Search

Remek Kinas reported that version 2.3.1 of the basal engine, released on October 21, has accelerated the performance of the basal-omni models. He wrote that in version 2.3, the engine with CUDA graphs began to fully support basal-omni (Gemma 4), previously available in fast mode via vLLM or slower PyTorch. Among the solutions implemented, he mentioned single calculations of document states for multiple questions as well as FlashAttention 3 and FlexAttention. In tests on a single H100 card, with 16 parallel queries, basal-omni-1.0-mini achieved 61.2 decisions per second compared to 10.0 in version 2.2, basal-omni-1.1-max — 16.2 compared to 4.8, and basal-omni-1.0-ultra — 6.1 compared to 2.0. For five questions about a long document, the results were 21.4 compared to 1.6, 10.6 compared to 0.72, and 4.8 compared to 0.28 decisions per second, respectively. Kinas noted that out of 10,606 test decisions, the best response matched version 2.2 in 99.85–99.99% of cases where the model was confident. The share of automated decisions at 1% error was expected to change by at most 0.1 points. Among the new features in version 2.3.1, he mentioned support for documents of up to 82,000 tokens by the ultra model on a single H100 card; previously, he noted that memory was insufficient above about 32,000. He also mentioned that in JevBench, basal did not provide 15 responses due to exceeding the limit. He added that agents and conversations only process the new part of the state, and the step for basal-omni-1.1-max was shortened from 0.79 to 0.27 seconds. In the case of multiple questions about a long document, the throughput of the mini model increased from 21.9 to 33.7 questions per second, a 54% increase. Kinas also wrote that Blackwell cards use the fast engine by default, and basal-2.0 (Bielik) in eager mode on H100 reduced response time from 591 to 52 ms. He listed clearer HTTP codes 422, 503, and 500, load available at /health, the basal-calibrate message about too few labels, and support for images and audio through vLLM for the mini model, which he described as 6–8 times faster. He also thanked the contributors and testers: Miłosz (Butterski), Grzegorz Brzezinka, from Piotrków, and Kamil (kamilrosa), as well as edurdias and the company BRAUN SI / NovaraX for testing on their own data. Special thanks were given to Damian Plewa and the IT SOL team for the basal-rs engine. #basal #AI #inference (translated)

Łukasz Warzecha criticized an author whom he described as an idol of the elites, and assessed her novels as verbose and grafomanic. He claimed that almost no one reads them, although few admit it. He accused her of only answering questions posed in a submissive manner, and in other cases refusing interviews. https://t.co/HsIJg1RHe4 #literatura #opinia #kultura (translated)

Karol Nawrocki: "Mam nadzieję, że krytycy literatury się ze mną zgodzą, ale Dziady Adama Mickiewicza, mógł napisać tylko Adam Mickiewicz"😂 Taki grafomański bełkot Karola Nawrockiego, mógł wypowiedzieć tylko Karol Nawrocki. https://t.co/qStl6d1hGD #literatura #Mickiewicz #krytyka

Jakby to napisać dyplomatycznie… Problemem JB jest jego przekonanie o nieomylności i posiadaniu tajemnej wiedzy w wielu dziedzinach na podstawie… no właśnie na podstawie w zasadzie czego? Ten człowiek przecież niczego nie bada twórczo, uczciwie nie zgłębia do jądra problemu, a jego książki to czysta grafomania i strumień samoświadomości. Piszę to po przeczytaniu pięciu prac, gdzie moja lista zarzutów jest znacznie dłuższa, ale to nie pora na to. Nie chce już wracać do znanego problemu przypisywania sobie cudzych ustaleń naukowych, ale zwyczajnie w świecie JB ma bardzo mgliste pojęcie o historii jako takiej. Może nawet *coś tam czytał*, ale jego mózg selektywnie wybiera sobie informacje pod określone tezy. No i istotnym czynnikiem jest twórcza modyfikacja tego w oparciu o głębokie przekonanie, że jego wnioski są słuszne i odkrywcze. Trudno JB nazwać naukowcem, bo nauki nie uprawia, ale nawet jako popularyzator po prostu przepada w przedbiegach. Ten człowiek autentycznie nie zna podstawowych faktów, notorycznie zmyśla, no i żyje sobie w bańce, że coś tam wykoncypował twórczego. Ale na obronę JB muszę powiedzieć, że to wszystko jest i tak jego wielkim sukcesem marketingowym. Idealnie wpisał się w model postkomunistycznego kraju z upośledzonymi neokolonialnymi elitami i aparatem państwowym. Dokładnie takich mamy „ekspertów” jaki mamy poziom nauki, akceptacji dla bylejakości i systemu kontroli społecznej, a właściwie jej braku. Witajcie w Polsce 2026, gdzie stary profesor komuch ma więcej w głowie niż cwaniaczek w skórzanej kurtce, który używa co drugie zdanie zwrotów, których nie rozumie 90% ludzi. #nauka #polska #kultura

