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.
79 lines
2.3 KiB
79 lines
2.3 KiB
WEBVTT |
|
|
|
00:00.830 --> 00:01.970 |
|
It's remarkable. |
|
|
|
00:01.970 --> 00:04.970 |
|
But you are now at the 95% point. |
|
|
|
00:04.970 --> 00:07.820 |
|
There's 5% remaining of this course. |
|
|
|
00:07.880 --> 00:13.850 |
|
Uh, maybe it's not remarkable because hopefully you're feeling like you've built up such strong skills |
|
|
|
00:13.850 --> 00:15.800 |
|
and you're getting really experienced. |
|
|
|
00:15.830 --> 00:21.050 |
|
You probably saw some of the problems I was having a moment ago, before I did, and knew exactly how |
|
|
|
00:21.050 --> 00:22.010 |
|
to sort them out. |
|
|
|
00:22.010 --> 00:25.880 |
|
So what you can now do is this long list of things. |
|
|
|
00:25.880 --> 00:29.150 |
|
I could no longer fit it into three bullets, and I struggled to fit it into four. |
|
|
|
00:29.180 --> 00:30.590 |
|
There's far too much stuff. |
|
|
|
00:30.590 --> 00:35.660 |
|
I had to put structured outputs in here in the first bullet, just to try and fit it in, but there's |
|
|
|
00:35.660 --> 00:41.090 |
|
so much that you can do across closed source and open source models, calling them in all sorts of ways |
|
|
|
00:41.090 --> 00:43.700 |
|
now, including productionizing them. |
|
|
|
00:43.850 --> 00:50.990 |
|
Uh, and now we change topic to talk about agent workflows Agentic frameworks. |
|
|
|
00:51.110 --> 00:55.100 |
|
This is something that we touched on in previous weeks. |
|
|
|
00:55.100 --> 01:00.380 |
|
Uh, we, we had, we used uh, function calling, which we said at the time is a kind of light version |
|
|
|
01:00.380 --> 01:07.370 |
|
of, of agent workflows where you do have, um, a task that's being divided into smaller tasks with |
|
|
|
01:07.370 --> 01:10.430 |
|
specific models carrying out tasks. |
|
|
|
01:10.430 --> 01:15.350 |
|
And as you take that up a notch and start to get to an environment like we've been talking about this |
|
|
|
01:15.350 --> 01:20.660 |
|
week, where you have multiple models that are all working independently, potentially being organized |
|
|
|
01:20.660 --> 01:27.170 |
|
by a planning task, a planning agent, and then put together in some sort of a framework which has |
|
|
|
01:27.170 --> 01:34.340 |
|
memory and which can, uh, live beyond outside the context of a human necessarily interacting with |
|
|
|
01:34.340 --> 01:34.670 |
|
it. |
|
|
|
01:34.700 --> 01:41.960 |
|
You're getting to a full on agentic workflow, and that is what we will be talking about next time. |
|
|
|
01:41.960 --> 01:43.310 |
|
And I'm really excited about it. |
|
|
|
01:43.310 --> 01:44.540 |
|
And I will see you then.
|
|
|