GenAItechLab.com

New Book: State of the Art in GenAI & LLMs — Creative Projects, with Solutions https://lnkd.in/g82c9Z8D With 23 top projects, 96 subprojects, and 6000 lines of Python code, this vendor-neutral coursebook is a goldmine for any analytic professional or AI/ML engineer interested in developing superior GenAI or LLM enterprise apps using ground-breaking technology. This is not another book discussing the same topics that you learn in bootcamps, college classes, Coursera, or at work. Instead, the focus is on implementing solutions that address and fix the main problems encountered in current applications. Using foundational redesign rather than patches such as prompt engineering to fix backend design flaws. You will learn how to quickly implement from scratch applications actually used by Fortune 100 companies, outperforming OpenAI and the likes by several order of magnitudes, in terms of quality, speed, memory requirements, costs, interpretability (explainable AI), security, latency, and training complexity. ➡️ Read more and get your copy at https://lnkd.in/g82c9Z8D

30 Features that Dramatically Improve LLM Performance – Part 3: https://lnkd.in/gBnbc6Cx This is the third and final article in this series, featuring some of the most powerful features to improve RAG/LLM performance. In particular: speed, latency, relevancy (hallucinations, lack of exhaustivity), memory use and bandwidth (cloud, GPU, training, number of parameters), security, explainability, as well as incremental value for the user. I implemented these features in my in-memory xLLM system. See details in my recent book, at https://lnkd.in/g82c9Z8D. The featured image shows the xLLM Web API, now live and available to everyone, based on anonymized sample enterprise corpus from Fortune 100 company. I will share the link when the documentation is finalized. Sign-up to my newsletter at https://lnkd.in/gvvF72aG, to not miss it. In this article, you will find: ➡️ Self-tuning and customization, with user able to fine-tune in real time and select intuitive parameter values ➡️ Local, global parameters, and real-time debugging offered to user ➡️ Displaying relevancy scores for each item in prompt results, and score customization by the user ➡️ Intuitive hyperparameters: system based on explainable AI ➡️ Sorted n-grams and token order preservation, further reducing the size of backend tables ➡️ Blending standard tokens with tokens from the knowledge graph ➡️ Boosted weights for knowledge-graph tokens ➡️ Versatile command prompt allowing you to fine-tune parameters, check the size of backend tables, and a lot more ➡️ Boost long multitokens and rare single tokens to reflect their importance and higher quality Read full content at https://lnkd.in/gBnbc6Cx.

Hyperfast Contextual Custom LLM with Agents, Multitokens, Explainable AI, and Distillation https://lnkd.in/eNB5zqZB New additions to this ground-breaking system include multi-token distillation when processing prompts, agents to meet user intent, more NLP, and a command prompt menu accepting both standard prompts and various actions. I also added several illustrations, featuring xLLM in action with a full session and sample commands to fine-tune in real-time. All the code, input sources (anonymized corporate corpus from fortune 100 company), contextual backend tables including embeddings, are on GitHub. My system has zero weight, no transformer, and no neural network. It relies on explainable AI, does not require training, is fully reproducible, and fits in memory. Yet your prompts can retrieve relevant full text entities from the corpus with no latency — including URLs, categories, titles, email addresses, and so on — thanks to well-designed architecture. Read more, get the code, paper and everything for free, at https://lnkd.in/eNB5zqZB

30 Features that Dramatically Improve LLM Performance - Part 1 https://lnkd.in/gDzSaHCC Many are ground-breaking innovations that make LLMs much faster and not prone to hallucinations. They reduce the cost, latency, and amount of computer resources (GPU, training) by several orders of magnitude. Some of them improve security, making your LLM more attractive to corporate clients. I introduced a few of these features in my previous article "New Trends in LLM Architecture". Now I offer a comprehensive list, based on the most recent developments.

New Random Generators for Large-Scale Reproducible AI https://lnkd.in/gXSE9d4Y Modern GenAI apps rely on billions if not trillions of pseudo-random numbers. You find them in the construction of latent variables in nearly all deep neural networks and almost all applications: computer vision, synthetization, and LLMs. Yet, few AI systems offer reproducibility, though those described in my recent book, do. When producing so many random numbers or for strong encryption, you need top grade generators. The most popular one — adopted by Numpy and other libraries — is the Mersenne twister. It is known for its flaws, with new ones discovered during my research, and shared with you. This paper has its origins in the development of a new foundational framework to prove the conjectured randomness and other statistical properties of the digits of infinitely many simple math constants, such as e or π. Here, I focus on three main areas. First, how to efficiently compute the digits of the mathematical constants in question to use them at scale. Then, new tests to compare two types of random numbers: those generated by Python, versus those from the math constants investigated here, and help decide which systems are best. Finally, I propose a new type of strongly random digits based on an incredibly simple formula (one small line of code) leading to fast computations. One of the benefits of my proposed random bit sequences, besides stronger randomness and fast implementation at scale, is to not rely on external libraries that may change over time. These libraries may get updated and render your results non-replicable in the long term if (say) Numpy decides to modify the internal parameters of its random generator. By combining billions of constants, each with its own seed, with billions of digits from each constant, it is impossible to guess what formula you used to generate your digits, when security is important. Some of my randomness tests involve predicting the value of a string given the values of previous strings in a sequence, a topic at the core of many large language models (LLMs). Methods based on neural networks — mines being an exception — are notorious for hiding the seeds used in the various random generators involved. It leads to non-replicable results. It is my hope that this article will raise awareness about this issue, while offering better generators that do not depend on which library version you use. Last but not least, the datasets used here are infinite, giving you the opportunity to work on truly big data and infinite numerical precision. And at the same time, get a glimpse at deep number theory results and concepts, explained in simple English. ➡️ To access full article with code: dowload paper #44 at https://lnkd.in/gvvF72aG 

