Update app.py

app.py

@@ -39,21 +39,27 @@ async def predict(request: Request):
             raise HTTPException(status_code=400, detail="No text provided")
         
         # Tokenize input
-        inputs = tokenizer(text)
+        inputs = tokenizer(
+            text,
+            return_tensors="np",
+            padding=False,  # Disable padding
+            truncation=False,  # Disable truncation
+            add_special_tokens=True  # Ensure CLS/SEP tokens
+        )
         
-        # Run model
-        outputs = session.run(None, {
-            "input_ids": inputs["input_ids"].astype(np.int64),
-            "attention_mask": inputs["attention_mask"].astype(np.int64)
-        })
+        onnx_inputs = {
+            "input_ids": np.array(inputs["input_ids"], dtype=np.int64),
+            "attention_mask": np.array(inputs["attention_mask"], dtype=np.int64)
+        }
+        
+        outputs = session.run(None, onnx_inputs)
         
         # Prepare response with converted types
-        response = {
-            "embedding": convert_output(outputs[0]),  # Process main output
-            "tokens": tokenizer.convert_ids_to_tokens(inputs["input_ids"][0])
+        return {
+            "embedding": outputs[0][0].astype(float).tolist(),
+            "input_ids": inputs["input_ids"][0].tolist(),
+            "attention_mask": inputs["attention_mask"][0].tolist()
         }
-        print("embeddings", response["embedding"])
-        return jsonable_encoder(response)
     
     except Exception as e:
         raise HTTPException(status_code=500, detail=str(e))