From the uDemy course on LLM engineering.
https://www.udemy.com/course/llm-engineering-master-ai-and-large-language-models
You can not select more than 25 topics
Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
217 lines
5.6 KiB
217 lines
5.6 KiB
WEBVTT |
|
|
|
00:00.560 --> 00:03.110 |
|
And now over to my Mac people. |
|
|
|
00:03.110 --> 00:04.130 |
|
And I have news for you. |
|
|
|
00:04.130 --> 00:05.570 |
|
It's exactly the same thing. |
|
|
|
00:05.570 --> 00:10.640 |
|
You go to a favorite browser, you go to Alarm.com, you'll see the same screen. |
|
|
|
00:10.640 --> 00:14.600 |
|
You press download and then download for Mac OS. |
|
|
|
00:14.630 --> 00:17.150 |
|
It will of course download it locally. |
|
|
|
00:17.150 --> 00:20.390 |
|
You will then run the installer and it will install it. |
|
|
|
00:20.390 --> 00:25.460 |
|
You then bring up a terminal window and you bring up a terminal window by going to applications and |
|
|
|
00:25.460 --> 00:26.990 |
|
utilities and then terminal. |
|
|
|
00:27.050 --> 00:32.960 |
|
And when you bring up that terminal window as it comes up here, you type the same instruction Olama |
|
|
|
00:33.830 --> 00:34.640 |
|
run. |
|
|
|
00:34.640 --> 00:42.320 |
|
And now the name of the model, which in our case is llama 3.2, the very latest small model from meta. |
|
|
|
00:42.320 --> 00:44.360 |
|
And it will then come up. |
|
|
|
00:44.420 --> 00:49.190 |
|
It will probably take a minute or two because it will need to download the 2 billion parameters associated |
|
|
|
00:49.190 --> 00:53.030 |
|
with with 3.2, but then it will be running just like this. |
|
|
|
00:53.030 --> 00:56.180 |
|
And now again we can try and put it to good use. |
|
|
|
00:56.270 --> 00:58.550 |
|
We can ask it this time will use a different language. |
|
|
|
00:58.550 --> 00:58.910 |
|
Why not? |
|
|
|
00:58.940 --> 01:04.470 |
|
Let's say I am trying to learn French. |
|
|
|
01:05.400 --> 01:07.620 |
|
I am a complete beginner. |
|
|
|
01:09.870 --> 01:17.400 |
|
Please have a conversation with me to teach me French. |
|
|
|
01:18.480 --> 01:19.980 |
|
And let's see how it does. |
|
|
|
01:20.010 --> 01:22.530 |
|
The first thing you'll notice is that it's a lot faster. |
|
|
|
01:22.530 --> 01:25.470 |
|
And that's because I'm actually running on a mac with an M1 chip. |
|
|
|
01:25.470 --> 01:27.870 |
|
And what we were seeing earlier was an emulation of a PC. |
|
|
|
01:27.870 --> 01:29.850 |
|
So of course this is a bit faster. |
|
|
|
01:29.940 --> 01:34.080 |
|
Your computer may be somewhere in between those two, depending on your architecture. |
|
|
|
01:34.350 --> 01:39.360 |
|
So off it's gone with the start and you can see that it's done a fab job. |
|
|
|
01:39.360 --> 01:42.000 |
|
It's clearly given a bunch of different options. |
|
|
|
01:42.000 --> 01:43.230 |
|
It's explained itself. |
|
|
|
01:43.230 --> 01:45.150 |
|
It's super impressive. |
|
|
|
01:45.150 --> 01:48.240 |
|
Uh, let's say uh, sure. |
|
|
|
01:49.890 --> 01:54.960 |
|
So I'm not actually answering his question, but there we go. |
|
|
|
01:54.960 --> 01:56.070 |
|
You get the idea. |
|
|
|
01:56.070 --> 02:01.770 |
|
And the point I wanted to make again is that this is, in fact, a paid commercial product on an app |
|
|
|
02:01.770 --> 02:02.520 |
|
that I'm using. |
|
|
|
02:02.520 --> 02:03.540 |
|
And here we have it. |
