xplai 0.1.9

xpl.ai client SDK.

Request early access from xpl.ai to get login credentials. ##CLI Usage

Authentication

Login.

xpl auth login

Options:
    -u, --username
    -s, --secret

Logout

    xpl auth logout

Tasks

Create task draft

xpl task draft

    Options:
    -n, --name
    -m, --modality: # 'image', 'video', 'audio', 'text'
                    # Currently supported: 'image'
    -c, --csv: # A path to csv file containing training examples that will be used to pretreain the model.

Initialize Active learning pipeline from the task.

• Model for the task will be assembled,
• Data pipelines will be initialized
• Sample examples for pretraining will be uploaded
• Training from provided examples will start

xpl task commit

List tasks

xpl task list

Describe task's detail information

xpl task describe

    Options:
    -t, --task_id: # an id of the task. Can be obtained from "xpl task list" command

Annotations

List active annotation jobs that are associated with the task. This will return list of urls, corresponding toi annotation jobs. Url directs to the annotation UI.

xpl annotations list

    Options:
    -t, --task_id: # an id of the task. Can be obtained from "xpl task list" command

Concepts

Search XPL Platform for the concepts by the lemma

xpl concept search

    Options:
    -l, --lemma: # A single word human-readable lemma that describes concept in English

##End-user integration The core of end-user integration is a Task object. ###Instantiate Task object

task = Task(task_id='b622d1e8bd0d43d0be184239bf9ce53d',
            task_api_key='a714e3ce19594872be7fb9042d02134b')
            # task_api_key can be obtained from cli: "xpl task describe"

Instantiating Task object will:

• verify validity of the task's api key,
• check the active version of the model
• if needed: fetch the model from storage using secured http channel and cache it on the disk
• load model and will be ready to accept data for inferences ###Run task on data of corresponding modality

from xplai import Task

image_path = '/data/image.png'
concepts_detected = task.run(source=image_path)

# do something useful
for concept in concepts_detected:
    print(f'detected {concept.display_name} in {image_path}')