[CodeStyle][py2] remove `six` package (part 1) (#46965)

paddle/fluid/inference/tests/api/full_pascalvoc_test_preprocess.py

@@ -19,7 +19,7 @@ import os
 import sys
 from paddle.dataset.common import download
 import tarfile
-from six.moves import StringIO
+from io import StringIO
 import hashlib
 import tarfile
 import argparse

python/paddle/dataset/cifar.py

@@ -31,8 +31,7 @@ import numpy
 import paddle.dataset.common
 import paddle.utils.deprecated as deprecated
 import tarfile
-import six
-from six.moves import cPickle as pickle
+import pickle
 
 __all__ = []
 
@@ -46,11 +45,10 @@ CIFAR100_MD5 = 'eb9058c3a382ffc7106e4002c42a8d85'
 def reader_creator(filename, sub_name, cycle=False):
 
     def read_batch(batch):
-        data = batch[six.b('data')]
-        labels = batch.get(six.b('labels'), batch.get(six.b('fine_labels'),
-                                                      None))
+        data = batch[b'data']
+        labels = batch.get(b'labels', batch.get(b'fine_labels', None))
         assert labels is not None
-        for sample, label in six.moves.zip(data, labels):
+        for sample, label in zip(data, labels):
             yield (sample / 255.0).astype(numpy.float32), int(label)
 
     def reader():

python/paddle/dataset/common.py

@@ -20,7 +20,7 @@ import shutil
 import sys
 import importlib
 import paddle.dataset
-import six.moves.cPickle as pickle
+import pickle
 import tempfile
 import glob
 import paddle

python/paddle/dataset/conll05.py

@@ -24,7 +24,6 @@ import tarfile
 import gzip
 import paddle.dataset.common
 import paddle.utils.deprecated as deprecated
-from six.moves import zip, range
 
 __all__ = []
 

python/paddle/dataset/image.py

@@ -30,39 +30,34 @@ the image layout as follows.
   be keep consistent between the training and inference period.
 """
 
-import six
 import numpy as np
 # FIXME(minqiyang): this is an ugly fix for the numpy bug reported here
 # https://github.com/numpy/numpy/issues/12497
-if six.PY3:
-    import subprocess
-    import sys
-    import os
-    interpreter = sys.executable
-    # Note(zhouwei): if use Python/C 'PyRun_SimpleString', 'sys.executable'
-    # will be the C++ execubable on Windows
-    if sys.platform == 'win32' and 'python.exe' not in interpreter:
-        interpreter = sys.exec_prefix + os.sep + 'python.exe'
-    import_cv2_proc = subprocess.Popen([interpreter, "-c", "import cv2"],
-                                       stdout=subprocess.PIPE,
-                                       stderr=subprocess.PIPE)
-    out, err = import_cv2_proc.communicate()
-    retcode = import_cv2_proc.poll()
-    if retcode != 0:
-        cv2 = None
-    else:
-        try:
-            import cv2
-        except ImportError:
-            cv2 = None
+import subprocess
+import sys
+import os
+
+interpreter = sys.executable
+# Note(zhouwei): if use Python/C 'PyRun_SimpleString', 'sys.executable'
+# will be the C++ execubable on Windows
+if sys.platform == 'win32' and 'python.exe' not in interpreter:
+    interpreter = sys.exec_prefix + os.sep + 'python.exe'
+import_cv2_proc = subprocess.Popen([interpreter, "-c", "import cv2"],
+                                   stdout=subprocess.PIPE,
+                                   stderr=subprocess.PIPE)
+out, err = import_cv2_proc.communicate()
+retcode = import_cv2_proc.poll()
+if retcode != 0:
+    cv2 = None
 else:
     try:
         import cv2
     except ImportError:
         cv2 = None
+
 import os
 import tarfile
-import six.moves.cPickle as pickle
+import pickle
 
 __all__ = []
 

python/paddle/dataset/imdb.py

@@ -26,7 +26,6 @@ import collections
 import tarfile
 import re
 import string
-import six
 
 __all__ = []
 
@@ -49,9 +48,8 @@ def tokenize(pattern):
         while tf != None:
             if bool(pattern.match(tf.name)):
                 # newline and punctuations removal and ad-hoc tokenization.
-                yield tarf.extractfile(tf).read().rstrip(
-                    six.b("\n\r")).translate(None, six.b(
-                        string.punctuation)).lower().split()
+                yield tarf.extractfile(tf).read().rstrip(b'\n\r').translate(
+                    None, string.punctuation.encode('latin-1')).lower().split()
             tf = tarf.next()
 
 
@@ -66,11 +64,11 @@ def build_dict(pattern, cutoff):
             word_freq[word] += 1
 
     # Not sure if we should prune less-frequent words here.
-    word_freq = [x for x in six.iteritems(word_freq) if x[1] > cutoff]
+    word_freq = [x for x in word_freq.items() if x[1] > cutoff]
 
     dictionary = sorted(word_freq, key=lambda x: (-x[1], x[0]))
     words, _ = list(zip(*dictionary))
-    word_idx = dict(list(zip(words, six.moves.range(len(words)))))
+    word_idx = dict(list(zip(words, range(len(words)))))
     word_idx['<unk>'] = len(words)
     return word_idx
 

python/paddle/dataset/imikolov.py

@@ -23,7 +23,6 @@ import paddle.dataset.common
 import paddle.utils.deprecated as deprecated
 import collections
 import tarfile
-import six
 
 __all__ = []
 
@@ -68,13 +67,11 @@ def build_dict(min_word_freq=50):
             # remove <unk> for now, since we will set it as last index
             del word_freq['<unk>']
 
-        word_freq = [
-            x for x in six.iteritems(word_freq) if x[1] > min_word_freq
-        ]
+        word_freq = [x for x in word_freq.items() if x[1] > min_word_freq]
 
         word_freq_sorted = sorted(word_freq, key=lambda x: (-x[1], x[0]))
         words, _ = list(zip(*word_freq_sorted))
-        word_idx = dict(list(zip(words, six.moves.range(len(words)))))
+        word_idx = dict(list(zip(words, range(len(words)))))
         word_idx['<unk>'] = len(words)
 
     return word_idx
@@ -96,7 +93,7 @@ def reader_creator(filename, word_idx, n, data_type):
                     l = ['<s>'] + l.strip().split() + ['<e>']
                     if len(l) >= n:
                         l = [word_idx.get(w, UNK) for w in l]
-                        for i in six.moves.range(n, len(l) + 1):
+                        for i in range(n, len(l) + 1):
                             yield tuple(l[i - n:i])
                 elif DataType.SEQ == data_type:
                     l = l.strip().split()

python/paddle/dataset/mnist.py

@@ -23,7 +23,6 @@ import paddle.utils.deprecated as deprecated
 import gzip
 import numpy
 import struct
-from six.moves import range
 
 __all__ = []
 

python/paddle/dataset/movielens.py

@@ -28,7 +28,6 @@ import paddle.dataset.common
 import paddle.utils.deprecated as deprecated
 import re
 import functools
-import six
 
 __all__ = []
 
@@ -206,7 +205,7 @@ def max_movie_id():
     Get the maximum value of movie id.
     """
     __initialize_meta_info__()
-    return six.moves.reduce(__max_index_info__, list(MOVIE_INFO.values())).index
+    return functools.reduce(__max_index_info__, list(MOVIE_INFO.values())).index
 
 
 @deprecated(
@@ -219,7 +218,7 @@ def max_user_id():
     Get the maximum value of user id.
     """
     __initialize_meta_info__()
-    return six.moves.reduce(__max_index_info__, list(USER_INFO.values())).index
+    return functools.reduce(__max_index_info__, list(USER_INFO.values())).index
 
 
 def __max_job_id_impl__(a, b):
@@ -239,7 +238,7 @@ def max_job_id():
     Get the maximum value of job id.
     """
     __initialize_meta_info__()
-    return six.moves.reduce(__max_job_id_impl__,
+    return functools.reduce(__max_job_id_impl__,
                             list(USER_INFO.values())).job_id
 
 

python/paddle/dataset/uci_housing.py

@@ -20,7 +20,6 @@ parse training set and test set into paddle reader creators.
 """
 
 import numpy as np
-import six
 import tempfile
 import tarfile
 import os
@@ -75,7 +74,7 @@ def load_data(filename, feature_num=14, ratio=0.8):
         axis=0), data.sum(axis=0) / data.shape[0]
     # if you want to print the distribution of input data, you could use function of feature_range
     #feature_range(maximums[:-1], minimums[:-1])
-    for i in six.moves.range(feature_num - 1):
+    for i in range(feature_num - 1):
         data[:, i] = (data[:, i] - avgs[i]) / (maximums[i] - minimums[i])
     offset = int(data.shape[0] * ratio)
     UCI_TRAIN_DATA = data[:offset]

python/paddle/dataset/wmt14.py

@@ -20,7 +20,6 @@ parse training set and test set into paddle reader creators.
 
 """
 
-import six
 import tarfile
 
 import paddle.dataset.common
@@ -173,8 +172,8 @@ def get_dict(dict_size, reverse=True):
     tar_file = paddle.dataset.common.download(URL_TRAIN, 'wmt14', MD5_TRAIN)
     src_dict, trg_dict = __read_to_dict(tar_file, dict_size)
     if reverse:
-        src_dict = {v: k for k, v in six.iteritems(src_dict)}
-        trg_dict = {v: k for k, v in six.iteritems(trg_dict)}
+        src_dict = {v: k for k, v in src_dict.items()}
+        trg_dict = {v: k for k, v in trg_dict.items()}
     return src_dict, trg_dict
 
 

python/paddle/dataset/wmt16.py

@@ -29,7 +29,6 @@ Multi30K: Multilingual English-German Image Descriptions.
 """
 
 import os
-import six
 import tarfile
 from collections import defaultdict
 
@@ -63,9 +62,7 @@ def __build_dict(tar_file, dict_size, save_path, lang):
     with open(save_path, "wb") as fout:
         fout.write(("%s\n%s\n%s\n" % (START_MARK, END_MARK, UNK_MARK)).encode())
         for idx, word in enumerate(
-                sorted(six.iteritems(word_dict),
-                       key=lambda x: x[1],
-                       reverse=True)):
+                sorted(word_dict.items(), key=lambda x: x[1], reverse=True)):
             if idx + 3 == dict_size: break
             fout.write(word[0].encode())
             fout.write(b'\n')

python/paddle/distributed/elastic.py

@@ -13,7 +13,6 @@
 # limitations under the License.
 
 import argparse
-import six
 import os
 
 
@@ -30,7 +29,7 @@ class Command(object):
         self.np_path = self.prefix + '/np'
 
     def set_np(self, np):
-        self.etcd.put(self.np_path, six.b('{}'.format(np)))
+        self.etcd.put(self.np_path, '{}'.format(np).encode('latin-1'))
 
     def scale_np(self, np):
         if self.etcd.get(self.np_path)[0] != None:

python/paddle/distributed/fleet/elastic/manager.py

@@ -262,7 +262,7 @@ class ElasticManager(object):
                         logger.info(
                             f"[lease_heartbeat] register host={self.curr_host}")
                         self.etcd.put(self.host_path,
-                                      six.b(self.curr_host),
+                                      self.curr_host.encode('latin-1'),
                                       lease=host_lease)
                 except Exception as e:
                     logger.error(
@@ -276,11 +276,15 @@ class ElasticManager(object):
                                              daemon=True)
         keepalived_thread.start()
 
-        self.etcd.put(self.host_path, six.b(self.curr_host), lease=host_lease)
+        self.etcd.put(self.host_path,
+                      self.curr_host.encode('latin-1'),
+                      lease=host_lease)
 
         # endpoints handle DISTRIBUTED_TRAINER_ENDPOINTS and PADDLE_TRAINERS
-        self.etcd.put(self.endpoints_path,
-                      six.b('{}|{}'.format(self.dist_endpoints, self.trainers)))
+        self.etcd.put(
+            self.endpoints_path,
+            '{}|{}'.format(self.dist_endpoints,
+                           self.trainers).encode('latin-1'))
 
         def endpoints_call_back(event):
             if not self.dist_endpoints:
@@ -433,7 +437,7 @@ class ElasticManager(object):
 
     def _update_endpoint(self, endpoints, hosts):
         self.etcd.put(self.endpoints_path,
-                      six.b('{}|{}'.format(endpoints, hosts)))
+                      '{}|{}'.format(endpoints, hosts).encode('latin-1'))
 
     def _update_fault_tolrance(self):
         rank = int(os.getenv('PADDLE_TRAINER_ID', -1))

python/paddle/distributed/fleet/launch.py

@@ -59,7 +59,6 @@ import sys
 import tempfile
 import os
 import time
-import six
 import copy
 import pathlib
 from argparse import ArgumentParser, REMAINDER
@@ -80,7 +79,7 @@ __all__ = []
 
 def _print_arguments(args):
     print("-----------  Configuration Arguments -----------")
-    for arg, value in sorted(six.iteritems(vars(args))):
+    for arg, value in sorted(vars(args).items()):
         print("%s: %s" % (arg, value))
     print("------------------------------------------------")
 

python/paddle/distributed/fleet/launch_utils.py

@@ -24,7 +24,6 @@ import shutil
 from contextlib import closing
 import multiprocessing
 import socket
-import six
 import struct
 import json
 
