WEBVTT 00:00.080 --> 00:01.040 Welcome back. 00:01.040 --> 00:07.920 Now let's start with the first step which is preparing the data set for our model. 00:07.960 --> 00:14.680 Open Google Colab using Colab Google URL and click Open Colab. 00:14.840 --> 00:17.000 Go to upload browse. 00:17.200 --> 00:23.480 And here we have inside this folder using TensorFlow. 00:23.800 --> 00:25.520 This is the file that we need. 00:25.520 --> 00:27.960 Number six click open. 00:28.120 --> 00:35.280 And by the way all those files are available in the resources folder under each section. 00:35.280 --> 00:40.960 Or you can download them from the resources folder inside the first section. 00:40.960 --> 00:50.440 So here we have this file I want from you to follow me step by step linear regression with TensorFlow. 00:50.440 --> 01:02.680 In the previous videos and in the previous notebooks, we used the sky learn NumPy, pandas, and matplotlib 01:02.680 --> 01:03.480 libraries. 01:03.520 --> 01:10.500 Now we're going to use some of those libraries, but we're going to introduce the TensorFlow library 01:10.500 --> 01:11.860 in this model. 01:11.860 --> 01:22.580 In the previous videos we get um, an equation a learned equation saying that price equals to 2.9 times 01:22.620 --> 01:24.660 size plus 2.7. 01:24.980 --> 01:29.020 This is the final conclusion from the previous model. 01:29.300 --> 01:36.420 If you if you remember now we're going to start with with TensorFlow from zero. 01:36.540 --> 01:40.540 We're going to not tell him this this equation. 01:40.540 --> 01:44.500 We need to see the and get to this equation. 01:44.500 --> 01:49.780 We need to conclude this equation as a final result okay. 01:49.820 --> 01:54.300 So let's start building our model with TensorFlow. 01:54.420 --> 01:59.300 The first step we need to do is to install TensorFlow. 01:59.580 --> 02:05.660 So here let's start with pip install TensorFlow. 02:05.860 --> 02:13.000 If you are using Ides like a PyCharm, Jupyter or others. 02:13.000 --> 02:18.040 You need to use pip install TensorFlow and not the exclamation mark. 02:18.080 --> 02:25.760 Let me run here we are using the Python package installer to install TensorFlow and here we go. 02:25.920 --> 02:33.760 By the way, those libraries are already installed because a requirement already satisfied by Google 02:33.760 --> 02:34.640 Colab. 02:34.640 --> 02:37.720 So there is no need to execute this step. 02:37.720 --> 02:45.480 But I'm showing you if you are using the pre the Ides on your computer like pie chart. 02:45.640 --> 02:48.240 Okay, so I'll clear the output. 02:48.240 --> 02:49.680 This is the first step. 02:49.920 --> 02:51.720 This is the second step. 02:51.720 --> 02:58.640 Installing TensorFlow by using import TensorFlow as tf. 02:59.200 --> 03:02.840 And we need to print the TensorFlow version. 03:02.840 --> 03:08.040 Let me run and here we go TensorFlow version is 2.9. 03:08.240 --> 03:16.820 Always use the TF dot version or any library version to check the version and to check if the library 03:16.820 --> 03:20.300 is already installed correctly or not. 03:20.420 --> 03:25.500 The second step is generating random data. 03:25.500 --> 03:30.900 So here installing TensorFlow and here generating random data. 03:30.940 --> 03:40.420 Let me start by importing some libraries that we are going to use numpy as np, import pandas as pd 03:40.900 --> 03:44.540 and import matplotlib.pyplot as plt. 03:44.780 --> 03:52.740 And as we did in the previous videos, we need to set random seeds for reproducibility. 03:52.740 --> 04:01.820 So NP random seed 42 and tf which is TensorFlow random seed 42. 04:01.860 --> 04:12.020 Anytime you call numpy random function like NP random dot rand or NP random dot shuffle, etc., you'll 04:12.020 --> 04:20.970 get the same random results and you will set the same number each run as long as the seed is fixed. 04:21.010 --> 04:23.130 Let me show you an example here. 04:23.170 --> 04:27.370 Print NP dot random dot rand. 04:27.610 --> 04:30.450 And here I need to pass number three. 04:30.690 --> 04:32.370 So let me run. 04:32.410 --> 04:32.930 Okay. 04:33.210 --> 04:34.490 Those are the numbers. 04:34.490 --> 04:35.890 Let me run again. 04:35.930 --> 04:37.450 Again and again. 04:37.650 --> 04:41.730 So every time you run this you get the same results. 04:41.770 --> 04:51.890 So those results would be and those functions are going to be used for generating random numbers. 04:51.890 --> 04:57.370 But then we're going to fix them and we're going to use them each time. 04:57.370 --> 05:01.730 So every time we run the application we'll get the same results. 05:01.770 --> 05:02.210 Okay. 05:02.250 --> 05:11.890 So those are used for getting the same results, the same random numbers each run as long as the seed 05:11.890 --> 05:12.650 is fixed. 05:12.770 --> 05:22.390 TensorFlow tf dot random dot set seed TensorFlow also has its own random number generator independent 05:22.390 --> 05:23.390 from numpy. 05:23.750 --> 05:33.190 This sets the seed for TensorFlow's random operation like random dot uniform or random dot normal random 05:33.190 --> 05:37.390 initialization of weights in neural networks. 05:37.430 --> 05:46.750 Okay, so if your project uses both NumPy and TensorFlow, setting seeds in both ensures your entire 05:46.750 --> 05:49.270 pipeline is reproducible. 05:49.270 --> 05:56.710 Data shuffling, weight initialization, and etc. etc. without this, every run might produce slightly 05:56.710 --> 06:01.070 different results, making debugging and comparison harder. 06:01.310 --> 06:04.990 Okay, so in short np.random.seed. 06:04.990 --> 06:15.510 And this is for controlling randomness from and py and this will control the randomness from TensorFlow.