TFC-pretraining
TFC-pretraining copied to clipboard
Where is 3-layer 1 D Resnet?
![image](https://user-images.githubusercontent.com/59809602/230899112-ae270b5c-619d-4f5f-a2e7-ce02f2b5662e.png)
![image](https://user-images.githubusercontent.com/59809602/230899158-ddf2c1ed-6f11-4759-9eeb-2a16e4dee72e.png)
In the paper, you mentioned using ResNet, but there is a Transformer in the code.
The built-in implementation of Pytorch is still used, where the input is supposed to be (seq_len, N, D), but you just (N, 1, seq_len) tensor.
Is this the reason why Transformer is not as good as ResNet?