diff --git a/deepspeech/frontend/normalizer.py b/deepspeech/frontend/normalizer.py
index cbb82b8e301a30e903b42f7b52d9348ac80bb35e..3b1fbe7c6faa7639061ab07d98b7601a2bdf6b35 100644
--- a/deepspeech/frontend/normalizer.py
+++ b/deepspeech/frontend/normalizer.py
@@ -13,7 +13,6 @@
 # limitations under the License.
 """Contains feature normalizers."""
 import json
-import random
 
 import numpy as np
 import paddle
@@ -27,18 +26,19 @@ from deepspeech.frontend.utility import read_manifest
 __all__ = ["FeatureNormalizer"]
 
 
+# https://github.com/PaddlePaddle/Paddle/pull/31481
 class CollateFunc(object):
-    ''' Collate function for AudioDataset
-    '''
-
-    def __init__(self):
-        pass
+    def __init__(self, feature_func):
+        self.feature_func = feature_func
 
     def __call__(self, batch):
         mean_stat = None
         var_stat = None
         number = 0
-        for feat in batch:
+        for item in batch:
+            audioseg = AudioSegment.from_file(item['feat'])
+            feat = self.feature_func(audioseg)  #(D, T)
+
             sums = np.sum(feat, axis=1)
             if mean_stat is None:
                 mean_stat = sums
@@ -52,30 +52,25 @@ class CollateFunc(object):
                 var_stat += square_sums
 
             number += feat.shape[1]
-        return paddle.to_tensor(number), paddle.to_tensor(
-            mean_stat), paddle.to_tensor(var_stat)
-        #return number, mean_stat, var_stat
+        return number, mean_stat, var_stat
 
 
 class AudioDataset(Dataset):
-    def __init__(self, manifest_path, feature_func, num_samples=-1, rng=None):
-        self.feature_func = feature_func
-        self._rng = rng
+    def __init__(self, manifest_path, num_samples=-1, rng=None, random_seed=0):
+        self._rng = rng if rng else np.random.RandomState(random_seed)
         manifest = read_manifest(manifest_path)
         if num_samples == -1:
             sampled_manifest = manifest
         else:
-            sampled_manifest = self._rng.sample(manifest, num_samples)
+            sampled_manifest = self._rng.choice(
+                manifest, num_samples, replace=False)
         self.items = sampled_manifest
 
     def __len__(self):
         return len(self.items)
 
     def __getitem__(self, idx):
-        key = self.items[idx]['feat']
-        audioseg = AudioSegment.from_file(key)
-        feat = self.feature_func(audioseg)  #(D, T)
-        return feat
+        return self.items[idx]
 
 
 class FeatureNormalizer(object):
@@ -112,7 +107,7 @@ class FeatureNormalizer(object):
             if not (manifest_path and featurize_func):
                 raise ValueError("If mean_std_filepath is None, meanifest_path "
                                  "and featurize_func should not be None.")
-            self._rng = random.Random(random_seed)
+            self._rng = np.random.RandomState(random_seed)
             self._compute_mean_std(manifest_path, featurize_func, num_samples,
                                    num_workers)
         else:
@@ -150,29 +145,11 @@ class FeatureNormalizer(object):
                           featurize_func,
                           num_samples,
                           num_workers,
+                          batch_size=64,
                           eps=1e-20):
         """Compute mean and std from randomly sampled instances."""
-        # manifest = read_manifest(manifest_path)
-        # if num_samples == -1:
-        #     sampled_manifest = manifest
-        # else:
-        #     sampled_manifest = self._rng.sample(manifest, num_samples)
-        # features = []
-        # for instance in sampled_manifest:
-        #     features.append(
-        #         featurize_func(AudioSegment.from_file(instance["feat"])))
-        # features = np.hstack(features)  #(D, T)
-        # self._mean = np.mean(features, axis=1)  #(D,)
-        # std = np.std(features, axis=1)  #(D,)
-        # std = np.clip(std, eps, None)
-        # self._istd = 1.0 / std
-
-        collate_func = CollateFunc()
-
-        dataset = AudioDataset(manifest_path, featurize_func, num_samples,
-                               self._rng)
-
-        batch_size = 20
+        collate_func = CollateFunc(featurize_func)
+        dataset = AudioDataset(manifest_path, num_samples, self._rng)
         data_loader = DataLoader(
             dataset,
             batch_size=batch_size,
@@ -185,9 +162,9 @@ class FeatureNormalizer(object):
             all_var_stat = None
             all_number = 0
             wav_number = 0
-            for batch in data_loader():
+            for i, batch in enumerate(data_loader):
                 number, mean_stat, var_stat = batch
-                if all_mean_stat is None:
+                if i == 0:
                     all_mean_stat = mean_stat
                     all_var_stat = var_stat
                 else:
@@ -198,12 +175,12 @@ class FeatureNormalizer(object):
 
                 if wav_number % 1000 == 0:
                     print('process {} wavs,{} frames'.format(wav_number,
-                                                             int(all_number)))
+                                                             all_number))
 
         self.cmvn_info = {
             'mean_stat': list(all_mean_stat.tolist()),
             'var_stat': list(all_var_stat.tolist()),
-            'frame_num': int(all_number),
+            'frame_num': all_number,
         }
 
         return self.cmvn_info
diff --git a/deepspeech/io/dataset.py b/deepspeech/io/dataset.py
index 4f58d41c6b61bfeda44f085d5124e9c78d7dbf65..333fc0c648fd9e28f94558e5ea5dd8d0101ebfd6 100644
--- a/deepspeech/io/dataset.py
+++ b/deepspeech/io/dataset.py
@@ -12,12 +12,12 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 import io
-import random
 import tarfile
 import time
 from collections import namedtuple
 from typing import Optional
 
+import numpy as np
 from paddle.io import Dataset
 from yacs.config import CfgNode
 
@@ -209,7 +209,7 @@ class ManifestDataset(Dataset):
             use_dB_normalization=use_dB_normalization,
             target_dB=target_dB)
 
-        self._rng = random.Random(random_seed)
+        self._rng = np.random.RandomState(random_seed)
         self._keep_transcription_text = keep_transcription_text
         # for caching tar files info
         self._local_data = namedtuple('local_data', ['tar2info', 'tar2object'])