Ludwig

Announcing Ludwig v0.8—the first #opensource low-code framework optimized for building and #finetuning LLMs on your data. 🎉 New features incl. fine-tuning, integrations w/ Deepspeed, parameter efficient fine-tuning (#LoRA), prompt templating and more! pbase.ai/3s6WCu4

The fine-tuned #SLM leaderboard has been updated to include #GPT4o-mini. So, who's the champ? 👑 Let's hear it for #opensource! Fine-tuned open-source models dominated—outperforming #GPT4 and GPT-4o-mini on 80% of tasks. 😵🥊

✨ Introducing the Fine-tuning Index! A comprehensive set of benchmarks for 13 fine-tuned #opensource #LLMs and leading models from #OpenAI across 31 diverse tasks. The index reports essential metrics, including: 📊 Performance ⚡ Speed 💰 Cost pbase.ai/3wI4S6o

Ready to fine-tune and serve your own #LLM that outperforms #GPT4 — for free? 🚀 Join our virtual #workshop to get hands on with open-source #Ludwig + #LoRAX and learn how to fine-tune and serve #LoRA adapters for customer support use cases! 🙋‍♀️🙋‍♂️ pbase.ai/4awR658

There's a common belief that #opensource LLMs aren't as performant as #GPT4. Well, oftentimes they're not... unless you #finetune! Join our upcoming virtual workshop to learn how you can fine-tune OSS #LLMs to outperform GPT-4: pbase.ai/43YD0Hk

Network and learn from #LLM developers working on the most popular #opensource projects—all while taking in a view of the SF Bay. Click the link to see the talks and save your spot. Speakers include: @predibase @llama_index @guardrails_ai @tryolabs. pbase.ai/49cy7v2

🎉 NEW RELEASE: #LoRAX v0.9 is out! This release has the most contributions to date. Thank you to our growing community of contributors!! 👩‍💻👨‍💻 What's new: 👉 Overhauled #LoRA memory manager 👉 New model architecture: Qwen 2 👉 Optional max_new_tokens pbase.ai/4cAXqtZ

☎️ New #LLM tutorial from the #Ludwig community: Streamlining Customer Support! ☎️ Check out this blog from Ludwig hackathon winners, Divyansh Mishra, Abhishek Kaushik and Ankit Patil to hear how they fine-tuned #Mistral for intent #classification. pbase.ai/43rrLXs

📣 Bay Area Friends: Join us Apr 4th to learn about the lastest open-source #LLM frameworks and tips for building #production-grade apps from the leaders at @predibase @llama_index, @tryolabs and @guardrails_ai! Save your spot and invite your friends! pbase.ai/3vkKG9Q

🎉 Excited to share the launch of the @DeepLearningAI course Efficiently Serving #LLMs developed w/ @predibase CTO @TravisAddair 🎉 ✅ Implement a modern inference stack w/ caching, batching & #quantization ✅ Intro to #LoRA adapters ✅...and much more pbase.ai/3IFlLB5

👩‍🏫 Question/Answering systems are a cornerstone use case for #LLMs. Check out this short tutorial from Ludwig hackathon winner, Yogesh Haribhau Kulkarni, to hear how he fine-tuned #Llama2 on FAQ data for Indian tax law to create a #chatbot. pbase.ai/4a8X63q

👏 We recently hosted a community #hackathon and we're excited to share the contributions from one of our winners, Iuliana Stroia, and her project, Assessing #Health Data with #ML 🏥 ➡️ Youtube: pbase.ai/3Is8Uly ➡️ Notebook from the project: pbase.ai/48LXfsk

📣 ICYMI: Our team of #LLM experts hosted an interactive discussion on how they fine-tuned and served 25+ #Mistral adapters that rival #GPT4 using #LoRAX. Check out the replay to learn about our approach and how you can get started! pbase.ai/49zXE2v

Off the shelf is good, fine-tuned is better. Here's everything you need to #efficiently fine-tune @Google's #Gemma7B for your task. 🚀 Best of all, you can do it with open-source #Ludwig! 🔥 pbase.ai/3wC5l9B

Announcing #Ludwig v0.10.0: 🧠 Support for @Google's #Gemma 2B / 7B ⚙️ Added U-Net encoder-decoder and image output feature 🎛️ Added #Phi2 to model presets 📝 Support for prompt lookup decoding during generation 🛠️ Multiple bug fixes Full release notes: pbase.ai/49PfBdb

🎉 New minor release for #LoRAX, the open-source framework for serving 100s of #LLMs on a single #GPU: LoRAX v0.8 now supports structured outputs (#JSON mode) courtesy of #Outlines!

Introducing #LoRA Land: 25 fine-tuned #mistral-7b models that outperform #gpt4 for specific tasks. You can prompt the fine-tuned #LLMs and compare their results to mistral-7b-instruct in real time! All powered by Ludwig and LoRAX. Check out LoRA Land: pbase.ai/49mjjef

Ready to migrate off of those #expensive oversized commercial #LLMs? 💸 Check out our Model #Distillation Playbook containing all of our best practices for turning large models into #smaller, faster, cheaper LLMs! pbase.ai/49fViFH

🎊 Congrats to our #LLM Hackathon Winners! 🎊 To celebrate Ludwig's 10k🌟 we ran a mini virtual hackathon. Projects include: ▶️ Automating data quality checks in Python ▶️ #QA for Indian Tax Code ▶️ #Classification for Customer Service and more! pbase.ai/3SGAiSD

Jan edition of our Fine-Tuned Newsletter: 🧠 Fine-Tune #Zephyr for Cust Support 🏎️ @caveminds Podcast: Smaller Cheaper #Finetuned #LLMs 👩‍💻 Coding for non-coders 🚀 OSS: Ludwig v0.9.3 / #LoRAX v0.6 💬 Product: #Prompting fine-tuned #endpoints ...and more! pbase.ai/3UlJNrE

🎉 Announcing #Ludwig v0.9.3! 🎉 ✅ Support for microsoft/phi-2 ✅ Ensure correct padding token for #Phi and Pythia models ✅ Cast LLMEncoder output to torch.float32, freeze final layer at init ✅ Enable IA3 adapters ✅ Add batch size tuning for #LLMs pbase.ai/47L8Ur2