I’m getting training accuracy and testing accuracy much greater than 1, with training accuracy around 64 and test accuracy around 32. I’m using gloss.SoftmaxCrossEntropyLoss and linear regression model with one hidden layer from chapter 3. My batch size is 256. I split the features and labels when doing part 3. Want to ask why this is happening?