LLMOps Space

We're hosting a 𝐡𝐚𝐧𝐝𝐬-𝐨𝐧 𝐰𝐨𝐫𝐤𝐬𝐡𝐨𝐩 about "𝐄𝐯𝐚𝐥𝐮𝐚𝐭𝐢𝐧𝐠 𝐑𝐀𝐆 & 𝐋𝐋𝐌 𝐀𝐩𝐩𝐬".🚀 In this session, Shir Chorev (CTO & Co-Founder) & Nadav Barak (Head of AI) from our team, lead a hands-on workshop on evaluating RAG and LLM-based applications. 💡 We will give all attendees access to the Deepchecks system during the workshop. 🤓 We will cover methodologies for assessing initial experiments, comparing versions, and performing ongoing evaluations in production using Deepchecks LLM Evaluation. ✅ ( ❤️ 300+ people have signed up for this one already) 🚀 𝐑𝐞𝐠𝐢𝐬𝐭𝐞𝐫 𝐡𝐞𝐫𝐞: https://lnkd.in/dscdK9HH 📆 𝐃𝐚𝐭𝐞 & 𝐓𝐢𝐦𝐞: Oct 8th, 2024 | 08:00 AM PST Feel free to drop your questions for speakers here in the comment section. 💬 #LLMOps #LLMs #AI #ML #GenAI #AISafety

😇Did you see OpenAI's announcement about #SearchGPT? OK, here is my bet - #SearchGPT won't be a big deal at first. They will see there are many missing features (like easy display of local businesses, easy transition to other websites with previews, etc), it will be a tiny improvement from ChatGPT. And won't really bite more into Google's user base than #ChatGPT already has... Then after a while, 1 of 2 things will happen. Either they will make major major changes to the experience, or they will take a step back and announce it didn't really work. Thoughts? ☝️𝐅𝐨𝐥𝐥𝐨𝐰 Philip Tannor 𝐟𝐨𝐫 𝐜𝐨𝐧𝐭𝐞𝐧𝐭 𝐚𝐛𝐨𝐮𝐭 #artificialintelligence , #llms 𝐚𝐧𝐝 #llmops, as well as announcements about Deepchecks' #opensource releases & the community at LLMOps Space.

The upcoming LLMOps Space event is about "𝐁𝐮𝐢𝐥𝐝𝐢𝐧𝐠 𝐭𝐡𝐞 𝐅𝐢𝐧𝐞-𝐓𝐮𝐧𝐢𝐧𝐠 𝐏𝐢𝐩𝐞𝐥𝐢𝐧𝐞 𝐟𝐨𝐫 𝐀𝐥𝐢𝐠𝐧𝐦𝐞𝐧𝐭 𝐨𝐟 𝐋𝐋𝐌𝐬". 🏗 In this session, Maksim Nekrashevich, ML & LLM Engineer from Nebius AI will discuss reinforcement learning with human feedback (RLHF), prompt tuning, and AI workflow management. 🚀 We will cover the key aspects of aligning LLMs and explore how to set up the necessary infrastructure to maintain a versatile alignment pipeline. ✅ 🚀 𝐑𝐞𝐠𝐢𝐬𝐭𝐞𝐫 𝐡𝐞𝐫𝐞: https://lnkd.in/dM4sj9ur 📅 𝐃𝐚𝐭𝐞 & 𝐓𝐢𝐦𝐞: July 11th, 2024 | 8.00 AM PST | 5.00 PM CET 📣 The session will be hosted by Philip Tannor, CEO and Co-Founder at Deepchecks. #LLMOps #MLOps #LLMs #GenAI #AI #ML

🌌 Upcoming webinar: Taming AI or How we build the alignment pipeline https://lnkd.in/dvpFvsAw Speaking at the LLMOps Space community's webinar will be Maksim Nekrashevich, ML & LLM Engineer at Nebius AI. Accompanied by Philip Tannor, CEO and Co-Founder at Deepchecks, Maxim will discuss: - Incorporating LLMs into the data collection for supervised fine-tuning (SFT) and reinforcement learning with human feedback (RLHF) to maximize efficiency. - Techniques for instilling desired behaviors in LLMs through the strategic use of prompt tuning. - An exploration of cutting-edge workflow management and how it facilitates rapid prototyping of highly-intensive distributed training procedures. When: Thursday, July 11 at 17:00 UTC+2 / 8:00 AM PST Where: Zoom Register: https://lnkd.in/dvpFvsAw #webinars #LLMs #alignment #SFT #RLHF #promptengineering

🔥 Two weeks ago we released #TextGrad, our new library for automated prompt optimization. The feedback we got since then has been amazing, with more than 600 stars on GitHub! ⭐ 📕 GitHub: https://lnkd.in/g9grWEwf 📕 Paper: https://lnkd.in/gYg3h7dP A short summary: 1️⃣ TextGrad is an "autograd for text" and provides an automated way to improve prompts with few lines of code. TextGrad's syntax is similar to PyTorch's, so it should all feel very familiar! 2️⃣ TextGrad implements an entire engine for backpropagation through text feedback provided by LLMs, strongly building on the gradient metaphor: We can optimize compound AI systems. 3️⃣ TextGrad can provide feedback on system prompts or coding solutions and optimize them! We have also applied TextGrad to molecule and treatment plan optimization! Amazing work led by Mert Yuksekgonul, Joseph Boen, Sheng Liu, Zhi Huang, Carlos Guestrin, and James Zou

