jamesbriggs

7 ms·

1.
by jamesbriggs 11mo ago
We used frameworks in the past, tried langchain, langgraph, and Openai's agents SDK pretty extensively. Now we roll our own, generally a much better and cleaner experience. We essentially built our own internal framework for our own us
2.
by jamesbriggs 1y ago
My bad, the link was wrong - you found the right one. I've updated it in the repo too, thanks. Let me know how it goes!
3.
LoRA Fine-Tuning Tiny LLMs as Expert Agents(youtube.com)
4 points by jamesbriggs 1y ago|3 comments
4.
by jamesbriggs 1y ago
Sharing my walkthrough on fine-tuning LLMs with LoRA using NVIDIA's NeMo microservices. The result is a llama-3.2-1b-instruct model fine-tuned to be really good at function-calling, making it ideal for agent-use. It was a ton of fun to
5.
by jamesbriggs 1y ago
Releasing this walkthrough on fine-tuning LLMs with LoRA using NVIDIA's NeMo Microservices (they sponsored the video, but with no reqs on what I do or say). We cover a ton on building prod AI applications, including: - LoRA fine-tuning
6.
by jamesbriggs 2y ago
I made a course covering everything you'd need to know to start building with LangChain. It assumes no level of expertise, you can be a complete beginner and by the end of the course be building AI agents with chat memory, streaming, a
7.
by jamesbriggs 3y ago
Tutorial showing how to do faster and more deterministic function calling using the open source Semantic Router library
8.
OSS for building more steerable AI agents(youtube.com)
2 points by jamesbriggs 3y ago|1 comments
9.
by jamesbriggs 3y ago
I and my team have been building a python library for improving the steerability of AI agents, the library also allows us to add an essentially unlimited number of tools to agents, add safety guardrails, etc. We have a lot more coming and I
10.
by jamesbriggs 3y ago
Not weighing on whether it is AGI-like or not, but my view on the search component is that it's like when we google info, we have a search term and we don't know whether that will return good results or not. So I suppose in order
11.
by jamesbriggs 3y ago
Yeah RAG has been around for some time, paper [2] being where I first stumbled onto it — I remember a 2-3 years ago building a RAG pipeline using either the model they trained in paper 2, or that was trained based on the same idea, with Dee
12.
by jamesbriggs 3y ago
I may be misunderstanding, but I'll try to answer — quantization typically means retrieval will be slower (if referring to techniques like product quantization), but that is the case whether you're at 10K vectors or 1B vectors, af
13.
by jamesbriggs 4y ago
Very cool, how are you doing the speech-to-text part, with Whisper?
14.
by jamesbriggs 4y ago
For Huberman Podcast I imagine he pays someone to do the annotations manually, so they're accurate. But on most videos I've found Whisper's annotations to be more accurate than YouTube's default annotations - not to bash
15.
by jamesbriggs 4y ago
I should add, Riley used the ada embedding model (rather than sentence transformers). Performance wise they should be similar (in ability to encode meaning accurately) but the ada model can encode a much larger chunk of text. I don't k
16.
by jamesbriggs 4y ago
you can return the chunks of text containing the answers, but not generate answers as that isn't what text-embedding-ada-002 is for. For that you need generation model (davinci in this case)
17.
by jamesbriggs 4y ago
I built something similar using a variety of YouTube channels focused on NLP, AI, etc. The app is here https://huggingface.co/spaces/jamescalam/ask-youtube - you can ask things like "what is a transformer mod
18.
Giving generative AI “memory” for Q&A(youtube.com)
1 points by jamesbriggs 4y ago|0 comments
19.
by jamesbriggs 4y ago
Hi, author of the article here. They act as a "position signal" that modifies the patch embedding. The learned signals are similar to other neighbouring position signals, and the later layers of the model will use the "simila
20.
Show HN: Ask YouTube – search for specific answers in videos(huggingface.co)
2 points by jamesbriggs 4y ago|0 comments
21.
Using OpenAI's Whisper to Rebuild YT Search(pinecone.io)
1 points by jamesbriggs 4y ago|0 comments
22.
Hugging Face Datasets 101(youtube.com)
1 points by jamesbriggs 4y ago|1 comments
23.
by jamesbriggs 4y ago
Intro to a series of content I'm producing on Hugging Face datasets, one of the most coolest libraries for ML
24.
Spotify's Podcast Search Explained(pinecone.io)
3 points by jamesbriggs 4y ago|0 comments
25.
by jamesbriggs 4y ago
Faiss is great, but I see there being several strong reasons for using Pinecone: 1. The people who built the Pinecone vector index include some of the top experts in information retrieval/vector search in the world, the chances of some
26.
by jamesbriggs 5y ago
I think the use case of chatbots is better solved with open domain Q&A (eg https://www.pinecone.io/learn/question-answering/ ). The focus of most chatbots seems to be on answering questions, but wrapping it up
27.
by jamesbriggs 5y ago
You can train a good sentence transformer on ~10K sentences using TSDAE (an unsupervised training approach), covered in Chapter 7 here: https://www.pinecone.io/learn/unsupervised-training-sentence...
28.
by jamesbriggs 5y ago
Agree that this would be super helpful
29.
by jamesbriggs 5y ago
The HuggingFace course is very good for NLP: https://huggingface.co/course/chapter1/1
30.
by jamesbriggs 5y ago
I don't remember the paper, but the authors working on this same problem suggested to do this + also include an overlap between spans.

More ›