RT @bnjmn_marie: Ich arbeere daran, GGUF-Unterstützung zu vLLM durch die Humming-Kernel hinzuzufügen. In nur 5 Stunden mit GPT-6 Astra habe ich es geschafft, vollständige Unterstützung für Quantisierungsformate von IQ/Q2 bis Q8 zu implementieren, die Geschwindigkeit von llama.cpp in meinen Benchmarks zu erreichen und es bei Batch-Größe 8 um etwa 30% zu übertreffen. Noch mehr auf Arint.info #Benchmarks #GGUF #Humming #LLM #Quantisierung #vLLM #arint_info https://x.com/bnjmn_marie/status/2113803691958599684
Related
📢 New #updates · 13 Nov #4.3· Bramble 0.18.0-desktop — Added duplicate entry, improved passkey handling, added language ...
📢 New #updates · 13 Nov #4.3· Bramble 0.18.0-desktop — Added duplicate entry, improved passkey handling, added language support, enhanced import…All updates live on https://selfhos...
In the end, the biggest risk Jensen Huang faced for Nvidia was himselfJensen Huang became the rock star of the AI revolu...
In the end, the biggest risk Jensen Huang faced for Nvidia was himselfJensen Huang became the rock star of the AI revolution. Then revelations about his personal life upended his l...
So many accusations of vibe-coding, so little time... Humble Bundle goes through a re-design that's being accused of mes...
So many accusations of vibe-coding, so little time... Humble Bundle goes through a re-design that's being accused of messy vibe-coding https://www.gamingonlinux.com/2026/11/humble-...