debug

.idea/workspace.xml

@@ -6,6 +6,7 @@
   <component name="ChangeListManager">
     <list default="true" id="23565123-73ab-4f40-a9ef-1086e0c9e1ec" name="Changes" comment="">
       <change beforePath="$PROJECT_DIR$/.idea/workspace.xml" beforeDir="false" afterPath="$PROJECT_DIR$/.idea/workspace.xml" afterDir="false" />
+      <change beforePath="$PROJECT_DIR$/preprocess.py" beforeDir="false" afterPath="$PROJECT_DIR$/preprocess.py" afterDir="false" />
       <change beforePath="$PROJECT_DIR$/script.py" beforeDir="false" afterPath="$PROJECT_DIR$/script.py" afterDir="false" />
     </list>
     <option name="SHOW_DIALOG" value="false" />
@@ -99,7 +100,7 @@
       <updated>1742573353560</updated>
       <workItem from="1742573355153" duration="587000" />
       <workItem from="1742806974298" duration="2741000" />
-      <workItem from="1742810431420" duration="9606000" />
+      <workItem from="1742810431420" duration="9891000" />
     </task>
     <servers />
   </component>

preprocess.py

@@ -13,7 +13,7 @@ def pad_audio(x, max_len=48000):
     return padded_x
 
 
-def preprocess_new(audio_file):
+def preprocess(audio_file):
     print(f'Preprocessing {audio_file}')
     # Load the audio file
     y, sr = librosa.load(audio_file, sr=16000)
@@ -33,7 +33,7 @@ def preprocess_new(audio_file):
     return tensor
 
 
-def preprocess(audio_file):
+def preprocess_old(audio_file):
     # Load the audio file
     y, sr = librosa.load(audio_file, sr=None)
     mfccs = librosa.feature.mfcc(y=y, sr=sr, n_mfcc=13)

script.py

@@ -54,15 +54,16 @@ config = {
 model = RawNet(config, device)
 
 print('Load model weights')
-model_path = '/tmp/checkpoints/RAWNET_ASVSPOOF_FOR_INTHEWILD_PURDUE.pth'
-model.load_state_dict(torch.load(model_path, map_location=device))
+# model_path = '/tmp/checkpoints/RAWNET_ASVSPOOF_FOR_INTHEWILD_PURDUE.pth'
+# model.load_state_dict(torch.load(model_path, map_location=device))
 
 model.eval()
 
 print('Loaded RawNet2 Weights')
 
-del model
-model = Model().to(device) # OLD MODEL
+# # EVALUATE OLD MODEL
+# del model
+# model = Model().to(device)
 
 # iterate over the dataset
 out = []
@@ -90,13 +91,13 @@ for el in tqdm.tqdm(dataset_remote):
             # positive score correspond to synthetic prediction
             # negative score correspond to pristine prediction
 
-            # OLD MODEL
-            score = model(tensor.to(device)).cpu().item()
-            # # RAWNNET2 MODEL
-            # score = model(tensor.to(device))[:, 1].cpu()
-            # print(f'SCORE OUT: {score}')
-            # score = score.mean().item()
-            # print(f'SCORE FINAL: {score}')
+            # # OLD MODEL
+            # score = model(tensor.to(device)).cpu().item()
+            # RAWNNET2 MODEL
+            score = model(tensor.to(device))[:, 1].cpu()
+            print(f'SCORE OUT: {score}')
+            score = score.mean().item()
+            print(f'SCORE FINAL: {score}')
 
             # we require a hard decision to be submited. so you need to pick a threshold
             pred = "generated" if score > model.threshold else "pristine"