpg --dri_video assets/examples/driving/d0.mp4 --cfg configs/trt_mp_infer.yaml
```

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

### Creating Python Virtual Environment on macOS

Source: https://github.com/warmshao/fasterliveportrait/blob/master/README.md

This snippet outlines the process of setting up a Python 3.10 virtual environment named 'flip' using Miniforge (Conda) on macOS. This isolated environment is recommended to manage project dependencies and avoid conflicts with system-wide Python installations.

```Shell
conda create -n flip python=3.10 && conda activate flip
```

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

### Running FasterLivePortrait with Paste-Back and Animal Model (Python)

Source: https://github.com/warmshao/fasterliveportrait/blob/master/README_ZH.md

This command executes the `run.py` script with specific parameters to enable the optimized `paste_back` function using `torchgeometry + cuda`, apply an animal model, and process a source image with a driving video using a TensorRT inference configuration.

```Python
python run.py --src_image assets/examples/source/s39.jpg --dri_video assets/examples/driving/d0.mp4 --cfg configs/trt_infer.yaml --paste_back --animal
```

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

### Downloading Animal ONNX Model (Shell)

Source: https://github.com/warmshao/fasterliveportrait/blob/master/README_ZH.md

This command downloads the ONNX file for the animal model from Hugging Face Hub to the local checkpoints directory, which is required before converting it to a TensorRT file for accelerated inference.

```Shell
huggingface-cli download warmshao/FasterLivePortrait --local-dir ./checkpoints
```

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

### Cloning Onnxruntime Repository for GPU Compilation

Source: https://github.com/warmshao/fasterliveportrait/blob/master/README.md

This command clones the official Microsoft ONNX Runtime GitHub repository. This is a prerequisite step for compiling onnxruntime-gpu from source, which is necessary to get CUDA support for grid_sample.

```shell
git clone https://github.com/microsoft/onnxruntime
```

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

### Updating Project Code and Dependencies

Source: https://github.com/warmshao/fasterliveportrait/blob/master/README.md

These commands ensure the project's source code is up-to-date and all required Python dependencies are installed. The `update.bat` script provides a convenient way for Windows users to perform these actions automatically.

```Shell
git pull origin master
```

```Shell
pip install requirements.txt
```

```Shell
update.bat
```

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

### Updating FasterLivePortrait Docker Image (Shell)

Source: https://github.com/warmshao/fasterliveportrait/blob/master/README_ZH.md

This command pulls the latest version of the FasterLivePortrait Docker image (v3), ensuring access to the most recent bug fixes and optimizations, including those related to Xpose ops.

```Shell
docker pull shaoguo/faster_liveportrait:v3
```

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

### Updating Code and Python Dependencies (Shell/Batch)

Source: https://github.com/warmshao/fasterliveportrait/blob/master/README_ZH.md

This set of commands updates the project's source code to the latest version and installs/updates all required Python dependencies. For Windows users, a batch script is provided for convenience.

```Shell
git pull origin master
```

```Python
pip install requirements.txt
```

```Batch
update.bat
```

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

### Converting ONNX to TensorRT on Windows (Batch)

Source: https://github.com/warmshao/fasterliveportrait/blob/master/README_ZH.md

This batch script converts ONNX model files to TensorRT format on Windows systems, which is a prerequisite for leveraging TensorRT's accelerated inference capabilities within the integrated package.

```Batch
scripts\all_onnx2trt.bat
```

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

### Testing Web UI with ONNX Mode on macOS

Source: https://github.com/warmshao/fasterliveportrait/blob/master/README.md

This snippet demonstrates how to run the FasterLivePortrait web user interface in ONNX mode on macOS. Executing `python webui.py --mode onnx` starts the web application, allowing users to interact with the model using ONNX runtime.

```Shell
python webui.py --mode onnx
```

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

### Running Realtime with Animal Model (Python)

Source: https://github.com/warmshao/fasterliveportrait/blob/master/README_ZH.md

This command runs the FasterLivePortrait application in real-time mode, using a specified source image and the animal model, configured for TensorRT inference. The `dri_video 0` typically indicates using a webcam as the driving source.

```Python
python run.py --src_image assets/examples/source/s39.jpg --dri_video 0 --cfg configs/trt_infer.yaml --realtime --animal
```

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

### Pulling Latest Docker Image

Source: https://github.com/warmshao/fasterliveportrait/blob/master/README.md

This command pulls the most recent version of the FasterLivePortrait Docker image. Using the latest image ensures access to optimized functions, bug fixes, and new features without requiring local environment setup.

```Shell
docker pull shaoguo/faster_liveportrait:v3
```

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

### Running FasterLivePortrait API Docker Container Locally

Source: https://github.com/warmshao/fasterliveportrait/blob/master/assets/docs/API.md

This command runs the FasterLivePortrait API Docker image in detached mode, utilizing all available GPUs. It maps a local checkpoint directory into the container, sets environment variables for the model path and server port, and exposes the service port. The container is configured to restart automatically.

