nllb-serve-slim

Dockerfile

@@ -3,14 +3,18 @@ FROM python:3.8.14-slim-buster
 RUN apt-get update && apt-get install -y \
     build-essential \
     && rm -rf /var/lib/apt/lists/*
-
+	
 WORKDIR /app
 
-RUN pip3 install --upgrade pip
-
 COPY ./requirements.txt ./requirements.txt
+COPY . /app
 
 # Install package from source code
-RUN pip3 install -r ./requirements.txt
+RUN pip install --upgrade pip && \
+    pip install torch==1.13.1+cpu torchvision==0.14.1+cpu -f https://download.pytorch.org/whl/torch_stable.html && \
+    pip install -r requirements.txt
+
+EXPOSE 6060
 
-EXPOSE 5001
+# Command to start the server
+CMD ["python", "-m", "nllb_serve"]

README.md

@@ -21,6 +21,11 @@ nllb-serve -h
 python -m nllb_serve -h
 ```
 
+## Docker
+
+    docker build -t nllb-serve-cpu .
+    docker run -p 6060:6060 --name nllb-cpu-trad nllb-serve-cpu
+
 ## Start server
 
 ```bash
@@ -131,6 +136,9 @@ options:
                         Batch size; number of sentences (default: 10)
 ```
 
+## Additional
+If you want to use this server to translate web pages with the Firefox/Chrome browser, you can use the extension   [linguist-NLLB-translator](https://github.com/MoonDragon-MD/linguist-NLLB-translator)
+
 ## License
 
 The code and model weights carry different licenses. 

nllb_serve/app.py

@@ -18,8 +18,8 @@
 from transformers import AutoModelForSeq2SeqLM, AutoTokenizer
 
 from . import log, DEF_MODEL_ID
-
-device = torch.device(torch.cuda.is_available() and 'cuda' or 'cpu')
+os.environ['CUDA_VISIBLE_DEVICES'] = ''
+device = torch.device('cpu')
 log.info(f'torch device={device}')
 
 #DEF_MODEL_ID = "facebook/nllb-200-distilled-600M"

requirements.txt

@@ -1,7 +1,8 @@
-transformers>=4.21
+torch==1.13.1+cpu
+transformers==4.37.0
 Flask==3.0.2
 Werkzeug>=3.0.0
-torch>=1.12
-gunicorn
+gunicorn==20.1.0
 gevent
 sentence-splitter
+sentencepiece==0.1.99