tensorflow implementation added

Author: graviraja

README.md

@@ -2,3 +2,4 @@
 ## This repo contains sample codes in pytorch
 
 1. basics.py : contains the simple 2 layer network implemented in numpy and pytorch
+2. static_graph.py : contains the same 2 layer network implemented in tensorflow

static_graph.py

@@ -0,0 +1,67 @@
+""" Implementing the 2 layer network defined in basics.py
+using tensorflow, for understanding the difference between
+dynamic graph (pytorch) and static graph (tensorflow).
+"""
+
+import tensorflow as tf
+import numpy as np
+
+# ################################################################### #
+# Implementing a 2 layer network with 1 hidden layer using tensorflow #
+# ################################################################### #
+
+N = 64 # batch size
+D_in = 1000 # input dimension
+H = 100 # hidden dimension
+D_out = 10 # output dimension
+
+# define the placeholders for the input and target data
+# these will be filled when the graph is executed.
+x = tf.placeholder(dtype=tf.float32, shape=(None, D_in), name="inputs")
+y = tf.placeholder(dtype=tf.float32, shape=(None, D_out), name="targets")
+
+# create the weights and initialize them with random data
+# A tensorflow variable persists its value across executions of graph.
+w1 = tf.Variable(tf.random_normal((D_in, H)), name="w1")
+w2 = tf.Variable(tf.random_normal((H, D_out)), name="w2")
+
+# forward pass
+# this doesn't perform any numeric operations
+# it merely sets up the computational graph that we will execute later.
+h = tf.matmul(x, w1)
+h_relu = tf.maximum(h, tf.zeros(1))
+y_pred = tf.matmul(h_relu, w2)
+
+# loss
+loss = tf.reduce_sum((y_pred - y) ** 2.0)
+
+# backward propagation
+# compute gradients
+w1_grad, w2_grad = tf.gradients(loss, [w1, w2])
+
+learning_rate = 1e-6
+
+# update the weights
+# to update the weights we need to evaluate new_w1 and new_w2 when executing the graph.
+# In tensorflow updating the weights is part of the computational graph.
+# In pytorch this happens outside the computational graph.
+new_w1 = w1.assign(w1 - (learning_rate * w1_grad))
+new_w2 = w2.assign(w2 - (learning_rate * w2_grad))
+
+print('-------------------------------------')
+print("Training the network using tensorflow")
+print('-------------------------------------')
+
+with tf.Session() as sess:
+ # initialize all the variables with their initializers.
+ sess.run(tf.global_variables_initializer())
+
+ # create the data
+ x_value = np.random.randn(N, D_in)
+ y_value = np.random.randn(N, D_out)
+
+ for epoch in range(500):
+ # feed the x_value and y_value to x, y placeholders.
+ loss_value, _, _ = sess.run([loss, new_w1, new_w2], feed_dict={x: x_value, y: y_value})
+ if epoch % 50 == 0:
+ print(f"epoch : {epoch}, loss : {loss_value}")