This is a cog model for the all-mpnet-base-v2 sentence-transformers embedding model. This embedding model is based on MPNet and fine-tuned on 1 billion sentence pairs (see here for details).
GPU machine. You'll need a Linux machine with an NVIDIA GPU attached and the NVIDIA Container Toolkit installed. If you don't already have access to a machine with a GPU, check out our guide to getting a GPU machine.
Docker. You'll be using the Cog command-line tool to build and push a model. Cog uses Docker to create containers for models.
First, install Cog:
sudo curl -o /usr/local/bin/cog -L "https://github.com/replicate/cog/releases/latest/download/cog_$(uname -s)_$(uname -m)" sudo chmod +x /usr/local/bin/cog
From the root directory of this repo, run:
chmod +x scripts/download_and_prepare_model.py
cog run python scripts/download_and_prepare_model.py --config model_setup.yaml
You can run the model locally to test it:
cog predict -i text="You may know a word by the company it keeps."
Make sure to specify "private" to keep the model private.
Replicate supports running models on CPU or a variety of GPUs. The default GPU type is a T4 and that may be sufficient; however, for maximal batch side and performance, you may want to consider more performant GPUs like A100s.
Alternatively, if you will only be encoding single documents and you want to minimze spend at the cost of latency, you can run this model on CPU. You'll observe higher latencies, but this may be acceptable for your use case.
Click on the "Settings" tab on your model page, scroll down to "GPU hardware", and select "A100". Then click "Save".
Log in to Replicate:
cog login Push the contents of your current directory to Replicate, using the model name you specified in step 3:
cog push r8.im/username/modelname Learn more about pushing models to Replicate.
Now that you've pushed the model to Replicate, you can run it from the website or with an API.
To use your model in the browser, go to your model page.
To use your model with an API, click on the "API" tab on your model page. You'll see commands to run the model with cURL, Python, etc.
To learn more about how to use Replicate, check out our documentation.