Thomas Qvist
@turbothomaten.bsky.social
Passion for software, østers og Paris
Reposted by Thomas Qvist
Reposted by Thomas Qvist
Et skriv om året der gik i tech med krypteret telefon-ballade, X-nedtur, Elon-optur, demaskering af OpenAI og masser af performativ digital harme. Løse tanker, kolde fakta og måske forslag til en lidt bedre fremtid
radar.dk/artikel/flyv...
radar.dk/artikel/flyv...
Flyvske tanker fra året i tech: Intet socialt medie er uundværligt, og hvor er det danske Moonshot-projekt ? | Radar
Vi blæser gennem året der gik i tech med krypteret telefon-ballade, X-nedtur, Elon-optur, demaskering af OpenAI og masser af performativ tech-harme.
radar.dk
December 27, 2024 at 7:44 AM
Et skriv om året der gik i tech med krypteret telefon-ballade, X-nedtur, Elon-optur, demaskering af OpenAI og masser af performativ digital harme. Løse tanker, kolde fakta og måske forslag til en lidt bedre fremtid
radar.dk/artikel/flyv...
radar.dk/artikel/flyv...
Reposted by Thomas Qvist
Bamba-9B "hybrid Mamba2" model delivers significant improvements in throughput and latency, enhancing real-time application performance.
Benchmarking with vLLM against Llama 3.1 8B for long contexts shows:
🔹 2.5x throughput improvement
🔹 2x lower latency
Repo: github.com/foundation-m...
Benchmarking with vLLM against Llama 3.1 8B for long contexts shows:
🔹 2.5x throughput improvement
🔹 2x lower latency
Repo: github.com/foundation-m...
GitHub - foundation-model-stack/bamba: Train, tune, and infer Bamba model
Train, tune, and infer Bamba model. Contribute to foundation-model-stack/bamba development by creating an account on GitHub.
github.com
December 18, 2024 at 10:00 PM
Bamba-9B "hybrid Mamba2" model delivers significant improvements in throughput and latency, enhancing real-time application performance.
Benchmarking with vLLM against Llama 3.1 8B for long contexts shows:
🔹 2.5x throughput improvement
🔹 2x lower latency
Repo: github.com/foundation-m...
Benchmarking with vLLM against Llama 3.1 8B for long contexts shows:
🔹 2.5x throughput improvement
🔹 2x lower latency
Repo: github.com/foundation-m...
Reposted by Thomas Qvist
Reposted by Thomas Qvist
TRL is a cornerstone of LLM post training and imo it's the default to learn.
There are great alternatives like Unsloth, Axolotl, and AutoTrain. But if you want a daily drive that does experimentation to production, it's TRL.
🧵 these community notebooks guide you through TRL's core:
There are great alternatives like Unsloth, Axolotl, and AutoTrain. But if you want a daily drive that does experimentation to production, it's TRL.
🧵 these community notebooks guide you through TRL's core:
November 25, 2024 at 10:16 AM
TRL is a cornerstone of LLM post training and imo it's the default to learn.
There are great alternatives like Unsloth, Axolotl, and AutoTrain. But if you want a daily drive that does experimentation to production, it's TRL.
🧵 these community notebooks guide you through TRL's core:
There are great alternatives like Unsloth, Axolotl, and AutoTrain. But if you want a daily drive that does experimentation to production, it's TRL.
🧵 these community notebooks guide you through TRL's core:
Reposted by Thomas Qvist
🔥 First open source reasoning model from @Alibaba_Qwen just dropped! 😮
Qwen QwQ 32B is in preview, does reasoning like o1 and Deepseek R1 but y'know, on your local machine!
🔗 Demo: https://thursdai.news/qwq-32b-prev
🌐 Model: https://thursdai.news/O9d9hi0
📃 Blog: https://thursdai.news/qwq-32b-pre
Qwen QwQ 32B is in preview, does reasoning like o1 and Deepseek R1 but y'know, on your local machine!
🔗 Demo: https://thursdai.news/qwq-32b-prev
🌐 Model: https://thursdai.news/O9d9hi0
📃 Blog: https://thursdai.news/qwq-32b-pre
November 27, 2024 at 7:40 PM
🔥 First open source reasoning model from @Alibaba_Qwen just dropped! 😮
Qwen QwQ 32B is in preview, does reasoning like o1 and Deepseek R1 but y'know, on your local machine!
🔗 Demo: https://thursdai.news/qwq-32b-prev
🌐 Model: https://thursdai.news/O9d9hi0
📃 Blog: https://thursdai.news/qwq-32b-pre
Qwen QwQ 32B is in preview, does reasoning like o1 and Deepseek R1 but y'know, on your local machine!
🔗 Demo: https://thursdai.news/qwq-32b-prev
🌐 Model: https://thursdai.news/O9d9hi0
📃 Blog: https://thursdai.news/qwq-32b-pre
Reposted by Thomas Qvist
Show me a group of employees who lack good ideas and I'll show you a boss who knows all the answers.
#leadership
#leadership
November 25, 2024 at 6:06 PM
Show me a group of employees who lack good ideas and I'll show you a boss who knows all the answers.
#leadership
#leadership
Reposted by Thomas Qvist
Så er der møde i sommerdæksklubben ❄️😆
November 23, 2024 at 7:18 PM
Så er der møde i sommerdæksklubben ❄️😆
Reposted by Thomas Qvist
Tiny robot ‘kidnaps’ 12 big bots from China showroom https://interestingengineering.com/innovation/ai-robot-kidnaps-12-robots-in-shanghai #AI #robots #China (LOL 😀)
November 23, 2024 at 5:14 AM
Tiny robot ‘kidnaps’ 12 big bots from China showroom https://interestingengineering.com/innovation/ai-robot-kidnaps-12-robots-in-shanghai #AI #robots #China (LOL 😀)
Reposted by Thomas Qvist
I am so excited about the migration from X that I made a whole YouTube video about what it takes to run a social media company in a responsible way.
Don't worry, I will be back to covering the normal topics and the war in Ukraine next time.
Don't worry, I will be back to covering the normal topics and the war in Ukraine next time.
Twitter and responsible social media
YouTube video by Anders Puck Nielsen
www.youtube.com
November 16, 2024 at 8:57 PM
I am so excited about the migration from X that I made a whole YouTube video about what it takes to run a social media company in a responsible way.
Don't worry, I will be back to covering the normal topics and the war in Ukraine next time.
Don't worry, I will be back to covering the normal topics and the war in Ukraine next time.