Grigory Malivenko

@Gasp34 I updated tracing module. You can try to convert your model with new version of the converter

Hello @NoushNabi. Yes, I can confirm that nn.PixelShuffle doesn't work properly. To convert your model, please, replace nn.PixelShuffle with this equivalent module: ``` class pixel_shuffle(nn.Module): def __init__(self, scale_factor): super(pixel_shuffle, self).__init__()...

Hello @NoushNabi. Can you show me a full trace of the error?

hello, @bendangnuksung! Yep, Transpose / Squeeze layers a bit different in Keras and in PyTorch, so, there are some workarounds for that. As far as i can see, your model...

Hello @bendangnuksung. I see the problem now. The recurrent layers don't converting properly in this version of converter. I will add support of them as soon as possible.

hello! Thank you for research! Yes, constant is a bit confusing layer. You can make a pull request if you want to contribute your changes. Or i can fix that...

ONNX still have a very bad support of recurrent layers.

Hello, @snakers4. I've tested the Inception v4 today, It seems more or less accurate but not all PyTorch parameters have Keras/TF equivalent. It seems like AvgPooling parameter `count_include_pad` may cause...

Hello, @aveysov. How are you using a `IlgSimplifiedNormalized` class? As I can see there is `nn.Dropout`, ``` self.classifier = nn.Sequential( nn.Linear(num_skip1+num_skip2+num_skip3+num_skip4, 1024), nn.ReLU(True), nn.Dropout(p=0.5), nn.Linear(1024, num_classes), ) ``` May be...