Questions about tcn checkpoint parameter count and conversion to PyTorch

Hi, I have some questions regarding the TCN model checkpoint: I converted the MLP part of TCN to PyTorch, but when saving the checkpoint from the original TCN version, I found the total parameter count is 18,432. According to the network architecture (147 → 64 → 64 → 64 → 4), the expected parameter count should be 18,052. So, there are about 380 extra parameters — could you please clarify what these extra parameters represent? Regarding how to convert the TCN checkpoint to the PyTorch format, is the parameter saving order something like [Layer0_Weights, Layer0_Bias, Layer1_Weights, Layer1_Bias, ...]? Also, is the memory layout "row-major"? <img width="1100" height="482" alt="Image" src="https://github.com/user-attachments/assets/e6958aa3-ff39-4214-bb9d-23621bb88b78" /> <img width="906" height="514" alt="Image" src="https://github.com/user-attachments/assets/4b32bce6-33e0-45a2-8628-6892a7734fb4" /> <img width="964" height="230" alt="Image" src="https://github.com/user-attachments/assets/39c62cc7-52ce-4dfd-b8bd-f02ecf7e8170" />