Update eduport_tts_mal.py

eduport_tts_mal.py

@@ -72,26 +72,32 @@ class SpeechDataset(Dataset):
 def collate_fn(batch):
     audio_inputs, text_inputs = zip(*batch)
     
+    # Ensure audio inputs are 3D
+    audio_inputs = [audio.squeeze(0) if audio.dim() == 3 else audio for audio in audio_inputs]
+    
     # Pad audio inputs to the maximum length in the batch
     max_audio_len = max([audio.size(1) for audio in audio_inputs])
     
-    # Create padded audio inputs
     audio_inputs_padded = []
     for audio in audio_inputs:
         if audio.size(1) < max_audio_len:
-            # Create padding tensor with the same number of channels
-            padding = torch.zeros(audio.size(0), max_audio_len - audio.size(1))
+            padding = torch.zeros(audio.size(0), max_audio_len - audio.size(1), device=audio.device)
             padded_audio = torch.cat([audio, padding], dim=1)
             audio_inputs_padded.append(padded_audio)
         else:
             audio_inputs_padded.append(audio[:, :max_audio_len])
     
-    # Convert to tensor
+    # Stack audio inputs
     audio_inputs_padded = torch.stack(audio_inputs_padded)
     
-    # Pad text inputs to the longest transcript length
+    # Pad text inputs
     max_text_len = max([text.size(0) for text in text_inputs])
-    text_inputs_padded = torch.stack([torch.cat([text, torch.tensor([0] * (max_text_len - text.size(0)))], dim=0) if text.size(0) < max_text_len else text[:max_text_len] for text in text_inputs])
+    text_inputs_padded = torch.stack([
+        torch.cat([text, torch.zeros(max_text_len - text.size(0), dtype=text.dtype)], dim=0) 
+        if text.size(0) < max_text_len 
+        else text[:max_text_len] 
+        for text in text_inputs
+    ])
     
     return audio_inputs_padded, text_inputs_padded
 
@@ -142,6 +148,10 @@ class SpeechRecognitionModel(torch.nn.Module):
         self.decoder.gradient_checkpointing_enable()
 
     def forward(self, audio_input, text_input):
+        # Ensure audio_input is 3D: [batch_size, channels, time]
+        if audio_input.dim() == 4:
+            audio_input = audio_input.squeeze(1)  # Remove extra dimension
+        
         # Extract encoder hidden states
         encoder_output = self.encoder(audio_input).last_hidden_state
         
@@ -224,7 +234,10 @@ def train_model(num_epochs=10, accumulation_steps=16):
         for batch_idx, (audio_input, text_input) in enumerate(train_progress):
 
             # Move tensors to device
-            audio_input = audio_input.squeeze(1).to(device)
+            audio_input = audio_input.squeeze.to(device)
+            # Squeeze or reshape if necessary
+            if audio_input.dim() == 4:
+                audio_input = audio_input.squeeze(1)  # Remove extra singleton dimension
             text_input = text_input.to(device)
 
             # Use autocast for mixed precision training