Modernizing the fundamentals of log management at Uber: How we used CLP to build a new logging infra that lets users view and analyze their logs seamlessly, at scale! 🪵🔍 Read more: https://lnkd.in/gUDr2CKe #logging #softwareengineering #developertools #CLP #UberEngineering #UberEng
Uber Engineering’s Post
More Relevant Posts
-
Definitely I will use it to share my ML pipelines :D! exited to apply this new knowledge.
Montserrat Peñaloza-Amion's Statement of Accomplishment | DataCamp
datacamp.com
To view or add a comment, sign in
-
Batch processing at scale is a requirement for the majority of AI-powered apps, from ETA prediction to image processing. Kubernetes, as great as it is, may fall short when having to deal with large datasets, support multi-tenancy and handle massively parallel executions. Check out the recent blog post from Haytham Abuelfutuh, our CTO, to learn more about how Union.ai (supercharged ⚡ by Flyte) delivers a platform that not only augments Kubernetes but delivers data awareness, job dependency management, and efficient resource utilization for batch workloads. 👇
Scaling patterns for batch workloads on K8s • Union.ai
union.ai
To view or add a comment, sign in
-
Interesting read on using Kong with Istio https://lnkd.in/emRxncfg.
Using Istio and Kong in Kubernetes Cluster
tanmaybatham.medium.com
To view or add a comment, sign in
-
Innovative Transformational Leader | Multi-Industry Experience | AI & SaaS Expert | Generative AI | DevOps, AIOps, SRE & Cloud Technologies | Experienced Writer | Essayist | Digital Content Creator | Author
v0.8.0 by laipz8200 via Release notes from dify ([Global] oracle cloud) URL: https://ift.tt/LmSAo8a ✨ What’s New in v0.8.0? ✨ Hey everyone, we’re excited to announce the release of version 0.8.0! This update brings a mix of new features, enhancements, and crucial bug fixes. Here’s a quick rundown: 🔥 Key Feature Parallel Execution of Nodes in Workflows by @takatost, @zxhlyh, and @YIXIAO0 in #8192. Nodes can now be executed in parallel within a workflow, greatly increasing the execution speed. This feature is especially beneficial for complex workflows that involve multiple steps or processes, allowing for quicker completion times and improved performance. Dive deeper into the details and unleash the full potential of these new features by exploring our latest blog post and documentation! 🚀 New Features Support gpt-4o-2024-08-06 and json_schema for Azure OpenAI Service: Support for the latest GPT-4o model and JSON schema for Azure OpenAI by @hjlarry in #7648 Support Oracle Cloud Infrastructure Generative AI Service: Oracle Cloud Infrastructure is now a supported model provider by @tmuife in #7775 Support Fish Audio TTS: Added support for Fish Audio Text-to-Speech models by @leng-yue in #7982 ⚠️ Deprecated Features Deprecate N to 1 Retrieval by @zxhlyh in #8134 The N-to-1 retrieval strategy is officially deprecated in this version, of which the entrance will be closed but applications that have selected this feature will still be retained. We recommend switching to the more flexible multi-path retrieval strategy to boost your application's retrieval efficiency. ⚙️ Enhancements Update App Published Time After Clicking Publish Button: The published time of an app now updates correctly when you click the publish button by @vicoooo26 in #7801 Return Page Number of PDF Documents Upon Retrieval: When retrieving PDF documents, the page number is now returned for better navigation by @jasonkang14 in #7749 🛠️ Bug Fixes Fix Notion Table Extract: Fixed issues with extracting data from Notion tables by @JohnJyong in #7925 Fix Nvidia Rerank Top N Missed: Addressed issues with Nvidia rerank top N functionality by @JohnJyong in #8185 Fix Claude Credential Validation: Resolved credential validation issues for Claude by @crazywoola in #8109 That’s it for this release! As always, we appreciate your feedback and contributions. Do it for you! 🚀 Upgrade Guide Docker compose deployments Warning The docker-compose.yaml has been refactored. If you've made any changes to the file, make sure to check out the "Upgrade to new docker compose deployment" section above for usage and migration tips. Back up your customized docker-compose YAML file (optional) cd docker cp docker-compose.yaml docker-compose.yaml.$(date +%s).bak Get the latest code from the main branch git checkout main git pull origin main Stop the service,Command, please execute in the docker directory docker compose down Back up data tar -cvf...
v0.8.0 by laipz8200 via Release notes from dify \(\[Global\] oracle cloud\) URL: https://ift.tt/LmSAo8a ✨ What’s New in v0.8.0? ✨ Hey everyone, we’re excited to announce the release of version 0.8.0! This update brings a mix of new features, enhancements, and crucial bug fixes. Here’s a quick rundown: 🔥 Key Feature Parallel Execution of Nodes in Workflows by \@takatost, \@zxhlyh, and...
