novasush

const functions = require("firebase-functions");
const cors = require("cors");
const express = require("express");
const bodyParser = require('body-parser');
const compression = require("compression");

// Express app config
const tasksApp = express();
tasksApp.use(compression());
tasksApp.use(bodyParser.json());

novasush

novasush

from tensorflow.python.compiler.tensorrt import trt_convert as trt

# Instantiate the TF-TRT converter
# Here saved model directory is the path to the saved model
# You can customise precision mode to FP32 FP16 or INT8

converter = trt.TrtGraphConverterV2(
   input_saved_model_dir=SAVED_MODEL_DIR,
   precision_mode=trt.TrtPrecisionMode.FP32
)

novasush

# Using base image provided by nginx unit
FROM nginx/unit:1.22.0-python3.9
# Alternatively you can use different tags from https://hub.docker.com/r/nginx/unit

COPY requirements.txt /fastapi/requirements.txt

RUN pip install -r /fastapi/requirements.txt

COPY config.json /docker-entrypoint.d/config.json

novasush

{
    "listeners": {
        "*:80": {
            "pass": "applications/fastapi"
        }
    },

    "applications": {
        "fastapi": {
            "type": "python 3.9",

novasush

from fastapi import FastAPI

app = FastAPI()


@app.get("/")
async def index():
    """
    A simple Hello World GET request
    """

novasush

class floatingrange(object):
    def __init__(self, start=None, stop=None, decimal=0):
        pass
            
    def __str__(self):
        pass
    
    def __repr__(self):
        pass
    

novasush

# Shuffle and batch the train_dataset. Use a buffer size of 1024
# for shuffling and a batch size 32 for batching. 
train_dataset = train_dataset.shuffle(1024).batch(32)
# Parallelize the loading by prefetching the train_dataset.
# Set the prefetching buffer size to tf.data.experimental.AUTOTUNE.
train_dataset = train_dataset.prefetch(tf.data.experimental.AUTOTUNE)

novasush

# Get the number of CPU cores. 
cores = multiprocessing.cpu_count()

print(cores)

# Parallelize the transformation of the train_dataset by using
# the map operation with the number of parallel calls set to
# the number of CPU cores.
train_dataset = train_dataset.map(read_tfrecord,num_parallel_calls=cores)

novasush

def read_tfrecord(serialized_example):
    
    # Create the feature description dictionary
    feature_description = {
        'image': tf.io.FixedLenFeature((), tf.string, ""),
        'label': tf.io.FixedLenFeature((), tf.int64, -1),
    }
    # Parse the serialized_example and decode the image
    example = tf.io.parse_single_example(serialized_example, feature_description)
    image = tf.io.decode_jpeg(example['image'], channels=3)