Peaks/troughs Not Large Enough

[justustulo]

I have had some promising results running the model on simple univariate time series data. I performed some hyperparameter tuning, however, my peaks are troughs are consistently too small (see below).

Is there a parameter(s) to better tune this to allow for more volatility? Any ideas are appreciated. Thank you!

[2ySong]

hello, I konw his dataset codes is

def __getsamples(self, data):
        X = torch.zeros((self.sample_num, self.window, self.var_num))
        Y = torch.zeros((self.sample_num, 1, self.var_num))

        for i in range(self.sample_num):
            start = i
            end = i + self.window
            X[i, :, :] = torch.from_numpy(data[start:end, :])
            Y[i, :, :] = torch.from_numpy(data[end + self.horizon - 1, :])

        return (X, Y)

here, he only predict a point. but you got 350 points. Could you explain it? Thank you!