github.com
To view or add a comment, sign in
-
Being able to debug and understand the flow in modern asynchronous applications is challenging but important. This article from Panchanan Panigrahi gives a good overview of what Distributed Tracing is and why you should be using some form of it. https://lnkd.in/evap5JvF
Unlocking the Power of Distributed Tracing: Navigating the Digital Cosmos🌌🔍✨
dev.to
To view or add a comment, sign in
-
Black Box vs White Box Observability in Kubernetes FROM the Google SRE Book: Chapter 6 The simplest way to think about black-box monitoring versus white-box monitoring is that black-box monitoring is symptom-oriented and represents active—not predicted—problems: "The system isn’t working correctly, right now." White-box monitoring depends on the ability to inspect the innards of the system, such as logs or HTTP endpoints, with instrumentation. White-box monitoring therefore allows detection of imminent problems, failures masked by retries, and so forth. The following post represents different layers in observability. Really nice start. I recommend it. https://lnkd.in/dEapuTxc #sre #kubernetes #observability
Black Box vs White Box Observability in Kubernetes
itnext.io
To view or add a comment, sign in
-
Check out my latest blog post on how XetHub brings software engineering best practices to ML development & MLOps. What do you think?
MLOps promises full automation of ML pipelines 🤖, but real world implementations can be difficult to navigate due to fragmented tools and poor observability. Tools originally built to support software development often can't scale to needs of ML workloads, leading to slow and inefficient development patterns. 🐌 So what can be done? Read our take on the challenges and a potential solution: https://lnkd.in/gZM4g2k8
XetHub | Turning MLOps challenges into modern ML workflows
To view or add a comment, sign in
-
Observability: From Zero to Hero In the last two years, I’ve set up Kubernetes-based observability for three different customers. While the setup isn’t overly complex, it’s easy to veer off the beaten track without proper guidance. To help others navigate this process, I’m starting a series called “Observability from Zero to Hero.” Over the next few weeks, I’ll share practical insights, key lessons, proven techniques and dark arts that have made these implementations successful. Here’s the first part of the series – a step-by-step guide to getting started with observability in Kubernetes. Read Part I here: https://lnkd.in/eMbsVDp6 Stay tuned for more! #observability #kubernetes
Observability From Hero to Zero - Part I: Up and Running
til.jingkaihe.com
To view or add a comment, sign in
-
Revolutionizing Kubernetes With K8sGPT: A Deep Dive Into AI-Driven Insights https://lnkd.in/eUyZ3-gs
K8sGPT: AI Insights for Kubernetes - DZone
To view or add a comment, sign in
-
Biomedical Engineer Unibg - ETH Zürich - MIT - Oxford University | MLOps Researcher | Software & ML Engineer
🚀 Exciting News! Introducing LLMLOps - Leveraging Language Models for Operationalizing Processes 🤖💬 Thrilled to share insights on this innovative approach combining MLops principles with the power of Large Language Models (LLM). 🌐✨ Key Points: - 🧠 Mindset Shift: Embrace a thinking methodology where LLMs drive decision-making and problem-solving. - 🔄 Workflow Integration: Seamlessly integrate LLMs into existing MLops pipelines for enhanced efficiency. - 🛠️ Tooling Evolution: Explore new tools tailored for managing LLM-based operations and workflows. - 🌐 Global Collaboration: Leverage LLMs for multilingual and culturally aware AI applications, fostering global collaboration. Let's shape the future of AI operations together! 🚀💡 (thanks Maria Vechtomova for the sharing) #LLMLOPS #AI #MLops #Innovation #LanguageModels #TechRevolution
This might be the first available hands-on LLMOps course out there! By Paul Iusztin, Pau Labarta Bajo, and Alexandru Răzvanț 👋. In the course, they walk you through the 3-step pipeline design 𝟭. 𝗧𝗿𝗮𝗶𝗻𝗶𝗻𝗴 𝗣𝗶𝗽𝗲𝗹𝗶𝗻𝗲 ➡ load a proprietary Q&A dataset ➡ fine-tunes an open-source LLM using QLoRA ➡ log the training experiments, visualize results, and register the model using Comet ➡ Deploy using Beam as a serverless GPU infrastructure. 𝟮. 𝗦𝘁𝗿𝗲𝗮𝗺𝗶𝗻𝗴 𝗥𝗲𝗮𝗹-𝘁𝗶𝗺𝗲 𝗣𝗶𝗽𝗲𝗹𝗶𝗻𝗲 ➡ ingest financial news from Alpaca ➡ clean and transform the news documents into embeddings in real-time using Bytewax ➡ store the embeddings into the Qdrant Vector DB ➡ deploy on an AWS EC2 machine using a GitHub Actions workflow. 𝟯. 𝗜𝗻𝗳𝗲𝗿𝗲𝗻𝗰𝗲 𝗣𝗶𝗽𝗲𝗹𝗶𝗻𝗲 ➡ download the fine-tuned model from Comet's model registry takes user questions as input ➡ query the Qdrant Vector DB and enhance the prompt with related financial news ➡ call the fine-tuned LLM for financial advice using the initial query, the context from the vector DB, and the chat history ➡ log the prompt & answer into Comet ML's LLMOps monitoring feature ➡ deploy using Beam as a serverless GPU infrastructure, as a RESTful API. 💡 Also, it is wrapped under a UI for demo purposes, implemented in Gradio. #llm #mlops Link to the course: https://lnkd.in/gZAHGfkA
To view or add a comment, sign in
185,147 followers