A brief guide to Acuitylite
Acuitylite is an end-to-end neural-network deployment tool for embedded systems.
Acuitylite support converting caffe/darknet/onnx/tensorflow/tflite models to TIM-VX/TFLite cases.
In addition, Acuitylite support asymmetric uint8 and symmetric int8 quantization.
Attention: We have introduced some important changes and updated the APIs that are not compatible with the version before Acuitylite6.21.0(include). Please read the document and demos carefully.
System Requirement
- OS:
Ubuntu Linux 20.04 LTS 64-bit(python3.8)
Ubuntu Linux 22.04 LTS 64-bit(python3.10)
Ubuntu Linux 24.04 LTS 64-bit(python3.12)
Install
1. build the recommended docker image and run a container
2. pip install acuitylite --no-deps
Document
Reference: https://verisilicon.github.io/acuitylite
Framework Support
Tips: You can export a TFLite app and using tflite-vx-delegate to run on TIM-VX if the exported TIM-VX app does not meet your requirements.
How to generate nbg and TIM-VX case
When you need generate TIM-VX case and nbg, please set the export() function's param pack_nbg_unify=True. Such as: TimVxExporter(model).export(pack_nbg_unify=True), it will use our default SDK. If you want to use your own SDK and licence, please set the param of export() viv_sdk, licence. Such as: TimVxExporter(model).export(pack_nbg_unify=True, viv_sdk=your_sdk_path, licence=path_of_licence_txt)
Attention: your sdk directory structure must strictly follow the directory structure of acuitylib/vsi_sdk!!! your sdk need satisfy the structure of "your_sdk_path/build/install", "your_sdk_path/prebuilt-sdk/x86_64_linux", otherwise the path may have problems. And the licence content is the device target which you want to use.
How to run TIM-VX case
The exported TIM-VX case supports both make and cmake.
Please set environment for build and run case:
- TIM_VX_DIR=/path/to/tim-vx/build/install
- VIVANTE_SDK_DIR=/path/to/tim-vx/prebuilt-sdk/x86_64_linux
- LD_LIBRARY_PATH=$TIM_VX_DIR/lib:$VIVANTE_SDK_DIR/lib
Attention: The TIM_VX_DIR path should include lib and header files of TIM-VX. You can refer TIM-VX to build TIM-VX.
How to generate nbg by Ovxlib
When you need generate nbg, please use OvxlibExporter class and set the export() function's param pack_nbg_only=True. Such as: OvxlibExporter(model).export(pack_nbg_only=True), it will use our default SDK. If you want to use your own SDK and licence, please set the "viv_sdk" and "licence" params of export() function. Such as: OvxlibExporter(model).export(pack_nbg_only=True, viv_sdk=your_sdk_path, licence=path_of_licence_txt)
Attention: your sdk directory structure must strictly follow the directory structure of acuitylib/vsi_sdk!!! your sdk need satisfy the structure of "your_sdk_path/prebuilt-sdk/x86_64_linux", otherwise the path may have problems. The content of licence is the device target which you want to use.
Support
Create issue on github or email to ML_Support@verisilicon.com
Metadata
Release files for acuitylite 6.57.1
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Actions Reset |
|---|---|---|---|---|
| acuitylite-6.57.1-cp310-cp310-manylinux2010_x86_64.whl | CPython 3.10 | CPython 3.10 | Linux glibc 2.12+ x86-64 | Details for acuitylite-6.57.1-cp310-cp310-manylinux2010_x86_64.whl |
Release files / acuitylite-6.57.1-cp310-cp310-manylinux2010_x86_64.whl
| Download URL | acuitylite-6.57.1-cp310-cp310-manylinux2010_x86_64.whl |
|---|---|
| Size | 246.7 MB |
| Tags | CPython 3.10 Linux glibc 2.12+ x86-64 |
|
SHA-256 checksum How to use checksums |
bc687bc67aa475e97f174dfa7bbbab05c1872ed116f5954810da858ce72c99b8
|
|
BLAKE2b-256 checksum How to use checksums |
f435857019b1073d196e9a1b816c8f8d85a983f9d66aab2807028fbcc5638165
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/4.0.2 CPython/3.10.12
|
Security for acuitylite 6.57.1
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
No provenance found
None of the files in this release were uploaded with provenance metadata.
A note on package security
PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.
To learn more about protecting yourself from malicious actors, visit our package security documentation.