In load_dataset function shown below:
Line 187 and 188, I'm confused these 2 lines may cause negative speed data.
And then I try to print the data in the dataloader in training process: (pytorch implement in dcrnn_supervisor.py)
I found a lot of speed data is negative: (As shown below)
The first column represent the speed and the second column is time.
Since loss function compute the mask after the scaler tranform operation, I think some of data input to the model are possibly wrong.
I hope to get answer to my question
In load_dataset function shown below: Line 187 and 188, I'm confused these 2 lines may cause negative speed data. And then I try to print the data in the dataloader in training process: (pytorch implement in dcrnn_supervisor.py) I found a lot of speed data is negative: (As shown below) The first column represent the speed and the second column is time. Since loss function compute the mask after the scaler tranform operation, I think some of data input to the model are possibly wrong. I hope to get answer to my question