Ruoqian Guo comments

Results 6 comments of


                                            Ruoqian Guo

🐛 [Bug] Advanced Indexing/GatherND compilation causes error: `isObject() INTERNAL ASSERT FAILED at "libtorch/_virtual_includes/ATen/ATen/core/ivalue_inl.h":123, please report a bug to PyTorch. Expected Object but got None`

@narendasan The complete function looks like that https://github.com/pytorch/TensorRT/pull/921#discussion_r823279468. I think we can address this problem and i will try to support several indices input.

🐛 [Bug] Advanced Indexing/GatherND compilation causes error: `isObject() INTERNAL ASSERT FAILED at "libtorch/_virtual_includes/ATen/ATen/core/ivalue_inl.h":123, please report a bug to PyTorch. Expected Object but got None`

In container `pytorch/pytorch:1.12.0-cuda11.3-cudnn8-devel`, I just tested this script on master `d1768aa3d2c7d7d91d9f061e3e5dc5f976124dfe`. And the results look like correct. Whether this results are related to container? ``` PyTorch: torch.Size([2, 4]) WARNING: [Torch-TensorRT]...

训练0712的时候显存变大

因为训练voc 07 时，为了节省显存，会把图片crop（保留中心区域）进行训练。而voc 0712数据中，有一张图片的长宽比很大(2.+)，而图片中的物体只出现在图片两侧，如果把图片crop保留中心位置，这张图片就没有物体了，训练会报错，所以在训练voc 0712数据时，不使用crop操作，没有节省现存的行为。

sh make.sh error

What is your version of pytorch?

locally shared weights in conv3 and conv4

@timemao , how did they shared the weights?

support python2 and get more reasonable mAP

hi, i changed resnet layer4 stride from 1 to 2, and got more reasonable mAP on Pascal VOC 2007(75.7) and 07 + 12(80.5)