NLP

SpaCy

Run SpaCy on GPU

Pre-install

Go to Install SpaCy and select your Hardware - GPU CUAD version
You should get the install command like pip install -U spacy[cuda92]
If you don't know which CUAD version using, use command nvcc --version to check
Or you can check the installed version on C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA in Windows

Runtime

import spacy
spacy.prefer_gpu()
nlp_model = spacy.load("en_core_web_sm")

HanLP

Run on GPU

In the offical document (maybe already out-dated), it mentioned that HanLP will look for the GPU on runtime
And no need for user to set whether HanLP using GPU or CPU
HanLP Documentation - Use GPUs There are no setting or validation for checking HanLP running on GPU
怎么查看GPU是否成功使用

Limit which GPU available

# macOS / Linux
export CUDA_VISIBLE_DEVICES=0,1
# Windows
set CUDA_VISIBLE_DEVICES=0,1

PyTorch / Tensorflow in HanLP

HanLP made dependency on PyTorch / Tensorflow, therefore, you need to set PyTorch and Tensorflow to use GPU as well

Troubleshoot

ModuleNotFoundError: No module named 'pip'

python -m ensurepip
python -m pip install --upgrade pip

Reference

Windows查看CUDA版本
Compare runtimes of spaCy NER pipelines using CPU and GPU
spaCy - Getting Started
解决 ModuleNotFoundError: No module named 'pip'