Job New SDK

Overview

When a job is created it returns info and the job_id. job_id is what's used to make changes to other objects that interact with the job.

SDK

Function

Python

dir = project.directory.get(name = 'Default')
member_list = project.get_member_list()
member_list_ids = [x['member_id'] for x in member_list]
job = project.job.new(
    name = "my job test pablo 222",
    instance_type = "box",
    share = "Project",
    sync_directories = [dir],
    label_schema_id = 1,
    tag_list = ["super", "apples", "something"],
    members_list_ids = member_list_ids,
    auto_launch = True
)

Input Files, Create a new Guide, and Launch a Job Example

On this example we do some file uploads, then create a guide and finally launch a job with this guide and the Default dataset attached to it. The job is automatically launched to start labeling.

Python

from diffgram import Project
from glob import glob

project = Project(
  project_string_id = "replace_with_project_string",
  client_id = "replace_with_client_id",
  client_secret = "replace_with_client_secret")


directory_path = "../replace_with_your_directory/*"
path_list = glob.glob(directory_path)

file_list = []

for path in path_list:

  # Replace with from_url() method if possible
	file = project.file.from_local(path)
	file_list.append(file)

guide = project.guide.new(
			name = "Traffic lights",
			description_markdown = "my description"
			)

dir = project.directory.get(name = 'Default')
member_list = project.get_member_list()
member_list_ids = [x['member_id'] for x in member_list]
job = project.job.new(
		name = "my job",
		instance_type = "box",
		share = "Project",
		file_list = file_list,
		guide = guide,
    sync_directories = [dir],
    members_list_ids = member_list_ids,
    auto_launch = True
		)

After creating your job it can be viewed on the Diffgram.com UI. It can be launched from there with the same effect.

Label Schema

By default, a newly created job will include the default Label Schema, or you can specify a schema with label_schema_id

Launch()

The job can be launched by calling job.launch()

or by calling Job Launch API.

Note that jobs are created in a draft state by default, and must be launched to start syncing files.

Attaching Datasets with Job New

Note that there is a dict with the first key being a list attached_directories_list.

json

attached_directories_dict: 
{
  attached_directories_list: [
  	 {
      "selected": "sync",
      "directory_id": 2
    }
  ]
}