Running Pytorch with KoBERT on MacOS Devices

安裝 Google sentencepiece

brew install cmake
git clone https://github.com/google/sentencepiece.git
mkdir build && cd build && cmake ..
make -j $(nproc)
sudo make install
sudo update_dyld_shared_cache
cd ../python
python setup.py bdist_wheel
pip install dist/sentencepiece*.whl

安裝依賴

pip install torch transformers gluonnlp pandas tqdm sentencepiece protobuf

安裝並運行

from transformers import AutoModel, AutoTokenizer
import torch

# Load the model and tokenizer
model = AutoModel.from_pretrained("monologg/kobert")
tokenizer = AutoTokenizer.from_pretrained("monologg/kobert", trust_remote_code=True)

# Example text (in Korean)
text = "안녕하세요, KoBERT를 사용해봅시다."

# Tokenize the text
inputs = tokenizer(text, return_tensors="pt")

# Get the model's output
with torch.no_grad():
    outputs = model(**inputs)

# Print the output
print(outputs)