WEBVTT 1 00:00:00.480 --> 00:00:02.700 Now I mentioned that in Power User mode, 2 00:00:02.700 --> 00:00:07.080 you got some extra options here below your responses. 3 00:00:07.080 --> 00:00:08.190 And not just there, 4 00:00:08.190 --> 00:00:09.930 we also got an extra option here, 5 00:00:09.930 --> 00:00:12.180 for example, when we send a message. 6 00:00:12.180 --> 00:00:16.203 This User button here is not there in regular User Mode. 7 00:00:17.040 --> 00:00:19.560 So what are these options all about? 8 00:00:19.560 --> 00:00:23.010 For example, we got this Branch option here, 9 00:00:23.010 --> 00:00:26.550 and this essentially allows you to fork the chat 10 00:00:26.550 --> 00:00:31.550 to create multiple simultaneous parallel chat histories 11 00:00:32.100 --> 00:00:33.930 after a given message. 12 00:00:33.930 --> 00:00:36.480 So, for example, if I click this button here, 13 00:00:36.480 --> 00:00:40.770 you'll see that two chats are created here. 14 00:00:40.770 --> 00:00:43.890 So it's very similar to Duplicate in the end, 15 00:00:43.890 --> 00:00:46.530 where I also duplicate a chat. 16 00:00:46.530 --> 00:00:50.850 It's simply a way for you to have multiple, 17 00:00:50.850 --> 00:00:52.500 parallel chat histories 18 00:00:52.500 --> 00:00:56.070 that are based on the same start of a chat, 19 00:00:56.070 --> 00:00:58.110 but then change from a certain point on, 20 00:00:58.110 --> 00:01:01.680 so where you branch out into different versions of the chat 21 00:01:01.680 --> 00:01:04.440 because you might have different follow up questions. 22 00:01:04.440 --> 00:01:06.330 So if I got a response by the model 23 00:01:06.330 --> 00:01:08.730 and now I have two different follow up questions, 24 00:01:08.730 --> 00:01:11.580 which I don't want to have in the same chat history, 25 00:01:11.580 --> 00:01:15.900 I could branch my chat after this message. 26 00:01:15.900 --> 00:01:18.150 We also have continuous Assistant message, 27 00:01:18.150 --> 00:01:23.150 which is useful if you got a response that seems cut off. 28 00:01:23.400 --> 00:01:24.810 Now that's not the case here, 29 00:01:24.810 --> 00:01:27.060 but if you've got a super long response, 30 00:01:27.060 --> 00:01:30.243 at some point, it may actually be too much 31 00:01:30.243 --> 00:01:32.070 for the model to handle. 32 00:01:32.070 --> 00:01:33.360 They have certain limits 33 00:01:33.360 --> 00:01:35.520 regarding the amount of response tokens 34 00:01:35.520 --> 00:01:37.440 they can generate in one go. 35 00:01:37.440 --> 00:01:40.350 And if that happens, if the response is cut off, 36 00:01:40.350 --> 00:01:41.640 you can always click this button 37 00:01:41.640 --> 00:01:44.051 to have the model continue with the generation 38 00:01:44.051 --> 00:01:47.490 and continue generating a response. 39 00:01:47.490 --> 00:01:50.010 Now, regarding the options down here, 40 00:01:50.010 --> 00:01:53.490 user allows you to actually toggle 41 00:01:53.490 --> 00:01:57.000 between the mode of the message you're about to send. 42 00:01:57.000 --> 00:01:58.560 Now the default is always User 43 00:01:58.560 --> 00:02:01.260 because you are the user of this application, 44 00:02:01.260 --> 00:02:04.230 you are the user interacting with the AI model. 45 00:02:04.230 --> 00:02:06.870 But for advanced use cases, 46 00:02:06.870 --> 00:02:11.310 you also might want to simulate a certain AI response. 47 00:02:11.310 --> 00:02:14.962 You might want to artificially create a chat history 48 00:02:14.962 --> 00:02:17.040 that never happened. 