@@ -835,9 +834,7 @@ def get_device_proc_info(args):
                 "gpus' number:{} mod args.nproc_per_node:{} must == 0".format(len(gpus), args.nproc_per_node)
 
             n = int(len(gpus) / int(args.nproc_per_node))
-            devices_per_proc = [
-                gpus[i:i + n] for i in six.moves.range(0, len(gpus), n)
-            ]
+            devices_per_proc = [gpus[i:i + n] for i in range(0, len(gpus), n)]
         else:
             devices_per_proc = gpus
     elif device_mode == DeviceMode.ASCEND_NPU:
@@ -847,9 +844,7 @@ def get_device_proc_info(args):
                 "npus' number:{} mod args.nproc_per_node:{} must == 0".format(len(npus), args.nproc_per_node)
 
             n = int(len(npus) / int(args.nproc_per_node))
-            devices_per_proc = [
-                npus[i:i + n] for i in six.moves.range(0, len(npus), n)
-            ]
+            devices_per_proc = [npus[i:i + n] for i in range(0, len(npus), n)]
         else:
             devices_per_proc = npus
     elif device_mode == DeviceMode.XPU:
@@ -859,9 +854,7 @@ def get_device_proc_info(args):
                 "xpus' number:{} mod args.nproc_per_node:{} must == 0".format(len(xpus), args.nproc_per_node)
 
             n = int(len(xpus) / int(args.nproc_per_node))
-            devices_per_proc = [
-                xpus[i:i + n] for i in six.moves.range(0, len(xpus), n)
-            ]
+            devices_per_proc = [xpus[i:i + n] for i in range(0, len(xpus), n)]
         else:
             devices_per_proc = xpus
     elif device_mode == DeviceMode.MLU:
@@ -871,9 +864,7 @@ def get_device_proc_info(args):
                 "mlus' number:{} mod args.nproc_per_node:{} must == 0".format(len(mlus), args.nproc_per_node)
 
             n = int(len(mlus) / int(args.nproc_per_node))
-            devices_per_proc = [
-                mlus[i:i + n] for i in six.moves.range(0, len(mlus), n)
-            ]
+            devices_per_proc = [mlus[i:i + n] for i in range(0, len(mlus), n)]
         else:
             devices_per_proc = mlus
     elif device_mode == DeviceMode.CPU:

python/paddle/distributed/fleet/utils/fs.py

@@ -17,7 +17,6 @@ import multiprocessing
 
 import re
 import time
-import six
 import abc
 from paddle.fluid import core
 import functools
@@ -456,7 +455,7 @@ class HDFSClient(FS):
         self.pre_commands.append(dfs)
 
         if configs:
-            for k, v in six.iteritems(configs):
+            for k, v in configs.items():
                 config_command = '-D%s=%s' % (k, v)
                 self.pre_commands.append(config_command)
 

python/paddle/distributed/launch/context/__init__.py

@@ -17,7 +17,6 @@ from paddle.distributed.launch import plugins
 from .node import Node
 from .status import Status
 from .args_envs import parse_args, fetch_envs, env_args_mapping
-import six
 
 import logging
 
@@ -43,7 +42,7 @@ class Context(object):
 
     def print(self):
         self.logger.info("-----------  Configuration  ----------------------")
-        for arg, value in sorted(six.iteritems(vars(self.args))):
+        for arg, value in sorted(vars(self.args).items()):
             self.logger.info("%s: %s" % (arg, value))
         self.logger.info("--------------------------------------------------")
 

python/paddle/distributed/launch/controllers/master.py

@@ -204,7 +204,7 @@ class ETCDMaster(Master):
         self.ctx.logger.debug("sync path {} value {}".format(path, value))
 
         while not self.ctx.status.is_done():
-            self.client.put(path, six.b(value))
+            self.client.put(path, value.encode('latin-1'))
 
             result = [i for i in self.client.get_prefix(prefix)]
             result = copy.deepcopy(result)
@@ -244,7 +244,7 @@ class ETCDMaster(Master):
         #self.client.delete_prefix(self.job_prefix)
 
         beat_path = "{}/{}".format(self.heartbeat_prefix, pod_id)
-        self.client.put(beat_path, six.b(pod_id), lease=lease)
+        self.client.put(beat_path, pod_id.encode('latin-1'), lease=lease)
 
         def _beat_watch(event):
             self.ctx.status.restart()
@@ -257,7 +257,9 @@ class ETCDMaster(Master):
                 try:
                     lease.refresh()
                     if pod_id not in self.fetch_peer_alive():
-                        self.client.put(beat_path, six.b(pod_id), lease=lease)
+                        self.client.put(beat_path,
+                                        pod_id.encode('latin-1'),
+                                        lease=lease)
                         self.ctx.logger.debug("Heartbeat register again")
                 except Exception as e:
                     self.ctx.logger.error("Heartbeat error {}".format(e))
@@ -307,7 +309,8 @@ class ETCDMaster(Master):
 
     def set_status(self, status):
         assert self.client.put(
-            self.job_prefix, six.b(status),
+            self.job_prefix,
+            status.encode('latin-1'),
             lease=self.client.lease(600)), "set status failed {}".format(status)
 
     def get_status(self):

python/paddle/distributed/launch/plugins/__init__.py

@@ -12,7 +12,6 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-import six
 import os
 
 __all__ = []
@@ -21,7 +20,7 @@ __all__ = []
 # print configuration after args are well filled in controller init
 def log(ctx):
     ctx.logger.info("-----------  Configuration  ----------------------")
-    for arg, value in sorted(six.iteritems(vars(ctx.args))):
+    for arg, value in sorted(vars(ctx.args).items()):
         ctx.logger.info("%s: %s" % (arg, value))
     ctx.logger.info("--------------------------------------------------")
 

python/paddle/distributed/ps/utils/public.py

@@ -18,7 +18,6 @@ import collections
 import os
 import warnings
 import logging
-import six
 import paddle.fluid as fluid
 from paddle.fluid import core
 import paddle.fluid.framework as framework
@@ -1029,7 +1028,7 @@ def _get_output_map_from_op(varmap, op):
 
 def get_varlist_from_op_map(var_map):
     var_list = []
-    for key, varlist in six.iteritems(var_map):
+    for key, varlist in var_map.items():
         if not isinstance(varlist, list):
             varlist = [varlist]
         for i in range(len(varlist)):
@@ -1079,7 +1078,7 @@ def block_append_op(program, origin_program, block, op):
     merge_ordereddict = origin_program.global_block().vars.copy()
     merge_ordereddict.update(block.vars)
     inputs = _get_input_map_from_op(merge_ordereddict, op)
-    for key, varlist in six.iteritems(inputs):
+    for key, varlist in inputs.items():
         if not isinstance(varlist, list):
             varlist = [varlist]
         for var in varlist:
@@ -1092,7 +1091,7 @@ def block_append_op(program, origin_program, block, op):
                     block._clone_variable(var, force_persistable=False)
 
     outputs = _get_output_map_from_op(origin_program.global_block().vars, op)
-    for key, varlist in six.iteritems(outputs):
+    for key, varlist in outputs.items():
         if not isinstance(varlist, list):
             varlist = [varlist]
         for var in varlist:

python/paddle/distributed/spawn.py

@@ -15,7 +15,6 @@
 import multiprocessing
 import os
 import signal
-import six
 import sys
 import warnings
 
@@ -171,7 +170,7 @@ def _get_subprocess_env_list(nprocs, options):
         env_devices = os.getenv("CUDA_VISIBLE_DEVICES", None)
         if env_devices is None or env_devices == "":
             env_devices_list = [
-                str(x) for x in six.moves.range(core.get_cuda_device_count())
+                str(x) for x in range(core.get_cuda_device_count())
             ]
         else:
             env_devices_list = env_devices.split(',')
@@ -206,7 +205,7 @@ def _get_subprocess_env_list(nprocs, options):
         env_devices = os.getenv("XPU_VISIBLE_DEVICES", None)
         if env_devices is None or env_devices == "":
             env_devices_list = [
-                str(x) for x in six.moves.range(core.get_xpu_device_count())
+                str(x) for x in range(core.get_xpu_device_count())
             ]
         else:
             env_devices_list = env_devices.split(',')
@@ -240,7 +239,7 @@ def _get_subprocess_env_list(nprocs, options):
         env_devices = os.getenv("MLU_VISIBLE_DEVICES", None)
         if env_devices is None or env_devices == "":
             env_devices_list = [
-                str(x) for x in six.moves.range(core.get_mlu_device_count())
+                str(x) for x in range(core.get_mlu_device_count())
             ]
         else:
             env_devices_list = env_devices.split(',')

python/paddle/distributed/utils/launch_utils.py

@@ -21,7 +21,6 @@ import subprocess
 from contextlib import closing
 import socket
 from distutils.util import strtobool
-import six
 
 from paddle.distributed.fleet.launch_utils import get_backend_by_compile_flag
 from ..utils.log_utils import get_logger
@@ -542,6 +541,6 @@ def watch_local_trainers(procs, nranks):
 
 def _print_arguments(args):
     print("-----------  Configuration Arguments -----------")
-    for arg, value in sorted(six.iteritems(vars(args))):
+    for arg, value in sorted(vars(args).items()):
         print("%s: %s" % (arg, value))
     print("------------------------------------------------")

python/paddle/fluid/backward.py

@@ -19,7 +19,6 @@ from paddle.fluid import program_guard
 from . import core
 import collections
 import copy
-import six
 import logging
 from .. import compat as cpt
 from . import unique_name
@@ -337,20 +336,18 @@ def _create_op_desc_(op_type, inputs, outputs, attrs):
     """
     op_desc = core.OpDesc()
     op_desc.set_type(op_type)
-    for para, args in six.iteritems(inputs):
+    for para, args in inputs.items():
         op_desc.set_input(
             para,
             list(
-                map(
-                    lambda arg: arg.decode()
-                    if isinstance(arg, six.binary_type) else arg, args)))
-    for para, args in six.iteritems(outputs):
+                map(lambda arg: arg.decode()
+                    if isinstance(arg, bytes) else arg, args)))
+    for para, args in outputs.items():
         op_desc.set_output(
             para,
             list(
-                map(
-                    lambda arg: arg.decode()
-                    if isinstance(arg, six.binary_type) else arg, args)))
+                map(lambda arg: arg.decode()
+                    if isinstance(arg, bytes) else arg, args)))
 
     op_role_attr_name = core.op_proto_and_checker_maker.kOpRoleAttrName()
     op_device_attr_name = core.op_proto_and_checker_maker.kOpDeviceAttrName()
@@ -360,7 +357,7 @@ def _create_op_desc_(op_type, inputs, outputs, attrs):
             op_role_attr_name] = core.op_proto_and_checker_maker.OpRole.Backward
     if op_device_attr_name not in attrs:
         attrs[op_device_attr_name] = ""
-    for name, val in six.iteritems(attrs):
+    for name, val in attrs.items():
         if isinstance(val, framework.Block):
             op_desc.set_block_attr(name, val.desc)
         else:
@@ -604,7 +601,7 @@ def _addup_repetitive_outputs_(op_descs,
                     # record the latest device
                     var_device[var_name] = op_device
 
-    for var_name, inputs in six.iteritems(renamed_vars):
+    for var_name, inputs in renamed_vars.items():
         if len(renamed_vars[var_name]) > 1:
             if len(renamed_vars[var_name]) > _MAX_ADD_NUM_:
                 _accumulate_gradients_by_sum_op_(var_name, renamed_vars,
@@ -823,7 +820,7 @@ def _find_not_need_ops(grad_op_descs, forward_ops, input_grad_names_set):
 
 def serialize_op_decs(op_desc):
     protostr = op_desc.serialize_to_string()
-    proto = framework_pb2.OpDesc.FromString(six.binary_type(protostr))
+    proto = framework_pb2.OpDesc.FromString(bytes(protostr))
     return proto.__str__()
 
 
@@ -1375,9 +1372,9 @@ def _find_parent_op_(sub_block):
         return None
 
     program = sub_block.program
-    for block_id in six.moves.range(program.num_blocks):
+    for block_id in range(program.num_blocks):
         block_desc = program.block(block_id).desc
-        for op_idx in six.moves.range(block_desc.op_size()):
+        for op_idx in range(block_desc.op_size()):
             op = block_desc.op(op_idx)
             if op.has_attr("sub_block") and op._block_attr_id(
                     "sub_block") == sub_block_id:
@@ -1510,7 +1507,7 @@ def _rename_grad_(block, start_op_idx, grad_to_var, target_grad_map):
                 op_desc._rename_output(name, new_name)
                 var_map[name] = new_name
 
-    for g, ng in six.iteritems(var_map):
+    for g, ng in var_map.items():
         if g in grad_to_var:
             grad_to_var[ng] = grad_to_var[g]
             grad_to_var.pop(g)
@@ -1548,7 +1545,7 @@ def _get_no_grad_set_name(no_grad_set):
             for i, no_grad_var in enumerate(no_grad_set):
                 if isinstance(no_grad_var, framework.Variable):
                     no_grad_set_name.add(no_grad_var.name)
-                elif isinstance(no_grad_var, six.string_types):
+                elif isinstance(no_grad_var, str):
                     no_grad_set_name.add(no_grad_var)
                 else:
                     raise TypeError(
@@ -1825,11 +1822,11 @@ def append_backward(loss,
         parameters = []
         for i, param in enumerate(parameter_list):
             check_type(param, 'parameter_list[%s]' % i,
-                       (framework.Variable, six.string_types),
+                       (framework.Variable, str),
                        'fluid.backward.append_backward')
             if isinstance(param, framework.Variable):
                 parameters.append(param.name)
-            elif isinstance(param, six.string_types):
+            elif isinstance(param, str):
                 parameters.append(param)
     else:
         params = program.global_block().all_parameters()

python/paddle/fluid/clip.py

@@ -13,7 +13,6 @@
 # limitations under the License.
 
 import copy
-import six
 import warnings
 
 import functools
@@ -829,7 +828,7 @@ def set_gradient_clip(clip, param_list=None, program=None):
 
     if param_list is None:
         param_list = program.block(0).all_parameters()
-    if all(isinstance(elem, six.string_types) for elem in param_list):
+    if all(isinstance(elem, str) for elem in param_list):
         param_list = [program.block(0).var(elem) for elem in param_list]
     if not all(isinstance(elem, framework.Parameter) for elem in param_list):
         raise TypeError(

python/paddle/fluid/compiler.py

@@ -14,7 +14,6 @@
 
 import multiprocessing
 import os
-import six
 import sys
 import warnings
 from . import framework
@@ -430,11 +429,11 @@ class CompiledProgram(object):
         self._persistable_vars = list(set(self._persistable_vars))
         self._persistable_vars.sort()
 
-        return core.ParallelExecutor(
-            places, self._persistable_vars,
-            self._loss_name if self._loss_name else six.u(''), self._scope,
-            self._local_scopes, self._exec_strategy, self._build_strategy,
-            self._graph)
+        return core.ParallelExecutor(places, self._persistable_vars,
+                                     self._loss_name if self._loss_name else '',
+                                     self._scope, self._local_scopes,
+                                     self._exec_strategy, self._build_strategy,
+                                     self._graph)
 
     def _compile_inference(self):
         return core.create_paddle_predictor(self._infer_config)

python/paddle/fluid/contrib/decoder/beam_search_decoder.py

@@ -22,7 +22,6 @@ This API is still under active development and may change drastically.
 
 from ...wrapped_decorator import signature_safe_contextmanager
 import numpy as np
-import six
 
 from ... import layers
 from ...framework import Variable
@@ -197,7 +196,7 @@ class StateCell(object):
         self._helper = LayerHelper('state_cell', name=name)
         self._cur_states = {}
         self._state_names = []
-        for state_name, state in six.iteritems(states):
+        for state_name, state in states.items():
             if not isinstance(state, InitState):
                 raise ValueError('state must be an InitState object.')
             self._cur_states[state_name] = state
@@ -352,7 +351,7 @@ class StateCell(object):
         if self._in_decoder and not self._switched_decoder:
             self._switch_decoder()
 
-        for input_name, input_value in six.iteritems(inputs):
+        for input_name, input_value in inputs.items():
             if input_name not in self._inputs:
                 raise ValueError('Unknown input %s. '
                                  'Please make sure %s in input '
@@ -367,7 +366,7 @@ class StateCell(object):
         if self._in_decoder and not self._switched_decoder:
             self._switched_decoder()
 
-        for state_name, decoder_state in six.iteritems(self._states_holder):
+        for state_name, decoder_state in self._states_holder.items():
             if id(self._cur_decoder_obj) not in decoder_state:
                 raise ValueError('Unknown decoder object, please make sure '
                                  'switch_decoder been invoked.')
@@ -687,7 +686,7 @@ class BeamSearchDecoder(object):
             feed_dict = {}
             update_dict = {}
 
-            for init_var_name, init_var in six.iteritems(self._input_var_dict):
+            for init_var_name, init_var in self._input_var_dict.items():
                 if init_var_name not in self.state_cell._inputs:
                     raise ValueError('Variable ' + init_var_name +
                                      ' not found in StateCell!\n')
@@ -735,8 +734,7 @@ class BeamSearchDecoder(object):
                     self.state_cell.update_states()
                     self.update_array(prev_ids, selected_ids)
                     self.update_array(prev_scores, selected_scores)
-                    for update_name, var_to_update in six.iteritems(
-                            update_dict):
+                    for update_name, var_to_update in update_dict.items():
                         self.update_array(var_to_update, feed_dict[update_name])
 
     def read_array(self, init, is_ids=False, is_scores=False):

python/paddle/fluid/contrib/layers/nn.py

@@ -16,7 +16,6 @@ Contrib layers just related to the neural network.
 """
 
 import os
-import six
 import warnings
 import inspect
 

python/paddle/fluid/contrib/memory_usage_calc.py

@@ -20,8 +20,6 @@ batch size to fully utilize a GPU.
 This API is still under active development and may change drastically.
 """
 
-import six
-
 from .. import core
 from ..framework import Program, Variable
 

python/paddle/fluid/contrib/slim/quantization/imperative/ptq_config.py

@@ -12,7 +12,6 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-import six
 import abc
 import copy
 

python/paddle/fluid/contrib/slim/quantization/imperative/ptq_quantizer.py

@@ -98,8 +98,7 @@ def combine_abs_max_and_hist(tensor, origin_max, origin_hist, bins,
         return new_max, new_hist
 
 
-@six.add_metaclass(abc.ABCMeta)
-class BaseQuantizer(object):
+class BaseQuantizer(metaclass=abc.ABCMeta):
     """
     Base quantizer for activation and weight.
     """
@@ -173,8 +172,7 @@ class PerChannelAbsmaxQuantizer(BaseQuantizer):
         self.thresholds = self.abs_max_vals
 
 
-@six.add_metaclass(abc.ABCMeta)
-class BaseHistQuantizer(BaseQuantizer):
+class BaseHistQuantizer(BaseQuantizer, metaclass=abc.ABCMeta):
     """
     """
 

python/paddle/fluid/contrib/slim/tests/quant2_int8_image_classification_comparison.py

@@ -18,7 +18,6 @@ import sys
 import argparse
 import logging
 import struct
-import six
 import numpy as np
 import time
 import paddle

python/paddle/fluid/contrib/slim/tests/quant_int8_image_classification_comparison.py

@@ -18,7 +18,6 @@ import sys
 import argparse
 import logging
 import struct
-import six
 import numpy as np
 import time
 import paddle

python/paddle/fluid/contrib/slim/tests/test_graph.py

@@ -13,7 +13,6 @@
 # limitations under the license.
 
 import os
-import six
 import numpy as np
 import unittest
 import paddle

python/paddle/fluid/contrib/slim/tests/test_quantization_mkldnn_pass.py

@@ -17,7 +17,6 @@ import unittest
 import random
 import numpy as np
 import paddle.fluid as fluid
-import six
 import paddle
 from paddle.fluid.framework import IrGraph
 from paddle.fluid.contrib.slim.quantization import QuantizationFreezePass

python/paddle/fluid/contrib/slim/tests/test_quantization_pass.py

@@ -17,7 +17,6 @@ import unittest
 import random
 import numpy as np
 import paddle.fluid as fluid
-import six
 import paddle
 from paddle.fluid.framework import IrGraph
 from paddle.fluid.contrib.slim.quantization import QuantizationTransformPass
@@ -38,7 +37,7 @@ def linear_fc(num):
     data = fluid.layers.data(name='image', shape=[1, 32, 32], dtype='float32')
     label = fluid.layers.data(name='label', shape=[1], dtype='int64')
     hidden = data
-    for _ in six.moves.xrange(num):
+    for _ in range(num):
         hidden = fluid.layers.fc(hidden, size=128, act='relu')
     loss = fluid.layers.cross_entropy(input=hidden, label=label)
     loss = paddle.mean(loss)
@@ -72,7 +71,7 @@ def residual_block(num, quant_skip_pattern=None):
                               dtype='int64',
                               append_batch_size=False)
     hidden = data
-    for _ in six.moves.xrange(num):
+    for _ in range(num):
         conv = conv_bn_layer(hidden, 16, 3, 1, 1, act=None, bias_attr=True)
         short = conv_bn_layer(hidden, 16, 1, 1, 0, act=None)
         hidden = fluid.layers.elementwise_add(x=conv, y=short, act='relu')
@@ -579,7 +578,7 @@ def quant_dequant_residual_block(num, quant_skip_pattern=None):
                               dtype='float32')
     label = fluid.layers.data(name='label', shape=[1], dtype='int64')
     hidden = data1
-    for _ in six.moves.xrange(num):
+    for _ in range(num):
         conv = conv_bn_layer(hidden, 16, 3, 1, 1, act=None, bias_attr=True)
         short = conv_bn_layer(hidden, 16, 1, 1, 0, act=None)
         hidden = fluid.layers.elementwise_add(x=conv, y=short, act='relu')

python/paddle/fluid/contrib/slim/tests/test_quantization_scale_pass.py

@@ -16,7 +16,6 @@ import os
 import unittest
 import random
 import numpy as np
-import six
 import tempfile
 import paddle.fluid as fluid
 import paddle

python/paddle/fluid/contrib/slim/tests/test_quantize_transpiler_v2.py

@@ -16,7 +16,6 @@ import os
 import unittest
 import random
 import numpy as np
-import six
 import paddle.fluid as fluid
 import paddle
 from paddle.fluid.framework import IrGraph

python/paddle/fluid/contrib/slim/tests/test_user_defined_quantization.py

@@ -17,7 +17,6 @@ import unittest
 import json
 import random
 import numpy as np
-import six
 import tempfile
 import paddle.fluid as fluid
 import paddle

python/paddle/fluid/contrib/tests/test_quantize_transpiler.py

@@ -13,7 +13,6 @@
 # limitations under the license.
 
 import numpy as np
-import six
 
 import unittest
 import paddle
@@ -29,7 +28,7 @@ def linear_fc(num):
     data = fluid.layers.data(name='image', shape=[1, 32, 32], dtype='float32')
     label = fluid.layers.data(name='label', shape=[1], dtype='int64')
     hidden = data
-    for _ in six.moves.xrange(num):
+    for _ in range(num):
         hidden = fluid.layers.fc(hidden, size=128, act='relu')
     loss = fluid.layers.cross_entropy(input=hidden, label=label)
     loss = paddle.mean(loss)
@@ -57,7 +56,7 @@ def residual_block(num):
     data = fluid.layers.data(name='image', shape=[1, 32, 32], dtype='float32')
     label = fluid.layers.data(name='label', shape=[1], dtype='int64')
     hidden = data
-    for _ in six.moves.xrange(num):
+    for _ in range(num):
         conv = conv_bn_layer(hidden, 16, 3, 1, 1, act=None, bias_attr=True)
         short = conv_bn_layer(hidden, 16, 1, 1, 0, act=None)
         hidden = fluid.layers.elementwise_add(x=conv, y=short, act='relu')

python/paddle/fluid/contrib/tests/test_weight_decay_extend.py

@@ -12,7 +12,6 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-import six
 import unittest
 from functools import partial
 import numpy as np
@@ -32,7 +31,7 @@ def fake_imdb_reader(word_dict_size,
                      class_dim=2):
 
     def __reader__():
-        for _ in six.moves.range(sample_num):
+        for _ in range(sample_num):
             length = np.random.random_integers(low=lower_seq_len,
                                                high=upper_seq_len,
                                                size=[1])[0]

python/paddle/fluid/data.py

@@ -13,7 +13,6 @@
 # limitations under the License.
 
 import numpy as np
-import six
 
 from paddle.fluid import core
 from paddle.fluid.layer_helper import LayerHelper
@@ -107,11 +106,11 @@ def data(name, shape, dtype='float32', lod_level=0):
     """
     helper = LayerHelper('data', **locals())
 
-    check_type(name, 'name', (six.binary_type, six.text_type), 'data')
+    check_type(name, 'name', (bytes, str), 'data')
     check_type(shape, 'shape', (list, tuple), 'data')
 
     shape = list(shape)
-    for i in six.moves.range(len(shape)):
+    for i in range(len(shape)):
         if shape[i] is None:
             shape[i] = -1
 

python/paddle/fluid/data_feeder.py

@@ -15,8 +15,6 @@
 from . import core
 import numpy as np
 import os
-import six
-from six.moves import zip, range, xrange
 import multiprocessing
 import warnings
 
@@ -188,7 +186,7 @@ class DataToLoDTensorConverter(object):
 
     def _reset(self):
         self.data = []
-        self.lod = [[] for _ in six.moves.range(self.lod_level)]
+        self.lod = [[] for _ in range(self.lod_level)]
 
     def feed(self, data):
         self._feed_impl_(data, self.lod, self.lod_level)
@@ -249,8 +247,7 @@ class BatchedTensorProvider(object):
     def __call__(self):
         idx = 0
         for each_sample in self.generator():
-            for each_slot, each_converter in six.moves.zip(
-                    each_sample, self.converters):
+            for each_slot, each_converter in zip(each_sample, self.converters):
                 each_converter.data.append(each_slot)
 
             idx += 1
@@ -333,7 +330,7 @@ class DataFeeder(object):
         if program is None:
             program = default_main_program()
         for each_var in feed_list:
-            if isinstance(each_var, six.string_types):
+            if isinstance(each_var, str):
                 each_var = program.block(0).var(each_var)
             if not isinstance(each_var, Variable):
                 raise TypeError("Feed list should contain a list of variable")
@@ -383,9 +380,8 @@ class DataFeeder(object):
 
         """
         converter = []
-        for lod_level, shape, dtype in six.moves.zip(self.feed_lod_level,
-                                                     self.feed_shapes,
-                                                     self.feed_dtypes):
+        for lod_level, shape, dtype in zip(self.feed_lod_level,
+                                           self.feed_shapes, self.feed_dtypes):
             converter.append(
                 DataToLoDTensorConverter(place=self.place,
                                          lod_level=lod_level,
@@ -396,12 +392,10 @@ class DataFeeder(object):
             assert len(each_sample) == len(converter), (
                 "The number of fields in data (%d) does not match " +
                 "len(feed_list) (%d)") % (len(each_sample), len(converter))
-            for each_converter, each_slot in six.moves.zip(
-                    converter, each_sample):
+            for each_converter, each_slot in zip(converter, each_sample):
                 each_converter.feed(each_slot)
         ret_dict = {}
-        for each_name, each_converter in six.moves.zip(self.feed_names,
-                                                       converter):
+        for each_name, each_converter in zip(self.feed_names, converter):
             ret_dict[each_name] = each_converter.done()
         return ret_dict
 
@@ -461,13 +455,13 @@ class DataFeeder(object):
         """
         if isinstance(self.place, core.CUDAPlace):
             places = [
-                core.CUDAPlace(i) for i in six.moves.xrange(
-                    self._get_number_of_places_(num_places))
+                core.CUDAPlace(i)
+                for i in range(self._get_number_of_places_(num_places))
             ]
         else:
             places = [
-                core.CPUPlace() for _ in six.moves.xrange(
-                    self._get_number_of_places_(num_places))
+                core.CPUPlace()
+                for _ in range(self._get_number_of_places_(num_places))
             ]
 
         if len(iterable) != len(places):
@@ -477,7 +471,7 @@ class DataFeeder(object):
                              "must be same.")
 
         place = self.place
-        for p, batch in six.moves.zip(places, iterable):
+        for p, batch in zip(places, iterable):
             self.place = p
             yield self.feed(batch)
         self.place = place

python/paddle/fluid/debugger.py

@@ -13,7 +13,6 @@
 # limitations under the License.
 
 import sys
-import six
 import random
 import os
 import re
@@ -235,7 +234,7 @@ def draw_block_graphviz(block, highlights=None, path="./temp.dot"):
     graph = GraphPreviewGenerator("some graph")
     # collect parameters and args
     protostr = block.desc.serialize_to_string()
-    desc = framework_pb2.BlockDesc.FromString(six.binary_type(protostr))
+    desc = framework_pb2.BlockDesc.FromString(bytes(protostr))
 
     def need_highlight(name):
         if highlights is None: return False

python/paddle/fluid/dygraph/dygraph_to_static/convert_call_func.py

@@ -23,6 +23,7 @@ import types
 
 import numpy
 import six
+import builtins
 
 from paddle.fluid.dygraph.container import Sequential
 from paddle.fluid.dygraph.dygraph_to_static.convert_operators import convert_len, convert_zip
@@ -73,7 +74,7 @@ def is_builtin(func, name=None):
 
     if isinstance(func, types.BuiltinFunctionType) and name_judge():
         return True
-    elif func in six.moves.builtins.__dict__.values() and name_judge():
+    elif func in builtins.__dict__.values() and name_judge():
         return True
     else:
         return False

python/paddle/fluid/dygraph/dygraph_to_static/error.py

@@ -13,7 +13,6 @@
 # limitations under the License.
 
