Module GPU test fixes (#369)

* Fixed GPU tests

torchaudio/transforms.py

@@ -51,7 +51,7 @@ class Spectrogram(torch.nn.Module):
         self.win_length = win_length if win_length is not None else n_fft
         self.hop_length = hop_length if hop_length is not None else self.win_length // 2
         window = window_fn(self.win_length) if wkwargs is None else window_fn(self.win_length, **wkwargs)
-        self.window = window
+        self.register_buffer('window', window)
         self.pad = pad
         self.power = power
         self.normalized = normalized
@@ -136,7 +136,7 @@ class MelScale(torch.nn.Module):
 
         fb = torch.empty(0) if n_stft is None else F.create_fb_matrix(
             n_stft, self.f_min, self.f_max, self.n_mels, self.sample_rate)
-        self.fb = fb
+        self.register_buffer('fb', fb)
 
     def forward(self, specgram):
         r"""
@@ -260,7 +260,7 @@ class MFCC(torch.nn.Module):
         if self.n_mfcc > self.MelSpectrogram.n_mels:
             raise ValueError('Cannot select more MFCC coefficients than # mel bins')
         dct_mat = F.create_dct(self.n_mfcc, self.MelSpectrogram.n_mels, self.norm)
-        self.dct_mat = dct_mat
+        self.register_buffer('dct_mat', dct_mat)
         self.log_mels = log_mels
 
     def forward(self, waveform):