NPU Compiler Installation*

1. Environment Setup*

Python Environment

Supports Linux with Python 2.7, Python 3.6, Python 3.7.

Note: If users need to compile and convert PyTorch models, the NPU compiler version must be at least 1.6.0b0 or above, and Python 3.7 must be used.
Frontend Deep Learning Framework Version Requirements
- TensorFlow 1.4 - 1.15
- PyTorch 1.10 - 1.13

2. Install the gxnpuc toolchain*

$ pip install npu_compiler

3. Update gxnpuc toolchain*

$ pip install --upgrade npu_compiler

4. Check the toolchain version*

After the installation or update is complete, you can use the following command to view the version number of the current toolchain.

$ gxnpuc --version

important

If the following error appears when running gxnpuc --version, please refer to the following two methods to solve it:

$ gxnpuc --version
bash: gxnpuc: command not found

Method 1: Confirm whether the gxnpuc file is in ~/.local/ bin, if yes, add ~/.local/bin to PATH In the directory, add method: add the following statement in ~/.bashrc
```
export PATH=~/.local/bin:$PATH
```

Method 2: Check the path of gxnpuc through import npu_complier:

$ python
Python 3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56)
[GCC 7.2.0] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> import npu_compiler as nc
>>> help(nc)

The output is as follows:

Help on package npu_compiler:

name
    npu_compiler - #coding: utf-8

PACKAGE CONTENTS
    __main__
    v100 (package)
    v150 (package)
    v160 (package)

DATA
    VERSION = '1.5.6'

FILE
    /home/jindg/anaconda3/lib/python3.6/site-packages/npu_compiler/__init__.py

Finally, change /home/jindg/anaconda3/lib in the log above to /home/jindg/anaconda3/bin</ font>, and then add it to PATH in ~/.bashrc.

Attention

If the problem still cannot be solved, please contact NationalChip engineers.