The only playground that lets you query dozens of open-source and proprietary models at once and compare quality, cost, throughput. Get started for free 👇
Our batch evaluation tool comes with a scoring model that is able to reproduce academic MMLU benchmark results very closely. Check it out :👇
https://t.co/XoLAcfp78U
The Thanksgiving chaos at OpenAI highlights the importance of open-source AI. Relying exclusively on OpenAI for your AI strategy is risky and unwise. @neutralino1's take below 👇
https://t.co/O87zkgMMcm
We're super excited to share our new free AI-assisted batch evaluation tool for LLMs.
You can evaluate, score, and compare LLMs for your specific application, on your own eval dataset.
Reach out for an invite link.
https://t.co/ccbOipbyPT
The gist of my talk at @MLOpsWorld was that there are two emerging classes of #LLMs: Oracle models which are very large, hard and pricy to train and run, and Worker Bee models which are cheap, fast, and good at specific tasks. Fine-tuned on synthetic data.
Check out Josh's talk at #RaySummit on how Sematic enables distributed compute via on-demand Ray clusters. cc @raydistributed@anyscalecompute
https://t.co/qPpuO1SB73
🚀 These YC companies make building with AI 10x easier!
🧑💻 An underrated aspect of @ycombinator is the wealth of resources the community has created for building with LLMs & AI.
🔖 Here's a list of tools you can start using today. You don’t want to miss bookmarking this!
🧵From testing & fine-tuning to infrastructure, the range of AI dev tools crafted by YC founders is incredible.
❓What’s a hard thing you wish was easy when building AI apps?
🥁 Orchestration
@SematicAI: The open-source orchestrator loved by ML teams. It enables end-to-end pipelines to reduce model turnaround time by 80%.
@DAGWorks’s Hamilton: Open-source micro-orchestration framework for describing data flows. Companies use it for modeling data and feature engineering pipelines, prompt engineering, and LLM application workflows.
Arakoo's EdgeChains: Open Source SDK that models generative AI applications as config management. Built on top of Jsonnet as the orchestration grammar.
We're excited to release our free and #opensource pipeline to fine-tune #Llama2, #FLANT5, and #GPTJ.
👇Check it out and reach out if you want to learn more about fine-tuning #LLMs.
https://t.co/bVqYF6RPB1
Exciting to see a potential alternative to transformers that yields lower latencies and costs, and on @AMD#GPUs.
This is what the #AI industry needs to move faster.
The main bottlenecks at this time are access to GPUs, training and inference costs, and inference latency. Seems like RetNet could address all of the above. Of course it doesn't come without tradeoffs, but these can be mitigated or even removed over time.
By abstracting away infrastructure, and guaranteeing visualizations, traceability, and observability, we’ve measured an 80% speed-up in #ML model development and retraining time with Sematic. The result? Faster innovation and competitive edge. #AI#Productivity
📢We're proud to publish our first Customer Case Study 📢
Voxel reported an 80% reduction in model training time thanks to Sematic's productivity gains.
Traceability, visualizations, scaling contribute to cutting down turnaround time for #MachineLearning.
https://t.co/Qbg4LppM9M
"At Cruise, we identified a wide gap between the complexity of #CloudInfrastructure and the needs of MLEs. Our goal was to let them iterate on model logic and visualize the impact of their changes quickly, without having to learn cloud infra."
https://t.co/fjwEf2LRPK
@neutralino1
answered insightful questions from
@jrdothoughts
for The Sequence newsletter. Topics: scaling production ML pipelines, #MLOps in the era of #GenerativeAI and Sematic's vision. Check it out 👇
https://t.co/fjwEf2MpFi
Check out my latest article: Streamlining HealthTech Machine Learning with @SematicAI End-to-End Pipeline Solution https://t.co/exQ6v7Kp8U via @LinkedIn