Issues
- 2
- 3
BlazeXlaOp为什么用异步实现?是否有性能问题?
#13 opened by striveth - 0
如何部署其他的模型?
#12 opened by hdhidcbe1 - 2
gpu训练报错 CPU->GPU Memcpy failed Aborted
#11 opened by Dengyingjie - 1
有测试效果吗,和TDM相比如何?
#10 opened by Dengyingjie - 1
bucket_size比较大的时候, state_ops.assign性能很差
#9 opened by ambjlon - 1
- 7
好像只支持Item2Item的检索,支持User2Item的检索吗?
#5 opened by treper - 1
如果候选集有500万items, 那么需要模型打分10万次?
#6 opened by ambjlon - 1
想了解下huge_const算子的作用是啥
#4 opened by houjincheng1992 - 1
为什么build_opt_graph.py发布模型使用了多个自定义算子?
#3 opened by ambjlon - 1
[Feat]: Any plan to refactor in C++?
#1 opened by yinze00 - 1
nann/util.py中的fast_argtopk实现有问题
#2 opened by YuhwaChoong