|
|
|
02:03.570 --> 02:10.120 |
|
We have built effectively a commercial project for free, using open source in a matter of minutes, |
|
|
|
02:10.150 --> 02:14.290 |
|
just immediately unleashing the power of an LLM on your computer. |
|
|
|
02:15.220 --> 02:18.670 |
|
So I now have an exercise for you right away. |
|
|
|
02:18.670 --> 02:20.380 |
|
I'd like you to of course, do this. |
|
|
|
02:20.410 --> 02:22.150 |
|
Install it, make sure that it works. |
|
|
|
02:22.150 --> 02:26.110 |
|
If you have any problems whatsoever, then you can always contact me at any point. |
|
|
|
02:26.140 --> 02:31.570 |
|
Feel free to message me or send me an email or LinkedIn with me and ask for help. |
|
|
|
02:31.570 --> 02:35.530 |
|
But hopefully this is a really easy install that will have you up and running. |
|
|
|
02:35.560 --> 02:36.370 |
|
Do this. |
|
|
|
02:36.370 --> 02:39.040 |
|
Have a quick experiment with a with a language. |
|
|
|
02:39.040 --> 02:41.950 |
|
Pick a language maybe one you don't know and try it out. |
|
|
|
02:42.040 --> 02:46.060 |
|
And then the next thing to try is I'd like you to experiment with different models. |
|
|
|
02:46.060 --> 02:50.650 |
|
So if we go back to the llama page, you'll see that there's a model heading up here. |
|
|
|
02:50.650 --> 02:53.710 |
|
And for both PC and Mac people, it's the same thing. |
|
|
|
02:53.710 --> 02:57.070 |
|
You can see some different models and read about the story behind them. |
|
|
|
02:57.070 --> 03:03.700 |
|
Llama 3.2 from Meta Jama from Google, which is its open source low parameter version. |
|
|
|
03:03.700 --> 03:10.130 |
|
Quan, which is a powerhouse of a model from Alibaba Cloud, which is less well known, but it is one |
|
|
|
03:10.130 --> 03:11.360 |
|
of the strongest. |
|
|
|
03:11.390 --> 03:16.970 |
|
You can click on a model and you will then see the different versions of it, and over here is the way |
|
|
|
03:16.970 --> 03:17.750 |
|
that you can run it. |
|
|
|
03:17.750 --> 03:23.030 |
|
You can simply press that button there to copy it to the clipboard, and then paste it into your PowerShell |
|
|
|
03:23.030 --> 03:24.170 |
|
or your terminal. |
|
|
|
03:24.260 --> 03:27.650 |
|
And these are the various different versions of it. |
|
|
|
03:27.650 --> 03:31.310 |
|
And you can just go back to models, browse through them. |
|
|
|
03:31.310 --> 03:34.250 |
|
Experiment 53.5 from Microsoft. |
|
|
|
03:34.400 --> 03:39.350 |
|
Also very powerful model and you can read more about them. |
|
|
|
03:39.350 --> 03:40.580 |
|
Uh, experiment with them. |
|
|
|
03:40.580 --> 03:43.580 |
|
And I would like to ask you to try this out for a new language. |
|
|
|
03:43.670 --> 03:49.550 |
|
Try exploring, maybe try and learn something completely different and use different models and see |
|
|
|
03:49.550 --> 03:52.820 |
|
if you can figure out which models are best for you. |
|
|
|
03:52.850 --> 03:57.590 |
|
As you pick some of the larger models, they may be slower on your computer, so find out which models |
|
|
|
03:57.590 --> 04:00.320 |
|
perform the best and give you the best results. |
|
|
|
04:00.320 --> 04:03.890 |
|
That's your exercise and then I will see you for the next time when we will. |
|
|
|
04:03.920 --> 04:06.920 |
|
Then do things like introductions and course and all that stuff. |
|
|
|
04:07.010 --> 04:07.700 |
|
See you then.
|
|
|