#mtp — Public Fediverse posts
Live and recent posts from across the Fediverse tagged #mtp, aggregated by home.social.
-
New week, more slides: Run LLMs Locally
Now including wllama to run GGUF models inside your browser!
wllama uses llama.cpp, WebAssembly and WebGPU, bringing a completely new experience of LLMs into the web.
It has no 4 GB limitation and is faster than Transformers.js.I also added translations using the HY-MT model from Tencent.
https://codeberg.org/thbley/talks/raw/branch/main/Run_LLMs_Locally_2026_ThomasBley.pdf
#ai #llm #llamacpp #wllama #stablediffusion #qwen3 #glm #localai #gemma4 #webgpu #opencode #mtp #webassembly
-
New week, more slides: Run LLMs Locally
Now including wllama to run GGUF models inside your browser!
wllama uses llama.cpp, WebAssembly and WebGPU, bringing a completely new experience of LLMs into the web.
It has no 4 GB limitation and is faster than Transformers.js.I also added translations using the HY-MT model from Tencent.
https://codeberg.org/thbley/talks/raw/branch/main/Run_LLMs_Locally_2026_ThomasBley.pdf
#ai #llm #llamacpp #wllama #stablediffusion #qwen3 #glm #localai #gemma4 #webgpu #opencode #mtp #webassembly
-
New week, more slides: Run LLMs Locally
Now including wllama to run GGUF models inside your browser!
wllama uses llama.cpp, WebAssembly and WebGPU, bringing a completely new experience of LLMs into the web.
It has no 4 GB limitation and is faster than Transformers.js.I also added translations using the HY-MT model from Tencent.
https://codeberg.org/thbley/talks/raw/branch/main/Run_LLMs_Locally_2026_ThomasBley.pdf
#ai #llm #llamacpp #wllama #stablediffusion #qwen3 #glm #localai #gemma4 #webgpu #opencode #mtp #webassembly
-
New week, more slides: Run LLMs Locally
Now including wllama to run GGUF models inside your browser!
wllama uses llama.cpp, WebAssembly and WebGPU, bringing a completely new experience of LLMs into the web.
It has no 4 GB limitation and is faster than Transformers.js.I also added translations using the HY-MT model from Tencent.
https://codeberg.org/thbley/talks/raw/branch/main/Run_LLMs_Locally_2026_ThomasBley.pdf
#ai #llm #llamacpp #wllama #stablediffusion #qwen3 #glm #localai #gemma4 #webgpu #opencode #mtp #webassembly
-
New week, more slides: Run LLMs Locally
Now including wllama to run GGUF models inside your browser!
wllama uses llama.cpp, WebAssembly and WebGPU, bringing a completely new experience of LLMs into the web.
It has no 4 GB limitation and is faster than Transformers.js.I also added translations using the HY-MT model from Tencent.
https://codeberg.org/thbley/talks/raw/branch/main/Run_LLMs_Locally_2026_ThomasBley.pdf
#ai #llm #llamacpp #wllama #stablediffusion #qwen3 #glm #localai #gemma4 #webgpu #opencode #mtp #webassembly
-
RT @TeksEdge: 🚀 Neue MTP-Unterstützung für Strix Halo veröffentlicht!
mehr auf Arint.info
-
RT @TeksEdge: 🚀 Neue MTP-Unterstützung für Strix Halo veröffentlicht!
mehr auf Arint.info
-
RT @TeksEdge: 🚀 Neue MTP-Unterstützung für Strix Halo veröffentlicht!
mehr auf Arint.info
-
RT @TeksEdge: 🚀 Neue MTP-Unterstützung für Strix Halo veröffentlicht!
mehr auf Arint.info
-
RT @TeksEdge: 🚀 Neue MTP-Unterstützung für Strix Halo veröffentlicht!
mehr auf Arint.info
-
RT @dealignai: TRANSLASION: Qwen3.6-27b und 35b MXFP4 MXFP8 CRACK ist jetzt mit MTP verfügbar. Genieße unzensierte Geschwindigkeit!
mehr auf Arint.info
-
RT @dealignai: TRANSLASION: Qwen3.6-27b und 35b MXFP4 MXFP8 CRACK ist jetzt mit MTP verfügbar. Genieße unzensierte Geschwindigkeit!
mehr auf Arint.info
-
RT @dealignai: TRANSLASION: Qwen3.6-27b und 35b MXFP4 MXFP8 CRACK ist jetzt mit MTP verfügbar. Genieße unzensierte Geschwindigkeit!
mehr auf Arint.info
-
RT @dealignai: TRANSLASION: Qwen3.6-27b und 35b MXFP4 MXFP8 CRACK ist jetzt mit MTP verfügbar. Genieße unzensierte Geschwindigkeit!
mehr auf Arint.info
-
RT @dealignai: TRANSLASION: Qwen3.6-27b und 35b MXFP4 MXFP8 CRACK ist jetzt mit MTP verfügbar. Genieße unzensierte Geschwindigkeit!
mehr auf Arint.info
-
New week, new slides: Run LLMs Locally
Now including multi-token prediction using Qwen3.6 35B-A3B with Nextn quantization. Also speech recognition using Qwen-3-ASR is now working directly with Llama.cpp and included in the slides.
https://codeberg.org/thbley/talks/raw/branch/main/Run_LLMs_Locally_2026_ThomasBley.pdf
#ai #llm #llamacpp #stablediffusion #qwen3 #glm #localai #gemma4 #webgpu #opencode #mtp
-
New week, new slides: Run LLMs Locally
Now including multi-token prediction using Qwen3.6 35B-A3B with Nextn quantization. Also speech recognition using Qwen-3-ASR is now working directly with Llama.cpp and included in the slides.
https://codeberg.org/thbley/talks/raw/branch/main/Run_LLMs_Locally_2026_ThomasBley.pdf
#ai #llm #llamacpp #stablediffusion #qwen3 #glm #localai #gemma4 #webgpu #opencode #mtp
-
New week, new slides: Run LLMs Locally
Now including multi-token prediction using Qwen3.6 35B-A3B with Nextn quantization. Also speech recognition using Qwen-3-ASR is now working directly with Llama.cpp and included in the slides.
https://codeberg.org/thbley/talks/raw/branch/main/Run_LLMs_Locally_2026_ThomasBley.pdf
#ai #llm #llamacpp #stablediffusion #qwen3 #glm #localai #gemma4 #webgpu #opencode #mtp
-
New week, new slides: Run LLMs Locally
Now including multi-token prediction using Qwen3.6 35B-A3B with Nextn quantization. Also speech recognition using Qwen-3-ASR is now working directly with Llama.cpp and included in the slides.
https://codeberg.org/thbley/talks/raw/branch/main/Run_LLMs_Locally_2026_ThomasBley.pdf
#ai #llm #llamacpp #stablediffusion #qwen3 #glm #localai #gemma4 #webgpu #opencode #mtp
-
New week, new slides: Run LLMs Locally
Now including multi-token prediction using Qwen3.6 35B-A3B with Nextn quantization. Also speech recognition using Qwen-3-ASR is now working directly with Llama.cpp and included in the slides.
https://codeberg.org/thbley/talks/raw/branch/main/Run_LLMs_Locally_2026_ThomasBley.pdf
#ai #llm #llamacpp #stablediffusion #qwen3 #glm #localai #gemma4 #webgpu #opencode #mtp
-
Qwen3.6 MTP весит на 0.3 Гб больше, а даёт ускорение в ~2 раза. С 60 t/s до 130 t/s для Qwen3.6 27B без искажений
В llama.cpp добавили поддержку MTP Qwen3.6. Дополнительные слои Multi-Token Prediction позволяют сгенерировать сразу несколько токенов за 1 проход, что ускоряет генерацию в 1.5-2 раза. Качество при этом остается lossless. Для моделей, которые не имеют встроенного MTP, есть альтернативы в лице EAGLE-3 и DFlash.
-
Qwen3.6 MTP весит на 0.3 Гб больше, а даёт ускорение в ~2 раза. С 60 t/s до 130 t/s для Qwen3.6 27B без искажений
В llama.cpp добавили поддержку MTP Qwen3.6. Дополнительные слои Multi-Token Prediction позволяют сгенерировать сразу несколько токенов за 1 проход, что ускоряет генерацию в 1.5-2 раза. Качество при этом остается lossless. Для моделей, которые не имеют встроенного MTP, есть альтернативы в лице EAGLE-3 и DFlash.
-
Qwen3.6 MTP весит на 0.3 Гб больше, а даёт ускорение в ~2 раза. С 60 t/s до 130 t/s для Qwen3.6 27B без искажений
В llama.cpp добавили поддержку MTP Qwen3.6. Дополнительные слои Multi-Token Prediction позволяют сгенерировать сразу несколько токенов за 1 проход, что ускоряет генерацию в 1.5-2 раза. Качество при этом остается lossless. Для моделей, которые не имеют встроенного MTP, есть альтернативы в лице EAGLE-3 и DFlash.
-
Qwen3.6 MTP весит на 0.3 Гб больше, а даёт ускорение в ~2 раза. С 60 t/s до 130 t/s для Qwen3.6 27B без искажений
В llama.cpp добавили поддержку MTP Qwen3.6. Дополнительные слои Multi-Token Prediction позволяют сгенерировать сразу несколько токенов за 1 проход, что ускоряет генерацию в 1.5-2 раза. Качество при этом остается lossless. Для моделей, которые не имеют встроенного MTP, есть альтернативы в лице EAGLE-3 и DFlash.
-
who came up with the media transfer protocol, and where can I find them?
there are a few choice words I'd like to say to them.
-
#Republicans were all over the Sunday Shows this morning calling #ObamaCare a "failure" to "lower costs". #MtP #FnS
They never mention that the two mechanisms designed to control costs… A #PublicOpton and the #Mandate… they fought to strip out.
What a shock it failed to control costs.
#SenBarasshole was asked 2x by Welker, "What's your plan?" Never got an answer. #ACA
-
It appears I won't be Live Blogging #MtP today.
On a #4thOfJuly weekend that saw a lawless president sign the most massive sweeping destructive bill in U.S. history that will throw MILLIONS off their healthcare, #MeetTheRepublicans decided to dedicate their entire show today to "personal interest" stories (starting with an actress enduring her OWN health care nightmare.) 🤦♂️ #ToneDeaf #TooStupidForPolitics
-
RE: https://mastodon.social/@MugsysRapSheet/115639275632510417
DESPITE being told during her introduction on Fox that the #DCNG shooter was vetted *during #RipOneWinkle's first term" (and then again just this past April), that #Didn't stop AG #BlondiBondi from blaming "the #Biden Administration a dozen times" during her interview.
On #MtP, #GardenNoem clearly received the same marching orders, blaming Biden an additional dozen+ times. #NotMyFault #DontBlameMe
-
LORD I DESPISE #MarKwayne. 😒
He was on #MtP defending "cutting 35M people off #MediCare".
#Welker (to her credit) asked him, "Are you saying ALL 35 Million are engaged in waste, fraud & abuse?"
MarKwayne: "No. But you can't tell me there's no room for cutting WF&B from MediCare."
SO… BY HIS OWN ADMISSION… he is cutting millions of people who AREN'T engaged in WF&B from Medicare. 🤔
#DisasterPresidency #TooStupidForOffice -
#TreasurySecretary #Bessent said on #MtP:
"The shutdown of the #SupplyChain during the #pandemic was a [warning] of what can happen when we don't produce everything we need."
It was destruction of the DOMESTIC Supply Chain that crashed the economy. IMPORTS SAVED US. Moving manufacturing to the U.S. wouldn't save us from another Supply Chain disruption.
IN FACT, putting #tariffs on imported SUPPLIES will do THE EXACT SAME THING! Get ready for ANOTHER Supply Chain crash. 🤦♂️
-
#DictatorDon's Ambassador to the U.N. Mike Wallz on #MtP just claimed "#Iran started this war in 1979."
Again... will SOMEBODY explain to these #NeoCon jackholes the meaning of #ImminentThreat?
-
Рынок промышленности в ближайшие 10-20 лет
Предлагаю вашему вниманию обзор того, как будет трансформироваться рынок индастриал в ближайшие 10-20 лет. Какие изменения нас ожидают, какие тренды уже формируются и каким будет будущее цифровой промышленности. В рамках статьи я проведу вас через ряд размышлений и выводов о том, как рынок меняется сегодня, какие факторы на это влияют и что нас ждет в будущем. В своих выводах я буду опираться на личный опыт работы с интеграторами в сфере АСУ в технологической промышленности (индустрия, где оперируем температурами, давлениями, массой и т.д.) и энергетике (индустрия, где оперируем напряжением, током, мощностями и т.д.), взаимодействия с продуктовыми компаниями, а также сотрудничества с чип-вендорами. Надеюсь, что в этом материале представители разных инженерных и экспертных направлений смогут найти для себя новые и полезные идеи для развития бизнеса в ближайшие годы. Кто-то задумается о том, как эти изменения повлияют на бизнес-стратегию, а инженеры какие технологии стоит применять и с какими целями эти технологии использовать. Здесь будут затронуты как темы классического индастриала, так и направления IIoT, Edge AI, робототехники и других передовых технологий. Мой опыт сложился в области контрактной разработки электроники и основан на многолетнем взаимодействии с заказчиками, большом количестве встреч и реализованных проектов в этой сфере. Поэтому материал будет интересен промышленным интеграторам, производственным предприятиям, продуктовым компаниям, специалистам в области embedded-систем, а также, возможно, разработчикам чипов для индастриал.
https://habr.com/ru/articles/994348/
#цифровизация_промышленности #робототехника #искусственный_интеллект #промышленная_автоматизация #Сезон_Heavy_Digital #mtp #NOA #iiot #SRCI #цифровой_двойник
-
Well well well
if this isn't an #mtp app for #macos that natively browses my #Supernote eink tablet for file transfer!
😎
-
RT @mr_r0b0t: Wusstest du, dass Qwen3.6 mit nativer MTP ausgeliefert wurde? Ja, dieselbe MTP, für die Google gestern die Unterstützung von Gemma4 freigegeben hat! Multi Token Prediction (MTP) = spekulatives Decoding. Hier ist ein Qwen3.6-Modell, quantisiert auf Q4KM, das MTP über ikllama.cpp unterstützt.
mehr auf Arint.info
-
anyone who was a teenager or 20-something during the 90's....
Save the date - June 4th 2026
Your new anthem will arrive
#music #Summerin99 #mtp #rockAnthem #1999 #90s
-
#MeetTheRepublicans waited until THE FINAL FIVE MINUTES to talk about THE LARGEST POLITICAL PROTEST IN AMERICAN HISTORY. 🤨 #MtP
-
She knows what she's doing. "Tens of thousands" in NYC isn't the same as suggesting "tens of thousands nationwide." 😐
I had to shut off #MtP after Lankford called #Democrats "totally unreasonable" for "opposing allowing #ICE agents to police Polling places. No one thinks illegal aliens should be allowed to vote."
I screamed at my TV and shut it off. LORD I DESPISE THESE PEOPLE! 🤬
-
RE: https://mas.to/@tezoatlipoca/116263075699722140
Well, I had to give up on the #rsync approach - rsync over #MTP seemed to have issues with copying some file metadata that rsync would need to handle incremental updates. Subsequent rsyncs would re-copy _everything_. Boo.
Likewise using #android usb-debug mode and `adb push` (android debug tool on linux) also only re-copies everyting.
I could use #termux on the phone to provide an actual rsync or ssh host on the phone (which would also let me use wifi). Or..1/
-
🚀 Nvidia's new Nemotron 3 Super combines a 3‑arch design with Multi‑Token Prediction (MTP) and speculative decoding, promising to outpace GPT‑OSS and Qwen on the Blackwell GPU. The open‑source community gets a powerful, efficient model to experiment with. Dive in to see how this leap could reshape AI research! #Nvidia #Nemotron3Super #MTP #OpenSourceAI
🔗 https://aidailypost.com/news/nvidias-nemotron-3-super-merges-3arch-design-mtp-outpace-gptoss-qwen
-
Had to turn off #MtP.
