TEST

app.py

@@ -48,7 +48,10 @@ def generate(
     device = "cuda"
     print(current_input)
     input_ids = tokenizer(current_input, return_tensors="pt").input_ids.to(device)
-    original_input_length = input_ids.shape[1]  # Remember the input length
+    
+    outputs = model.generate(**inputs, max_new_tokens=100)
+
+    print(tokenizer.decode(outputs[0], skip_special_tokens=False))
 
 
     if len(input_ids) > MAX_INPUT_TOKEN_LENGTH: