WEBVTT

1
00:00:00.600 --> 00:00:01.800
<v Maximilian>At this point,</v>

2
00:00:01.800 --> 00:00:06.800
we covered a lot of important features offered by LM Studio,

3
00:00:06.960 --> 00:00:10.170
but we have barely scratched the surface of what you can do

4
00:00:10.170 --> 00:00:11.700
as a power user

5
00:00:11.700 --> 00:00:13.320
because as a Power User,

6
00:00:13.320 --> 00:00:17.160
you can tweak the settings of your chat

7
00:00:17.160 --> 00:00:19.650
and how the model generates output.

8
00:00:19.650 --> 00:00:22.770
And I did already cover the System Prompt

9
00:00:22.770 --> 00:00:26.850
and that you can store a System Prompt in a preset,

10
00:00:26.850 --> 00:00:28.680
but that's not all you can do here.

11
00:00:28.680 --> 00:00:31.710
Instead, you can also tweak the model settings

12
00:00:31.710 --> 00:00:34.050
or how the model will behave here

13
00:00:34.050 --> 00:00:36.840
in this chat or in this preset.

14
00:00:36.840 --> 00:00:38.580
If you go to Model,

15
00:00:38.580 --> 00:00:41.190
you got a bunch of settings you can change here,

16
00:00:41.190 --> 00:00:42.750
and whatever you will change here

17
00:00:42.750 --> 00:00:44.910
can also be saved to a preset.

18
00:00:44.910 --> 00:00:46.740
So if you do save a preset,

19
00:00:46.740 --> 00:00:50.940
it does include both your system prompt if you set any

20
00:00:50.940 --> 00:00:53.190
and your settings here.

21
00:00:53.190 --> 00:00:55.290
So what are the settings we have her?

22
00:00:55.290 --> 00:00:57.540
Under Settings, most importantly,

23
00:00:57.540 --> 00:01:00.543
we can set the temperature of our model here.

24
00:01:01.380 --> 00:01:03.840
There are more settings we can enable.

25
00:01:03.840 --> 00:01:05.220
For example, what should happen

26
00:01:05.220 --> 00:01:07.680
if we leave our Context window

27
00:01:07.680 --> 00:01:10.080
if there's too much content being generated

28
00:01:10.080 --> 00:01:11.910
or sent to the model by us?

29
00:01:11.910 --> 00:01:14.880
But I'll get back to the Context window a little bit later.

30
00:01:14.880 --> 00:01:15.990
We can also control

31
00:01:15.990 --> 00:01:18.570
how many CPU threads will be used during inference,

32
00:01:18.570 --> 00:01:22.140
but typically, the default setting is okay here.

33
00:01:22.140 --> 00:01:24.210
But what's that Temperature all about?

34
00:01:24.210 --> 00:01:25.830
Well, the Temperature,

35
00:01:25.830 --> 00:01:28.020
as this tooltip tells us here,

36
00:01:28.020 --> 00:01:30.060
controls how much randomness

37
00:01:30.060 --> 00:01:33.240
will be in the generated output.

38
00:01:33.240 --> 00:01:36.300
And this Temperature here actually works together

39
00:01:36.300 --> 00:01:38.550
with the Sampling settings.

40
00:01:38.550 --> 00:01:43.550
There, we got Top K, Repeat Penalty and Top P and Min P.

41
00:01:44.820 --> 00:01:48.240
And what are Temperature Top K, Top P,

42
00:01:48.240 --> 00:01:50.343
and Min P all about now?