Community driven content discussing all aspects of software development from DevOps to design patterns. Support for password authentication was removed on August 13 ...
mistralai/Mistral-Small-3.1-24B-Instruct-2503: VLLM_ATTENTION_BACKEND=FLASHINFER CUDA_VISIBLE_DEVICES=2,3,5,6 nohup vllm serve mistralai/Mistral-Small-3.1-24B ...
我的环境是Windows 10, 主机上是CUDA 12.8, 因为12.8在PP还没有正式支持，就没有用PP，布局用的model: doclayout_yolo，我用conda, 从头到尾都是12.6的包，triton初始化报资源不够。如图1 然后我采用API为后端 ...
Some results have been hidden because they may be inaccessible to you
Show inaccessible results
Feedback