Here in this example, we created a simple classification model for detecting Skin Cancer patients by classifying their skin specimen Image as Benign (Non-Cancer) or Malignant (Possibly cancerous) one. 

Training Specimen :

Output:

Steps

1) Search for 'inject' node which looks like this: 

 Search for the 'function' node which looks like this :

The 'console' node is optional and can be used for viewing console output when development is needed, search and drop in the editor. This node looks like this :

2) Drop and connect these nodes to create flow as shown in figure (for this example, we named:  to host our main flow, Here We renamed "Python function" to "Dir-maker" which can be done by double-clicking and giving it name using "Name" : 

3) Prior to training a model (In this case of the classification model), we need to create a directory structure for the project which is as : 

Here we are making a python script that simplifies the process of generating a directory for our models, specimens, and results.

Dir-Maker

import os

path = './public/private/Image_Classification/'
project_name = path + "skin-cancer"
required_dir_struct = ['Dataset','Temp','Executed_NB','Models','Results','QueryImages']
for folder in required_dir_struct:
try:
os.mkdir(project_name+'/'+folder)
except OSError as error:
print(error)
status = 'Dir struct has been created !'
msg['payload'] = status
return status

This will create Directories as a script for viewing directory click on:

We can see the directory structure Generated as:

4) Create a new node with the same 'inject', 'Python 3' (Trainer) Node, 'HTTP output', 'console' and connect them as shown in the figure :Here "Trainer" same node renamed from "Python 3 function".

The Fastai framework is used for training a model for this project as the Fastai background works on Jupyter notebook.

Code in: Trainer

import subprocess
import papermill as pm
import os
from os import path
# Fastai : 1.0.61 : path : /interplay_v2/public/private/Fastai-ENV/bin/python
# Dependency : p2j , papermill , python kernal
# jupyter lab : https://github.com/jupyterlab/debugger
#---------------------------------------------------------------------------------------------------+
# PATHS
#
#---------------------------------------------------------------------------------------------------+
def execute():
result_json=pm.execute_notebook("/interplay_v2/public/private/Image_Classification/skin-cancer/ImgClassification-v2
-Generic.ipynb","/interplay_v2/public/private/Image_Classification/skin-cancer/Executed_NB/executed-img-class-v2.ipynb",
parameters = dict(
CYCLE = 1 ,
PROJECT_NAME = "skin-cancer",
INTERPLAY_ROOT =
"/interplay_v2/public/private/Image_Classification/",
))
return result_json
final_execute = execute()
 msg['payload'] = final_execute

5) Click on "Deploy", after Successful deployment notification, we can trigger nodes as we desire,

6) Click on the Trigger point of the inject node (Highlighted) :

7) After triggering the second node initiates the Jupyter notebook consisting of Training AI with predefined parameters, we can see the console parameter output similar to this :

after triggering the third flow (Predictor) wait for some time for classification output and segregated as Benign and Malnigant one are stored in our working directory in the 'Results' folder.