Step-by-Step Guide to Install ModelScope and Perform NLP Inference in Python & PHP

ModelScope Overview

ModelScope is an open‑source Model‑as‑a‑Service platform that provides a unified API for AI models, enabling developers to load, run and fine‑tune models with a single call.

Environment Setup

Python and Conda

Download and install Anaconda, then create an isolated Conda environment for ModelScope.

wget https://repo.anaconda.com/archive/Anaconda3-2024.02-1-Linux-x86_64.sh
sh Anaconda3-2024.02-1-Linux-x86_64.sh
conda create -n modelscope-env python=3.10
conda activate modelscope-env
# If "conda activate" fails, run:
source ~/.bashrc

PyTorch

Install the PyTorch stack (CPU or CUDA wheels are selected automatically).

pip3 install torch torchvision torchaudio

ModelScope Python library

pip3 install modelscope
pip install setuptools_scm
# Install NLP extras and point to the ModelScope release index
pip3 install "modelscope[nlp]" -f https://modelscope.oss-cn-beijing.aliyuncs.com/releases/repo.html

Verification

Run a quick inference to confirm that the library and its dependencies are functional.

python -c "from modelscope.pipelines import pipeline; print(pipeline('word-segmentation')('今天天气不错，适合出去游玩，你说呢'))"

Text Classification with ModelScope

Typical classification scenarios include:

Sentiment analysis – classify a sentence as positive, negative or neutral.

Sentence similarity – decide whether two sentences convey the same meaning.

Natural Language Inference (NLI) – determine logical entailment, contradiction or neutrality between a pair of sentences.

Custom label mapping – assign user‑defined categories to input texts.

ModelScope performs inference through the pipeline helper.

from modelscope.pipelines import pipeline
pipeline_ins = pipeline(task='nli', model='damo/nlp_structbert_nli_chinese-base')
result = pipeline_ins(input=('四川商务职业学院和四川财经职业学院哪个好？',
                          '四川商务职业学院商务管理在哪个校区？'))
print(result)
# Expected output (truncated)
# {'scores': [0.8383, 0.0834, 0.0784], 'labels': ['中立', '矛盾', '蕴涵']}

Calling ModelScope from PHP via PHPY

Build and install PHPY

PHPY is a PHP extension that embeds a Python interpreter, allowing PHP code to import and call Python modules such as ModelScope.

# Clone the repository
git clone https://github.com/swoole/phpy.git
cd phpy/

# Generate the extension build configuration
/usr/local/php-8.2.14/bin/phpize --with-php-config=/usr/local/php-8.2.14/bin/php-config

# Configure the build, pointing to the Conda environment that contains ModelScope
./configure --with-php-config=/usr/local/php-8.2.14/bin/php-config \
    --with-python-dir=/home/www/anaconda3/envs/modelscope-env \
    --with-python-version=3.10

# Compile and install
make -j4
sudo make install

# Enable the extension
echo "extension=phpy.so" >> /usr/local/php-8.2.14/etc/php.ini

# Verify the installation
/usr/local/php-8.2.14/bin/php -m | grep phpy

PHP script for NLI inference

<?php
$pipeline = PyCore::import('modelscope.pipelines')->pipeline;
$pipeline_ins = $pipeline('nli', 'damo/nlp_structbert_nli_chinese-base');

$inputs = [
    "四川商务职业学院和四川财经职业学院哪个好？",
    "四川商务职业学院商务管理在哪个校区？"
];

echo $pipeline_ins(new \PyTuple($inputs));
?>

The script prints a dictionary with scores and labels identical to the Python example, demonstrating that ModelScope can be accessed from PHP.