WEBVTT

00:00.800 --> 00:02.000
Hello everyone!

00:02.680 --> 00:05.480
In today's video we will learn about Llama Guard.

00:06.440 --> 00:07.880
What is llama guard?

00:08.280 --> 00:11.560
Llama guard three, which is the third version of the original?

00:11.560 --> 00:20.760
Llama guard is a llama 3.18 billion parameter pre-trained model fine tuned for content safety classification.

00:21.240 --> 00:22.480
It acts as an LLM.

00:23.680 --> 00:30.680
It generates text in its output that indicates whether a given prompt or response is safe or unsafe,

00:31.200 --> 00:32.360
and if unsafe.

00:32.680 --> 00:36.040
It also lists the content categorized violated.

00:36.960 --> 00:43.960
It provides content moderation in eight languages and was optimized for support, safety, and security

00:43.960 --> 00:46.160
for search and code interpreter tools.

00:46.160 --> 00:46.760
Calls.

00:47.440 --> 00:51.440
Now, what are the categories that Llama Guard three supports?

00:51.880 --> 01:00.680
Supports violent crimes, sex related crimes, defamation, privacy, weapons, suicide, and all the

01:00.680 --> 01:01.040
others.

01:01.040 --> 01:05.960
Here 14 different categories that have been listed on their documentation.

01:06.840 --> 01:12.560
So Llama Guard two had about eight, which now has been updated to 14.

01:12.600 --> 01:14.080
In Llama Guard three.

01:14.640 --> 01:19.240
So now let's go ahead and understand how to execute Llama Guard in collab.