Skip to content

Latest commit

 

History

History
104 lines (70 loc) · 2.95 KB

README.md

File metadata and controls

104 lines (70 loc) · 2.95 KB

Example task containers

This folder contains example task containers showing how to build your own task container.

Examples

Folder Description
/python-template A simple Python task container that returns the json input as output.
/knn A task container that uses the K-Nearest Neighbors algorithm to classify data.
/stablediffusion A task container that uses a prompt to generate an image.
/sgx-task A basic SGX container that runs code in a Trusted Execution Environment.

Structure

Each folder must contain all of the files listed in the structure as follows:

- folder
  - Dockerfile
  - src
  - description.txt
  - name.txt
  - example_input.bin
  - example_output.bin
  - config.json

Building a task container

1. App that handles requests

The app will receive a post request at localhost:8080\ and return the result. The input format is any json object and the output format is a string. After the app starts up, print "meca-init-done".

For example, we create a flask app:

  from flask import Flask, request
  import json

  app = Flask(__name__)

  @app.route('/', methods=['POST'])
  def hello_world():
      data = request.json
      return data['input']

  print("meca-init-done")

2. Dockerfile

Expose port 8080 to the app of the request handler.

Following the flask app created above, the Dockerfile may look like this:

FROM python:3.9-slim

WORKDIR /app

COPY flask_app.py flask_app.py

EXPOSE 8080

CMD ["python", "-m", "flask", "--app", "flask_app.py", "run", "--host=0.0.0.0", "--port=8080"]

3. config.json fields

The config.json file contains the configuration of the task, which it will use to build its container, with the following default fields:

DEFAULTS:

{
  "resource": {
    "cpu": 1,
    "mem": 128,
    "use_gpu": false, 
    "gpu_count": 0
  }
}
field description
resource.cpu The number of CPUs to allocate to the task. This field will limit the users that can run the task to those with the same or more CPUs available.
resource.mem The amount of memory to allocate to the task in MB. This field will limit the users that can run the task to those with the same or more memory available.
resource.use_gpu Whether the task requires a GPU to run.
resource.gpu_count The number of GPUs to allocate to the task, if use_gpu is true.

4. Describe your task

Fill in the description.txt and name.txt files with the description and name of your task. In the example_input.bin and example_output.bin files, provide examples of the input and output of your task exactly as they are read in the body of a HTTP request and reply.

5. Using and uploading the container

Build the image and push it to IPFS via the MECAnywhere CLI as a task developer. You will be compensated with the task fee that you list for each task executed by a MECAnywhere client.

Test your task folder structure by running the test function in the MECAnywhere CLI.