🔥The 9th Round of Easy Loan, Earn $40 Reward is in progress❗️ ⏰ Promotion Period: January 15th - Feburary 15th, 2025 👉 Register now and check more details at gate.io/campaigns/358

Happy to OSS gpt-fast, a fast and hackable implementation of transformer inference in <1000 lines of native PyTorch with support for quantization, speculative decoding, TP, Nvidia/AMD support, and more! Code: github.com/pytorch-labs/g… Blog: pytorch.org/blog/accelerat… (1/12)

I'm excited to ship marker - a pdf to markdown converter that is 10x faster than nougat, more accurate outside arXiv, and has low hallucination risk. Marker is optimized for throughput, like converting LLM pretrain data. Find it here - github.com/VikParuchuri/m… .

🔥Open-source, open-science, and data curation for the win! Meet Notus 7B, a new LLM tuned with DPO on a new curated UltraFeedback dataset, surpassing Zephyr and Claude 2 on AlpacaEval. Built on the shoulders of giants: 🙌Hugging Face Alignment Handbook argilla.io/blog/notus7b

Now that Mixtral 8x7b is available in🤗Transformers, you might be wondering what the heck is a Mixture of Experts? We wrote an explainer on: ❓ How MoEs differ to conventional Transformers 🏋️‍♀️ How they're trained 🏎️ Subtleties with inference Enjoy! huggingface.co/blog/moe

Since Mixture of Expert (MoE) LLMs are all the rage as of this weekend, thanks to the Mixtral-8x-7B release, here's a quick explainer. The figure below shows the architecture behind the Switch Transformer (arxiv.org/abs/2101.03961), a great intro to MoEs. The model depicted in

New RAG technique alert 🚨 We’ve come up with an advanced RAG technique in LlamaIndex 🦙 that lets you ask structured questions over many documents ✨: 1. Model each document as a metadata dictionary - store more attributes beyond a simple text summary. (e.g. a row in SQL

RAG parameters are not one-size-fits-all, and you should tune it! Ravi Theja created a comprehensive walkthrough on how you should tune your retrieval parameters (specifically around hybrid search) based on properties of your query dataset. For some queries, having a low α

Exciting update! Cohere R+ on Azure, with its advanced RAG capabilities enables our AI assistant Atom deliver more accurate and verified answers with enterprise grade security. Looking forward to unlocking more use cases with Embed and Rerank. cohere Atomicwork

Thrilled to be partnering with cohere on launch of their R+ model in Microsoft Azure to enhance the enterprise AI capabilities of Atomicwork solution. Command R+ is the most advanced model for Enterprise RAG use cases!✨🚀 #AI #B2B #SaaS #RAG

This is 1000% true. I want to share a personal story from Atomicwork journey. At the start of Q4’23 till Q1’24 end, we asked our entire team at Atomicwork to go into war mode for intensity, consistency and urgency to ensure that we are ready for large enterprise

For AI apps, UX is very key. One issue that bothered us at Atomicwork for a long time was slow or missed slot-fill in forms due to SOTA models latency until we moved to Command-R from cohere ✨🚀 Now the worst case latency is down by 2X and less than 3s for even complex

Powered by leading enterprise AI platform cohere, we're excited to launch our digital workplace experience solution for modern enterprises! atomicwork.com/blog/digital-w…

Team Atomicwork kicking things off at @SITS_UK If you're around, come by stand #595, ask Aparna Chugh, Padmapriya Murali, Vijay Rayapati, and Vijay Shankar about the future of ITSM. And don't miss Aparna and Vijay's upcoming session on “AI Guardrails for IT Leaders” in ‘Theatre 3’!

Kicking off Day 2 at @SITS_UK. If you missed meeting our team yesterday, swing by stand #595. Ask Aparna Chugh, Padmapriya Murali, Vijay Shankar, and Vijay Rayapati what makes the Atomicwork approach "modern" in a sea of bolt-on AI ITSM products. Grab some purple swag while at it 💜

Teaming up with cohere, we have launched our new digital workplace experience solution, designed to revolutionize productivity and support for modern enterprises. What do we mean by that? Read on. 🧵 (1/6)

📣 Our Spring updates are here! Presenting a quick roundup of all the features we’ve built in the past two months to help you level up your #ITSM. 🧵

The Atomicwork team had a front-row seat at @SITS_UK in London last month. After countless chats and keen observations, one thing became abundantly clear: #AI is the future of #ITSM. 🌟 1/7

Don't miss out on the opportunity to learn from Cohere Co-Founder, Ivan Zhang, and Atomicwork CEO, Vijay Rayapati, as they explore GenAI's potential to revolutionize workflow automation. Join us on July 23 at 2:00 pm ET: info.cohere.ai/enterprise-wor…

At Atomicwork , we are committed to improving the accuracy of our AI assistant, Atom. With this goal, we experimented with ragas and Weaviate • vector database to improve the precision of our assistants' responses. Shared more details in this article: atomicwork.com/blog/ragas-imp…

Wrapped up Stanford CS336 (Language Models from Scratch), taught with an amazing team Tatsunori Hashimoto Marcel Rød Neil Band Rohith Kuditipudi. Researchers are becoming detached from the technical details of how LMs work. In CS336, we try to fix that by having students build everything: