revert random permutation from numpy to paddle (#792)

Author: HydrogenSulfate

examples/epnn/conf/epnn.yaml

@@ -41,12 +41,11 @@ TRAIN:
   iters_per_epoch: 1
   save_freq: 50
   eval_during_train: true
-  eval_with_no_grad: true
   lr_scheduler:
     epochs: ${TRAIN.epochs}
     iters_per_epoch: ${TRAIN.iters_per_epoch}
     gamma: 0.97
-    decay_steps: 1
+    decay_steps: 10000000
   pretrained_model_path: null
   checkpoint_path: null
 

examples/epnn/epnn.py

@@ -114,7 +114,7 @@ def _transform_in_stress(_in):
         save_freq=cfg.TRAIN.save_freq,
         eval_during_train=cfg.TRAIN.eval_during_train,
         validator=validator_pde,
-        eval_with_no_grad=cfg.TRAIN.eval_with_no_grad,
+        eval_with_no_grad=cfg.EVAL.eval_with_no_grad,
     )
 
     # train model

examples/epnn/functions.py

@@ -248,7 +248,9 @@ def get(self, epochs=1):
         label_dict_train = {"dummy_loss": []}
         label_dict_val = {"dummy_loss": []}
         for i in range(epochs):
-            shuffled_indices = np.random.permutation(self.data_state.x_train.shape[0])
+            shuffled_indices = paddle.randperm(
+                n=self.data_state.x_train.shape[0]
+            ).numpy()
             input_dict_train["state_x"].append(
                 self.data_state.x_train[shuffled_indices[0 : self.itrain]]
             )
@@ -263,7 +265,7 @@ def get(self, epochs=1):
             )
             label_dict_train["dummy_loss"].append(0.0)
 
-        shuffled_indices = np.random.permutation(self.data_state.x_valid.shape[0])
+        shuffled_indices = paddle.randperm(n=self.data_state.x_valid.shape[0]).numpy()
         input_dict_val["state_x"].append(
             self.data_state.x_valid[shuffled_indices[0 : self.itrain]]
         )
@@ -296,7 +298,7 @@ def __init__(self, dataset_path, train_p=0.6, cross_valid_p=0.2, test_p=0.2):
     def get_shuffled_data(self):
         # Need to set the seed, otherwise the loss will not match the precision
         ppsci.utils.misc.set_random_seed(seed=10)
-        shuffled_indices = np.random.permutation(self.x.shape[0])
+        shuffled_indices = paddle.randperm(n=self.x.shape[0]).numpy()
         n_train = math.floor(self.train_p * self.x.shape[0])
         n_cross_valid = math.floor(self.cross_valid_p * self.x.shape[0])
         n_test = math.floor(self.test_p * self.x.shape[0])