aribornstein

from azureml.core.webservice import AciWebservice, Webservice
from azureml.core.model import Model

aspect_lex = Model(ws, 'aspect_lex')
opinion_lex = Model(ws, 'opinion_lex')    

service = Model.deploy(workspace=ws,
                       name='absa-srvc', 
                       models=[aspect_lex, opinion_lex],
                       inference_config=inference_config, 

aribornstein

from azureml.core.model import InferenceConfig
inference_config = InferenceConfig(environment=deploy_env,
                                   entry_script="score.py")

aribornstein

from azureml.core import Environment
deploy_env = Environment.from_conda_specification('absa_env', "myenv.yml")
deploy_env.environment_variables={'NLP_ARCHITECT_BE': 'CPU'}

aribornstein

from azureml.core.conda_dependencies import CondaDependencies 

pip = ["azureml-defaults", "azureml-monitoring", "git+https://github.com/NervanaSystems/nlp-architect.git@absa"]

myenv = CondaDependencies.create(pip_packages=pip)

with open("myenv.yml","w") as f:
    f.write(myenv.serialize_to_string())

aribornstein

from azureml.core.webservice import AciWebservice

aciconfig = AciWebservice.deploy_configuration(cpu_cores=1, 
                                               memory_gb=1,  
                                               tags={"data": "text",  "method" : "NLP Architcet ABSA"}, 
                                               description='Predict ABSA with NLP Architect')
                                               
                                               

aribornstein

from azureml.core.compute import ComputeTarget, AmlCompute
from azureml.core.compute_target import ComputeTargetException

cluster_name = "cluster"
compute_config = AmlCompute.provisioning_configuration(vm_size='STANDARD_D3_V2', # VM Size N series have GPU
                                                       vm_priority='lowpriority', # Cheap Computer
                                                       min_nodes=0, # Auto Scale Min Limit
                                                       max_nodes=5) #Auto Scale Max Limit
cluster = ComputeTarget.create(ws, cluster_name, compute_config)
cluster.wait_for_completion(show_output=True)

aribornstein

from azureml.train.estimator import Estimator

script_params = {
    '--data_folder': ds,
}

nlp_est = Estimator(source_directory='.',
                   script_params=script_params,
                   compute_target=cluster,
                   entry_script='train.py',

aribornstein

########### Python Form Recognizer Train #############
from requests import post as http_post

# Endpoint URL
base_url = r"<Endpoint>" + "/formrecognizer/v1.0-preview/custom"
source = r"<SAS URL>"
headers = {
    # Request headers
    'Content-Type': 'application/json',
    'Ocp-Apim-Subscription-Key': '<Subscription Key>',

aribornstein

import requests
# If you are using a Jupyter notebook, uncomment the following line.
#%matplotlib inline
import matplotlib.pyplot as plt
from matplotlib.patches import Rectangle
from PIL import Image
from io import BytesIO

# Replace <Subscription Key> with your valid subscription key.
subscription_key = ""

aribornstein

//http://www.easyrgb.com/en/math.php

function tfRGB2LAB(rgb) {
  let a_con = tf.pow(rgb.add(0.055).div(1.055), 2.4);
  let b_con = rgb.div(12.92);
  let rgb_2 = tf.where(rgb.greater(0.04045), a_con, b_con);

  const [r, g, b] = rgb_2.split(3, 2);

  let x = r