WEBVTT 00:00.080 --> 00:01.040 Welcome back. 00:01.040 --> 00:04.160 Now let's create the Loadmodel function. 00:04.200 --> 00:10.200 Outside the initialize function, create a new function named as load mode. 00:10.200 --> 00:12.840 This function takes a string. 00:13.040 --> 00:18.880 Name it as file name as a parameter and returns byte buffer. 00:18.920 --> 00:26.000 This function loads a file from assets and returns it as a byte buffer. 00:26.040 --> 00:27.840 Let me write this comment below. 00:28.000 --> 00:33.200 Returns null if loading fails, so I use this question mark. 00:33.400 --> 00:39.280 So here we need to load the model and returns it as byte buffer. 00:39.520 --> 00:44.120 Uses memory mapping for efficient file access. 00:44.160 --> 00:48.600 Let me start by opening the model file. 00:48.760 --> 00:52.200 So here return return try. 00:52.240 --> 00:58.480 Let me try to open the file model file or my model whatever you want. 00:58.600 --> 01:04.200 Context dot dot open folder. 01:04.240 --> 01:05.720 What's the name of the folder? 01:05.720 --> 01:13.740 Is the file name parameter that it's best to do the conversion of the file to bytebuffer. 01:14.020 --> 01:25.380 We need to start with the input stream equals to file input stream and set the model file dot file descriptor. 01:25.420 --> 01:31.060 The first line appends the file from the app's assets folder. 01:31.060 --> 01:40.460 Using assets file descriptor provides access to the file descriptor, start offset, and declared length, 01:40.460 --> 01:47.500 while the second line creates a file input stream from the file descriptor. 01:47.780 --> 01:52.180 This allows reading the file as a stream of bytes. 01:52.180 --> 01:53.060 So here. 01:53.220 --> 02:00.180 Step number two reading the model file as stream byte stream of bytes. 02:00.300 --> 02:12.460 The third step is the file channel val file channel equals to input stream dot channel file channel 02:12.580 --> 02:13.980 dot map. 02:14.140 --> 02:23.250 We need to pass three parameters the file channel, the start of set and the declared length or size 02:23.250 --> 02:24.490 of the file channel. 02:24.530 --> 02:27.090 Alt+ enter to import the file channel. 02:27.090 --> 02:36.330 Obtains a file channel from the input stream, and the file channel provides methods for file operations, 02:36.330 --> 02:38.530 including memory mapping. 02:38.650 --> 02:45.890 I want from you to write those those comments below because they are very important okay. 02:45.930 --> 02:51.210 And helps you to understand the code, the file channel. 02:51.250 --> 02:52.250 Here we are. 02:53.410 --> 02:56.810 We are making a memory map of the file. 02:57.170 --> 03:00.330 Mapping the file to the memory okay. 03:00.450 --> 03:10.530 File channel dot map mode read only maps the file as read only model file dot start offset starting 03:10.570 --> 03:16.130 position in the file usually zero and the declared length. 03:16.170 --> 03:19.450 The length of the file to the map. 03:19.450 --> 03:29.110 Returns a map byte buffer that directly represents the file content in the Memory. 03:29.150 --> 03:37.150 Why we need memory mapping we need for file content is loaded on demand. 03:37.150 --> 03:46.590 Lazy loading, which is efficient memory mapping is used for fast access direct memory access without 03:46.590 --> 03:56.150 copying data and OS managed because operating system handles paging and caching. 03:56.310 --> 04:02.230 So those are the advantages of mapping and memory mapping. 04:02.390 --> 04:11.630 Okay, now we need to catch any errors that may result from this mapping and this conversion. 04:11.630 --> 04:18.990 So for that we need to start with catch here E exception. 04:18.990 --> 04:27.950 And we need to log e tag and load it model print the message and return null okay. 04:28.150 --> 04:32.270 So this is the load model function. 04:32.270 --> 04:37.270 This is how we load the model and convert it as byte buffer.