u

app.py

@@ -116,10 +116,26 @@ if __name__ == "__main__":
         device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
         torch_dtype = torch.float16 if torch.cuda.is_available() else torch.float32
 
+        MODEL_NAME = "alakxender/wav2vec2-large-mms-1b-dv-syn-md" # Trained on 100% Synthetic Data (130-150 Hours)
+        # MODEL_NAME = "alakxender/wav2vec2-large-mms-1b-cv" # Trained on Common Voice Data (Unknown Hours)
+        # MODEL_NAME =  "alakxender/whisper-small-dv-syn-md" # Trained on 100% Synthetic Data (150 Hours)   
+        # MODEL_NAME = "alakxender/whisper-small-cv" # Trained on Common Voice Data (Unknown Hours)
+        # MODEL_NAME = "alakxender/whisper-medium-dv-syn-md" # Trained on 100% Synthetic Data (150 Hours)
+        # MODEL_NAME = "alakxender/whisper-medium-cv" # Trained on Common Voice Data (Unknown Hours)
+        # MODEL_NAME = "alakxender/whisper-large-v3-dv-syn-md" # Trained on 100% Synthetic Data (150 Hours)
+        # MODEL_NAME = "alakxender/whisper-large-v3-cv" # Trained on Common Voice Data (Unknown Hours)
+        # MODEL_NAME = "alakxender/whisper-large-v3-calls-md" # Trained on phone calls (65 Hours)
+        # MODEL_NAME = "alakxender/wav2vec2-large-mms-1b-calls-md" # Trained on phone calls (65 Hours)
+        # MODEL_NAME = "alakxender/wav2vec2-large-xlsr-calls-md" # Trained on phone calls (23 Hours)
+        # MODEL_NAME = "alakxender/wav2vec2-large-xlsr-dv-syn-md" # Trained on 100% Synthetic Data (80 Hours)
+        # MODEL_NAME = "alakxender/dhivehi-asr-full-ctc" # Trained on multiple datasets (350+ Hours)
+        # MODEL_NAME = "alakxender/dhivehi-asr-full-ctc-v2" # Trained on multiple datasets (350+ Hours)
+        # MODEL_NAME = "alakxender/dhivehi-asr-full-whisper-v3" # Trained on multiple datasets (350+ Hours)
+
         # Load model and processor with LM
-        processor = Wav2Vec2ProcessorWithLM.from_pretrained("alakxender/wav2vec2-large-mms-1b-dv-syn-md")
+        processor = Wav2Vec2ProcessorWithLM.from_pretrained(MODEL_NAME)
         model = Wav2Vec2ForCTC.from_pretrained(
-            "alakxender/wav2vec2-large-mms-1b-dv-syn-md",
+            MODEL_NAME,
             torch_dtype=torch_dtype
         ).to(device)