Really enjoyed listening to Ep 8 of @ZetaVector's podcast on ColBERT and ColBERTv2 — seems to be their most popular episode! Nit: 𝗖𝗼𝗹𝗕𝗘𝗥𝗧 stands for 𝗖𝗼ntextual 𝗹ate interaction over 𝗕𝗘𝗥𝗧, an oft-missed double pun on the 'late' night show. youtube.com/watch?v=1hDK7g…

🎯Intel Extension for Transformers demonstrated the leadership inference performance on CPU in MLPerf v3.0! It also showcased the extreme inference efficiency (throughput/INT8 TOPs w/o sparsity) among the main submitters! Data source: MLPerf v3.0 results

Watch now: ‘Few-shot Learning in Production’ by Moshe Wasserblat, NLP Research Manager INTEL presented at the #Healthcare #NLPSummit 2023 hubs.ly/Q01Q9RX80

An Efficient Sparse Inference Software Accelerator for Transformer-based Language Models on CPUs paper page: huggingface.co/papers/2306.16… In recent years, Transformer-based language models have become the standard approach for natural language processing tasks. However, stringent

📢The initial sample code to run INT4 model using ONNX Runtime is available: github.com/intel/intel-ex… 👉Working on the performance improvement. Stay tuned!

I can't believe that my two beautiful daughters are on these posters. It's been a full month. I miss them so much. Please help me get them back💔

🔥Thanks to WorldofAI for publishing an amazing video to introduce NeuralChat, the most powerful 7B model crated by Intel AI ranked Top-1 in Hugging Face LLM open leaderboard! 🎯Video: youtube.com/watch?v=bWhZ1u… #intelai Wei Li Huma Abidi @KeDing Moshe Wasserblat

They stole them. They stole our babies. They took them from their homes. They took them in their pyjamas. They took our babies from their cribs. They took our children from their beds. They took our sisters. Our brothers. They took our mums and dads. They took our

🎁Thrilled to share Intel Neural Compressor v2.4 is out on a nice snowy day in SH, a special release for model quantization/compression for LLMs, helping to bring AI everywhere. 👨‍💻Release notes: github.com/intel/neural-c… 🚀Code: github.com/intel/neural-c… #iamintel #intelai #oneapi

🎯Quantization + Speculative decoding shows significant speedup up to 7.3x on Xeon using Intel AI SWs: 📢IPEX: github.com/intel/intel-ex… ITREX: github.com/intel/intel-ex… 🤗Blog: huggingface.co/blog/intel-sta… Congrats to Intel AI and Hugging Face team! Moshe Wasserblat Huma Abidi

Don't miss the opportunity to hear from Moshe Wasserblat, NLP Research Manager Intel, who will be presenting at the #Healthcare #NLPSummit 2024 on April 2-3. His talk is titled ‘Intel’s fastRAG for Medical Use-cases’ Read more: hubs.li/Q02pYlsS0

Tune in to Moshe Wasserblat to learn about Intel's fastRAG framework. Discover how this innovative toolset integrates state-of-the-art retrieval and generative models to enhance efficiency in medical NLP applications.

Thank you, Moshe Wasserblat, for a great talk and for explaining the integration of diverse retrieval and generative models, empowering NLP researchers and developers to drive breakthroughs in healthcare technology

🚀 Dicta-LM 2.0 - מודל שפה גנרטיבי גדול בעברית, פתוח, וחזק! 🎉 דמו של גרסת הצ'אט: shorturl.at/BPY09 אתר החדש של דיקטה עבור תרגום מכונה אנגלית<->עברית שמבוסס על המודל: translate.dicta.org.il קישור לבלוג המלא: dicta.org.il/dicta-lm 🧵>>

We need more benchmarks outside of English, to account for linguistic diversity and evaluate languages on their specificities! Hebrew, for ex, is morphologically rich. דיקטה DICTA & Tal Geva collaborate to make a new Open LLM Leaderboard for Hebrew! 🔥 huggingface.co/blog/leaderboa…

Speculative decoding is fast. DISCO improves it further by **10-100%** 🤯 It's so consistent that it now is the default in `transformers` when you trigger assisted generation/speculative decoding. The folks behind DISCO released a blog post today, briefly explaining how it works

Label a few examples; train a small classifier; outperform LLMs 500x larger & slower for your classification task. Sounds difficult? It's not. Check out this blogpost using Argilla, distilabel, and SetFit to try it yourself: huggingface.co/blog/sdiazlor/…

2/ Huge thanks to the dream team! Jonathan Mamou, Oren Pereg, João Gante, Lewis Tunstall, Daniel Korat, Moshe Berchansky, Moshe Wasserblat, Roy Schwartz 🌟 Blog: huggingface.co/blog/dynamic_s… Paper at NeurIPS ENLSP Workshop: arxiv.org/pdf/2405.04304 Find more in this thread... 💡