 import os
-import six
 import sys
 import traceback
 import linecache
@@ -380,14 +379,11 @@ class ErrorData(object):
             raise
 
         new_exception = self.create_exception()
-        if six.PY3:
-            # NOTE(liym27):
-            # 1. Why `raise new_exception from None`?
-            #   In Python 3, by default, an new exception is raised with trace information of the caught exception.
-            #   This only raises new_exception and hides unwanted implementation details from tracebacks of the
-            #   caught exception.
-            # 2. Use exec to bypass syntax error checking in Python 2.
-
-            six.exec_("raise new_exception from None")
-        else:
-            raise new_exception
+        # NOTE(liym27):
+        # Why `raise new_exception from None`?
+        #
+        # In Python 3, by default, an new exception is raised with trace information of the caught exception.
+        # This only raises new_exception and hides unwanted implementation details from tracebacks of the
+        # caught exception.
+
+        raise new_exception from None

python/paddle/fluid/dygraph/dygraph_to_static/function_spec.py

@@ -12,7 +12,6 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-import six
 import inspect
 import numpy as np
 import collections
@@ -83,7 +82,7 @@ class FunctionSpec(object):
 
         args = list(args)
 
-        for i in six.moves.range(len(args), len(self._arg_names)):
+        for i in range(len(args), len(self._arg_names)):
             arg_name = self._arg_names[i]
             if arg_name in kwargs:
                 args.append(kwargs[arg_name])
@@ -315,7 +314,7 @@ def convert_to_input_spec(inputs, input_spec):
     elif isinstance(input_spec, dict):
         input_with_spec = {}
         check_type_and_len(inputs, input_spec, True)
-        for name, input in six.iteritems(inputs):
+        for name, input in inputs.items():
             if name in input_spec:
                 input_with_spec[name] = convert_to_input_spec(
                     input, input_spec[name])
@@ -380,7 +379,7 @@ def _replace_spec_name(name, input_spec):
         return processed_specs
     elif isinstance(input_spec, dict):
         processed_specs = {}
-        for key, spec in six.iteritems(input_spec):
+        for key, spec in input_spec.items():
             processed_specs[key] = _replace_spec_name(key, spec)
         return processed_specs
     else:

python/paddle/fluid/dygraph/dygraph_to_static/ifelse_transformer.py

@@ -12,7 +12,6 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-import six
 import copy
 import textwrap
 from collections import defaultdict
@@ -270,7 +269,7 @@ class NameVisitor(gast.NodeVisitor):
         return False
 
     def _update_name_ids(self, new_name_ids):
-        for name_id, ctxs in six.iteritems(new_name_ids):
+        for name_id, ctxs in new_name_ids.items():
             self.name_ids[name_id] = ctxs + self.name_ids[name_id]
 
 

python/paddle/fluid/dygraph/dygraph_to_static/logging_utils.py

@@ -15,7 +15,6 @@
 import os
 import threading
 
-import six
 from paddle.fluid import log_helper
 from paddle.fluid.dygraph.dygraph_to_static.utils import ast_to_source_code
 
@@ -117,7 +116,7 @@ class TranslatorLogger(object):
         self._need_to_echo_code_to_stdout = code_to_stdout
 
     def check_level(self, level):
-        if isinstance(level, (six.integer_types, type(None))):
+        if isinstance(level, (int, type(None))):
             rv = level
         else:
             raise TypeError("Level is not an integer: {}".format(level))

python/paddle/fluid/dygraph/dygraph_to_static/partial_program.py

@@ -13,7 +13,6 @@
 # limitations under the License.
 
 import numpy as np
-import six
 
 import paddle
 from paddle.fluid import framework, backward, core, program_guard
@@ -873,7 +872,7 @@ class PartialProgramLayer:
             param_and_buffer_names_set.add(var.name)
 
         for block in main_program.blocks:
-            for name, var in six.iteritems(block.vars):
+            for name, var in block.vars.items():
                 if isinstance(var, framework.Parameter):
                     if name not in param_and_buffer_names_set:
                         raise ValueError(

python/paddle/fluid/dygraph/dygraph_to_static/program_translator.py

@@ -15,7 +15,6 @@
 import collections
 from paddle.utils import gast
 import inspect
-import six
 import textwrap
 import threading
 import weakref
@@ -983,7 +982,7 @@ class ProgramCache(object):
         return len(self._caches)
 
     def concrete_programs(self):
-        return [cp for key, (cp, _) in six.iteritems(self._caches)]
+        return [cp for key, (cp, _) in self._caches.items()]
 
 
 def synchronized(func):

python/paddle/fluid/dygraph/dygraph_to_static/utils.py

@@ -19,9 +19,9 @@ import copy
 import collections
 from paddle.utils import gast
 import inspect
-import os, sys
+import os
+import sys
 import shutil
-import six
 import tempfile
 import textwrap
 import numpy as np
@@ -135,7 +135,7 @@ def data_layer_not_check(name, shape, dtype='float32', lod_level=0):
     """
     helper = LayerHelper('data', **locals())
     shape = list(shape)
-    for i in six.moves.range(len(shape)):
+    for i in range(len(shape)):
         if shape[i] is None:
             shape[i] = -1
 
@@ -488,7 +488,7 @@ def generate_name_node(name_ids, ctx=gast.Load(), gen_tuple_if_single=False):
 
     This function is used at several gast.Return statements.
     """
-    if isinstance(name_ids, six.string_types):
+    if isinstance(name_ids, str):
         name_ids = [name_ids]
     if not isinstance(name_ids, (list, tuple, set)):
         raise TypeError(
@@ -640,7 +640,7 @@ def recover_globals_attribute(src_obj, dst_obj):
     src_globals = getattr(src_obj, attr_name, {})
     dst_globals = getattr(dst_obj, attr_name, {})
 
-    for k, v in six.iteritems(src_globals):
+    for k, v in src_globals.items():
         # ignore builtin attribute.
         if not (k.startswith('__') and k.endswith('__')):
             dst_globals[k] = v
@@ -889,7 +889,7 @@ class IsControlFlowVisitor(gast.NodeVisitor):
 
         # Look up the node_var_type_map by name_id.
         if self.node_var_type_map:
-            if name_id and isinstance(name_id, six.string_types):
+            if name_id and isinstance(name_id, str):
                 var_type = self.node_var_type_map.get(name_id, None)
                 if var_type and var_type & NodeVarType.TENSOR_TYPES:
                     return True

python/paddle/fluid/dygraph/dygraph_to_static/variable_trans_func.py

@@ -12,7 +12,6 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-import six
 import paddle
 import textwrap
 from paddle.utils import gast
@@ -59,7 +58,7 @@ def to_static_variable(x):
         return paddle.full(shape=[1], dtype='bool', fill_value=x)
     if isinstance(x, float):
         return paddle.full(shape=[1], dtype='float64', fill_value=x)
-    if isinstance(x, six.integer_types):
+    if isinstance(x, int):
         return paddle.full(shape=[1], dtype='int64', fill_value=x)
     if isinstance(x, UndefinedVar) or x is None:
         """

python/paddle/fluid/dygraph/io.py

@@ -13,7 +13,6 @@
 # limitations under the License.
 
 import os
-import six
 import pickle
 import numpy as np
 
@@ -69,17 +68,17 @@ def _is_persistable(var_desc):
 def _is_parameter(persistable_var_desc, program_desc):
     # 1. firstly, param should be input of op
     input_ops = []  # op can be repeated
-    for block_idx in six.moves.range(program_desc.num_blocks()):
+    for block_idx in range(program_desc.num_blocks()):
         block = program_desc.block(block_idx)
-        for op_idx in six.moves.range(block.op_size()):
+        for op_idx in range(block.op_size()):
             op = block.op(op_idx)
             # NOTE: parameter is the input of a certain op
             if persistable_var_desc.name() in op.input_arg_names():
                 input_ops.append(op)
     # 2. secondly, param should not be output of op or be same op's output
-    for block_idx in six.moves.range(program_desc.num_blocks()):
+    for block_idx in range(program_desc.num_blocks()):
         block = program_desc.block(block_idx)
-        for op_idx in six.moves.range(block.op_size()):
+        for op_idx in range(block.op_size()):
             op = block.op(op_idx)
             if persistable_var_desc.name() in op.output_arg_names():
                 # such as batch_norm_op
@@ -92,7 +91,7 @@ def _is_parameter(persistable_var_desc, program_desc):
 
 def _get_persistable_vars(program_desc):
     persistable_vars = []
-    for i in six.moves.range(program_desc.num_blocks()):
+    for i in range(program_desc.num_blocks()):
         block = program_desc.block(i)
         persistable_vars.extend(list(filter(_is_persistable, block.all_vars())))
     return persistable_vars
@@ -111,7 +110,7 @@ def _get_persistable_var_names(program_desc):
 
 def _get_all_var_names(program_desc):
     all_var_names = set()
-    for i in six.moves.range(program_desc.num_blocks()):
+    for i in range(program_desc.num_blocks()):
         block = program_desc.block(i)
         for var in block.all_vars():
             all_var_names.add(var.name())
@@ -142,10 +141,10 @@ def _append_loaded_suffix_to_var(program_desc):
         new_name = _append_loaded_suffix(var_desc.name())
         suffix_varname_dict[new_name] = old_name
         var_desc.set_name(new_name)
-        for block_idx in six.moves.range(program_desc.num_blocks()):
+        for block_idx in range(program_desc.num_blocks()):
             block = program_desc.block(block_idx)
             block._rename_var(old_name.encode(), new_name.encode())
-            for op_idx in six.moves.range(block.op_size()):
+            for op_idx in range(block.op_size()):
                 op = block.op(op_idx)
                 op._rename_input(old_name, new_name)
                 op._rename_output(old_name, new_name)
@@ -191,7 +190,7 @@ def _rename_var_program_desc(program_desc, include=None, exclude=None):
     dict_rename_var_new_old = dict()
     old_names = []
     # Store all old names
-    for b_idx in six.moves.range(program_desc.num_blocks()):
+    for b_idx in range(program_desc.num_blocks()):
         cur_block = program_desc.block(b_idx)
         for var in cur_block.all_vars():
             old_names.append(var.name())
@@ -199,7 +198,7 @@ def _rename_var_program_desc(program_desc, include=None, exclude=None):
     # Create dict_rename_var_new_old and dict_rename_var_old_new for non double
     # grad variables
     has_double_grad = False
-    for b_idx in six.moves.range(program_desc.num_blocks()):
+    for b_idx in range(program_desc.num_blocks()):
         cur_block = program_desc.block(b_idx)
         for var_idx, var in enumerate(cur_block.all_vars()):
             name_old = var.name()
@@ -232,7 +231,7 @@ def _rename_var_program_desc(program_desc, include=None, exclude=None):
     if has_double_grad:
         double_grad_rename_dict = {}
         for name_old in dict_rename_var_old_new:
-            for b_idx in six.moves.range(program_desc.num_blocks()):
+            for b_idx in range(program_desc.num_blocks()):
                 cur_block = program_desc.block(b_idx)
                 for var_idx, var in enumerate(cur_block.all_vars()):
                     var_name = var.name()
@@ -247,9 +246,9 @@ def _rename_var_program_desc(program_desc, include=None, exclude=None):
                 double_grad_rename_dict[var_name]] = var_name
 
     # Rename on program desc
-    for b_idx in six.moves.range(program_desc.num_blocks()):
+    for b_idx in range(program_desc.num_blocks()):
         cur_block = program_desc.block(b_idx)
-        for op_idx in six.moves.range(cur_block.op_size()):
+        for op_idx in range(cur_block.op_size()):
             op = cur_block.op(op_idx)
             for input_arg_name in op.input_arg_names():
                 if input_arg_name in dict_rename_var_old_new:
@@ -283,8 +282,7 @@ def _build_program_by_desc(program_desc):
     prog = framework.Program()
     prog.desc = program_desc
     prog.blocks = [
-        framework.Block(prog, i)
-        for i in six.moves.range(prog.desc.num_blocks())
+        framework.Block(prog, i) for i in range(prog.desc.num_blocks())
     ]
     prog._sync_with_cpp()
     return prog
@@ -292,9 +290,9 @@ def _build_program_by_desc(program_desc):
 
 def _change_is_test_status(program_desc, is_test):
     # change all `is_test` attributes
-    for i in six.moves.range(program_desc.num_blocks()):
+    for i in range(program_desc.num_blocks()):
         block = program_desc.block(i)
-        for j in six.moves.range(block.op_size()):
+        for j in range(block.op_size()):
             op = block.op(j)
             if op.has_attr('is_test'):
                 op._set_attr('is_test', is_test)
@@ -407,7 +405,7 @@ class _ProgramHolder(object):
         # remove feed, fetch and scale-1 op, remove op_callstack attr
         ops_to_remove = []
         root_block = program_desc.block(0)
-        for i in six.moves.range(root_block.op_size()):
+        for i in range(root_block.op_size()):
             op = root_block.op(i)
             if op.type() == 'feed':
                 ops_to_remove.append(i)
@@ -511,7 +509,7 @@ class _ProgramHolder(object):
         program = _build_program_by_desc(program_desc_copy)
         # 3. Add the outputs which is only used for training and not saved in
         # inference program.
-        for block_idx in six.moves.range(program.num_blocks):
+        for block_idx in range(program.num_blocks):
             block = program.block(block_idx)
             for op in block.ops:
                 if op.type == "batch_norm":

python/paddle/fluid/dygraph/jit.py

@@ -22,7 +22,6 @@ import inspect
 import threading
 from typing import Text, Tuple, Any, List
 
-import six
 import paddle
 from paddle.fluid import core, dygraph
 from paddle.fluid.compiler import BuildStrategy, CompiledProgram, ExecutionStrategy
@@ -338,7 +337,7 @@ class _SaveLoadConfig(object):
     def model_filename(self, filename):
         if filename is None:
             return
-        if not isinstance(filename, six.string_types):
+        if not isinstance(filename, str):
             raise TypeError(
                 "The config `model_filename` should be str, but received input's type is %s."
                 % type(filename))
@@ -354,7 +353,7 @@ class _SaveLoadConfig(object):
     def params_filename(self, filename):
         if filename is None:
             return
-        if not isinstance(filename, six.string_types):
+        if not isinstance(filename, str):
             raise TypeError(
                 "The config `params_filename` should be str, but received input's type is %s."
                 % type(filename))
@@ -985,7 +984,7 @@ def save(layer, path, input_spec=None, **configs):
             # we only record the state_dict variable's structured name
             state_names_dict = dict()
             state_var_dict = dict()
-            for structured_name, var in six.iteritems(dygraph_state_dict):
+            for structured_name, var in dygraph_state_dict.items():
                 state_names_dict[var.name] = structured_name
                 state_var_dict[var.name] = var
 

python/paddle/fluid/dygraph/layer_hooks.py

@@ -12,7 +12,6 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-import six
 import warnings
 
 from paddle.fluid.framework import default_main_program, _non_static_mode
@@ -64,7 +63,7 @@ def set_op_customized_attrs_post_hook(layer, inputs, outputs):
         layer._op_recorder.ops = ops
 
         for op in ops:
-            for attr_name, val in six.iteritems(layer._customized_attrs):
+            for attr_name, val in layer._customized_attrs.items():
                 op._set_attr(attr_name, val)
 
