E-mail writer

Usage

Create models directory on your project root then download this trained model on AESLC dataset and put it in the models directory. Then, you can run it as a container:

• Build the image docker build -f email_writer.Dockerfile -t email_writer . and run it with docker run -p 6060:6060 --name email_writer email_writer. To enable gpu, install container-toolki and run the container with docker run --gpus all -p 6060:6060 --name email_writer email_writer.
• Or you can run it direct by setting up the repo (preferably inside a venv) with pip install -r requirements.txt and pip install -e email_writer
  • Then, run the command uvicorn email_writer.main:app --port 6060 --host 0.0.0.0

To call the inference endpoint, just call http://0.0.0.0:6060/email with a GET request and header of Content-Type=application/json with the needed json body or directly by using swagger at http://0.0.0.0:6060/docs#/default/read_root_email_get.

Endpoint docs

• subject: subject of the email
• from: sender of the email
• to: receiver of the email
• salutation: salutation of the sender
• temperature: temperature used by model inference
• n_gen: number of examples generated

Example request:

{
"subject": "Interview Challenge",
"salutation": "Giovani nice to meet you last week",
"from": "[email protected]",
"to": "[email protected]",
"temperature": 0.7,
"n_gen": 4,
}