dipanjanS

import base64
import requests 

with open('sneaker.jpg', "rb") as imageFile:
    img_b64enc = base64.b64encode(imageFile.read())
    data = {'b64_img': img_b64enc}

API_URL = 'http://0.0.0.0:5000/apparel_classifier/api/v1/model2_predict'
# sending post request and saving response as response object
r = requests.post(url=API_URL, data=data)

dipanjanS

r = requests.get('http://0.0.0.0:5000/apparel_classifier/api/v1/liveness')
r.status_code, r.text

dipanjanS

# load dependencies
import warnings
warnings.simplefilter(action='ignore', category=FutureWarning)
import numpy as np
import requests
import base64
import json
from io import BytesIO
from flask import Flask, request, jsonify
from flask_cors import CORS

dipanjanS

# create serving function
INPUT_SHAPE_RN = (32, 32, 3)
model2 = create_cnn_architecture_model2(input_shape=INPUT_SHAPE_RN)
model2.load_weights('./model_weights/cnn_model2_wt.h5')

def predict_apparel_model2_regular(img, img_dims=(32,32), label_map=class_names):
    
    sample_img_processed = (np.array([resize_image_array(img, 
                                                         img_size_dims=img_dims) 
                                                for img in np.stack([[img]]*3, 

dipanjanS

# create serving function
def predict_apparel_model2_serving(img, img_dims=(32,32), label_map=class_names):
    
    sample_img_processed = (np.array([resize_image_array(img, 
                                                         img_size_dims=img_dims) 
                                                for img in np.stack([[img]]*3, 
                                                                    axis=-1)])) / 255.
    data = json.dumps({"signature_name": "serving_default", 
                       "instances": sample_img_processed.tolist()})
    

dipanjanS

%%time

sample_test_data = test_images
sample_test_labels = test_labels
IMG_DIMS = (32, 32)
sample_test_data_processed = (np.array([resize_image_array(img, 
                                                           img_size_dims=IMG_DIMS) 
                                            for img in np.stack([sample_test_data]*3, 
                                                                axis=-1)])) / 255.
data = json.dumps({"signature_name": "serving_default", 

dipanjanS

warmup_data = np.load('serve_warmup_data.npy')
warmup_model2_serve(warmup_data)

dipanjanS

%%bash --bg 
docker run --runtime=nvidia -p 8501:8501 \
  --mount type=bind,source=/home/jupyter/tensorflow_serving/tf_saved_models,target=/home/jupyter/tensorflow_serving/tf_saved_models \
  --mount type=bind,source=/home/jupyter/tensorflow_serving/models.conf,target=/home/jupyter/tensorflow_serving/models.conf \
  -t tensorflow/serving:latest-gpu --model_config_file=/home/jupyter/tensorflow_serving/models.conf

dipanjanS

%%time

sample_test_data = test_images
sample_test_labels = test_labels
IMG_DIMS = (32, 32)
sample_test_data_processed = (np.array([resize_image_array(img, 
                                                           img_size_dims=IMG_DIMS) 
                                            for img in np.stack([sample_test_data]*3, 
                                                                axis=-1)])) / 255.
data = json.dumps({"signature_name": "serving_default", 

dipanjanS

%%time

sample_test_data = test_images
sample_test_labels = test_labels
sample_test_data_processed = np.expand_dims(sample_test_data / 255., axis=3)
data = json.dumps({"signature_name": "serving_default", 
                   "instances": sample_test_data_processed.tolist()})

HEADERS = {'content-type': 'application/json'}
MODEL1_API_URL = 'http://localhost:8501/v1/models/fashion_model_serving/versions/1:predict'