WEBVTT 00:00.080 --> 00:01.800 We trained our model. 00:01.800 --> 00:04.800 Now let's evaluate the model performance. 00:05.040 --> 00:10.400 Here we're going to create a new cell for plotting the training history. 00:10.400 --> 00:12.640 So plot training history. 00:12.920 --> 00:17.920 I'll start with a variable called history plot history. 00:17.920 --> 00:23.760 And if we go back to here let me see the train. 00:23.760 --> 00:25.240 The model history. 00:25.440 --> 00:27.880 History is the trained model. 00:27.880 --> 00:33.280 So let me scroll down and pass it to the plot history function. 00:33.320 --> 00:39.200 Here I created a new function called plot history and passed the history as its parameter. 00:39.240 --> 00:39.680 Okay. 00:39.720 --> 00:42.560 Plot training and validation loss values. 00:42.560 --> 00:44.400 Let's start with plt. 00:44.520 --> 00:56.400 As we did before, PLT is used from the library that used for plotting and making diagrams 12 by two 00:56.680 --> 01:03.720 figure size equals Matplot library and plt subplot. 01:03.760 --> 01:10.360 1 to 1 plot historyhistory the last and make the label. 01:10.400 --> 01:12.560 Set the label to training class. 01:12.600 --> 01:15.920 Make another plot for the values. 01:16.080 --> 01:18.680 Make the label for validation loss. 01:18.680 --> 01:21.960 So this is for the validation loss and this is the loss. 01:21.960 --> 01:29.520 And also set the X label to the apex Y label to the loss MSE. 01:29.880 --> 01:31.720 Let's make it here MSE. 01:32.040 --> 01:37.320 Make a legend and a title for this diagram. 01:37.360 --> 01:40.240 Training and validation loss okay. 01:40.400 --> 01:43.600 Also let me create another plot. 01:43.920 --> 01:45.240 This is 1 to 1. 01:45.240 --> 01:49.520 This is the left side for the left left diagram. 01:49.560 --> 01:54.040 We talked about this library a lot in the previous videos. 01:54.040 --> 01:54.520 Guys. 01:54.760 --> 01:55.960 It's very simple. 01:56.000 --> 01:58.440 I'm making x and y axis. 01:58.680 --> 02:00.280 X axis is x. 02:00.680 --> 02:03.800 X and y axis is the loss in MSE. 02:03.800 --> 02:13.680 I am plotting the from the model that I trained called history loss and get the history of the history 02:13.680 --> 02:14.200 model. 02:14.200 --> 02:16.080 Get the loss from it. 02:16.080 --> 02:20.760 Let me print the Ma and see the ma. 02:20.800 --> 02:24.600 So plot subplot take the right diagram. 02:24.640 --> 02:35.880 I get the Ma, make the history, the val, ma and validation ma set the x label to epics y label to 02:35.920 --> 02:37.520 mean absolute error. 02:37.560 --> 02:42.160 Ma get the legend and title for it. 02:42.400 --> 02:42.920 Okay? 02:43.240 --> 02:48.320 Make this plot tight layout and show the plot. 02:48.360 --> 02:51.480 Let me run and call this function plot history. 02:51.520 --> 02:57.520 Let me at first run all we get this error in plot history. 02:57.560 --> 02:58.600 Let me see. 02:58.800 --> 03:01.520 Unknown argument label. 03:01.520 --> 03:02.720 Let me scroll up. 03:02.960 --> 03:03.320 Oops. 03:03.680 --> 03:04.160 Um. 03:04.760 --> 03:05.840 Here the plot. 03:06.040 --> 03:09.720 They a label with a small letter. 03:09.800 --> 03:13.960 So let me see the label y label. 03:14.240 --> 03:19.520 Okay, so there is no argument called capital letter L. 03:19.680 --> 03:20.440 Label. 03:20.480 --> 03:22.040 Small letter L. 03:22.200 --> 03:23.760 Scroll down to here. 03:23.800 --> 03:28.640 Label and label run again and here we go. 03:28.680 --> 03:30.200 Congratulations guys. 03:30.200 --> 03:34.040 This is our training and validation loss plots. 03:34.080 --> 03:39.720 Okay so here we have the validation loss in orange. 03:39.880 --> 03:41.920 The training loss in blue. 03:42.080 --> 03:54.680 If we see for the last MSE as epochs increase the MSE drops down and decrease to around as table one. 03:54.800 --> 04:03.760 And here, if the validation, the validation loss and training loss decreased and stable, then there 04:03.800 --> 04:06.320 is no need to create more epochs. 04:06.600 --> 04:09.400 For this we use the early access. 04:09.680 --> 04:14.840 And by the way this is a good and this is a good visualization. 04:14.840 --> 04:24.600 We uh, we end the previous videos, compared and analyzed the MSA and Maa in the data like those. 04:24.760 --> 04:31.120 But we, uh, we plotted this in order to understand it better. 04:31.160 --> 04:33.240 Also, this is true for the Maa. 04:33.440 --> 04:37.560 As epochs increase, that decreased okay. 04:37.920 --> 04:45.600 This is how we validate the data and how and how to evaluate the model performance.