Docker images for Huggingface transformers.

Available versions:

• 4.43.1 (CUDA 12.1, CUDA 12.1 langchain, CUDA 12.1 langchain/optional RAG, CUDA 12.1 MMS)
• 4.42.4-post (CUDA 12.1, CUDA 12.1 langchain)
• 4.42.3 (CUDA 12.1, CUDA 12.1 langchain)
• 4.40.2 (CUDA 12.1, CUDA 12.1 langchain)
• 4.40.0 (CUDA 11.7, CUDA 11.7 langchain)
• 4.36.0 (CUDA 11.7, CUDA 11.7 Mistral, CUDA 11.7 Text classification)
• 4.35.0 (CUDA 11.7, CUDA 11.7 Mistral)
• 4.31.0 (CUDA 11.7, CUDA 11.7 with falcontune, CUDA 11.7 Llama2 (8bit), CUDA 11.7 QLoRA, CUDA 11.7 RemBERT, CUDA 11.7 translate)
• 4.7.0 (CUDA 11.1, CUDA 11.1 with finetune-gpt2xl)

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

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

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