1. Prerequisites
1.1. Host Requirements
The host meets the requirements as described in Section 1.4, System Requirements
Resource | Note |
CPU | Intel ® CoreTM i5-6500 CPU@3.2GHz x4 with support of the Intel® Advanced Vector Extensions. |
GPU | (Optional) NVIDIA® GPU cards with CUDA® architectures. A compatible version of CUDA and cuDNN is installed for TensorFlow |
RAM | 8 GB at least |
Disk | 160 GB |
OS | Ubuntu 22.04 LTS 64-bit with Python 3.10, Ubuntu 20.04 LTS 64-bit with Python 3.8 (recommended), or Ubuntu 18.04 LTS 64-bit with Python 3.6 Note: Other Ubuntu versions are not recommended. |
1.2. Python Requirements
Ubuntu 22.04 with Python 3.10, Ubuntu 20.04 with Python 3.8, or Ubuntu 18.04 with Python 3.6 is set up.
1.3. Vivante IDE
The VivanteIDE is installed correctly if you want to generate NBG cases from the ACUITY Toolkit. Please refer to this document: NN IDE Install
2. Version Information
The NPU driver Vivante ACUITY Toolkit and Vivante IDE need to match the version
NPU Kernel Driver | Vivante ACUITY Toolkit | Vivante IDE |
v6.4.13.8 | 6.18.1 | 5.7.2 |
v6.4.15.9 | 6.21.1 | 5.8.2 |
3. Binary Toolkit
3.1. Install Procedure
Extract the ACUITY package, Verisilicon_Tool_Acuity_Toolkit_<version>.tgz, to a destination directory. There are two versions of the toolkit, Please confirm the version that needs to be installed:
acuity-toolkit-binary-6.18.1, matches the VIP9000-NPU kernel driver v6.4.13.8
acuity-toolkit-binary-6.21.1, matches the VIP9000-NPU kernel driver v6.4.15.9
From this directory, locate and unpack the binary package acuity-toolkit-binary-<version>.tgz. When the unpacking is completed, the installation finishes.
3.2. Toolkit Verification
Switch to the ./acuity-toolkit-binary-<version>/bin directory.
Set the environment variable ACUITY_PATH as follows, which will used to execute the ACUITY binary tools
Execute binary file
user@linux:~/c3v/acuity-toolkit-binary-6.21.1/bin$ ./pegasus -h
yang.luo@ti-d630:~/c3v/acuity-toolkit-binary-6.21.1/bin$ ./pegasus
usage: pegasus [-h] {import,export,generate,prune,inference,quantize,train,dump,measure,help} ...
Pegasus commands.
positional arguments:
{import,export,generate,prune,inference,quantize,train,dump,measure,help}
import Import models.
export Export models.
generate Generate metas.
prune prune models.
inference Inference model and get result.
quantize Quantize model.
train Train model.
dump Dump model activations.
measure Get amount of calculation, parameter and activation.
help Print a synopsis and a list of commands.
optional arguments:
-h, --help show this help message and exit
4. Wheel Toolkit
4.1. Install Procedure
Install requirement packages
acuity-toolkit-6.18.1. Matches the VIP9000-NPU kernel driver v6.4.13.8
tensorflow==2.10.0 scipy networkx==1.11 lmdb==0.93 onnxoptimizer==0.3.1 onnx==1.12.0 dill==0.2.8.2 ruamel.yaml==0.15.81 ply==3.11 torch==1.5.1
acuity-toolkit-6.21.1. Matches the VIP9000-NPU kernel driver v6.4.15.9
tensorflow==2.12.0 scipy networkx==1.11 lmdb==0.93 protobuf==3.20.3 onnxoptimizer==0.3.13 onnx==1.14.0 dill==0.2.8.2 ruamel.yaml==0.15.81 ply==3.11 torch==1.5.1
Extract the ACUITY package, Verisilicon_Tool_Acuity_Toolkit_<version>.tgz, to a destination directory. There are two versions of the toolkit, Please confirm the toolkit version matches the Python version that needs to be installed：
acuity-toolkit- whl-6.18.1-python-xxx
acuity-toolkit- whl-6.21.1-python-xxx
From this directory, locate and unpack the binary package acuity-toolkit-whl-<version>.tgz. select and install a Wheel package with the following commands:
pip3 install ./acuity-*.whl
4.2. Toolkit Verification
Switch to the ./acuity-toolkit-whl-<version>/bin directory.
Execute python file
user@linux:~/c3v/acuity-toolkit-whl-6.21.1-python3.8.10/bin$ python pegasus.py -h
usage: pegasus [-h] {import,export,generate,prune,inference,quantize,train,dump,measure,help} ...
Pegasus commands.
positional arguments:
{import,export,generate,prune,inference,quantize,train,dump,measure,help}
import Import models.
export Export models.
generate Generate metas.
prune prune models.
inference Inference model and get result.
quantize Quantize model.
train Train model.
dump Dump model activations.
measure Get amount of calculation, parameter and activation.
help Print a synopsis and a list of commands.
optional arguments:
-h, --help show this help message and exit