Custom Enterprise LLM/RAG with Real-Time Fine-Tuning https://lnkd.in/gaEi8-xn This article features an application of xLLM to extract information from a corporate corpus, using prompts referred to as “queries”. The goal is to serve the business user — typically an employee of the company or someone allowed access — with condensed, relevant pieces of information including links, examples, PDFs, tables, charts, definitions and so on, to professional queries. My custom sub-LLM designed from scratch does not rely on any Python library or API, and performs better than search tools available on the market, in terms of speed and results relevancy. It offers the user the ability to fine-tune parameters in real time, and can detect user intent to deliver appropriate output. The good performance comes from the quality of the well-structured input sources, combined with smart crawling to retrieve the embedded knowledge graph and integrate it into the backend tables. Traditional tools rely mostly on tokens, embeddings, billions of parameters and frontend tricks such as prompt engineering to fix backend issues. To the contrary, my approach focuses on building a solid backend foundational architecture from the ground up. Tokens and embeddings are not the most important components, by a long shot. Cosine similarity and dot products are replaced by pointwise mutual information. There is no neural network, no training, and a small number of explainable parameters, easy to fine-tune. When you think about it, the average human being has a vocabulary of 30,000 words. Even if you added variations and other pieces of information (typos, plural, grammatical tenses, product IDs, street names, and so on), you end up with a few millions at most, not trillions. Indeed, in expensive multi-billion systems, most tokens and weights are just noise: most are rarely fetched to serve an answer. This noise is a source of hallucinations. Read more, access the code and data, at https://lnkd.in/gaEi8-xn

Podcast: Creating Custom LLMs https://lnkd.in/g7_3uQFD Despite GPT, Claude, Gemini, LLama and the other host of LLMs that we have access to, a variety of organizations are still exploring their options when it comes to custom LLMs. Logging in to ChatGPT is easy enough, and so is creating a ‘custom’ openAI GPT, but what does it take to create a truly custom LLM? When and why might this be useful, and will it be worth the effort? Watch Vincent Granville discussing this topic, and why and how he created his own LLM system from scratch. 

LLM/RAG: Knowledge Graphs, Multi-Agents, Ultrafast Fine-tuning, No Latency https://lnkd.in/gpDysYT6 In this presentation, I explain all the components of a ground-breaking architecture with applications to local, enterprise LLMs and high-quality search targeted to advanced users and busy professionals. Some of the main features: o Multi-LLM with 2000 specialized, small sub-LLMs covering the entire human knowledge. o LLM router as top layer to decide which sub-LLM to call, or let the user choose. o Smart crawling to recover taxonomies and knowledge graphs embedded in carefully selected, high-quality input sources. Augmented with synonyms and abbreviation maps based on glossaries, indexes, and so on. o Ultrafast: No neural network but instead parametric weights governed by few explainable parameters rather than optimizing billions of weights (the neural network approach). o Customizable relevancy scores attached to each item returned to the user (URLs, related concepts, tables, and so on). To help the user decide on what to look for.  o Self-tuning (global or local to a sub-LLM) based on favorite hyperparameters chosen by users, and customized results. Local self-tuning is very fast, and a first step before global optimization. o Fast embedding search with probabilistic algorithm. Variable-length embeddings with contextual and multi-tokens. No dot product or cosine distance, but better metrics instead.  o Using the model evaluation metric as your loss function to achieve better relevancy. Introducing the concept of adaptive loss function. o Augmentation and refinement based on integrating user prompt elements in back-end tables. o Application to content clustering and predictive analytics based on text only. Using nested hashes that leverage the sparsity in keyword association tables (no huge, sparse similarity matrix involved). o Model evaluation based on knowledge graph reconstruction (category assignments) and comparison with the native one. ➡️ Download the free PowerPoint presentation from https://lnkd.in/gpDysYT6. With links to full source code on GitHub, datasets, documentation, related books & articles, and free training on the topic.

The New Generation of RAG and LLM Architectures: Access all the material at https://lnkd.in/g58xUvg8 In particular, my high-level PowerPoint presentation on the topic, featuring multi-tokens, contextual tokens, LLM routers, evaluation metrics used as adaptive loss function for better results, enterprise LLMs, fast-tuning like LoRA, auto-tuning, mixture of experts, knowledge graphs, LLM for search / clustering / predictive analytics, LLM with no transformer, variable-length embeddings, and alternative to vector and graph databases. 

GenAI Evaluation Metrics: Your Best Loss Functions to Boost Quality https://lnkd.in/gyUbtS8e Whether dealing with LLM, computer vision, clustering, predictive analytics, synthetization, or any other AI problem, the goal is to deliver high quality results in as little time as possible. Typically, you assess the output quality after producing the results, using model evaluation metrics. These metrics are also used to compare various models, or to measure improvement over the baseline. In unsupervised learning such as LLM or clustering, evaluation is not trivial. But in many cases, the task is straightforward. Yet you need to choose the best possible metric for quality assessment. Otherwise, it results in bad output rated as good. The best evaluation metrics may be hard to implement and compute. At the same time, pretty much all modern techniques rely on minimizing a loss function to achieve good performance. In particular, all neural networks are massive gradient descent algorithms that aim at minimizing a loss function. The loss function is usually basic (for instance, sums of squared differences) because it must be updated extremely fast each time a neuron gets activated and a weight is modified. There may be trillions of changes needed before getting a stable solution. In practice, the loss function is a proxy to the model evaluation metric: the lower the loss, the better the evaluation. At least, that’s the expectation [..] Continue reading, get the code, and see how to not get stuck in a local optimum: ➡️ https://lnkd.in/gyUbtS8e