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.
85 lines
2.5 KiB
85 lines
2.5 KiB
WEBVTT |
|
|
|
00:00.440 --> 00:07.280 |
|
Before we move on, let me show you one more time this fabulous slide that describes the simple three |
|
|
|
00:07.310 --> 00:13.130 |
|
abstractions and one method call to build a rag pipeline the abstractions. |
|
|
|
00:13.160 --> 00:16.190 |
|
Again, you create the LM, you create your memory. |
|
|
|
00:16.220 --> 00:20.690 |
|
You create your retriever just by calling your chroma vector store or whatever vector store you're working |
|
|
|
00:20.690 --> 00:22.520 |
|
with dot as retriever. |
|
|
|
00:22.520 --> 00:28.190 |
|
And then you can simply create your conversation retrieval chain with this one line passing in your |
|
|
|
00:28.190 --> 00:30.770 |
|
LM, your retriever, your memory, and you're done. |
|
|
|
00:30.770 --> 00:36.080 |
|
And of course, with this conversation chain you call invoke and you pass in a dictionary with your |
|
|
|
00:36.080 --> 00:43.400 |
|
question and with what you get back, you can simply call you look up the answer key and you will get |
|
|
|
00:43.400 --> 00:46.010 |
|
back the response from the model. |
|
|
|
00:46.340 --> 00:51.770 |
|
And so with that I would say that you have leveled up your skills. |
|
|
|
00:51.770 --> 00:53.540 |
|
You are upskilled. |
|
|
|
00:53.540 --> 00:59.030 |
|
You are now accomplished in the art of building rag pipelines. |
|
|
|
00:59.030 --> 01:00.380 |
|
Congratulations! |
|
|
|
01:00.740 --> 01:06.170 |
|
You can build your own rag knowledge worker not just for a fictitious company like Endure Elm, but |
|
|
|
01:06.170 --> 01:11.060 |
|
hopefully for real companies too, including your own using lamp chain. |
|
|
|
01:11.060 --> 01:12.200 |
|
And you can use chroma. |
|
|
|
01:12.230 --> 01:19.370 |
|
You could also see how easy it is to switch that up for other vector data stores, and using models |
|
|
|
01:19.370 --> 01:23.480 |
|
other than OpenAI, Lang chain makes that very simple indeed. |
|
|
|
01:23.570 --> 01:29.810 |
|
Next time we're going to talk about lang chains declarative language briefly, it's not something we're |
|
|
|
01:29.810 --> 01:31.100 |
|
going to use, but I'll show it to you. |
|
|
|
01:31.130 --> 01:32.900 |
|
So you could use it if you wish to. |
|
|
|
01:33.200 --> 01:38.390 |
|
Uh, I'm going to talk a bit about how Lang Chain works under the covers, just so you have more insight. |
|
|
|
01:38.390 --> 01:44.750 |
|
And in particular, we're going to look at a common problem with Rag, diagnose it and then talk about |
|
|
|
01:44.750 --> 01:45.770 |
|
how to fix it. |
|
|
|
01:45.770 --> 01:53.300 |
|
So that will all, uh, equip you to be able to use rag in anger in production projects as you'll really |
|
|
|
01:53.300 --> 01:54.950 |
|
understand the inner workings. |
|
|
|
01:55.040 --> 01:57.200 |
|
And with that, I will see you next time.
|
|
|