diff --git a/ppocr/postprocess/rec_postprocess.py b/ppocr/postprocess/rec_postprocess.py
index fbf8b93e3d11121c99ce5b2dcbf2149e15453d4a..5547f005759b6585252c62d1e8a9d468e9880a2f 100644
--- a/ppocr/postprocess/rec_postprocess.py
+++ b/ppocr/postprocess/rec_postprocess.py
@@ -891,7 +891,7 @@ class VLLabelDecode(BaseRecLabelDecode):
             ) + length[i])].topk(1)[0][:, 0]
             preds_prob = paddle.exp(
                 paddle.log(preds_prob).sum() / (preds_prob.shape[0] + 1e-6))
-            text.append((preds_text, preds_prob.numpy()[0]))
+            text.append((preds_text, float(preds_prob)))
         if label is None:
             return text
         label = self.decode(label)
diff --git a/test_tipc/prepare.sh b/test_tipc/prepare.sh
index 02ee8a24d241195d1330ea42fc05ed35dd7a87b7..e184b14ddc0718aa169229b3c95d2dc84d46a183 100644
--- a/test_tipc/prepare.sh
+++ b/test_tipc/prepare.sh
@@ -150,7 +150,9 @@ if [ ${MODE} = "lite_train_lite_infer" ];then
     # pretrain lite train data
     wget -nc -P  ./pretrain_models/ https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/MobileNetV3_large_x0_5_pretrained.pdparams  --no-check-certificate
     wget -nc -P ./pretrain_models/  https://paddleocr.bj.bcebos.com/dygraph_v2.0/en/det_mv3_db_v2.0_train.tar  --no-check-certificate
-    cd ./pretrain_models/ && tar xf det_mv3_db_v2.0_train.tar && cd ../
+    cd ./pretrain_models/
+    tar xf det_mv3_db_v2.0_train.tar
+    cd ../
     if [[ ${model_name} =~ "ch_PP-OCRv2_det" ]];then
         wget -nc -P ./pretrain_models/ https://paddleocr.bj.bcebos.com/PP-OCRv2/chinese/ch_PP-OCRv2_det_distill_train.tar --no-check-certificate
         cd ./pretrain_models/ && tar xf ch_PP-OCRv2_det_distill_train.tar && cd ../
diff --git a/test_tipc/supplementary/train.py b/test_tipc/supplementary/train.py
index e632d1d1803a85144bc750c3ff6ff51b1eb65973..a15a99ff85e6667700f7e57800e1feeb013da869 100644
--- a/test_tipc/supplementary/train.py
+++ b/test_tipc/supplementary/train.py
@@ -168,22 +168,22 @@ def train(config, scaler=None):
             if idx % 10 == 0:
                 et = time.time()
                 strs = f"epoch: [{epoch}/{EPOCH}], iter: [{idx}/{data_num}], "
-                strs += f"loss: {avg_loss.numpy()[0]}"
-                strs += f", acc_topk1: {acc['top1'].numpy()[0]}, acc_top5: {acc['top5'].numpy()[0]}"
+                strs += f"loss: {float(avg_loss)}"
+                strs += f", acc_topk1: {float(acc['top1'])}, acc_top5: {float(acc['top5'])}"
                 strs += f", batch_time: {round(et-st, 4)} s"
                 logger.info(strs)
                 st = time.time()
 
