updates for onnx

infer_intent.py

@@ -41,20 +41,25 @@ class IntentClassifier:
                 truncation=True,       # Truncate if the text is too long
                 max_length=64)
 
-        self.intent_model.eval()
-        with torch.no_grad():
-            outputs = self.intent_model(**inputs)
-            logits = outputs.logits
-            prediction = torch.argmax(logits, dim=1).item()
-            probabilities = torch.softmax(logits, dim=1)
-            rounded_probabilities = torch.round(probabilities, decimals=3)
-            
-            pred_result = self.id2label[prediction]
-            proba_result = dict(zip(self.label2id.keys(), rounded_probabilities.tolist()[0]))
-            if verbose:
-                print(sequence + " -> " + pred_result)
-                print(proba_result,  "\n")
-            return pred_result, proba_result
+        # Convert inputs to NumPy arrays
+        onnx_inputs = {k: v for k, v in inputs.items()}
+
+        # Run the ONNX model
+        logits = self.ort_session.run(None, onnx_inputs)[0]
+
+        # Get the prediction
+        prediction = np.argmax(logits, axis=1)[0]
+        probabilities = np.exp(logits) / np.sum(np.exp(logits), axis=1, keepdims=True)
+        rounded_probabilities = np.round(probabilities, decimals=3)
+
+        pred_result = self.id2label[prediction]
+        proba_result = dict(zip(self.label2id.keys(), rounded_probabilities[0].tolist()))
+        
+        if verbose:
+            print(sequence + " -> " + pred_result)
+            print(proba_result, "\n")
+        
+        return pred_result, proba_result
 
 
 def main():