xtuner

Docker images for the XTuner library for tuning large language models.

Available versions:

• 0.1.18 (CUDA 11.7)
• 0.1.15 (CUDA 11.7)
• 2024-02-19 (CUDA 11.7)
• 0.1.14 (CUDA 11.7)

Huggingface restricted access

In case models or datasets require being logged into Huggingface, you can give your Docker container access via an access token.

Create access token

In order to create an access token, do the following:

• Log into https://huggingface.co
• Go to Settings -> Access tokens
• Create a token (read access is sufficient, unless you want to push models back to huggingface)
• Copy the token onto the clipboard
• Save the token in a .env file, using HF_TOKEN as the variable name

Provide token to container

Add the following parameter to make all the environment variables stored in the .env file in the current directory available to your Docker container:

--env-file=`pwd`/.env

Log into Huggingface

With the HF_TOKEN environment variable set, you can now log into Huggingface inside your Docker container using the following command:

huggingface-cli login --token=$HF_TOKEN