         # remove pre-hook and post-hook

python/paddle/fluid/dygraph/layer_object_helper.py

@@ -13,11 +13,10 @@
 # limitations under the License.
 
 import copy
-import six
 from ..framework import Parameter, _non_static_mode, _global_flags
 from ..param_attr import ParamAttr
 from .. import core
-from six.moves import zip
+
 from ..layer_helper_base import LayerHelperBase
 from ..dygraph_utils import _append_activation_in_dygraph
 
@@ -78,7 +77,7 @@ class LayerObjectHelper(LayerHelperBase):
                 self.name))
         elif len(param_attr) == 1 and length != 1:
             tmp = [None] * length
-            for i in six.moves.range(length):
+            for i in range(length):
                 tmp[i] = copy.deepcopy(param_attr[0])
             param_attr = tmp
         return param_attr
@@ -150,7 +149,7 @@ class LayerObjectHelper(LayerHelperBase):
         act = act
         if act is None:
             return input_var
-        if isinstance(act, six.string_types):
+        if isinstance(act, str):
             act = {'type': act}
         else:
             raise TypeError(

python/paddle/fluid/dygraph/layers.py

@@ -16,7 +16,6 @@ import collections
 import contextlib
 import sys
 import numpy as np
-import six
 import re
 import copy
 import weakref
@@ -441,7 +440,7 @@ class Layer(object):
 
         """
         temp_attr = copy.deepcopy(attr)
-        if isinstance(temp_attr, six.string_types) and temp_attr == "":
+        if isinstance(temp_attr, str) and temp_attr == "":
             temp_attr = None
         return self._helper.create_parameter(temp_attr, shape, dtype, is_bias,
                                              default_initializer)
@@ -770,7 +769,7 @@ class Layer(object):
         if '_buffers' not in self.__dict__:
             raise ValueError(
                 "super(YourLayer, self).__init__() should be called first")
-        elif not isinstance(name, six.string_types):
+        elif not isinstance(name, str):
             raise TypeError(
                 "The name of buffer should be a string, but received {}.".
                 format(type(name).__name__))
@@ -1038,7 +1037,7 @@ class Layer(object):
         if '_parameters' not in self.__dict__:
             raise RuntimeError(
                 "super(YourLayer, self).__init__() should be called firstly.")
-        elif not isinstance(name, six.string_types):
+        elif not isinstance(name, str):
             raise TypeError(
                 "The name of parameter should be a string, but received {}.".
                 format(type(name).__name__))

python/paddle/fluid/dygraph/nn.py

@@ -13,7 +13,6 @@
 # limitations under the License.
 
 import paddle
-from six.moves import reduce
 from .. import core
 from ..layers import utils
 from ..layers import nn as F

python/paddle/fluid/dygraph/parallel.py

@@ -13,7 +13,6 @@
 # limitations under the License.
 
 import os
-import six
 import numpy as np
 import warnings
 from collections import OrderedDict

python/paddle/fluid/dygraph/tracer.py

@@ -12,8 +12,6 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-import six
-
 from collections import defaultdict
 from paddle.fluid import core
 from paddle.fluid import framework

python/paddle/fluid/executor.py

@@ -384,8 +384,8 @@ def _add_feed_fetch_ops(program,
                                fetch_op):
         for i, var in enumerate(fetch_list):
             assert isinstance(var, Variable) or isinstance(
-                var, six.string_types), ("Wrong type for fetch_list[%s]: %s" %
-                                         (i, type(var)))
+                var, str), ("Wrong type for fetch_list[%s]: %s" %
+                            (i, type(var)))
             global_block.append_op(type=fetch_op,
                                    inputs={'X': [var]},
                                    outputs={'Out': [fetch_var]},
@@ -429,7 +429,7 @@ def _fetch_var(name, scope=None, return_numpy=True):
     Returns:
        LodTensor|numpy.ndarray
     """
-    assert isinstance(name, six.string_types)
+    assert isinstance(name, str)
     if scope is None:
         scope = global_scope()
     assert isinstance(scope, core._Scope)
@@ -452,7 +452,7 @@ def _to_name_str(var):
             return var.desc.name()
         elif isinstance(var, str):
             return var
-        elif isinstance(var, six.string_types):
+        elif isinstance(var, str):
             return str(var)
         elif isinstance(var, Operator):
             return str(id(var))
@@ -987,7 +987,7 @@ class Executor(object):
     def _fetch_data(self, fetch_list, fetch_var_name, scope):
         outs = [
             core.get_fetch_variable(scope, fetch_var_name, i)
-            for i in six.moves.range(len(fetch_list))
+            for i in range(len(fetch_list))
         ]
         return outs
 
@@ -1017,7 +1017,7 @@ class Executor(object):
                     raise TypeError(
                         "The operator in fetch_list is not an optimize_op")
             elif isinstance(item, Variable) or isinstance(
-                    item, str) or isinstance(item, six.string_types):
+                    item, str) or isinstance(item, str):
                 _fetch_list.append(item)
             else:
                 raise TypeError(
@@ -1808,8 +1808,7 @@ class Executor(object):
         return exe.run(feed)
 
     def _check_fetch_list(self, fetch_list):
-        is_fetch_var = lambda var: isinstance(var,
-                                              (Variable, str, six.string_types))
+        is_fetch_var = lambda var: isinstance(var, (Variable, str))
         is_tuple_list = lambda var: isinstance(var, (tuple, list))
 
         if fetch_list is None: return []
@@ -2404,9 +2403,8 @@ class Executor(object):
                                    fetch_op):
             for i, var in enumerate(fetch_list):
                 assert isinstance(var, Variable) or isinstance(
-                    var,
-                    six.string_types), ("Wrong type for fetch_list[%s]: %s" %
-                                        (i, type(var)))
+                    var, str), ("Wrong type for fetch_list[%s]: %s" %
+                                (i, type(var)))
                 global_block.append_op(type=fetch_op,
                                        inputs={'X': [var]},
                                        outputs={'Out': [fetch_var]},

python/paddle/fluid/framework.py

@@ -20,7 +20,6 @@ from .wrapped_decorator import signature_safe_contextmanager, wrap_decorator
 import os
 import re
 import traceback
-import six
 import copy
 from types import MethodType, FunctionType
 
@@ -448,7 +447,7 @@ def require_version(min_version, max_version=None):
     zero_version = ['0', '0', '0', '0']
 
     def version_cmp(ver_a, ver_b):
-        for i in six.moves.range(len(ver_a)):
+        for i in range(len(ver_a)):
             if int(ver_a[i]) > int(ver_b[i]):
                 return 1
             elif int(ver_a[i]) < int(ver_b[i]):
@@ -688,7 +687,7 @@ def _cuda_ids():
     if gpus_env:
         device_ids = [int(s) for s in gpus_env.split(",")]
     else:
-        device_ids = six.moves.range(core.get_cuda_device_count())
+        device_ids = range(core.get_cuda_device_count())
     return device_ids
 
 
@@ -697,7 +696,7 @@ def _xpu_ids():
     if xpus_env:
         device_ids = [int(s) for s in xpus_env.split(",")]
     else:
-        device_ids = six.moves.range(core.get_xpu_device_count())
+        device_ids = range(core.get_xpu_device_count())
     return device_ids
 
 
@@ -706,7 +705,7 @@ def _npu_ids():
     if npus_env:
         device_ids = [int(s) for s in npus_env.split(",")]
     else:
-        device_ids = six.moves.range(core.get_npu_device_count())
+        device_ids = range(core.get_npu_device_count())
     return device_ids
 
 
@@ -715,7 +714,7 @@ def _mlu_ids():
     if mlus_env:
         device_ids = [int(s) for s in mlus_env.split(",")]
     else:
-        device_ids = six.moves.range(core.get_mlu_device_count())
+        device_ids = range(core.get_mlu_device_count())
     return device_ids
 
 
@@ -1322,8 +1321,7 @@ class ParameterMetaClass(VariableMetaClass):
             return issubclass(t, Parameter)
 
 
-@six.add_metaclass(VariableMetaClass)
-class Variable(object):
+class Variable(metaclass=VariableMetaClass):
     """
     **Notes**:
         **The constructor of Variable should not be invoked directly.**
@@ -1763,7 +1761,7 @@ class Variable(object):
         assert isinstance(throw_on_error, bool) and isinstance(
             with_details, bool)
         protostr = self.desc.serialize_to_string()
-        proto = framework_pb2.VarDesc.FromString(six.binary_type(protostr))
+        proto = framework_pb2.VarDesc.FromString(bytes(protostr))
         res_str = _debug_string_(proto, throw_on_error)
         if with_details:
             additional_attr = ("error_clip", )
@@ -2579,7 +2577,7 @@ def get_all_op_protos():
     protostrs = core.get_all_op_protos()
     ret_values = []
     for pbstr in protostrs:
-        op_proto = framework_pb2.OpProto.FromString(six.binary_type(pbstr))
+        op_proto = framework_pb2.OpProto.FromString(bytes(pbstr))
         ret_values.append(op_proto)
     return ret_values
 
@@ -2812,9 +2810,9 @@ class Operator(object):
                                 % (in_proto.name, len(in_args)))
                         in_arg_names = []
                         for index, arg in enumerate(in_args):
-                            if isinstance(arg, six.string_types):
+                            if isinstance(arg, str):
                                 in_arg_names.append(arg)
-                            elif isinstance(arg, six.binary_type):
+                            elif isinstance(arg, bytes):
                                 in_arg_names.append(arg.decode())
                             elif isinstance(arg, (Variable, core.VarBase)):
                                 in_arg_names.append(arg.name)
@@ -2850,13 +2848,13 @@ class Operator(object):
                             % (out_proto.name, len(out_args)))
                     out_arg_names = []
                     for arg in out_args:
-                        if isinstance(arg, six.string_types):
+                        if isinstance(arg, str):
                             out_arg_names.append(arg)
                         else:
                             out_arg_names.append(arg.name)
                         # TODO(minqiyang): could we remove variable's op in static mode?
                         if not _non_static_mode():
-                            if isinstance(arg, six.string_types):
+                            if isinstance(arg, str):
                                 block.var(arg).op = self
                             else:
                                 arg.op = self
@@ -2911,7 +2909,7 @@ class Operator(object):
 
         """
         protostr = self.desc.serialize_to_string()
-        proto = framework_pb2.OpDesc.FromString(six.binary_type(protostr))
+        proto = framework_pb2.OpDesc.FromString(bytes(protostr))
         return _debug_string_(proto, throw_on_error)
 
     def _to_readable_code(self, skip_op_callstack=True):
@@ -3528,8 +3526,7 @@ class Block(object):
             res_str += "\n}"
         else:
             protostr = self.desc.serialize_to_string()
-            proto = framework_pb2.BlockDesc.FromString(
-                six.binary_type(protostr))
+            proto = framework_pb2.BlockDesc.FromString(bytes(protostr))
             res_str = _debug_string_(proto, throw_on_error)
         return res_str
 
@@ -3581,7 +3578,7 @@ class Block(object):
         Returns:
             Variable: the Variable with the giving name.
         """
-        if not isinstance(name, six.string_types):
+        if not isinstance(name, str):
             raise TypeError(
                 "var require string as parameter, but get %s instead." %
                 (type(name)))
@@ -3650,7 +3647,7 @@ class Block(object):
         return list(self.iter_parameters())
 
     def iter_parameters(self):
-        return (item[1] for item in six.iteritems(self.vars)
+        return (item[1] for item in self.vars.items()
                 if isinstance(item[1], Parameter))
 
     def create_var(self, *args, **kwargs):
@@ -4736,14 +4733,14 @@ class IrGraph(object):
         """
         op_desc = core.OpDesc()
         op_desc.set_type(op_type)
-        for attr, value in six.iteritems(attrs):
+        for attr, value in attrs.items():
             self._update_desc_attr(op_desc, attr, value)
-        for input_name, var_nodes in six.iteritems(inputs):
+        for input_name, var_nodes in inputs.items():
             if not isinstance(var_nodes, list):
                 var_nodes = [var_nodes]
             op_desc.set_input(input_name,
                               [var_node.name() for var_node in var_nodes])
-        for output_name, var_nodes in six.iteritems(outputs):
+        for output_name, var_nodes in outputs.items():
             if not isinstance(var_nodes, list):
                 var_nodes = [var_nodes]
             op_desc.set_output(output_name,
@@ -4889,7 +4886,7 @@ class IrGraph(object):
         """
         adj_list = core.build_adjacency_list(self.graph)
         wrapped_adj_list = dict()
-        for k, v in six.iteritems(adj_list):
+        for k, v in adj_list.items():
             wrapped_adj_list[IrNode(k)] = {IrNode(n) for n in v}
         return wrapped_adj_list
 
@@ -5461,8 +5458,7 @@ class Program(object):
                 res_str += block.to_string(throw_on_error, with_details)
         else:
             protostr = self.desc.serialize_to_string()
-            proto = framework_pb2.ProgramDesc.FromString(
-                six.binary_type(protostr))
+            proto = framework_pb2.ProgramDesc.FromString(bytes(protostr))
             res_str = _debug_string_(proto, throw_on_error)
         return res_str
 
@@ -5545,7 +5541,7 @@ class Program(object):
                         print("op type is {}".format(op.type))
                         print("op inputs are {}".format(op.input_arg_names))
                         print("op outputs are {}".format(op.output_arg_names))
-                        for key, value in sorted(six.iteritems(op.all_attrs())):
+                        for key, value in sorted(op.all_attrs().items()):
                             if key not in ['op_callstack', 'op_role_var']:
                                 print(" [ attrs: {}:   {} ]".format(key, value))
 
@@ -5553,7 +5549,6 @@ class Program(object):
             1. To clone a test program, the sample code is:
                 .. code-block:: python
 
