Aditya Agrawal adityaiitb

Create binding for a simple function

#include<pybind11/pybind11.h>

int add(int i, int j) {
    return i + j;
}

List of TF ops

A graph def has nodes. Every node has

Op: E.g. AddV2, Placeholder, Const.
Name (namescope): This is also the name of the output tensor(s). To refer to a specific index, you can say "name:0".
Inputs (optional): E.g. Const and Placeholder ops do not have any inputs.

Placeholder ops are fed from the user input? Const ops are fed from the variables folder?

Documentation.

import tensorflow as tf

raw_dataset = tf.data.TFRecordDataset("foo.tfrecord")

for raw_record in raw_dataset.take(1):
    example = tf.train.Example()
 example.ParseFromString(raw_record.numpy())

The json.tool module provides a simple command line interface to validate and pretty-print JSON objects. link

python -m json.tool foo.json

Frozen graph is deprecated in TF2. Using Saved model is the new approach. Instead of creating a configuration file and using bazel in steps 1 and 2 described here, simply use the saved_model_cli. The final compilation step of creating a binary can be performed using a makefile instead of bazel.

Create a TF saved_model using tf.saved_model.save.
Use the saved_model_cli to perform XLA AOT compilation.

saved_model_cli aot_compile_cpu \
    --dir resnet/1 \
    --tag_set serve \
 --signature_def_key serving_default \

tf.function compiles a function into a callable TensorFlow graph. Documentation

@tf.function
def f(x):
  return x + 1

To obtain a graph / graphdef

Use the get_concrete_function method of the callable created by tf.function. It returns a tf.Graph object.

	# Data file
	# 1 10
	# 2 15
	# 3 50
	# 4 ...
	# ....

	# Can replace x2ticlabel with xticlabel as well
	# When column(1) modulo 8 is zero, use "2^{....}" else ""

	# line styles for ColorBrewer Set3
	# for use with qualitative/categorical data
	# provides 8 colors that are more saturated than the Pastels but less so than Set2
	# compatible with gnuplot >=4.2
	# author: Anna Schneider
	# Modified by Aditya

	# line styles
	set style line 1 lc rgb '#a6cee3'
	set style line 2 lc rgb '#1f78b4'

	# Show all rows
	pd.set_option('display.max_rows', None)

	# Reset index after an operation
	df.sort_values(['country', 'capital']).reset_index(drop=True)

	#!/usr/bin/env python3

	import tensorflow as tf
	from tensorflow_serving.apis import predict_pb2
	from tensorflow_serving.apis import prediction_service_pb2_grpc
	import numpy as np
	import grpc

	channel = grpc.insecure_channel("localhost:8500")
	stub = prediction_service_pb2_grpc.PredictionServiceStub(channel)