mpaepper

8 ms·

61.
by mpaepper 3y ago
The big difference is that you get the results immediately and iterations take minutes not days
62.
by mpaepper 4y ago
Glad to hear that as this was exactly my goal: make this more accessible, because the actual things needed to run it are super simple. To be fair to langchain, though: I like it a lot and it offers a lot more than my repo, but yeah definite
63.
by mpaepper 4y ago
I thought about that as well. Some kind of ensemble agent which is more robust, might play with that idea.
64.
by mpaepper 4y ago
It might be that it generated both the Thought and Final answer, but my current implementation skips to the end once the final answer is parsed. So even though there was a thought, I didn't print it
65.
by mpaepper 4y ago
Basically the tools the agent is using in this repository are like ChatGPT plugins.
66.
by mpaepper 4y ago
That's a great idea!
67.
by mpaepper 4y ago
Check out my agent implementation here which is just for fun and to get a better understanding. Maybe you want to contribute? https://github.com/mpaepper/llm_agents
68.
by mpaepper 4y ago
I agree that the agents are underappreciated. To make them more accessible I rewrote them in ~200 lines of code, so you can easily understand how it works. They have access to a python console, Google search and hacker news search: https:&
69.
Build your own agents which are controlled by LLMs(github.com)
218 points by mpaepper 4y ago|35 comments
70.
by mpaepper 4y ago
You can do it yourself with this repo and only pay the OpenAI API fees: https://github.com/mpaepper/content-chatbot
71.
by mpaepper 4y ago
This is awesome work and they also provide their 9B OpenFlamingo model which is based on Llama: https://huggingface.co/openflamingo/OpenFlamingo-9B
72.
Open Flamingo – open framework to train multimodal LLMs(laion.ai)
265 points by mpaepper 4y ago|25 comments
73.
by mpaepper 4y ago
I built this repo to do this for your own website content, that should get you a good starting point: https://github.com/mpaepper/content-chatbot
74.
by mpaepper 4y ago
Is there a research paper / arxiv which describes it in detail?
75.
by mpaepper 4y ago
Woah, that's a huge site! Should be fine, though, as it iterates over it, it creates embeddings and then stores them in the FAISS store ( https://github.com/facebookresearch/faiss ) which was created to handle a lar
76.
by mpaepper 4y ago
Thanks for the suggestion, but for my fun small experiment, FAISS was more than enough.
77.
by mpaepper 4y ago
Yes, this will be an interesting next experiment - adding agents with additional tools (also for example access to internal APIs) will be quite powerful.
78.
by mpaepper 4y ago
Currently, it only splits documents linearly, so if you have information which is written backwards or things like that, it will likely not work so well.
79.
by mpaepper 4y ago
Go for it, it's quite easy to do, you only need to adapt the ` https://github.com/mpaepper/content-chatbot/blob/main/create... ` file a bit to match the way your data is represented and then you are g
80.
by mpaepper 4y ago
I don't have larger scale experience on this at the moment, but I can tell you what I observed during my trials (also see my related blog entry for some more info: https://www.paepper.com/blog/posts/build-q-an
81.
by mpaepper 4y ago
Yep, that's the way it's currently implemented in langchain. The 4 is a hyperparameter you can change, though, so you could set it to 10 as well. The way it works is that it first looks up the N most relevant documents (N being 4
82.
Leave secret messages for Bing chat on web pages(twitter.com)
2 points by mpaepper 4y ago|1 comments
83.
Your website's content -> Q&A bot / chatbot(github.com)
84 points by mpaepper 4y ago|27 comments
84.
Build a Q&A bot from your website's content using langchain(paepper.com)
1 points by mpaepper 4y ago|0 comments
85.
Interactive visualization of Stable Diffusion image embeddings(paepper.com)
3 points by mpaepper 4y ago|0 comments
86.
Preview of New Bing Chatbot(twitter.com)
1 points by mpaepper 4y ago|0 comments
87.
by mpaepper 4y ago
Awesome, thank you so much. That is solid advice :)
88.
Ask HN: How to keep small kids busy?
3 points by mpaepper 4y ago|2 comments
89.
by mpaepper 4y ago
Can you showcase any projects you've worked at in the mean time? That way you could fill some gaps with project stories which might make it easier for you. Good luck with the search!
90.
Two Russian rockets hit Poland killing two(telegraph.co.uk)
24 points by mpaepper 4y ago|5 comments

More ›