🔥The 9th Round of Easy Loan, Earn $40 Reward is in progress❗️ ⏰ Promotion Period: January 15th - Feburary 15th, 2025 👉 Register now and check more details at gate.io/campaigns/358

На видеонаблюдението в секция 010200004 в Белица се вижда как пред всички останали членове на комисията жена попълва празни бюлетини, като отбелязва номерата 8 и 18 - това са съответно "ДПС - Ново начало" и ГЕРБ-СДС По информация на БНТ жената е от квотата на Ново начало.

This is the way!

An interesting thing about writing frontend compared to backend code with an LLM is that "taste" becomes much more important All good LLMs write me more or less the same Python function. For HTML/CSS/JavaScript I tend to pick Sonnet 3.5 because I really like its aesthetic taste!

Every time I screw up my code with silent bugs, I’m reminded of these words of wisdom from Karpathy Unfortunately I tend to immediately forget the lesson and go back to YOLO-mode 🥲

Docs on finetuning embedding models: sbert.net/docs/sentence_…

A blogpost we wrote 2 years ago: Fine-tuning with Low Budget and High Expectations jina.ai/news/fine-tuni…

Изборна справка от ромските махали. 1. жк Филиповци, най-западната махала на София. Април 2021 (вътрешен министър Терзийски/ГЕРБ) там гласуват 1100 души, повечето за ГЕРБ. След това четири поредни избора със служебни министри Рашков и Демердиев там активността беше вяла

If you are looking for some additional details on Haiku 3.5, the model card is the best I've found - assets.anthropic.com/m/1cd9d098ac3e…

Visualisation as a super power.

I just released Sentence Transformers v3.3.0 & it's huge! 4.5x speedup for CPU with OpenVINO int8 static quantization, training with prompts for a free perf. boost, PEFT integration, evaluation on NanoBEIR, and more! Full release notes: github.com/UKPLab/sentenc… Details in 🧵

TIL: With the release of `sentence-transformers==3.3.0`, you can now train and load LoRA adapters for embedding or ranking models! 👀 This looks very promising and awesome for domain-specific RAG applications! 🚀

Less and less reasons for staying on Twitter. Used to be the place to get the latest ML/NLP vibe. Now Linkedin (Linkedin!!!) does almost as well and various Discords/reddits do better. I am gradually moving towards the blue butterfly. Same handle. See you there.

Sentence Transformers v3.3.0 now allows you to train embedding models with prompts. In my quick tests, this improves my NDCG@10 by 0.66% and 0.90% - just by prepending "query: " and "document: " in front of your queries & docs. Info in 🧵

Разбира се, Кирил е виновен за всичко в държавата. И извън нея.

If a package fails to build during resolution or installation, uv will now tell you _why_ that package was included in the first place. I think this is such a cool error-message improvement.

450 идеи за града - това са вашите идеи за нашата София! Изключително щастлив съм, че имаме възможност да разгледаме и подготвим за гласуване рекорден брой предложения по Програма „Граждански бюджети на София“, след като отпуснахме и рекорден бюджет - 5 300 000 лева. Именно

We've released a technical report detailing our pre-deployment testing of Anthropic's upgraded Claude 3.5 Model with the U.S. AI Safety Institute. Read our blog for a high-level overview. aisi.gov.uk/work/pre-deplo…

I love doing these kind of recommendations! Stop using EasyOCR (github.com/JaidedAI/EasyO…), it's easy to use yes but a bit outdated (4/5 years old). Try MGP-STR instead, a ViT based model available on Hugging Face for running OCR at the edge! huggingface.co/alibaba-damo/m…

sqlite-vec v0.1.6 is now out, with metadata support! SQLite vector search w/ metadata filters 👀 - Perform extra filtering w/ WHERE clause in KNN queries - Internally shard vector indexes with partition keys - Aux columns for easy lookups read more: alexgarcia.xyz/blog/2024/sqli…

Wait - is the new GPT-4o a smaller and less intelligent model? We have completed running our independent evals on OpenAI’s GPT-4o release yesterday and are consistently measuring materially lower eval scores than the August release of GPT-4o. GPT-4o (Nov) vs GPT-4o (Aug): ➤