-                    import six
                     import paddle
                     import paddle.static as static
                     import paddle.utils as utils
@@ -5562,13 +5557,13 @@ class Program(object):
                     paddle.enable_static()
 
                     def print_prog(prog):
-                        for name, value in sorted(six.iteritems(prog.block(0).vars)):
+                        for name, value in sorted(prog.block(0).vars.items()):
                             print(value)
                         for op in prog.block(0).ops:
                             print("op type is {}".format(op.type))
                             print("op inputs are {}".format(op.input_arg_names))
                             print("op outputs are {}".format(op.output_arg_names))
-                            for key, value in sorted(six.iteritems(op.all_attrs())):
+                            for key, value in sorted(op.all_attrs().items()):
                                 if key not in ['op_callstack', 'op_role_var']:
                                     print(" [ attrs: {}:   {} ]".format(key, value))
 
@@ -5606,7 +5601,6 @@ class Program(object):
             2. The clone method can be avoid if you create program for training and program for testing individually.
                 .. code-block:: python
 
-                    import six
                     import paddle
                     import paddle.static as static
                     import paddle.utils as utils
@@ -5615,13 +5609,13 @@ class Program(object):
                     paddle.enable_static()
 
                     def print_prog(prog):
-                        for name, value in sorted(six.iteritems(prog.block(0).vars)):
+                        for name, value in sorted(prog.block(0).vars.items()):
                             print(value)
                         for op in prog.block(0).ops:
                             print("op type is {}".format(op.type))
                             print("op inputs are {}".format(op.input_arg_names))
                             print("op outputs are {}".format(op.output_arg_names))
-                            for key, value in sorted(six.iteritems(op.all_attrs())):
+                            for key, value in sorted(op.all_attrs().items()):
                                 if key not in ['op_callstack', 'op_role_var']:
                                     print(" [ attrs: {}:   {} ]".format(key, value))
 
@@ -5663,7 +5657,7 @@ class Program(object):
                 self.desc)
             forward_prog.blocks = [
                 Block(forward_prog, i)
-                for i in six.moves.range(forward_prog.desc.num_blocks())
+                for i in range(forward_prog.desc.num_blocks())
             ]
             forward_prog._sync_with_cpp()
             p = forward_prog._inference_optimize(prune_read_op=False)
@@ -5672,9 +5666,7 @@ class Program(object):
             p.current_block_idx = self.current_block_idx
             p._seed = self._seed
             p.desc = core.ProgramDesc(self.desc)
-            p.blocks = [
-                Block(p, i) for i in six.moves.range(self.desc.num_blocks())
-            ]
+            p.blocks = [Block(p, i) for i in range(self.desc.num_blocks())]
 
             p._current_role = self._current_role
             p.__op_role_var = self.__op_role_var
@@ -5737,7 +5729,7 @@ class Program(object):
             targets = [targets]
 
         for var in feeded_var_names:
-            if not isinstance(var, six.string_types):
+            if not isinstance(var, str):
                 raise ValueError(
                     "All feeded_var_names of Program._prune_with_input() can only be "
                     "str, but received %s." % type(var))
@@ -5763,7 +5755,7 @@ class Program(object):
             if not isinstance(t, Operator):
                 if isinstance(t, Variable):
                     name = t.name
-                elif isinstance(t, six.string_types):
+                elif isinstance(t, str):
                     name = str(t)
                 else:
                     raise ValueError(
@@ -5803,9 +5795,7 @@ class Program(object):
         res = Program()
         res.desc, pruned_origin_block_id_map = core.prune(
             self.desc, set(feeded_var_names), targets_idx)
-        res.blocks = [
-            Block(res, i) for i in six.moves.range(res.desc.num_blocks())
-        ]
+        res.blocks = [Block(res, i) for i in range(res.desc.num_blocks())]
         res._sync_with_cpp()
 
         res._copy_param_info_from(self)
@@ -5854,18 +5844,16 @@ class Program(object):
                     root_block._remove_var(var.name().encode())
 
         # change all `is_test` attributes to True
-        for i in six.moves.range(res.desc.num_blocks()):
+        for i in range(res.desc.num_blocks()):
             block = res.desc.block(i)
-            for j in six.moves.range(block.op_size()):
+            for j in range(block.op_size()):
                 op = block.op(j)
                 if op.has_attr('is_test'):
                     op._set_attr('is_test', True)
                 if op.type() == "batch_norm":
                     # Remove the output ReserveSpace of batch_norm if exists.
                     op.remove_output("ReserveSpace")
-        res.blocks = [
-            Block(res, i) for i in six.moves.range(res.desc.num_blocks())
-        ]
+        res.blocks = [Block(res, i) for i in range(res.desc.num_blocks())]
         res._sync_with_cpp()
         return res
 
@@ -5884,16 +5872,14 @@ class Program(object):
         res = Program()
         res.desc = core.ProgramDesc(self.desc)
 
-        res.blocks = [
-            Block(res, i) for i in six.moves.range(res.desc.num_blocks())
-        ]
+        res.blocks = [Block(res, i) for i in range(res.desc.num_blocks())]
         res._sync_with_cpp()
 
         # Note: The op_role and op_role_var cann't be deleted currently,
         # and we will try to remove them in the future.
         common_clipped_attrs_list = ['op_callstack', 'with_quant_attr']
 
-        for i in six.moves.range(res.desc.num_blocks()):
+        for i in range(res.desc.num_blocks()):
             block = res.desc.block(i)
             for var in block.all_vars():
                 var.clear_is_parameter()
@@ -6016,7 +6002,7 @@ class Program(object):
         """
         p = Program()
         p.desc = core.ProgramDesc(binary_str)
-        p.blocks = [Block(p, i) for i in six.moves.range(p.desc.num_blocks())]
+        p.blocks = [Block(p, i) for i in range(p.desc.num_blocks())]
         p._sync_with_cpp()
         return p
 
@@ -6033,7 +6019,7 @@ class Program(object):
         """
         p = Program()
         p.desc = desc
-        p.blocks = [Block(p, i) for i in six.moves.range(p.desc.num_blocks())]
+        p.blocks = [Block(p, i) for i in range(p.desc.num_blocks())]
         p._sync_with_cpp()
         return p
 
@@ -6304,7 +6290,7 @@ class Program(object):
         if not pruned_origin_block_id_map:
             pruned_origin_block_id_map = {
                 i: i
-                for i in six.moves.range(self.desc.num_blocks())
+                for i in range(self.desc.num_blocks())
             }
 
         # NOTE(zhiqiu): All vars in cloned program exist in original program.
@@ -6556,8 +6542,7 @@ class Program(object):
                      .format(name)))
 
 
-@six.add_metaclass(ParameterMetaClass)
-class Parameter(Variable):
+class Parameter(Variable, metaclass=ParameterMetaClass):
     """
     Parameter is derived from Variable. A parameter is a persistable
     Variable, and will be updated by optimizers after each iteration.

python/paddle/fluid/graphviz.py

@@ -14,14 +14,13 @@
 
 import os
 import random
-import six
 import functools
 import subprocess
 import logging
 
 
 def crepr(v):
-    if isinstance(v, six.string_types):
+    if isinstance(v, str):
         return '"%s"' % v
     return str(v)
 
@@ -104,7 +103,7 @@ class Graph(object):
                          stderr=subprocess.PIPE)
 
     def _rank_repr(self):
-        ranks = sorted(six.iteritems(self.rank_groups),
+        ranks = sorted(self.rank_groups.items(),
                        key=functools.cmp_to_key(
                            lambda a, b: a[1].priority > b[1].priority))
         repr = []
@@ -149,7 +148,7 @@ class Node(object):
             name=self.name,
             label=self.label,
             extra=',' + ','.join("%s=%s" % (key, crepr(value))
-                                 for key, value in six.iteritems(self.attrs))
+                                 for key, value in self.attrs.items())
             if self.attrs else "")
         return reprs
 
@@ -174,7 +173,7 @@ class Edge(object):
             target=self.target.name,
             extra="" if not self.attrs else "[" +
             ','.join("{}={}".format(attr[0], crepr(attr[1]))
-                     for attr in six.iteritems(self.attrs)) + "]")
+                     for attr in self.attrs.items()) + "]")
         return repr
 
 

python/paddle/fluid/incubate/checkpoint/auto_checkpoint.py

@@ -17,7 +17,6 @@ import logging
 import hashlib
 import json
 import os
-import six
 import time
 import collections
 from threading import Thread, current_thread
@@ -405,7 +404,7 @@ class TrainEpochRange(SerializableBase):
         # registerd exes
         d["exe_status"] = {}
         e = d["exe_status"]
-        for k, t in six.iteritems(self._exe_status):
+        for k, t in self._exe_status.items():
             e[t._key] = t._serialize()
         return json.dumps(d)
 
@@ -427,7 +426,7 @@ class TrainEpochRange(SerializableBase):
 
         # exes status
         e = d["exe_status"]
-        for k, v in six.iteritems(e):
+        for k, v in e.items():
             t = ExeTrainStatus()
             t._deserialize(v)
             self._exe_status[k] = t
@@ -480,7 +479,7 @@ class TrainEpochRange(SerializableBase):
             return
 
         e = self._exe_status
-        for k, t in six.iteritems(self._exe_status):
+        for k, t in self._exe_status.items():
             m = PaddleModel(t._exe, t._program)
             p = self._checker.get_exe_checkpoint_path(t._hash_key)
             t._epoch_no = self.get()

python/paddle/fluid/incubate/fleet/collective/__init__.py

@@ -32,7 +32,6 @@ import paddle
 
 import os
 import sys
-import six
 import json
 import re
 import shutil
@@ -285,7 +284,7 @@ class CollectiveOptimizer(DistributedOptimizer):
         return self._optimizer.apply_gradients(params_grads)
 
     def _check_condition(self, name, **kwargs):
-        for k, v in six.iteritems(kwargs):
+        for k, v in kwargs.items():
             if v is True:
                 assert False, "you can't use %s and %s together" % (name, k)
 

python/paddle/fluid/incubate/fleet/parameter_server/ir/pserver_pass.py

@@ -13,7 +13,6 @@
 # limitations under the License.
 
 import collections
-import six
 
 from paddle.fluid import core
 from paddle.fluid.framework import Block
@@ -95,7 +94,7 @@ def _append_pserver_non_opt_ops(optimize_block, opt_op, origin_program, config):
         """
 
         grad_block = None
-        for _, g in six.iteritems(var_dict):
+        for _, g in var_dict.items():
             if _orig_varname(g.name) == _orig_varname(var.name):
                 # skip per trainer vars
                 if g.name.find(".trainer_") == -1:
@@ -113,7 +112,7 @@ def _append_pserver_non_opt_ops(optimize_block, opt_op, origin_program, config):
     program = optimize_block.program
     # Append the ops for parameters that do not need to be optimized / updated
     inputs = _get_input_map_from_op(origin_program.global_block().vars, opt_op)
-    for key, varlist in six.iteritems(inputs):
+    for key, varlist in inputs.items():
         if not isinstance(varlist, list):
             varlist = [varlist]
         for i in range(len(varlist)):
@@ -134,7 +133,7 @@ def _append_pserver_non_opt_ops(optimize_block, opt_op, origin_program, config):
 
     outputs = _get_output_map_from_op(origin_program.global_block().vars,
                                       opt_op)
-    for key, varlist in six.iteritems(outputs):
+    for key, varlist in outputs.items():
         if not isinstance(varlist, list):
             varlist = [varlist]
         for i in range(len(varlist)):
@@ -913,7 +912,7 @@ def build_pserver_startup_program_pass(program, p_main_program, config):
     pserver_vars = p_main_program.global_block().vars
 
     created_var_map = collections.OrderedDict()
-    for _, var in six.iteritems(pserver_vars):
+    for _, var in pserver_vars.items():
         tmpvar = program.global_block()._clone_variable(var)
         created_var_map[var.name] = tmpvar
 

python/paddle/fluid/incubate/fleet/parameter_server/ir/public.py

@@ -779,7 +779,7 @@ class CompileTimeStrategy(object):
                 block_map[varname] = []
             block_map[varname].append((int(offset), int(size)))
 
-        for varname, split in six.iteritems(block_map):
+        for varname, split in block_map.items():
             orig_var = self.merged_variable_map[varname]
 
             if len(split) == 1:
@@ -829,7 +829,7 @@ class CompileTimeStrategy(object):
     def _dispatcher(self):
         ps_dispatcher = RoundRobin(self.get_ps_endpoints())
         ps_dispatcher.reset()
-        grad_var_mapping_items = list(six.iteritems(self.grad_var_mapping))
+        grad_var_mapping_items = list(self.grad_var_mapping.items())
 
         sparse_gradnames = [grad.name for _, grad in self.origin_sparse_pairs]
 

python/paddle/fluid/incubate/fleet/parameter_server/ir/trainer_pass.py

@@ -14,7 +14,6 @@
 # limitations under the License.
 
 import os
-import six
 import collections
 import warnings
 import math
@@ -1834,7 +1833,7 @@ def block_append_op(program, origin_program, block, op):
     merge_ordereddict = origin_program.global_block().vars.copy()
     merge_ordereddict.update(block.vars)
     inputs = _get_input_map_from_op(merge_ordereddict, op)
-    for key, varlist in six.iteritems(inputs):
+    for key, varlist in inputs.items():
         if not isinstance(varlist, list):
             varlist = [varlist]
         for var in varlist:
@@ -1847,7 +1846,7 @@ def block_append_op(program, origin_program, block, op):
                     block._clone_variable(var, force_persistable=False)
 
     outputs = _get_output_map_from_op(origin_program.global_block().vars, op)
-    for key, varlist in six.iteritems(outputs):
+    for key, varlist in outputs.items():
         if not isinstance(varlist, list):
             varlist = [varlist]
         for var in varlist:
@@ -1898,7 +1897,7 @@ def add_vars_by_var_list(var_name_list, origin_program, program, block):
 
 def get_varlist_from_op_map(var_map):
     var_list = []
-    for key, varlist in six.iteritems(var_map):
+    for key, varlist in var_map.items():
         if not isinstance(varlist, list):
             varlist = [varlist]
         for i in range(len(varlist)):

python/paddle/fluid/incubate/fleet/utils/hdfs.py

@@ -24,7 +24,6 @@ import copy
 import errno
 import time
 import logging
-import six
 #from . import fs
 from paddle.distributed.fleet.utils.fs import FS, LocalFS, FSFileExistsError, FSFileNotExistsError, ExecuteError, FSTimeOut, FSShellCmdAborted
 from paddle.fluid import core
@@ -91,7 +90,7 @@ class HDFSClient(FS):
         self.pre_commands.append(dfs)
 
         if configs:
-            for k, v in six.iteritems(configs):
+            for k, v in configs.items():
                 config_command = '-D%s=%s' % (k, v)
                 self.pre_commands.append(config_command)
 

python/paddle/fluid/io.py

@@ -15,7 +15,6 @@
 import os
 import errno
 import warnings
-import six
 import logging
 import pickle
 import contextlib
@@ -1310,14 +1309,14 @@ def save_inference_model(dirname,
             # "./infer_model".
 
