This week, we explore what FlashMLA is, why it matters, and how it builds on key innovations like Mixture of Experts, infinite context attention, and context caching! https://
louisbouchard.substack.com/p/how-flashmla
-cuts-kv-cache-memory
…
@whats_ai
-
FlashMLA: Cutting KV Cache Memory in LLMs
By
–
-
DeepSeek FlashMLA: Revolutionizing LLM Efficiency with 93% Memory Reduction
By
–
New video!! DeepSeek's FlashMLA Explained! DeepSeek’s Game-Changer for LLM Efficiency, cutting KV Cache Memory to 6.7%! (yes, to 6.7% of traditional methods, not by 6.7%!!). Learn more in the video: https://
youtu.be/PTNW3SLP8W8 #flashmla #deepseek #llms -
Nvidia’s Keynote Tackles Multiple Industries with Exciting Releases
By
–
Jensen’s keynote this year was once again Nvidia is just going to tackle every industry. So many releases and cool plans. Exciting for the future! Hyped for the rest of the talks and exhibits this week. Will share the coolest stuff 🙂 Thanks @nvidia for the amazing seats
-
OpenAI’s Reinforcement Fine-Tuning: Customize Language Models Efficiently
By
–
Good morning! Have you ever wanted to take a language model and make it answer the way you want without needing a mountain of data? Well, OpenAI’s got something for us: Reinforcement Fine-Tuning, or RFT, and it changes how we customize AI models. Instead of retraining it with
-
Reinforcement Fine-Tuning vs Supervised Fine-Tuning Explained
By
–
New video! We are diving into Reinforcement Fine-Tuning (RFT) What it is, how it works, when to use it and more… Let's demystify supervised fine-tuning vs. reinforcement fine-tuning. Both essentially have their use that we can discuss in one line: • Supervised
-
Reasoning systems for idea clustering and proposition grouping
By
–
Now, onto the cool part: after extracting propositions, we need a system that can reason about each one and decide whether it fits into an existing chunk or if it should start a new one. This grouping process mimics how we naturally group ideas when reading or taking notes.
-
Agentic Chunking: LLMs Merging Related Content Intelligently
By
–
Unlike traditional chunking that often isolates scattered information, agentic chunking leverages LLMs to merge related content, maintaining the natural connections between chapters, paragraphs, and clauses.
-
Semantic Chunking for Optimized RAG Retrieval and Context
By
–
Instead of simply splitting text at every sentence break, this approach groups ideas much like we naturally do when reading or taking notes. It preserves context and helps organize your LLM content when retrieving the RAG chunks.
-
Agentic Chunking: Semantic Text Decomposition Method
By
–
Following yesterday's post sharing chunking tips, I think "Agentic Chunking" is an awesome idea! Agentic chunking is a method for breaking down text into semantically coherent pieces (transforming your chunks) by extracting clear, stand-alone statements—called propositions.
-
Clarifying Implicit References in AI Technical Communication
By
–
Implicit references can also trip things up. So, instead of "It revolutionized NLP," say something clearer like, "The 'Attention Is All You Need' paper revolutionized NLP by introducing Transformers."