change noise scheduling for wavegrad. Compute beta values externally to enable better flexibility

TTS/bin/tune_wavegrad.py

@@ -59,11 +59,13 @@ if args.use_cuda:
 
 # setup optimization parameters
 base_values = sorted(np.random.uniform(high=10, size=args.search_depth))
+exponents = 10 ** np.linspace(-6, -2, num=args.num_iter)
 best_error = float('inf')
 best_schedule = None
 total_search_iter = len(base_values)**args.num_iter
 for base in tqdm(cartesian_product(base_values, repeat=args.num_iter), total=total_search_iter):
-    model.compute_noise_level(num_steps=args.num_iter, min_val=1e-6, max_val=1e-1, base_vals=base)
+    beta = exponents * base
+    model.compute_noise_level(beta)
     for data in loader:
         mel, audio = data
         y_hat = model.inference(mel.cuda() if args.use_cuda else mel)
@@ -78,11 +80,11 @@ for base in tqdm(cartesian_product(base_values, repeat=args.num_iter), total=tot
             mel_hat.append(torch.from_numpy(m))
 
         mel_hat = torch.stack(mel_hat)
-        mse = torch.sum((mel - mel_hat) ** 2)
+        mse = torch.sum((mel - mel_hat) ** 2).mean()
         if mse.item() < best_error:
             best_error = mse.item()
-            best_schedule = {'num_steps': args.num_iter, 'min_val':1e-6, 'max_val':1e-1, 'base_vals':base}
-            print(" > Found a better schedule.")
+            best_schedule = {'beta': beta}
+            print(f" > Found a better schedule. - MSE: {mse.item()}")
             np.save(args.output_path, best_schedule)
 
 

TTS/vocoder/configs/wavegrad_libritts.json

@@ -72,6 +72,8 @@
 
     // TRAINING
     "batch_size": 96,      // Batch size for training.
+
+    // NOISE SCHEDULE PARAMS - Only effective at training time.
     "train_noise_schedule":{
         "min_val": 1e-6,
         "max_val": 1e-2,

TTS/vocoder/models/wavegrad.py

@@ -79,8 +79,8 @@ class Wavegrad(nn.Module):
         return x
 
     def load_noise_schedule(self, path):
-        sched = np.load(path, allow_pickle=True).item()
-        self.compute_noise_level(**sched)
+        beta = np.load(path, allow_pickle=True).item()['beta']
+        self.compute_noise_level(beta)
 
     @torch.no_grad()
     def inference(self, x, y_n=None):
@@ -113,16 +113,13 @@ class Wavegrad(nn.Module):
         noisy_audio = noise_scale * y_0 + (1.0 - noise_scale**2)**0.5 * noise
         return noise.unsqueeze(1), noisy_audio.unsqueeze(1), noise_scale[:, 0]
 
-    def compute_noise_level(self, num_steps, min_val, max_val, base_vals=None):
+    def compute_noise_level(self, beta):
         """Compute noise schedule parameters"""
-        beta = np.linspace(min_val, max_val, num_steps)
-        if base_vals is not None:
-            beta *= base_vals
+        self.num_steps = len(beta)
         alpha = 1 - beta
         alpha_hat = np.cumprod(alpha)
         noise_level = np.concatenate([[1.0], alpha_hat ** 0.5], axis=0)
 
-        self.num_steps = num_steps
         # pylint: disable=not-callable
         self.beta = torch.tensor(beta.astype(np.float32))
         self.alpha = torch.tensor(alpha.astype(np.float32))