This project has the source code for MIVIsionX model compiler in mv_compile.cpp
mv_deploy consists of a model-compiler and necessary header/.cpp files which are required to run inference for a specific NeuralNet model
The "mv_compile" will be built as part of MIVisionX package installer To build and application using mv_compile, the user can use the deployment api from mv_deploy.h. The entire use of the mv_compile and deployment is shown in mv_objdetectsample The sample demonstrates the use of mv_compile utility to do video decoding and inference.
- Ubuntu
20.04
/22.04
or CentOS7
/8
- ROCm supported hardware
- AMD Radeon GPU or APU required
- ROCm
- Build & Install MIVisionX
- MIVisionX installs model compiler at
/opt/rocm/libexec/mivisionx
- mv_compile installs at
/opt/rocm/bin
and mvdeploy_api.h installs at/opt/rocm/include/mivisionx
- MIVisionX installs model compiler at
The mv_compile utility generates deployment library, header files, and .cpp files required to run inference for the specified model.
- Usage:
mv_compile
--model <model_name: name of the trained model with path> [required]
--install_folder <install_folder: the location for compiled model> [required]
--input_dims <input_dims: n,c,h,w - batch size, channels, height, width> [required]
--backend <backend: name of the backend for compilation> [optional - default:OpenVX_Rocm_GPU]
--fuse_cba <fuse_cba: enable or disable Convolution_bias_activation fuse mode (0/1)> [optional - default: 0]
--quant_mode <quant_mode: fp32/fp16 - quantization_mode for the model: if enabled the model and weights would be converted [optional -default: fp32]
- Sample Usage:
Caffe
./mv_compile --model models/model.caffemodel --install_folder install_folder --input_dims 1,3,224,224
ONNX
./mv_compile --model models/model.onnx --install_folder install_folder --input_dims 1,3,224,224
NNEF
./mv_compile --model models/model.nnef --install_folder install_folder --input_dims 1,3,224,224
This project is licensed under the MIT License - see the LICENSE.md file for details
Rajy Rawther - [email protected]