Swin-Transformer
Swin-Transformer copied to clipboard
microsoft
This is an official implementation for "Swin Transformer: Hierarchical Vision Transformer using Shifted Windows".
[help]How to export swin model to ONNX? Problem: Node (Concat_246) Op (Concat) [ShapeInferenceError]
I exported my trained model into ONNX by the following code: ``` torch.onnx.export(model, input_tensor, onnx_name, verbose=True, opset_version=12, input_names=['images'], output_names=['output'], use_external_data_format=False) ``` But when running onnx model, I got the following...
https://github.com/microsoft/Swin-Transformer/blob/2622619f70760b60a42b996f5fcbe7c9d2e7ca57/models/swin_transformer.py#L583 ``` where self.avgpool=self.norm: self.num_features * self.patches_resolution[0] * self.patches_resolution[1] // ((2 ** (self.num_layers - 1)) ** 2) so flops += self.num_features * self.patches_resolution[0] * self.patches_resolution[1] // (2 ** self.num_layers) //...
Dear authors, Sorry for repeating [this issue](https://github.com/microsoft/Swin-Transformer/issues/55). It will be great if you could release an ImageNet-22K pretrained Swin-T model (if possible). I believe this will greatly help researchers in...
Hi there, It's so exciting to know the post-norm can stabilize the training process. We tried to reimpl the swint w/ post-norm, just like reported in Table6 81.6 top1-acc in...
https://github.com/microsoft/Swin-Transformer/blob/main/optimizer.py#:~:text=def%20set_weight_decay(model,no_decay%2C%20%27weight_decay%27%3A%200.%7D%5D, in this code snippet you set some layers weight decay to zero. Is there any reason why you do this? Does it has a high impact on training results?...
Hi, We've released the code of LAVT (CVPR2022), which is based on Swin Transformer for Referring Image Segmentation. Would it be possible for you to add this to the "Third-party...
Hi, the ``relative_position_bias`` seems to be a constant index of ``self.relative_position_bias_table``, why should it be computed in forward() every time? For the code, why the following code can't be computed...
I train the `swin_tiny_patch4_window7_224` with one million classes and 100 million images with softmax loss and adamw, the batch size is 600 and train for 400,000 iterations but the model...
Changes come from the issue I was having here: https://github.com/microsoft/Swin-Transformer/issues/181 Added a check to remove encoder prefixes as done in https://github.com/SwinTransformer/Swin-Transformer-Object-Detection Added a head check to reinit if the head...
Hi @zeliu98 Thanks for this great work. I am trying to reproduce the results reported in the paper for Swin Small architecture using the exact same hyper-parameters as published in...