         if epoch % 10 == 0:
             acc = eval(config, model)
-            if len(best_acc) < 1 or acc['top5'].numpy()[0] > best_acc['top5']:
+            if len(best_acc) < 1 or float(acc['top5']) > best_acc['top5']:
                 best_acc = acc
                 best_acc['epoch'] = epoch
                 is_best = True
             else:
                 is_best = False
             logger.info(
-                f"The best acc: acc_topk1: {best_acc['top1'].numpy()[0]}, acc_top5: {best_acc['top5'].numpy()[0]}, best_epoch: {best_acc['epoch']}"
+                f"The best acc: acc_topk1: {float(best_acc['top1'])}, acc_top5: {float(best_acc['top5'])}, best_epoch: {best_acc['epoch']}"
             )
             save_model(
                 model,
@@ -276,22 +276,22 @@ def train_distill(config, scaler=None):
             if idx % 10 == 0:
                 et = time.time()
                 strs = f"epoch: [{epoch}/{EPOCH}], iter: [{idx}/{data_num}], "
-                strs += f"loss: {avg_loss.numpy()[0]}"
-                strs += f", acc_topk1: {acc['top1'].numpy()[0]}, acc_top5: {acc['top5'].numpy()[0]}"
+                strs += f"loss: {float(avg_loss)}"
+                strs += f", acc_topk1: {float(acc['top1'])}, acc_top5: {float(acc['top5'])}"
                 strs += f", batch_time: {round(et-st, 4)} s"
                 logger.info(strs)
                 st = time.time()
 
         if epoch % 10 == 0:
             acc = eval(config, model._layers.student)
-            if len(best_acc) < 1 or acc['top5'].numpy()[0] > best_acc['top5']:
+            if len(best_acc) < 1 or float(acc['top5']) > best_acc['top5']:
                 best_acc = acc
                 best_acc['epoch'] = epoch
                 is_best = True
             else:
                 is_best = False
             logger.info(
-                f"The best acc: acc_topk1: {best_acc['top1'].numpy()[0]}, acc_top5: {best_acc['top5'].numpy()[0]}, best_epoch: {best_acc['epoch']}"
+                f"The best acc: acc_topk1: {float(best_acc['top1'])}, acc_top5: {float(best_acc['top5'])}, best_epoch: {best_acc['epoch']}"
             )
 
             save_model(
@@ -401,22 +401,22 @@ def train_distill_multiopt(config, scaler=None):
             if idx % 10 == 0:
                 et = time.time()
                 strs = f"epoch: [{epoch}/{EPOCH}], iter: [{idx}/{data_num}], "
-                strs += f"loss: {avg_loss.numpy()[0]}, loss1: {avg_loss1.numpy()[0]}"
-                strs += f", acc_topk1: {acc['top1'].numpy()[0]}, acc_top5: {acc['top5'].numpy()[0]}"
+                strs += f"loss: {float(avg_loss)}, loss1: {float(avg_loss1)}"
+                strs += f", acc_topk1: {float(acc['top1'])}, acc_top5: {float(acc['top5'])}"
                 strs += f", batch_time: {round(et-st, 4)} s"
                 logger.info(strs)
                 st = time.time()
 
         if epoch % 10 == 0:
             acc = eval(config, model._layers.student)
-            if len(best_acc) < 1 or acc['top5'].numpy()[0] > best_acc['top5']:
+            if len(best_acc) < 1 or float(acc['top5']) > best_acc['top5']:
                 best_acc = acc
                 best_acc['epoch'] = epoch
                 is_best = True
             else:
                 is_best = False
             logger.info(
-                f"The best acc: acc_topk1: {best_acc['top1'].numpy()[0]}, acc_top5: {best_acc['top5'].numpy()[0]}, best_epoch: {best_acc['epoch']}"
+                f"The best acc: acc_topk1: {float(best_acc['top1'])}, acc_top5: {float(best_acc['top5'])}, best_epoch: {best_acc['epoch']}"
             )
             save_model(
                 model, [optimizer, optimizer1],
@@ -450,7 +450,7 @@ def eval(config, model):
     labels = paddle.concat(labels, axis=0)
     acc = metric_func(outs, labels)
 
-    strs = f"The metric are as follows: acc_topk1: {acc['top1'].numpy()[0]}, acc_top5: {acc['top5'].numpy()[0]}"
+    strs = f"The metric are as follows: acc_topk1: {float(acc['top1'])}, acc_top5: {float(acc['top5'])}"
     logger.info(strs)
     return acc