在nvidia-docker容器中执行pip3 install kerosene时遇UnicodeDecodeError

Hey there, let's work through this UnicodeDecodeError you're running into when trying to pip3 install kerosene in your nvidia-docker container. The error stems from the fuel dependency (required by kerosene) attempting to read its long description file using ASCII encoding, but the file contains non-ASCII characters—something Python 3.5's default file handling doesn't handle gracefully.

Here are a few straightforward fixes to resolve this:

1. Set UTF-8 Environment Variables Before Installing #

The quickest fix is to force Python to use UTF-8 encoding by setting environment variables right before your pip command. Run this in your container:

LC_ALL=C.UTF-8 pip3 install kerosene

Or if you prefer to set the variables first (useful if you're running multiple commands):

export LC_ALL=C.UTF-8
export LANG=C.UTF-8
pip3 install kerosene

This tells Python to use UTF-8 instead of ASCII when reading text files, which should bypass the decoding error entirely.

2. Manually Fix the Fuel Setup.py (If Environment Variables Don't Work) #

If the first method fails, you can directly modify the fuel package's setup script to specify the correct encoding:

1. Clone the fuel repository:

   git clone https://github.com/mila-udem/fuel.git
   cd fuel
   

2. Update the setup.py file to explicitly use UTF-8 when reading the description:

   sed -i 's/f.read().strip()/f.read(encoding="utf-8").strip()/' setup.py
   

3. Install the modified fuel package:

   pip3 install .
   

4. Now install kerosene:

   pip3 install kerosene
   

3. Upgrade Python (If Possible) #

Python 3.5 is pretty old, and newer Python versions (3.6+) handle file encoding more intelligently (defaulting to UTF-8 in most cases). If your container setup allows it, upgrading to a newer Python base image might prevent this and other similar issues down the line.

Give these steps a shot—option 1 is almost guaranteed to work since encoding mismatches like this are super common in older container environments.

内容的提问来源于stack exchange，提问作者Austin