Downloading an Open Source LLM

First we need to download the Google Flan model from Huggingface and then store it in our S3 bucket for the model server to access.

We will use a workbench and the data connection that you created in the Storage Data Connections section..

Prerequisites

  • You created the data connection My Storage.

    Workbench data connection form

  • You added the My Storage data connection to your workbench.

    Workbench edit form

Procedure

First, let’s navigate to the relevant notebooks.

  1. Navigate to test-drive/llm

  1. In your notebook environment, open the file 1_download_save.ipynb

  2. Follow the instructions directly in the notebook.

  3. The instructions will guide you through the process of downloading the model from Huggingface and then uploading it to your models bucket.

Jupyter Notebook 1

Verification

When you have completed the notebook instructions, you should see files listed in the directory/prefix: models/flan-t5-small

models/flan-t5-small/README.md
models/flan-t5-small/config.json
models/flan-t5-small/flax_model.msgpack
models/flan-t5-small/generation_config.json
models/flan-t5-small/model.safetensors
models/flan-t5-small/pytorch_model.bin
models/flan-t5-small/special_tokens_map.json
models/flan-t5-small/spiece.model
models/flan-t5-small/tf_model.h5
models/flan-t5-small/tokenizer.json
models/flan-t5-small/tokenizer_config.json