     """
-    if isinstance(feeded_var_names, six.string_types):
+    if isinstance(feeded_var_names, str):
         feeded_var_names = [feeded_var_names]
     elif export_for_deployment:
         if len(feeded_var_names) > 0:
             # TODO(paddle-dev): polish these code blocks
-            if not (bool(feeded_var_names) and all(
-                    isinstance(name, six.string_types)
-                    for name in feeded_var_names)):
+            if not (bool(feeded_var_names)
+                    and all(isinstance(name, str)
+                            for name in feeded_var_names)):
                 raise ValueError("'feed_var_names' should be a list of str.")
 
     if isinstance(target_vars, Variable):

python/paddle/fluid/layer_helper.py

@@ -13,14 +13,13 @@
 # limitations under the License.
 
 import copy
-import six
 
 from .framework import Parameter, dtype_is_floating, _non_static_mode, OpProtoHolder, _global_flags
 from . import unique_name
 from paddle.fluid.initializer import Constant, Xavier
 from .param_attr import ParamAttr
 from . import core
-from six.moves import zip
+
 from .layer_helper_base import LayerHelperBase
 from .dygraph_utils import _append_activation_in_dygraph
 
@@ -76,7 +75,7 @@ class LayerHelper(LayerHelperBase):
             raise ValueError("parameter number mismatch")
         elif len(param_attr) == 1 and length != 1:
             tmp = [None] * length
-            for i in six.moves.range(length):
+            for i in range(length):
                 tmp[i] = copy.deepcopy(param_attr[0])
             param_attr = tmp
         return param_attr
@@ -143,7 +142,7 @@ class LayerHelper(LayerHelperBase):
         act = self.kwargs.get('act', None)
         if act is None:
             return input_var
-        if isinstance(act, six.string_types):
+        if isinstance(act, str):
             act = {'type': act}
         else:
             raise TypeError(str(act) + " should be unicode or str")

python/paddle/fluid/layers/control_flow.py

@@ -23,7 +23,6 @@ from .nn import logical_and, logical_not, logical_or
 from .utils import assert_same_structure, map_structure, hold_mutable_vars, copy_mutable_vars, padding_to_same_structure, is_sequence, pack_sequence_as, flatten, to_sequence
 import numpy
 import warnings
-import six
 from functools import reduce, partial
 from ..data_feeder import convert_dtype, check_variable_and_dtype, check_type, check_dtype
 from ... import compat as cpt
@@ -143,7 +142,7 @@ def select_input_with_buildin_type(inputs, mask, name):
             raise RuntimeError(
                 f"Exceptions throwed while doing select_input on {name}:\n{e}")
 
-    elif (isinstance(false_var, (support_ret_buildin_type))
+    elif (isinstance(false_var, support_ret_buildin_type)
           and isinstance(false_var, type(true_var))):
         if false_var == true_var:
             return false_var
@@ -962,7 +961,7 @@ class StaticRNN(object):
         boot_memories = []
         pre_memories = []
         memories = []
-        for _, mem in six.iteritems(self.memories):
+        for _, mem in self.memories.items():
             boot_memories.append(mem.init)
             pre_memories.append(mem.pre_mem.name)
             assert mem.mem is not None, "%s should be updated in every step." % (
@@ -1205,7 +1204,7 @@ class While(object):
             })
 
 
-support_ret_buildin_type = (bool, float, six.integer_types)
+support_ret_buildin_type = (bool, float, int)
 
 
 def assign_skip_lod_tensor_array(input, output):

python/paddle/fluid/layers/detection.py

@@ -29,7 +29,6 @@ from . import ops
 from ... import compat as cpt
 from ..data_feeder import check_variable_and_dtype, check_type, check_dtype
 import math
-import six
 import numpy as np
 from functools import reduce
 from ..data_feeder import convert_dtype, check_variable_and_dtype, check_type, check_dtype
@@ -2199,7 +2198,7 @@ def multi_box_head(inputs,
                   min_sizes = []
                   max_sizes = []
                   step = int(math.floor(((max_ratio - min_ratio)) / (num_layer - 2)))
-                  for ratio in six.moves.range(min_ratio, max_ratio + 1, step):
+                  for ratio in range(min_ratio, max_ratio + 1, step):
                       min_sizes.append(base_size * ratio / 100.)
                       max_sizes.append(base_size * (ratio + step) / 100.)
                       min_sizes = [base_size * .10] + min_sizes
@@ -2338,7 +2337,7 @@ def multi_box_head(inputs,
         min_sizes = []
         max_sizes = []
         step = int(math.floor(((max_ratio - min_ratio)) / (num_layer - 2)))
-        for ratio in six.moves.range(min_ratio, max_ratio + 1, step):
+        for ratio in range(min_ratio, max_ratio + 1, step):
             min_sizes.append(base_size * ratio / 100.)
             max_sizes.append(base_size * (ratio + step) / 100.)
         min_sizes = [base_size * .10] + min_sizes

python/paddle/fluid/layers/io.py

@@ -110,11 +110,11 @@ def data(name,
     """
     helper = LayerHelper('data', **locals())
 
-    check_type(name, 'name', (six.binary_type, six.text_type), 'data')
+    check_type(name, 'name', (bytes, str), 'data')
     check_type(shape, 'shape', (list, tuple), 'data')
 
     shape = list(shape)
-    for i in six.moves.range(len(shape)):
+    for i in range(len(shape)):
         if shape[i] is None:
             shape[i] = -1
             append_batch_size = False

python/paddle/fluid/layers/layer_function_generator.py

@@ -17,7 +17,7 @@ import functools
 import warnings
 import string
 
-from six.moves import cStringIO
+from io import StringIO
 from ..proto import framework_pb2
 from ..framework import OpProtoHolder, Variable, core, convert_np_dtype_to_dtype_, _non_static_mode, in_dygraph_mode, _in_legacy_dygraph
 from ..layer_helper import LayerHelper
@@ -80,7 +80,7 @@ def _generate_doc_string_(op_proto,
     if not isinstance(op_proto, framework_pb2.OpProto):
         raise TypeError("OpProto should be `framework_pb2.OpProto`")
 
-    buf = cStringIO()
+    buf = StringIO()
     buf.write(escape_math(op_proto.comment))
     buf.write('\nArgs:\n')
     for each_input in op_proto.inputs:

python/paddle/fluid/layers/nn.py

@@ -20,7 +20,6 @@ import inspect
 import warnings
 
 import numpy as np
-import six
 
 import paddle
 from ..layer_helper import LayerHelper
@@ -6336,7 +6335,7 @@ def one_hot(input, depth, allow_out_of_range=False):
 
     helper = LayerHelper("one_hot", **locals())
     check_variable_and_dtype(input, 'input', ['int32', 'int64'], 'one_hot')
-    check_type(depth, 'depth', (six.integer_types, Variable), 'one_hot')
+    check_type(depth, 'depth', (int, Variable), 'one_hot')
     one_hot_out = helper.create_variable_for_type_inference(dtype='float32')
 
     if not isinstance(depth, Variable):

python/paddle/fluid/layers/rnn.py

@@ -151,8 +151,7 @@ class RNNCell(object):
             # TODO: Add check for the illegal
             if isinstance(seq, dict):
                 return True
-            return (isinstance(seq, Sequence)
-                    and not isinstance(seq, six.string_types))
+            return (isinstance(seq, Sequence) and not isinstance(seq, str))
 
         class Shape(object):
 

python/paddle/fluid/layers/utils.py

@@ -82,7 +82,7 @@ def is_sequence(seq):
     """
     if isinstance(seq, dict):
         return True
-    return (isinstance(seq, Sequence) and not isinstance(seq, six.string_types))
+    return (isinstance(seq, Sequence) and not isinstance(seq, str))
 
 
 def _hash_with_id(*args):
@@ -99,7 +99,7 @@ def _sorted(dict_):
     Returns a sorted list of the dict keys, with error if keys not sortable.
     """
     try:
-        return sorted(six.iterkeys(dict_))
+        return sorted(dict_.keys())
     except TypeError:
         raise TypeError("nest only supports dicts with sortable keys.")
 
@@ -159,11 +159,10 @@ def _sequence_like(instance, args):
         # ordered and plain dicts (e.g., flattening a dict but using a
         # corresponding `OrderedDict` to pack it back).
         result = dict(zip(_sorted(instance), args))
-        return type(instance)(
-            (key, result[key]) for key in six.iterkeys(instance))
+        return type(instance)((key, result[key]) for key in instance.keys())
     elif (isinstance(instance, tuple) and hasattr(instance, "_fields")
           and isinstance(instance._fields, Sequence)
-          and all(isinstance(f, six.string_types) for f in instance._fields)):
+          and all(isinstance(f, str) for f in instance._fields)):
         # This is a namedtuple
         return type(instance)(*args)
     else:
@@ -256,8 +255,8 @@ def _recursive_assert_same_structure(nest1, nest2, check_types):
                 "structure has type %s, while second structure has type %s." %
                 (type_nest1, type_nest2))
         if isinstance(nest1, dict):
-            keys1 = set(six.iterkeys(nest1))
-            keys2 = set(six.iterkeys(nest2))
+            keys1 = set(nest1.keys())
+            keys2 = set(nest2.keys())
             if keys1 != keys2:
                 raise ValueError(
                     "The two dictionaries don't have the same set of keys. First "
@@ -382,7 +381,7 @@ def _convert_to_tensor_list(old_list, dtype="int32"):
             ele.stop_gradient = True
             new_list_tensor.append(ele)
         else:
-            assert isinstance(ele, six.integer_types)
+            assert isinstance(ele, int)
             temp_out = fill_constant([1], dtype, ele, force_cpu=True)
             new_list_tensor.append(temp_out)
     return new_list_tensor
@@ -414,7 +413,7 @@ def check_shape(shape):
                     raise ValueError(
                         "All elements in ``shape`` must be positive when it's a list or tuple"
                     )
-                if not isinstance(ele, six.integer_types):
+                if not isinstance(ele, int):
                     raise TypeError(
                         "All elements in ``shape`` must be integers when it's a list or tuple"
                     )

python/paddle/fluid/metrics.py

@@ -17,8 +17,6 @@ Fluid Metrics
 
 import numpy as np
 import copy
-import warnings
-import six
 
 from .layer_helper import LayerHelper
 from .initializer import Constant
@@ -118,10 +116,9 @@ class MetricBase(object):
         """
         states = {
             attr: value
-            for attr, value in six.iteritems(self.__dict__)
-            if not attr.startswith("_")
+            for attr, value in self.__dict__.items() if not attr.startswith("_")
         }
-        for attr, value in six.iteritems(states):
+        for attr, value in states.items():
             if isinstance(value, int):
                 setattr(self, attr, 0)
             elif isinstance(value, float):
@@ -147,8 +144,7 @@ class MetricBase(object):
         """
         states = {
             attr: value
-            for attr, value in six.iteritems(self.__dict__)
-            if not attr.startswith("_")
+            for attr, value in self.__dict__.items() if not attr.startswith("_")
         }
         config = {}
         config.update({"name": self._name, "states": copy.deepcopy(states)})

python/paddle/fluid/nets.py

@@ -12,7 +12,6 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-import six
 from . import layers
 from .data_feeder import check_variable_and_dtype, convert_dtype
 from ..utils import deprecated
@@ -227,7 +226,7 @@ def img_conv_group(input,
     conv_with_batchnorm = __extend_list__(conv_with_batchnorm)
     conv_batchnorm_drop_rate = __extend_list__(conv_batchnorm_drop_rate)
 
-    for i in six.moves.range(len(conv_num_filter)):
+    for i in range(len(conv_num_filter)):
         local_conv_act = conv_act
         if conv_with_batchnorm[i]:
             local_conv_act = None

python/paddle/fluid/op.py

@@ -13,7 +13,6 @@
 # limitations under the License.
 
 import numpy as np
-import six
 
 import paddle.fluid.core as core
 import paddle.fluid.proto.framework_pb2 as framework_pb2
@@ -27,13 +26,13 @@ def get_all_op_protos():
     protostrs = core.get_all_op_protos()
     ret_values = []
     for pbstr in protostrs:
-        op_proto = framework_pb2.OpProto.FromString(six.binary_type(pbstr))
+        op_proto = framework_pb2.OpProto.FromString(bytes(pbstr))
         ret_values.append(op_proto)
     return ret_values
 
 
 def is_str(s):
-    return isinstance(s, six.string_types)
+    return isinstance(s, str)
 
 
 class OpDescCreationMethod(object):

python/paddle/fluid/optimizer.py

@@ -13,7 +13,6 @@
 # limitations under the License.
 
 import numpy as np
-import six
 import os
 import logging
 from collections import defaultdict
@@ -6227,8 +6226,7 @@ class RecomputeOptimizer(Optimizer):
         ), "_checkpoints should be a list of Variable or a list of String"
         for ckpt in checkpoints:
             assert (
-                isinstance(ckpt, six.string_types)
-                or isinstance(ckpt, Variable)
+                isinstance(ckpt, str) or isinstance(ckpt, Variable)
             ), "_checkpoints should be a list of Variable or a list of String"
         self._checkpoints = checkpoints
 

python/paddle/fluid/param_attr.py

@@ -12,10 +12,6 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-import six
-import warnings
-import sys
-
 from .initializer import Initializer, Xavier, Constant
 from .regularizer import WeightDecayRegularizer
 from paddle.fluid.data_feeder import check_type
@@ -172,7 +168,7 @@ class ParamAttr(object):
             return [ParamAttr._to_attr(a) for a in arg]
         elif isinstance(arg, ParamAttr):
             return arg
-        elif isinstance(arg, six.string_types):
+        elif isinstance(arg, str):
             return ParamAttr(name=arg)
         elif isinstance(arg, Initializer):
             return ParamAttr(initializer=arg)

python/paddle/fluid/profiler.py

@@ -15,7 +15,6 @@
 from . import core
 from .wrapped_decorator import signature_safe_contextmanager
 import os
-import six
 import sys
 
 from paddle.utils.deprecated import deprecated

python/paddle/fluid/tests/unittests/benchmark.py

@@ -14,7 +14,6 @@
 
 import numpy as np
 import time
-import six
 
 from paddle.fluid.op import Operator
 from op_test import OpTest
@@ -38,8 +37,7 @@ class BenchmarkSuite(OpTest):
             expect_t = np.array(item_cpu_out)
             actual = item_gpu_out
             actual_t = np.array(item_gpu_out)
-            var_name = variable if isinstance(
-                variable, six.string_types) else variable.name
+            var_name = variable if isinstance(variable, str) else variable.name
             np.testing.assert_allclose(actual_t,
                                        expect_t,
                                        rtol=1e-05,
@@ -49,7 +47,7 @@ class BenchmarkSuite(OpTest):
 
     def _get_input_names(self):
         inputs = []
-        for name, value in six.iteritems(self.inputs):
+        for name, value in self.inputs.items():
             if isinstance(value, list):
                 inputs.extend([sub_name for sub_name, _ in value])
             inputs.append(name)
@@ -57,7 +55,7 @@ class BenchmarkSuite(OpTest):
 
     def _get_output_names(self):
         outputs = []
-        for var_name, var in six.iteritems(self.outputs):
+        for var_name, var in self.outputs.items():
             if isinstance(var, list):
                 for sub_var_name, sub_var in var:
                     outputs.append(sub_var_name)

python/paddle/fluid/tests/unittests/collective/fleet/parallel_dygraph_transformer.py