🚨 Agora! Flamengo diz à Justiça que Gerson, do Cruzeiro, está 'desesperado'. Em petição enviada à Justiça na semana anterior ao jogo contra o Cruzeiro, o Fla disse que o volante e seu pai, Marcão, estão "desesperados", em processo onde as partes discutem o pagamento de uma multa de R$ 42,7 milhões. Na petição, o Rubro-Negro acusou Gerson e seu pai de tentarem atrasar o processo com pedidos de perícias e documentos "irrelevantes". Para o clube carioca, trata-se de uma "tentativa desesperada de protelar o desfecho que os réus sabem ser desfavorável". Gerson pediu uma perícia grafotécnica em uma carta manuscrita de demissão. O Flamengo alegou má-fé, já que o jogador havia admitido ter assinado o documento. A Justiça REJEITOU o pedido do jogador, classificando como DESNECESSÁRIO. O Fla cobra R$ 42,7 milhões pela rescisão antecipada do contrato de imagem de Gerson. O valor corresponde a 57 meses não cumpridos (R$ 750 mil por mês). Gerson, por sua vez, afirma que foi orientado pelo próprio Flamengo a escrever a carta de demissão para viabilizar sua ida ao Zenit. Ele classificou o processo como "COVARDE". 🗞️ @ESPNBrasil 📸 Gustavo Martins/CEC #Flamengo #Gerson #Justiça

Sukob Marka Livaje s trenerom Gonzalom Garcijom i sportskim direktorom Robertom Grafom postao je glavna tema u hrvatskim medijima, uz brojne kritike na račun Garcije koji je prema mišljenju mnogih pogrešno reagirao prema Livaji, što bi njegovo eventualno odlazak iz Hajduka mogao značiti početak kraja kluba. #Hajduk #MarkoLivaja #GonzaloGarcia

EPISÓDIO 1 — Achei que eram só lâmpadas... 💡 Há investigações que começam com um milhão de euros. Esta começou... com uma lâmpada. Ao fim de meia hora já tinha um Excel e diagramas de grafo. Ao fim de dois dias já tinha dezenas de empresas. Ao fim de uma semana já tinha sociedades chamadas: 👉 I-Quatro 👉 I-Seis 👉 I-Vinte 👉 I-Vinte e Seis 👉 I-Trinta e Um 👉 I-Trinta e Oito... "Quem é que dá estes nomes às empresas? Um contabilista a jogar bingo?" Algumas empresas passaram a chamar-se... Smartlight. Depois apareceram contratos de 5 milhões... 10 milhões... 15 milhões... 38 milhões de euros. Depois começaram a repetir-se os mesmos nomes, alguns ligados à política. Depois apareceram municípios. E foi aí que percebi que isto já não era sobre lâmpadas. Ainda não sei o que tenho em mãos dada a complexidade dos dados, mas uma coisa é certa... já não consigo largar este fio. E vocês também vão querer saber onde ele vai dar. 🧶 Continua... #Investigação #Empresas #Mistério

Já tady nejvíc proruského poslance v EP blokuju, protože mě ta jeho kremelská grafomanie otravovala, ale je to poprvé, co někdo jako argumentační faul vůči mě použil abecedu. K tabuli jsem chodil první už od první třídy, ale ani tehdy mi to spolužáci nevyčítali 🤷‍♂️ https://t.co/JEeEAH33aa #politika #eu #kremel