MonolithFoundation

@youkaichao @bnellnm Hello, would consider fix this for torch2.3 as soon as possible, am current install both with pip from pypi and build source all same error. Note I have...

@bnellnm you just mentioned torch2.3.1 have c10:optional difference from older version, does this is ok? Does vllm master branch can build with torch2.3.1? I am pretty sure I didn't have...

@bnellnm Hi, does this sombol defined? undefined symbol: _ZN5torch7LibraryC1ENS0_4KindESsSt8optionalIN3c1011DispatchKeyEEPKcj'

No, but I plan to release a brand new small models, way better on multilingual support compare with moondream, estimate open tomorrow.

这个是不是相当于是spatil merge写死了，并且和后面的PatchMerge也是耦合死的？ 我原来觉得对于原始size，直接用for去算每张图的feature，再组合反而更简单，反正用batch和用for并没有延迟太多

How do u think about the coordinate processing way?

@HumanZhong Hi, when training, does the coordinates processed on the fly? I mean, in a batch, each images couldn't be exactly same size and ratio, since you keeping the ratio,...

哎

@yuyun2000 你有了没

@csukuangfj kuangge请考虑支持