Giskard’s Post

New video demoing Databricks MLflow integration ✨🎥 This partnership combines Giskard's LLM evaluation capabilities with MLflow's model management features. It allows you to automatically evaluate model vulnerabilities to detect hidden issues like hallucinations and robustness problems, generate customize tests, and compare model performance across versions. Watch the demo to learn how to: 🔹 Integrate Giskard with MLflow. 🔹 Detect model vulnerabilities. 🔹 Customize domain-specific tests. Watch it now 👉 https://lnkd.in/e6iYEGqv #LLMeval #Databricks #MLflow #MLOps

The process of getting your models to production is not without its obstacles. You should expect many iterations, and there are a number of tools that can help. We have compiled a list of younger players in the world of MLOps, each with their different and unique approaches to the same problem. https://lnkd.in/dauuFphh #MLOps #DataOps #DataScience #ModelDeployment #AIDevelopment

Databricks recently released DBRX Instruct, an open, general-purpose LLM. Across a range of standard benchmarks, DBRX sets a new state-of-the-art for established open LLMs. Moreover, it provides the open community and enterprises building their own LLMs with capabilities that were previously limited to closed model APIs. #databricks #LLM # ModernDataPlatform https://lnkd.in/dEa_Mdnr

I attended the Databricks #PartnerTechTalk this morning and got the scoop on their latest gift to the open-source GenAI community—their Mixture of Experts model (MoE), #DBRX. Here’s the #TLDR, so you don't have to: 1. DBRX is a team effort, with the AI community chipping in to create this beast. It has 132B total parameters, but thanks to its MoE design, only 36B parameters are in play for any given input. 2. There are two flavors of DBRX: the base model, perfect for tweaking with your own data, and the Instruct model, which is already primed to tackle questions and follow instructions right out of the gate. 3. Research is pointing to the pitfalls of testing models against static benchmarks. DBRX scores excellently on the WildBench benchmark that tests language models on challenging tasks from real users. 4. Excited to play with DBRX ? Check it out on Hugging Face’s Databricks space, Databricks API or Playground, or on GitHub. Short, sweet, and everything you need to know! Thanks Tim Lortz for talking us through it.

DBRX 132B is wild! 🤯 Databricks introduces DBRX, a new 132B-parameter open LLM 🔥 Here are some of the technical details: • It utilizes a fine-grained mixture-of-experts (MoE) architecture with a total of 132B parameters. • Trained on 12 trillion tokens of text and code data. • 32K context size. • Outperforms Grok-1, Mixtral, etc. • Mixture of Experts, 16 experts. • Utilizes rotary position encodings (RoPE), gated linear units (GLU), and grouped query attention (GQA). • Offers open access to Base and Instruct models. The Instruct model is now available on the Clarifai Platform! Try it out here: https://lnkd.in/gPjtS3fP #DBRX #genererativeai #llm

I'm thrilled to announce that I've recently completed the Gen AI & LLM on Databricks Partner certification! Databricks Partner Badge, covered a comprehensive range of topics including: - Best practices for engaging customers in their GenAI and LLM strategies - Capalibilities of DBRX,Vector DB & RAG - Databricks AI security framework, Governance, compliance, and security - Technical demos and technology integrations #Databricks , #GenAI, #LLM, #DataScience

⏱ I took just 10 minutes to test drive the latest LLM from #Databricks. 🎥 Check out this 15-second clip of me experimenting with using the LLM #DBRX to analyze customer surveys in the Databricks playground. For this simple task, it was really fast, accuracy was good. It's easy to get overwhelmed by the constant stream of tech news. But taking that extra time to get hands on can really help you form your own opinions. #DBRX #Databricks #TechInsights #ExploreTech

Our LLM + RAG chatbot demo just got faster response time, cheaper inferences, and better evaluation metrics with one simple change. What did we do? Simple! We just updated our Chatbot to use Databricks DBRX Instruct, the new State Of The Art LLM trained by Databricks. DBRX outperforms most of the other LLMs in traditional benchmarks, and provides excellent Flop/price performance! Want to try it yourself with Langchain? Here are all the notebooks, guiding you step by step! https://lnkd.in/eu4Cf2HA

Here's a short lists of MLOps open-source tools that you can use for your ML pipelines! https://lnkd.in/gQf8wb_Y

Build your own RAG chatbot with #Databricks #DBRX... The demo notebooks will help to build end to end #llmops platform on Databricks.. And enhance your applications easily with the latest innovations like DBRX..