Update app.py

app.py

@@ -28,37 +28,22 @@ async def predict(request: PredictionRequest):
         logger.info(f"Loading model: {request.model}")
         model_path = MODELS[request.model]
         
-        # Load tokenizer and model
-        tokenizer = AutoTokenizer.from_pretrained(
-            model_path,
-            token=HF_TOKEN,
-        )
-        
-        model = AutoModelForSeq2SeqLM.from_pretrained(
-            model_path,
-            token=HF_TOKEN,
-            device_map="auto"
-        )
+        tokenizer = AutoTokenizer.from_pretrained(model_path, token=HF_TOKEN)
+        model = AutoModelForSeq2SeqLM.from_pretrained(model_path, token=HF_TOKEN)
         
         full_input = "Interpret this dream: " + request.inputs
-        logger.info(f"Processing: {full_input}")
+        logger.info(f"Processing input: {full_input}")
         
         inputs = tokenizer(
             full_input,
             return_tensors="pt",
             truncation=True,
-            max_length=512,
-            padding=True
-        )
-        
-        outputs = model.generate(
-            **inputs,
-            max_length=200,
-            num_beams=4,
-            no_repeat_ngram_size=2
+            max_length=512
         )
         
+        outputs = model.generate(**inputs, max_length=200)
         result = tokenizer.decode(outputs[0], skip_special_tokens=True)
+        
         return PredictionResponse(generated_text=result)
         
     except Exception as e: