chinakook

it could really have not been implemented as the assertion

I think the mini-batch size cannot be set in clstm until now, so the GPU version may not be faster than the CPU one. If you want to accelerate the...

A tensorflow implementation of popular library CLSTM and OCROPY line reader https://github.com/ferjad/tlstm Does anyone want to modify the repository to the cudnnLSTM or cudnnPersistentRNN implementation, by which the training could...

I also have a solution to train faster rcnn with no-valid-bbox img. I found that, the data transformers may change the label(as box)'s value so you need get the -1...

@Aktcob In the faster rcnn, the sampler would treat [-1,-1,-1,-1,-1] as ignore box, bug treat [200,200,200,200,-1] as negative box which will be trained ( with only softmax entropy, and without...

But I found no code to verify the area. The line below would assign the [200,200,200,200,-1] to negative, not ignore. https://github.com/dmlc/gluon-cv/blob/0f0b226e71519493e0744017f2af3b71ad77eb1c/gluoncv/model_zoo/rcnn/faster_rcnn/rcnn_target.py#L68

> @chinakook ye. u are right. should do some changes in the target generator of faster rcnn. But right now, training yolov3 is ok. Did solve the nan problem?

@sufeidechabei You can use the symbol.Pooling op to emulate Same pad as it has param `pooling_convention={'full', 'same', 'valid'}`

@sufeidechabei You can wrap a hybridblock to it or you can use hybridlambda.

You can install mxnet by pip install ...