RT @vllm_project: TRANSLASATION: vLLM v0.24.0 ist da! 571 Commits von 256 Mitwirkenden (77 neue). 🎉 Highlights: MiniMax-M3-Unterstützung (FP8/MXFP4 + breite AMD-Optimierung), DeepSeek-V4 reift weiter (FlashInfer Sparse-Index-Cache, Prefill-Chunk-Planning, jetzt auf SM120), Model Runner V2 verarbeitet nun standardmäßig quantisierte Modelle, eine neue einheitliche Streaming-Parser-Engine für Tool-Aufrufe und Reasoning, DiffusionGemma, DeepEP v2 für breite Expert-Parallelität und ein ausgereifter Rust-Frontend. Thread 👇 Video mehr auf Arint.info #AI #DeepSeek #MachineLearning #MiniMax #OpenSource #vLLM #arint_info https://x.com/vllm_project/status/2072159562992619991#m
Related
A quick AI storyThis is just a quick story about a time when an AI confidently gave me the wrong answers. Amazingly, tho...
A quick AI storyThis is just a quick story about a time when an AI confidently gave me the wrong answers. Amazingly, though, at one point it corrected itself.https://rodstephensboo...
ExpressVPN's password manager just got a big upgrade with secure sharing, passkey support, and moreAvailable today, Expr...
ExpressVPN's password manager just got a big upgrade with secure sharing, passkey support, and moreAvailable today, ExpressKeys now delivers more secure sharing, passkey support, a...
Bananas that #Meta admitted that they bought billions of dollars more of #AI compute than it turns out they needed, and ...
Bananas that #Meta admitted that they bought billions of dollars more of #AI compute than it turns out they needed, and their stock went up 9% at the news.https://youtu.be/okIHWPwR...