feat: 发票重新生成

main.py

@@ -50,7 +50,7 @@ if __name__ == "__main__":
         default="dataset/发票审查.json",
         help="val dataset path",
     )
-    parser.add_argument("--prompt_path", type=str, default="prompt/发票审查.txt", help="prompts dir")
+    parser.add_argument("--prompt_path", type=str, default="prompt/发票审查-init.txt", help="prompts dir")
     parser.add_argument(
         "--output_dir", type=str, default="output_dir", help="Path to output dir"
     )

util.py

@@ -18,7 +18,7 @@ def eval_sample(item, eval_fn, model):
     try:
         response = model(x)
         eval_output_variable = eval_fn(inputs=dict(prediction=response, ground_truth_answer=y))
-        return 1 if '正确' in eval_output_variable.value else 0
+        return 1 if '你的判断结果为[正确]' in eval_output_variable.value else 0
     except Exception as e:
         print('error in eval_sample: ', e)
         return 0