```Shell
docker run -d --gpus=all \
--name faster_liveportrait_api \
-v E:\my_projects\FasterLivePortrait\checkpoints:/root/FasterLivePortrait/checkpoints \
-e CHECKPOINT_DIR=/root/FasterLivePortrait/checkpoints \
-e SERVER_PORT=9871 \
-p 9871:9871 \
--restart=always \
shaoguo/faster_liveportrait_api:v1.0 \
/bin/bash
```

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

### Downloading JoyVASA Integration Models

Source: https://github.com/warmshao/fasterliveportrait/blob/master/README.md

These commands download the necessary models from Hugging Face for integrating JoyVASA functionality. These models enable audio-driven video or image generation, enhancing the project's capabilities.

```Shell
huggingface-cli download TencentGameMate/chinese-hubert-base --local-dir .\checkpoints\chinese-hubert-base
```

```Shell
huggingface-cli download jdh-algo/JoyVASA --local-dir ./checkpoints/JoyVASA
```

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

### Compiling TensorRT Plugin

Source: https://github.com/warmshao/fasterliveportrait/blob/master/README.md

This command initiates the compilation process for the `grid-sample3d-trt-plugin` using the configured Makefile. It builds the plugin's shared library (`.so` file) which is essential for TensorRT to support the 5D grid sample operation.

```shell
make
```

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

### Downloading Onnx Models with Hugging Face CLI

Source: https://github.com/warmshao/fasterliveportrait/blob/master/README.md

This command downloads the pre-converted ONNX model files for FasterLivePortrait from Hugging Face Hub to the specified local directory. This is the first step required before performing any inference.

```shell
huggingface-cli download warmshao/FasterLivePortrait --local-dir ./checkpoints
```

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

### Deploying FasterLivePortrait API

Source: https://github.com/warmshao/fasterliveportrait/blob/master/README.md

This command launches the API server for the FasterLivePortrait project, enabling programmatic access to its functionalities. It's designed for integration into other applications or services.

```Python
python api.py
```

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

### Running Main Script with Pickle Saving

Source: https://github.com/warmshao/fasterliveportrait/blob/master/README.md

Executing the main `run.py` script with this command now automatically saves the corresponding pickle file to the same directory as the driving video. This feature allows for direct reuse of generated data, streamlining workflows.

```Python
python run.py
```

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

### Launching Gradio WebUI with ONNX Runtime

Source: https://github.com/warmshao/fasterliveportrait/blob/master/README.md

This command initiates the Gradio web user interface, configuring it to use the ONNX Runtime for model inference. It provides a local web interface for interacting with the FasterLivePortrait features.

```Python
python webui.py --mode onnx
```

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

### Creating and Entering Build Directory

Source: https://github.com/warmshao/fasterliveportrait/blob/master/README.md

This command sequence creates a new directory named `build` and then changes the current working directory into it. This is a common practice in software compilation workflows to keep build artifacts separate from source files.

```shell
mkdir build && cd build
```

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

### Downloading ONNX Model Files on macOS

Source: https://github.com/warmshao/fasterliveportrait/blob/master/README.md

This snippet provides the command to download the required ONNX model files for FasterLivePortrait from Hugging Face. The `huggingface-cli` tool is used to fetch the models from the 'warmshao/FasterLivePortrait' repository and save them locally in a directory named `checkpoints`.

```Shell
huggingface-cli download warmshao/FasterLivePortrait --local-dir ./checkpoints
```

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

### Checking Out Specific Branch for Onnxruntime GPU

Source: https://github.com/warmshao/fasterliveportrait/blob/master/README.md

This command switches the cloned ONNX Runtime repository to a specific branch (liqun/ImageDecoder-cuda) that includes a custom CUDA implementation for grid_sample, which is crucial for GPU acceleration in this project.

```shell
git checkout liqun/ImageDecoder-cuda
```

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

### Running WebUI with MediaPipe Models

Source: https://github.com/warmshao/fasterliveportrait/blob/master/README.md

These commands launch the Gradio web UI, incorporating MediaPipe models to replace InsightFace for improved face detection and landmarking. Users can choose between TensorRT or ONNX inference backends.

```Python
python webui.py --mode trt --mp
```

```Python
python webui.py --mode onnx --mp
```

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

### Downloading Kokoro-82M Model

Source: https://github.com/warmshao/fasterliveportrait/blob/master/README.md

This command downloads the Kokoro-82M model from Hugging Face, which is essential for enabling text-driven video or image generation within the FasterLivePortrait framework. The model is saved to the specified local directory.

```Shell
huggingface-cli download hexgrad/Kokoro-82M --local-dir .\checkpoints\Kokoro-82M
```

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

### Compiling Onnxruntime GPU from Source

Source: https://github.com/warmshao/fasterliveportrait/blob/master/README.md

This comprehensive shell command compiles onnxruntime-gpu from source with specific CUDA and cuDNN configurations. Users must adjust cuda_version and CMAKE_CUDA_ARCHITECTURES to match their system, noting that cuDNN 9.x is incompatible.