49 00:02:17.040 --> 00:02:19.830 And you can do that by switching between roles. 50 00:02:19.830 --> 00:02:21.720 So if I create a new chat here, 51 00:02:21.720 --> 00:02:24.570 I could start with an Assistant message 52 00:02:24.570 --> 00:02:28.620 which says, "Hi, I am an assistant 53 00:02:28.620 --> 00:02:33.620 and I can help you with using open LLMs locally." 54 00:02:35.910 --> 00:02:37.170 And we could now simulate 55 00:02:37.170 --> 00:02:39.480 that the AI sent back this response 56 00:02:39.480 --> 00:02:41.520 even though it never did. 57 00:02:41.520 --> 00:02:43.140 So if I now sent this, 58 00:02:43.140 --> 00:02:46.500 I still get a response by the large language model. 59 00:02:46.500 --> 00:02:50.550 So it still treats this as if a regular user sent it, 60 00:02:50.550 --> 00:02:53.880 but internally, this will be stored as an Assistant message, 61 00:02:53.880 --> 00:02:55.920 as you also see here. 62 00:02:55.920 --> 00:02:59.490 And that may be interesting for more advanced use cases, 63 00:02:59.490 --> 00:03:02.430 where you then, for example, later down in the chat, 64 00:03:02.430 --> 00:03:04.950 you might want to ask the model what it can do for you. 65 00:03:04.950 --> 00:03:07.860 And it will refer back to its earlier responses, 66 00:03:07.860 --> 00:03:10.860 including responses that it actually never gave you, 67 00:03:10.860 --> 00:03:15.573 but that were faked by you by assigning such a fake role. 68 00:03:16.470 --> 00:03:18.990 Now, this is definitely a niche feature. 69 00:03:18.990 --> 00:03:22.200 I just wanted to mention it since this button is here. 70 00:03:22.200 --> 00:03:24.723 You won't have it in regular User mode. 71 00:03:25.650 --> 00:03:27.090 In Power User mode, 72 00:03:27.090 --> 00:03:30.540 you'll also find an extra sidebar here to the left. 73 00:03:30.540 --> 00:03:32.220 There, you can access your chats, 74 00:03:32.220 --> 00:03:35.400 which is the only option you have in User mode, 75 00:03:35.400 --> 00:03:37.650 which is why you don't even have a sidebar there, 76 00:03:37.650 --> 00:03:39.750 an extra sidebar I mean. 77 00:03:39.750 --> 00:03:41.340 But in Power User mode, 78 00:03:41.340 --> 00:03:44.520 you also got the Developer view here. 79 00:03:44.520 --> 00:03:46.590 So you don't just have that in Developer mode, 80 00:03:46.590 --> 00:03:48.063 but also in Power User mode. 81 00:03:49.080 --> 00:03:50.970 And in there, you'll find more information 82 00:03:50.970 --> 00:03:52.800 about the model that's being loaded. 83 00:03:52.800 --> 00:03:55.230 You can tweak some advanced settings here, 84 00:03:55.230 --> 00:03:57.453 and I'll get back to those later. 85 00:03:58.740 --> 00:04:00.240 And you'll find information 86 00:04:00.240 --> 00:04:03.240 about the web endpoints you could target 87 00:04:03.240 --> 00:04:05.430 if you were building a separate application 88 00:04:05.430 --> 00:04:08.340 that wants to communicate with this loaded model. 89 00:04:08.340 --> 00:04:10.380 You also find potential errors down here. 90 00:04:10.380 --> 00:04:12.600 For example, here for me, it logged an error, 91 00:04:12.600 --> 00:04:14.550 which I got during one response 92 00:04:14.550 --> 00:04:16.980 that was generated by the model. 93 00:04:16.980 --> 00:04:18.780 And you also find some debug logs, 94 00:04:18.780 --> 00:04:20.700 so not just error logs, 95 00:04:20.700 --> 00:04:23.100 that simply print out some information 96 00:04:23.100 --> 00:04:27.030 about the latest generation that was performed by the model. 