changes for squeezewave and non cuda

FastSpeech/synthesis.py

@@ -13,6 +13,9 @@ import utils
 import audio as Audio
 import glow
 import waveglow
+import time
+
+
 
 device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
 
@@ -21,7 +24,7 @@ def get_FastSpeech(num):
     checkpoint_path = "checkpoint_" + str(num) + ".pth.tar"
     model = nn.DataParallel(FastSpeech()).to(device)
     model.load_state_dict(torch.load(os.path.join(
-        hp.checkpoint_path, checkpoint_path))['model'])
+        hp.checkpoint_path, checkpoint_path),map_location=device)['model'])
     model.eval()
 
     return model
@@ -35,11 +38,15 @@ def synthesis(model, text, alpha=1.0):
     src_pos = np.stack([src_pos])
     with torch.no_grad():
         sequence = torch.autograd.Variable(
-            torch.from_numpy(text)).cuda().long()
+            torch.from_numpy(text)).long()
         src_pos = torch.autograd.Variable(
-            torch.from_numpy(src_pos)).cuda().long()
+            torch.from_numpy(src_pos)).long()
 
         mel, mel_postnet = model.module.forward(sequence, src_pos, alpha=alpha)
+        
+        #script for generating torch script
+        #traced_script_module = torch.jit.trace(model,(sequence,src_pos))
+        #traced_script_module.save("traced_fastspeech_model.pt")
 
         return mel[0].cpu().transpose(0, 1), \
             mel_postnet[0].cpu().transpose(0, 1), \
@@ -54,11 +61,15 @@ if __name__ == "__main__":
     model = get_FastSpeech(num)
     words = "Let’s go out to the airport. The plane landed ten minutes ago."
 
+    start = time.time()
     mel, mel_postnet, mel_torch, mel_postnet_torch = synthesis(
         model, words, alpha=alpha)
 
     if not os.path.exists("results"):
         os.mkdir("results")
+    
+    #do not use any vocoder , mel file generated will be passed to squeezewave vocoder. 
+    """
     Audio.tools.inv_mel_spec(mel_postnet, os.path.join(
         "results", words + "_" + str(num) + "_griffin_lim.wav"))
 
@@ -70,5 +81,13 @@ if __name__ == "__main__":
     mel_tac2, _, _ = utils.load_data_from_tacotron2(words, tacotron2)
     waveglow.inference.inference(torch.stack([torch.from_numpy(
         mel_tac2).cuda()]), wave_glow, os.path.join("results", "tacotron2.wav"))
-
     utils.plot_data([mel.numpy(), mel_postnet.numpy(), mel_tac2])
+
+    """
+    #melspec = torch.squeeze(mel_postnet_torch, 0)
+    torch.save(mel_postnet_torch, "../SqueezeWave/mel_spectrograms/test.pt")
+    end = time.time()
+    print("MEL Calculation:")
+    print(end-start)
+
+