A general-purpose language model can only process relatively simple visual tasks such as answering basic questions about an image or generating short captions. 🤓 This is primarily due to the lack of access to detailed pixel-level information, object segmentation data, and other granular annotations that would allow the model to precisely understand and reason about the various elements, relationships, and context within an image. 👾 ✅ Fine-tuning LMMs on domain-specific data can significantly improve their performance for targeted tasks. Learn how to 𝐟𝐢𝐧𝐞-𝐭𝐮𝐧𝐞 𝐚𝐧𝐝 𝐝𝐞𝐩𝐥𝐨𝐲 𝐭𝐡𝐞 𝐋𝐋𝐚𝐕𝐀 𝐦𝐨𝐝𝐞𝐥 on Amazon SageMaker. 👩💻 𝐋𝐢𝐧𝐤 𝐭𝐨 𝐆𝐢𝐭𝐇𝐮𝐛: https://lnkd.in/d3VgZFyr 👉 𝐋𝐢𝐧𝐤 𝐭𝐨 𝐚𝐫𝐭𝐢𝐜𝐥𝐞: https://lnkd.in/gccN6ycM Changsha Ma, Alfred Shen, Amazon Web Services (AWS) #LLMOps #MLOps #LLMs #AWS #GenAI #LLaVa #AI #ML

𝐔𝐧𝐢𝐪𝐮𝐞3𝐃: An Open Source model framework that converts static images into high-quality 3D models efficiently. 🔥 Unique3D significantly outperforms other image-to-3D baselines in terms of geometric and textural details. This framework uses a multi-view diffusion model with a corresponding normal diffusion model to generate multi-view images. 💡 It uses a multi-level upscale process to progressively improve the resolution of generated orthographic multi-views, as well as an instant and consistent mesh reconstruction algorithm called ISOMER, which fully integrates the color and geometric priors into mesh results. 🚀 𝐓𝐫𝐲 𝐲𝐨𝐮𝐫𝐬𝐞𝐥𝐟 𝐡𝐞𝐫𝐞 𝐨𝐧 𝐇𝐮𝐠𝐠𝐢𝐧𝐠 𝐅𝐚𝐜𝐞:  https://lnkd.in/gzZAYNGv 𝐋𝐢𝐧𝐤 𝐭𝐨 𝐭𝐡𝐞 𝐩𝐚𝐩𝐞𝐫: https://lnkd.in/d9vpX2BN 𝐋𝐢𝐧𝐤 𝐭𝐨 𝐆𝐢𝐭𝐇𝐮𝐛: https://lnkd.in/gmVRJ4e5 LLMOps Space Hugging Face #LLMs #AI #ML #LLMOps #GenAI #GenerativeAI

The recording of our previous webinar with Langfuse (YC W23) about "Traceability and Observability in Multi-Step LLM Systems" is now available. 📣 In this session, Marc Klingen, CEO & Co-Founder at Langfuse, talked about advanced techniques and best practices for implementing traceability and observability in multi-step LLM systems. 🏗 Topics that were covered: ✅ 𝐌𝐮𝐥𝐭𝐢-𝐒𝐭𝐞𝐩 𝐋𝐋𝐌 𝐀𝐩𝐩𝐥𝐢𝐜𝐚𝐭𝐢𝐨𝐧𝐬: Understanding the complexity of LLM systems that involve multiple steps, agents, and chains. ✅ 𝐈𝐧𝐭𝐞𝐠𝐫𝐚𝐭𝐢𝐨𝐧 𝐄𝐱𝐚𝐦𝐩𝐥𝐞 (𝐑𝐀𝐆): A detailed examination of how RAG systems work. ✅ 𝐓𝐫𝐚𝐜𝐞𝐬 𝐟𝐨𝐫 𝐎𝐛𝐬𝐞𝐫𝐯𝐚𝐛𝐢𝐥𝐢𝐭𝐲: The role of traces in observability and how they serve as a powerful abstraction for logging and monitoring in both production and development environments. 🚀 𝐖𝐚𝐭𝐜𝐡 𝐡𝐞𝐫𝐞: https://lnkd.in/dev5VzE5 💡 𝐉𝐨𝐢𝐧 LLMOps Space 𝐃𝐢𝐬𝐜𝐨𝐫𝐝 𝐜𝐨𝐦𝐦𝐮𝐧𝐢𝐭𝐲: https://lnkd.in/gKv2VYXN #llmops #mlops #llms #ai #ml #opensource #openai #genai #vectors

🚨Event Alert Join us this Thursday: Implementing Traceability and Observability in Multi-Step LLM systems Marc Klingen will talk about advanced techniques and best practices for implementing traceability and observability in multi-step LLM systems. We'll discuss how leveraging OpenTelemetry (OTel) can provide deep insights into RAG systems 🔗 Link to register in the comments. ⏰ Date & Time: This Thursday, June 13th, 2024 | 8.00 AM PST | 5.00 PM CET Topics that will be covered: ✅ Multi-Step LLM Applications: Understanding the complexity of LLM systems that involve multiple steps, agents, and chains. ✅ Integration Example (RAG): A detailed examination of how RAG systems work. We'll demonstrate how user input, message history, and additional memory are combined to produce an aggregated response with accompanying sources. ✅ Traces for Observability: Learn about the role of traces in observability and how they serve as a powerful abstraction for logging and monitoring in both production and development environment. 👇 Registration in the comments cc Philip Tannor & LLMOps Space