```shell
./build.sh --parallel \
--build_shared_lib --use_cuda \
--cuda_version 11.8 \
--cuda_home /usr/local/cuda --cudnn_home /usr/local/cuda/ \
--config Release --build_wheel --skip_tests \
--cmake_extra_defines CMAKE_CUDA_ARCHITECTURES="60;70;75;80;86" \
--cmake_extra_defines CMAKE_CUDA_COMPILER=/usr/local/cuda/bin/nvcc \
--disable_contrib_ops \
--allow_running_as_root
```

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

### Running FasterLivePortrait Docker Container

Source: https://github.com/warmshao/fasterliveportrait/blob/master/README.md

This command executes the FasterLivePortrait Docker image, mapping the local project directory into the container and exposing port 9870. It allocates all available GPUs for accelerated inference and ensures the container restarts automatically.

```shell
docker run -it --gpus=all \
--name faster_liveportrait \
-v $FasterLivePortrait_ROOT:/root/FasterLivePortrait \
--restart=always \
-p 9870:9870 \
shaoguo/faster_liveportrait:v3 \
/bin/bash
```

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

### Running Real-time Inference with Animal Model

Source: https://github.com/warmshao/fasterliveportrait/blob/master/README.md

This command executes the `run.py` script for real-time processing, typically using a webcam as the driving video source (indicated by `0`). It leverages TensorRT inference and activates the animal model, suitable for live applications.

```Python
python run.py --src_image assets/examples/source/s39.jpg --dri_video 0 --cfg configs/trt_infer.yaml --realtime --animal
```

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

### Downloading Animal ONNX Model

Source: https://github.com/warmshao/fasterliveportrait/blob/master/README.md

This command downloads the animal ONNX model from the FasterLivePortrait Hugging Face repository. After downloading, this ONNX file typically needs to be converted to the TensorRT format for optimized performance.

```Shell
huggingface-cli download warmshao/FasterLivePortrait --local-dir ./checkpoints
```

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

### Exporting CUDA Binaries to PATH

Source: https://github.com/warmshao/fasterliveportrait/blob/master/README.md

This command adds the CUDA toolkit's binary directory to the system's PATH environment variable. This allows the system to locate CUDA executables and libraries required for compiling and running CUDA-dependent applications, such as the TensorRT plugin.

```shell
export PATH=/usr/local/cuda/bin:$PATH
```

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

### Testing FasterLivePortrait with TensorRT Inference

Source: https://github.com/warmshao/fasterliveportrait/blob/master/README.md

This command executes the `run.py` script to test the FasterLivePortrait pipeline using TensorRT for inference. It specifies a source image, a driving video, and a configuration file for TensorRT inference, demonstrating the core functionality.

```shell
python run.py \
--src_image assets/examples/source/s10.jpg \
--dri_video assets/examples/driving/d14.mp4 \
--cfg configs/trt_infer.yaml
```

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

### Running Local Webcam with MediaPipe and TRT

Source: https://github.com/warmshao/fasterliveportrait/blob/master/README.md

This command executes the `run.py` script for local webcam usage, specifically configured to use MediaPipe models for processing. It utilizes TensorRT inference for efficient real-time performance with a live video input.

```Python
python run.py --src_image assets/examples/source/s12.jpg --dri_video 0 --cfg configs/trt_mp_infer.yaml
```

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

### Running FasterLivePortrait Real-time via Camera

Source: https://github.com/warmshao/fasterliveportrait/blob/master/README.md

This command runs the FasterLivePortrait application in real-time using a connected camera as the driving video source. It utilizes a specified source image and the TensorRT inference configuration, enabling live portrait animation.

```shell
python run.py \
--src_image assets/examples/source/s10.jpg \
--dri_video 0 \
--cfg configs/trt_infer.yaml \
--realtime
```

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

### Converting ONNX Models to TensorRT/OnnxruntimeGPU on Windows

Source: https://github.com/warmshao/fasterliveportrait/blob/master/README.md

This snippet describes how to convert ONNX model files to TensorRT or OnnxruntimeGPU format on Windows. This process is initiated by double-clicking the `all_onnx2trt.bat` script located in the `scripts` directory, which may take some time to complete.

```Batch
scripts\all_onnx2trt.bat
```

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

### Modifying CMakeLists.txt for CUDA Architectures

Source: https://github.com/warmshao/fasterliveportrait/blob/master/README.md

This line modifies the `CMakeLists.txt` file for the `grid-sample3d-trt-plugin` to specify compatible CUDA architectures. This is necessary to ensure the plugin compiles correctly for various NVIDIA GPU generations, supporting the required 5D input for grid sample operations.

```cmake
set_target_properties(${PROJECT_NAME} PROPERTIES CUDA_ARCHITECTURES "60;70;75;80;86")
```

=== COMPLETE CONTENT === This response contains all available snippets from this library. No additional content exists. Do not make further requests.