― Очень трепетно отношусь к своему времени и уважаю ваше. Не ответила — не судьба (может, я вам психику сохранила — ищите плюсы!).
― Что я ищу от сайта — без понятия, буду импровизировать.
― Интересны получающие удовольствие от жизни.
Вас много, холопы, а она одна, королева!
Сайты знакомств — ядрёный рассадник мужененавистничества, лишающий веры в человечество. Женщины могут открыто и безнаказанно не уважать мужчин, почувствовав себя Мисс мира, не выходя из хрущёвки.
― Мы пара МЖ. Мужчины-одиночки, [мы] отвечаем ТОЛЬКО на конкретные предложения об организации встречи! На «Привет. Встретимся? Познакомимся? Как дела?» не отвечаем.
И #продающеефото. И некая «поддержка» в интересах (с прочим досугом типа бани, ресторанов, шашлыков. нудизма 
). Заплати куколду за то, чтобы трахнуть (не) его тёлку? Маразм.
И хоть куколды должны страдать (раз не умеют создавать гетеросексуальные моногамные отношения), это нельзя популяризировать; но... 
Tencent's 'Hunyuan-T1'–The First Mamba-Powered Ultra-Large Model
Hunyuan T1 Mamba Reasoning model beats R1 on speed and metrics
New mammoth bone finds provide insights into the everyday life of Ice Age hunters. OeAI archaeologist have discovered the remains of at least 5 mammoths in Lower Austria! Stone tools, dismembered bones and tusk fragments indicate the utilisation of the meat and the processing of ivory by mammoth hunters 25,000 years ago.
#Archaeology #Mammoth #Prehistory #MAMBA #Langmannersdorf
https://www.oeaw.ac.at/en/oeai/media/news-archive/news-detail/langmannersdorf-en
Well, turns out it's a bit picky, but I did find a few cool games I could actually play on Windows 3.0.
We are #opensource developers, maintainers and consultants working on scientific and data-heavy software!
We are in part, or fully responsible for critical tools used by millions for research and tech across industries, like #Jupyter, #condaforge, #Mamba, or #ApacheArrow. We help navigate the complex data stack, use these tools efficiently, or develop and upstream contributions.
Follow us for light tech posts, opinions, links to deeper content as well as company updates.
@python_discussions An absolutely great blog post about the pain of environment management in #python.
It summaries the experiences I made over the years: Breaking my system python environment with plain #pip, moving to #conda for easy #CUDA installation, frustrated by its slow dependency solution and moving to #mamba; starting to like #poetry for its dependency locking but also annoyed by its undocumented C/ C++ Extensions support.
I thought about giving #pixi a try, maybe next time ...
@villares I would not use miniconda - use #mamba (nearly 100% conda compatible) - faster and free package manager - works perfectly fine with conda-forge https://mamba.readthedocs.io/ and for CI/CD or embedded environments there is #micromamba https://mamba.readthedocs.io/en/latest/user_guide/micromamba.html
#Conda is still driving me crazy.
The confusion is actually perfect...
Conda is actually a package manager.
But then there is also #Anaconda and #Miniconda as package managers.
Both Miniconda and Anaconda install conda.
Then there is #mamba A reimplementation of Conda in C++. There is also #Micromamba...but no #Minimamba...
@almenal99 not sure if this applies to you, but apparently recent versions of #conda use #mamba to resolve dependencies, speeding things up massively! Made a huge difference for me!^^
Don’t miss Wolf Vollprecht’s keynote at #EuroSciPy2024!
Wolf @wolfv, CEO of @prefix, creator of #mamba, and core member of @conda , will discuss making open-source contributions more accessible.
Join us for his talk, "Just contribute?!", and gain insights from his five-year journey in open source.
Get your tickets: https://ow.ly/uH7N50SUaHM
Oh, it is hard to keep up with Mistral. Beside NeMo 12B, they also released two other models:
* https://mistral.ai/news/mathstral/
* https://mistral.ai/news/codestral-mamba/
Codestral Mamba is exciting. I was not aware that they were also experimenting with Mamba2 models.
My understanding is that the Mamba architecture (https://arxiv.org/abs/2312.00752) is especially promising for problem domains that require large context lengths, so using it for programming makes sense. A fundamental difference compared to transformers (as introduced 2017 in the famous "Attention Is All You Need" paper) is that it does not have attention. Thus, it avoids the quadratic complexity that typically comes it. Another difference is that it is again a recurrent model; so, it is more like the once popular LSTM that fell out of fashion after the success of transformers.
Here is a longer interview with Albert Gu, one of the researchers behind Mamba, where he shares insights on his work on Mamba and Mamba2 and his design philosophy:
Finally the "#mamba for audio" papers are coming, appearing on arxiv. Surprisingly, they're based on spectrogram patches rather than raw audio. Why, when Mamba is for sequences? Probably because "spectrogram patch embeddings" are the best feature representation of the moment, and progressively they'll be replaced (?) as the early feature-extraction step.
Durante l'AI Festival ho parlato di #Mamba, una tecnologia rivoluzionaria in grado di superare i limiti di lunghezza delle sequenze e di efficienza computazionale. AI21labs ha creato #Jamba: il primo modello in produzione che combina Mamba con l'architettura dei Transformer tradizionali. 
Risultato: un sistema che ha la capacità di selezionare e "ricordare" informazioni importanti da lunghe sequenze, con una maggior capacità di comprensione del contesto.
#Mamba (2018)
After being left for dead at the hands of her brother, a rogue assassin goes after his most trusted employee, 'The Gatekeeper', to stake information and leave his empire in ruins. Unbeknownst to our heroine, a third player joins the game and another enemy is added to her growing list.
#GirlsWithGuns #SonaliiCastillo
#FilmMastodon 
Will #mamba overtake #Transformers in the #LLM race?
https://thegradient.pub/mamba-explained/
Don't know what rattler-build is yet? 
Cross-platform relocatable binaries/packages
Simple recipe format inspired by conda-build & boa
Bonus: rattler-build is a standalone tool with no dependencies on conda-build or Python! 
