Dun Liang comments

Results 31 comments of


                                            Dun Liang

pytorch version and manifold part

Jittor has a similar interface with PyTorch, and it is much faster than PyTorch, Jittor has a team answer users' questions on QQ group , have a try~

Segfault

Could you please provide the full log for debugging this issue? Jittor has a similar interface with PyTorch, and it is much faster than PyTorch, Jittor has a team answer...

support --disk-cache-size flag

It is worked for me. hope this PR could help!

无法反馈，cannot bp

Thanks for your feedback, several reasons will cause no grad: 1. model.31.bias doesn't use during forward, so gradient cannot backpropagate to it too. 2. model.31.bias is stop grad 3. model...

如何在单卡训练的时候设置要使用的GPU？

谢谢反馈，我们会尽快增加此特性😁

Feeature request: Tag the source to synchronize with pypi release

Thanks for your advice, now tag is added in git `^_^`, see https://github.com/Jittor/jittor/tags

jittor mpi 同步和 torch 不太一致

您好，您这边的报错信息是怎样的呢

希望能添加广播索引机制

您好，jittor已经在最新版本1.3.3.8中添加了以上支持，相关commit：https://github.com/Jittor/jittor/commit/0b13930ed3b6c6d7ccc0e80adf610e4c28866a7f

where()函数与torch行为不一致

谢谢您的反馈，最新版本jittor1.3.2.3已经与torch保持一致，目前jt.where同时支持一个参数和三个参数，其中一个参数的版本和numpy保持一致，三个参数的版本和torch保持一致

FusedOp cannot contain both cpu and cuda ops

您好，最新版本的jittor 1.3.1.53 已经彻底解决了该问题 `^_^`