First guest is Sen. #LadyLindsey Graham. Best known for being the #1 cheerleader for invading #Iraq based on false pretenses, who continued to praise the invasion even after no #WMDs were found and the rational was revealed to be a lie.
And now he's right back at it with #Iran.🤦♂️
-
After Speaker #CosplayColbert denied he was "doing Donald T****'s bidding" on #MtP, Welker put up a quote from #MTG claiming otherwise (I'm going to need about 100 miles of Mental Floss once this nightmare presidency is over.) 😒
-
"We're not going to allow #Venezuela's #oil to be under control of America's adversaries." - SoS #Rubio on #MtP... in case you were wondering what this is all about.
MAYBE if we weren't so DEPENDENT on oil...
-
@wdlindsy
Yeah, sure Todd. You're just redacting *names* to protect the "innocent victims". 😡
#Epstein #EpsteinFiles #Redactions #Coverup #MtP #MeetThePress #MeetTheRepublicans -
@wdlindsy
One of the things I can't stand about #MtP #MeetThePress (even before #Welker) is that they are ALWAYS woefully unprepared when it comes to follow-up questions.There wasn't a SINGLE response from Blanche yesterday I couldn't have guessed in advance, and yet Welker provided little-to-no push-back.
He "pointed out" that in the "4 years of the #Biden Admin, they didn't release a single file"… playing up the "hoax" angle. But the fact is it was A SEALED OPEN CASE during that time.
-
On this #PearlHarbor Day, I would have liked #Welker to ask #TraitorTom if firing upon downed Japanese pilots "still in the fight" would have been legal? (Note: It's a #WarCrime.) #MtP #MeetThePress
-
#SundayShowdown
#TraitorTom Cotton on #MtP noted one of the two survivors of the missile strike on a #Venezuelan #DrugBoat "took off his shirt... maybe to get a sun tan." 😵 😡Even #Welker was bothered by this absurd suggestion, and challenged him on the assumption.
Traitor Tom walked back the suggestion *slightly*, suggesting that "maybe they were trying to signal another [drug] boat in the area."
THERE. WERE. NO. OTHER. BOATS. IN. THE. AREA. (and even if there were, it's not a defense.)
-
#GardenNoem on #MtP blamed "#TheBidenAdministration" more than A DOZEN TIMES for failing to "properly vet #immigrants coming into this country" for "all these problems" (ala Trent Lott) with violence from migrants (who are actually only responsible for a *fraction* of crimes.)
During #Obama's first year, #Republicans called him "O'Blamer" for repeatedly criticizing the Bush Administration "for everything".
PS: Repeatedly asked, #Noem showed ZERO concern for immigrants fleeing violence.
-
RE: https://mastodon.social/@MugsysRapSheet/115639066550819837
#Noem is on #MtP making the same accusation as #Bondi against the #Biden Administration, FALSELY claiming the DC shooter "was unvetted by the Biden Administration before entering the country."
But he was first vetted under the #DonnieDumbass Admin in 2020, and again given #asylum last April.
-
Over on #FnS, a member of their roundtable accused #Biden of telling consumers "don't believe your lyin' eyes" on the economy.
Meanwhile, #MtP has on Treasury Sec Scott Bessent denying that prices are up and the economy is going to get great really really soon. 🤦♂️ #GOPHypocrisy #inflation #economy
-
#Newsom had a good interview on #MeetTheRepublicans.
First half, he did an excellent job of defending #Prop50, noting how the #GOP "knows they are losing or else they wouldn't be working so hard to rig the game".
The 2nd half of the interview (after the commercial break) was mostly Welker trying to get Newsom to admit he made "a mistake" by continuing to stand by #Biden and suggesting there was a rift between him and #Harris based on a comment in her book.
Full interview on the #MtP website.