efemarai 0.3.1

A CLI and SDK for interacting with the Efemarai ML testing platform.

A CLI and SDK for interacting with the Efemarai ML testing platform.

Setup

Install with

pip install -U efemarai

then run

$ efemarai init

and following the instructions to connect your account.

Example Usage

Working with projects

import efemarai as ef

session = ef.Session()

# List all projects
for project in session.projects:
    print(project.name)

# Create a new project
project = session.create_project(
    name="Aircraft Detection",
    description="Exmample object detection project",
    problem_type="ObjectDetection",
)

# Load an existing one
project = session.project("Aircraft Detection")

# Create new dataset
dataset = project.create_dataset(
    name="Example COCO Dataset",
    stage=ef.DatasetStage.Test,
    format=ef.DatasetFormat.COCO,
    data_url=root,
    annotations_url=annotations_url,
)

# Define the path to the model yaml
local_model_config = ef.Session._load_config_file("model.yaml")

## Create model
model = project.create_model(**model_config)

# List project models
for model in project.models:
    print(model.name)

# List project datasets
for dataset in project.datasets:
    print(dataset.name)

Working with stress tests

The best way to create a domain is to use the UI and manually inspect the various transformations and how they affect the images.

# Create a new stress test
test = project.create_stress_test(
    name="Test via SDK",
    model=project.model("COCO instances RCNN-R50"),
    domain=project.domain("Example Domain"),
    dataset="Example COCO Dataset", # Just a name also works
)

# Load an existing stress test
test = project.stress_test("Test via SDK")

# Download dataset with discovered vulnerabilities
dataset_filepath = test.vulnerabilities_dataset()

# Check test run state
print(f"Running: {test.running} Failed: {test.failed} Finished: {test.finished}")

Models, domains and datasets can be easily created programatically, but they require quite a few configuration paramaters to be provided. That's why the most convenient way to create a project with multiple models, domains and datasets is to put everything into a config file (see e.g. examples/aircraft_project.yaml) and then just load it with:

result = ef.Session().load("examples/aircraft_project.yaml")

# access the created entities
project = result["project"]
models = result["models"]
domains = result["domains"]
datasets = result["datasets"]