I greatly appreciate your work, and I strongly believe that this implementation can help regression models from becoming similar to persistence models during training to become a forecasting utility.
I need some help in creating a loss module with the help of your library, here is my current design (simple):
class CustomLoss(nn.Module):
def __init__(self):
super(CustomLoss, self).__init__()
def forward(self, predicted, target):
# Calculate the dynamic time warping distance using FASTDTW
distance, _ = fastdtw(predicted.detach().numpy(), target.detach().numpy(), dist=euclidean)
# Convert the distance to a PyTorch tensor
distance = torch.tensor(distance, dtype=torch.float32, requires_grad=True)
# Return the distance as the loss
return distance
criterion = CustomLoss()
Could I ask if this is the way to integrate your module into a custom loss module?
Hi repository owner(s)!
I greatly appreciate your work, and I strongly believe that this implementation can help regression models from becoming similar to persistence models during training to become a forecasting utility.
I need some help in creating a loss module with the help of your library, here is my current design (simple):
Could I ask if this is the way to integrate your module into a custom loss module?
Thank you!