telesto-base

Base Docker image and tools for telesto.ai models.

Instructions

telesto-base contains a pip-installable Python package and a Docker image, allowing you to easily package your models for telesto.ai competitions.

telesto-base package

To install the module, you can simply use pip:

pip install telesto-base

If you would like to use the latest not yet released version, you can install the one in the develop branch.

pip install git+https://github.com/telesto-ai/telesto-base.git@develop

The base image

The base image contains the pre-installed telesto-base module. Your submissions will use this as a base, so you'll only have to worry about the algorithms and not the packaging. To use it locally, you can pull the image from Docker Hub:

docker pull telestoai/model-api-base:latest

Alternatively, the image can also be built locally with the command

docker build -t telestoai/model-api-base -f Dockerfile .

An example model

If you are stuck on how to prepare your model for submission, we have prepared a concrete example for you. The example is available in the telesto-models repository with further instructions on the usage.

Test classification model API

Build and start a container

docker build -t telestoai/model-api-base -f Dockerfile .
docker run -p 9876:9876 --name model-api-base --rm --env USE_FALLBACK_MODEL=1 \
    telestoai/model-api-base classification

Send a sample input

curl -X POST -H "Content-Type:application/json" --data-binary @tests/data/class/example-input.json -i \
    http://localhost:9876/
...
{
    "predictions": [
        {"probs": {"cat": 0.32015, "dog": 0.67985}, "prediction": "dog"},
        {"probs": {"cat": 0.81545, "dog": 0.18455}, "prediction": "cat"}
    ]
}

Test segmentation model API

Build and start a container

docker build -t telestoai/model-api-base -f Dockerfile .
docker run -p 9876:9876 --name model-api-base --rm --env USE_FALLBACK_MODEL=1 \
    telestoai/model-api-base segmentation

Post a sample input

curl -X POST -H "Content-Type:application/json" --data-binary @tests/data/segm/example-input.json -i \
    http://localhost:9876/jobs
...
{
    "job_id": "b741bd19767441f6b7abd022744083c9"
}

Get the result

curl -H "Content-Type:application/json" -i http://localhost:9876/jobs/b741bd19767441f6b7abd022744083c9
...
{
    "mask": {
        "content": "<BASE_64_IMAGE>"
    }
}