97 00:04:27.030 --> 00:04:29.040 Now again, I'll get back to these more advanced settings 98 00:04:29.040 --> 00:04:30.600 a little bit later. 99 00:04:30.600 --> 00:04:32.880 You also have a My Models folder here, 100 00:04:32.880 --> 00:04:35.040 which shows you the different models 101 00:04:35.040 --> 00:04:36.150 that have been loaded. 102 00:04:36.150 --> 00:04:39.378 Here, you also find the path to the folder on your system 103 00:04:39.378 --> 00:04:42.150 where these models are stored. 104 00:04:42.150 --> 00:04:44.070 And you can change that path, 105 00:04:44.070 --> 00:04:47.040 open it in Finder or Windows Explorer, 106 00:04:47.040 --> 00:04:49.290 and always reset it back to the default path 107 00:04:49.290 --> 00:04:51.270 if you did change it. 108 00:04:51.270 --> 00:04:53.340 And of course, you also find a list of all the models 109 00:04:53.340 --> 00:04:56.310 that have been downloaded into this directory. 110 00:04:56.310 --> 00:04:58.620 And here you see that we got large language models 111 00:04:58.620 --> 00:04:59.880 or embedding models. 112 00:04:59.880 --> 00:05:02.430 Now embedding models are not the focus of this course. 113 00:05:02.430 --> 00:05:04.620 In the end, embedding models can be used 114 00:05:04.620 --> 00:05:07.050 to create so-called vector embeddings, 115 00:05:07.050 --> 00:05:11.490 which are numerical representations of words and sentences 116 00:05:11.490 --> 00:05:14.730 that also model relationships between words 117 00:05:14.730 --> 00:05:17.580 to describe it in a simple, brief way. 118 00:05:17.580 --> 00:05:19.500 And you can use such models 119 00:05:19.500 --> 00:05:23.310 that create such vector embeddings here in LM Studio 120 00:05:23.310 --> 00:05:25.500 because that could, for example, be irrelevant 121 00:05:25.500 --> 00:05:28.620 if you were building your own RAG application 122 00:05:28.620 --> 00:05:30.090 locally on your system, 123 00:05:30.090 --> 00:05:33.104 your own retrieval augmented generation application, 124 00:05:33.104 --> 00:05:34.560 for example, 125 00:05:34.560 --> 00:05:37.470 which is an advanced way of using generative AI, 126 00:05:37.470 --> 00:05:41.520 which I will cover in my main generative AI course, 127 00:05:41.520 --> 00:05:42.810 but not in this course, 128 00:05:42.810 --> 00:05:45.810 which is about running models locally. 129 00:05:45.810 --> 00:05:48.090 So here you see your large language models, 130 00:05:48.090 --> 00:05:51.150 and you can also manage these download models here. 131 00:05:51.150 --> 00:05:55.110 So you can tweak their default parameters 132 00:05:55.110 --> 00:05:56.610 that will be used as a default 133 00:05:56.610 --> 00:05:59.820 whenever you load such a model into LM Studio. 134 00:05:59.820 --> 00:06:03.210 And again, I'll get back to those later. 135 00:06:03.210 --> 00:06:07.230 And you can also view this specific model 136 00:06:07.230 --> 00:06:10.110 in your Finder or Windows Explorer, 137 00:06:10.110 --> 00:06:12.750 copy the path, view it on Hugging Face, 138 00:06:12.750 --> 00:06:17.310 or also delete it to free up the space on your system. 139 00:06:17.310 --> 00:06:18.690 So that's that. 140 00:06:18.690 --> 00:06:21.120 Last but not least, you got this Discover page, 141 00:06:21.120 --> 00:06:23.340 which gives you this model search page 142 00:06:23.340 --> 00:06:26.073 where you can discover and download new models.