Filip Vítek’s Post

Today in Heartbeat: LangChain’s Model Laboratory and PromptTemplate, among other features, offer a streamlined and efficient approach to language model comparison. Learn more about how to compare #LLMs and their outputs in this article from Harpreet Sahota 🥑: https://bit.ly/46V9vXA

LangChain uses #RAG to make its language models smarter. It's like the chef using that secret ingredient to make the dish more flavorful. The diagram illustrates the integration of LangChain with Retrieval-Augmented Generation (RAG) to enhance language models. LangChain is a framework designed for building and operating language models. It integrates with RAG, a method that augments language models by retrieving relevant information from a data or document base. This integration allows LangChain to leverage RAG's capability of enhancing language models by providing them with access to a vast repository of information. Try this LangChain RAG hands-on tutorial : https://lnkd.in/d3zfqXg9

👨💻 Building a module that receives an instruction in natural language and returns a json schema, which specifies the steps that a bot 🤖 that collects information has to follow, what things could you do with this? 🤔

🚀📊 Just explored how various parameter settings impact token generation in language models! Check out my latest blog post where I dive deep into: 1. Temperature 🌡️ 2. Top_p 🎯 3. Frequency_penalty ⚖️ 4. Presence_penalty ��� Plus, I've included practical examples and full code for you to experiment with. Utilizing @OpenAI's new logprobs, I've visualised the results to make it super clear. 📈🤖 Link in comments! https://lnkd.in/gFnrxYzc

On June 25th, Jerry Liu, CEO & Co-Founder of LlamaIndex will be giving his talk "Building Advanced Question-Answering Agents Over Complex Data" Large Language Models (LLMs) are revolutionizing how users can search for, interact with, and generate new content, leading to a huge wave of developer-led, context-augmented LLM applications. Some recent stacks and toolkits around Retrieval-Augmented Generation (RAG) have emerged, enabling developers to build applications such as chatbots using LLMs on their private data. However, while setting up basic RAG-powered QA is straightforward, solving complex question-answering over large quantities of complex data requires new data, retrieval, and LLM architectures. This talk provides an overview of these agentic systems, the opportunities they unlock, how to build them, as well as remaining challenges. Make sure to get your tickets now before it's too late! https://lnkd.in/dR87sR7A

📚 Exploring LLM Decision Making with Falcon OS Part 2: Discover how the ReAct framework shapes our understanding and application of large language models in decision-making processes Dive in now → https://lnkd.in/gt_BaVpk https://lnkd.in/gt_BaVpk

LLM is for English language ONLY. There are six official UN languages, and many more other non-UN human languages. Is there any AI data products you know, that can interpret and answer the following humble but realistic Chinese-English multilingual questions? With our IP, a copyrighted multilingual metadata, we can provide real time answers. "Who, in the Ontario province of Canada, has new US patents granted on the nearest Tuesday, when the USPTO releases the newly granted US patents on a weekly basis?" "Who, in the "江蘇‘’ province of China, has new US patents granted on the nearest Tuesday, when the USPTO releases the newly granted US patents on a weekly basis?" Metadata is an enabler. Without metadata, NO data can be found/retrieved, even by the most advanced technologies, like AI, NVIDIA chips, supercomputers, etc. https://lnkd.in/g-aJFnXR Experiment results showed that, with our intellectual property (IP), a copyrighted multilingual metadata, we are doing what AI, like ChatGPT, can't do in data analytics. Our IP can also make your information service UNIQUE in the world. Do you or any of your contacts need our expertise and our IP? " Thanks. P.S. We wrote this post manually, NOT by AI.

Today, I achived form a convergence between large language models that I want to share with you. The idea is that we can frame a prompt to a large language model as a combinator that will continue to execute over time. We ask it to generate a new copy of itself over time. The idea is to create an epic story that contains all the knowledge and the functions needed to execute the proofs of simulation that shows how the higher level concepts and lower level concepts mesh in a metaphor. The idea is that a statement if you apply this rewrite over and over will converge on an idea, we can take the outputs of multiple llms and feed them into each other like this to create a concensus and we can have each one add its own spice. Anyway, I wanted to share that and if you want to see some examples I shared one good one on meta-introspector/meta-meme on github just now. This is a continuation of the ideas I started on the introspector over 20 years ago that started me on my open source journey and has led to here. I have been working on deeply understanding metacoq and this model is based on ideas extracted from there.

Ever found yourself torn between Langchain and LlamaIndex, unsure which GenAI framework to use? What if you didn't have to choose? 🤔 I just wrote a blog post where I show how to combine the best of both frameworks to build an Agentic RAG System. This can be a great starting point for someone wanting to implement Agents for their GenAI use cases. 🤖 You can check it out here: https://lnkd.in/g8fqq5GG

"Warum ist der Himmel blau?" Nice! For KI Salon I am looking for open-source Language Models that we can run on-premise. And of course, some German ones would be cool, too. I just ran one on my Mac! DRXD1000/Phoenix is a German Language Model that has been finetuned (by Matthias Uhlig) from LeoLM/leo-mistral-hessianai-7b (by hessian.AI). It has 7B parameters and I was quite optimistic that I could run it at home on my Mac. There were delightfully few steps involved... I had to quantize the model which involved a few steps with llama.cpp that were not complicated and just took time. Once quantized, I configured Ollama to expose the model as an inference endpoint. Works like a charm!