@@ -13,7 +13,6 @@
 # limitations under the License.
 
 import numpy as np
-import six
 
 import paddle
 import paddle.fluid as fluid
@@ -824,7 +823,7 @@ def fake_data_reader():
 
     def __reader__():
         iteration = TrainTaskConfig.batch_size * batch_num
-        for _ in six.moves.range(iteration):
+        for _ in range(iteration):
             # random data
             np.random.seed = 90
             src_word_np = np.arange(1, seq_len + 1).reshape([seq_len

python/paddle/fluid/tests/unittests/dist_save_load.py

@@ -17,7 +17,6 @@ import sys
 
 import numpy as np
 import pickle
-import six
 
 import paddle
 import paddle.fluid as fluid
@@ -149,7 +148,7 @@ class TestDistSaveLoad2x2(TestDistSimnetBow2x2):
 
         if save_mode == "LOCAL":
             if need_save:
-                for _ in six.moves.xrange(RUN_STEP):
+                for _ in range(RUN_STEP):
                     loss, = exe.run(fetch_list=[avg_cost.name],
                                     feed=feeder.feed(get_data()))
                 if need_save and model_dir:
@@ -163,14 +162,14 @@ class TestDistSaveLoad2x2(TestDistSimnetBow2x2):
             skip_steps = int(os.getenv("SKIP_STEPS"))
             loss = None
             if need_save:
-                for idx in six.moves.xrange(8):
+                for idx in range(8):
                     loss, = exe.run(fetch_list=[avg_cost.name],
                                     feed=feeder.feed(get_data()))
                     if need_save and model_dir and idx == skip_steps and args.trainer_id == 0:
                         io.save_persistables(startup_exe, model_dir,
                                              trainer_prog)
             else:
-                for idx in six.moves.xrange(8):
+                for idx in range(8):
                     data = get_data()
                     if idx <= skip_steps:
                         continue

python/paddle/fluid/tests/unittests/dist_text_classification.py

@@ -15,7 +15,6 @@
 import paddle
 import paddle.fluid as fluid
 import os
-import six
 import tarfile
 import string
 import re
@@ -154,9 +153,8 @@ def tokenize(pattern):
         while tf != None:
             if bool(pattern.match(tf.name)):
                 # newline and punctuations removal and ad-hoc tokenization.
-                yield tarf.extractfile(tf).read().rstrip(
-                    six.b("\n\r")).translate(None, six.b(
-                        string.punctuation)).lower().split()
+                yield tarf.extractfile(tf).read().rstrip(b'\n\r').translate(
+                    None, string.punctuation.encode('latin-1')).lower().split()
             tf = tarf.next()
 
 

python/paddle/fluid/tests/unittests/dist_transformer.py

@@ -15,7 +15,7 @@
 import numpy as np
 import time
 import os
-import six
+import functools
 import time
 from functools import partial
 from os.path import expanduser
@@ -306,7 +306,7 @@ def pad_batch_data(insts,
     """
     return_list = []
     max_len = max(len(inst) for inst in insts)
-    num_token = six.moves.reduce(lambda x, y: x + y,
+    num_token = functools.reduce(lambda x, y: x + y,
                                  [len(inst)
                                   for inst in insts]) if return_num_token else 0
     # Any token included in dict can be used to pad, since the paddings' loss
@@ -547,7 +547,7 @@ def train_loop(exe, train_progm, dev_count, sum_cost, avg_cost, lr_scheduler,
                         np.log(TrainTaskConfig.label_smooth_eps /
                                (ModelHyperParams.trg_vocab_size - 1) + 1e-20))
     init = False
-    for pass_id in six.moves.xrange(TrainTaskConfig.pass_num):
+    for pass_id in range(TrainTaskConfig.pass_num):
         pass_start_time = time.time()
         for batch_id, data in enumerate(train_data()):
             if batch_id >= RUN_STEP:

python/paddle/fluid/tests/unittests/dygraph_to_static/test_dict.py

@@ -12,7 +12,6 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-import six
 import numpy as np
 import unittest
 
@@ -106,7 +105,7 @@ class MainNetWithDict(fluid.dygraph.Layer):
 
 # Test to call function defined outside of class.
 def update_cache(cache):
-    for k, val in six.iteritems(cache):
+    for k, val in cache.items():
         cache[k] = fluid.layers.softmax(val)
 
     return cache

python/paddle/fluid/tests/unittests/dygraph_to_static/test_op_attr.py

@@ -12,7 +12,6 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-import six
 import paddle
 import unittest
 
@@ -105,8 +104,7 @@ class CheckOpAttr(unittest.TestCase):
             ops = cur_block.ops
             for op in ops:
                 if op.type not in self.infos: continue
-                for attr_name, expect_vals in six.iteritems(
-                        self.infos[op.type]):
+                for attr_name, expect_vals in self.infos[op.type].items():
                     op_vals = op.desc.attr(attr_name)
                     if not isinstance(expect_vals, list):
                         expect_vals = [expect_vals]

python/paddle/fluid/tests/unittests/dygraph_to_static/transformer_util.py

@@ -14,7 +14,6 @@
 
 import pickle
 import warnings
-import six
 from functools import partial
 import numpy as np
 
@@ -271,11 +270,10 @@ def load(program, model_path, executor=None, var_list=None):
             "An UnicodeDecodeError is catched, which might be caused by loading "
             "a python2 saved model. Encoding of pickle.load would be set and "
             "load again automatically.")
-        if six.PY3:
-            load_bak = pickle.load
-            pickle.load = partial(load_bak, encoding="latin1")
-            fluid.load(program, model_path, executor, var_list)
-            pickle.load = load_bak
+        load_bak = pickle.load
+        pickle.load = partial(load_bak, encoding="latin1")
+        fluid.load(program, model_path, executor, var_list)
+        pickle.load = load_bak
 
 
 def load_dygraph(model_path, keep_name_table=False):
@@ -291,10 +289,9 @@ def load_dygraph(model_path, keep_name_table=False):
             "An UnicodeDecodeError is catched, which might be caused by loading "
             "a python2 saved model. Encoding of pickle.load would be set and "
             "load again automatically.")
-        if six.PY3:
-            load_bak = pickle.load
-            pickle.load = partial(load_bak, encoding="latin1")
-            para_dict, opti_dict = fluid.load_dygraph(
-                model_path, keep_name_table=keep_name_table)
-            pickle.load = load_bak
-            return para_dict, opti_dict
+        load_bak = pickle.load
+        pickle.load = partial(load_bak, encoding="latin1")
+        para_dict, opti_dict = fluid.load_dygraph(
+            model_path, keep_name_table=keep_name_table)
+        pickle.load = load_bak
+        return para_dict, opti_dict

python/paddle/fluid/tests/unittests/fake_reader.py

@@ -13,7 +13,6 @@
 # limitations under the License.
 
 import numpy as np
-import six
 
 
 def fake_imdb_reader(word_dict_size,
@@ -23,7 +22,7 @@ def fake_imdb_reader(word_dict_size,
                      class_dim=2):
 
     def __reader__():
-        for _ in six.moves.range(sample_num):
+        for _ in range(sample_num):
             length = np.random.random_integers(low=lower_seq_len,
                                                high=upper_seq_len,
                                                size=[1])[0]

python/paddle/fluid/tests/unittests/feed_data_reader.py

@@ -12,7 +12,6 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-import six
 import paddle.fluid as fluid
 from paddle.fluid.framework import Variable
 
@@ -44,13 +43,13 @@ class FeedDataReader(object):
         next_data = next(self._iter)
         feed_data = dict()
         assert len(self._feed_list) == len(next_data)
-        for key, value in six.moves.zip(self._feed_list, next_data):
+        for key, value in zip(self._feed_list, next_data):
             feed_data[key] = value
         return feed_data
 
     def _feed_parallel_executor(self, device_num):
         feed_data = []
-        for _ in six.moves.range(device_num):
+        for _ in range(device_num):
             feed_data.append(self._feed_executor())
 
         return feed_data

python/paddle/fluid/tests/unittests/gradient_checker.py

@@ -13,7 +13,6 @@
 # limitations under the License.
 """This is the lib for gradient checker unittest."""
 
-import six
 import numpy as np
 from itertools import product
 import paddle
@@ -139,7 +138,7 @@ def _compute_numerical_jacobian(program, x, y, place, scope, delta):
     np_type = dtype_to_np_dtype(x.dtype)
     jacobian = [make_jacobian(x, _product(yi.shape), np_type) for yi in y]
 
-    for i in six.moves.xrange(x_size):
+    for i in range(x_size):
         orig = _get_item(x_t, i, np_type)
         x_pos = orig + delta
         _set_item(x_t, i, x_pos, np_type)
@@ -151,7 +150,7 @@ def _compute_numerical_jacobian(program, x, y, place, scope, delta):
 
         _set_item(x_t, i, orig, np_type)
 
-        for j in six.moves.xrange(len(y)):
+        for j in range(len(y)):
             jacobian[j][i, :] = (y_pos[j] - y_neg[j]) / delta / 2.
 
     return jacobian
@@ -204,12 +203,12 @@ def _compute_analytical_jacobian(program, x, y, place, scope):
     filted = [(i, dxi) for i, dxi in enumerate(dx) if dxi is not None]
     filted_idx, filted_dx = zip(*filted)
 
-    for i in six.moves.xrange(y_size):
+    for i in range(y_size):
         _set_item(dy_t, i, 1, np_type)
 
         dx_res = exe.run(program, scope=scope, fetch_list=filted_dx)
 
-        for j in six.moves.xrange(len(filted_dx)):
+        for j in range(len(filted_dx)):
             dx_idx = filted_idx[j]
             if dx_res[j] is not None:
                 jacobian[dx_idx][:, i] = dx_res[j].flatten()
@@ -551,7 +550,7 @@ def get_static_double_grad(x,
         program = fluid.default_main_program()
     scope = fluid.executor.global_scope()
     y_grads = []
-    for i in six.moves.xrange(len(y)):
+    for i in range(len(y)):
         yi = y[i]
         dyi_name = _append_grad_suffix_(yi.name)
         np_type = dtype_to_np_dtype(yi.dtype)
@@ -768,7 +767,7 @@ def double_grad_check_for_dygraph(func,
             "please check the python api unit test used."
         raise RuntimeError(msg)
 
-    for i in six.moves.xrange(len(static_double_grad)):
+    for i in range(len(static_double_grad)):
         if not np.allclose(static_double_grad[i], eager_double_grad[i], rtol,
                            atol):
             msg = 'Check eager double result fail. Mismatch between static_graph double grad ' \
@@ -802,7 +801,7 @@ def get_static_triple_grad(x,
         program = fluid.default_main_program()
     scope = fluid.executor.global_scope()
     y_grads = []
-    for i in six.moves.xrange(len(y)):
+    for i in range(len(y)):
         yi = y[i]
         dyi_name = _append_grad_suffix_(yi.name)
         np_type = dtype_to_np_dtype(yi.dtype)
@@ -937,7 +936,7 @@ def triple_grad_check_for_dygraph(func,
             "please check the python api unit test used."
         raise RuntimeError(msg)
 
-    for i in six.moves.xrange(len(static_triple_grad)):
+    for i in range(len(static_triple_grad)):
         if not np.allclose(static_triple_grad[i], eager_triple_grad[i], rtol,
                            atol):
             msg = 'Check eager double result fail. Mismatch between static_graph double grad ' \

python/paddle/fluid/tests/unittests/ir/pass_test.py

@@ -12,7 +12,6 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-import six
 import random
 import unittest
 import warnings
@@ -104,8 +103,7 @@ class PassTest(unittest.TestCase):
         for p in pass_builder.all_passes():
             p.apply(graph)
         opt_program.blocks = [
-            Block(opt_program, i)
-            for i in six.moves.range(opt_program.desc.num_blocks())
+            Block(opt_program, i) for i in range(opt_program.desc.num_blocks())
         ]
         opt_program._sync_with_cpp()
         return opt_program
@@ -147,7 +145,7 @@ class PassTest(unittest.TestCase):
             len(self.fetch_list) == len(outs_opt),
             "Checking the number of fetchs failed. Expected: {}, Received: {}".
             format(len(self.fetch_list), len(outs_opt)))
-        for i in six.moves.xrange(len(self.fetch_list)):
+        for i in range(len(self.fetch_list)):
             is_allclose = np.allclose(outs_opt[i], outs[i], atol=atol)
             if not is_allclose:
                 a = outs_opt[i]
@@ -207,7 +205,7 @@ class PassTest(unittest.TestCase):
                 self.main_program.num_blocks, program.num_blocks))
 
         is_different = False
-        for i in six.moves.xrange(program.num_blocks):
+        for i in range(program.num_blocks):
             if len(self.main_program.block(i).ops) != len(program.block(i).ops):
                 # The number of ops in the block i of the origin program and
                 # the optimized program is different.

python/paddle/fluid/tests/unittests/ir/test_ir_subgraph_python_interface.py

@@ -15,7 +15,6 @@
 import unittest
 import paddle
 import paddle.fluid as fluid
-import six
 
 from paddle.fluid.framework import IrGraph
 from paddle.fluid.tests.unittests.op_test import OpTestTool
@@ -37,7 +36,7 @@ class TestQuantizationSubGraph(unittest.TestCase):
                                      dtype='float32')
             label = fluid.layers.data(name='label', shape=[1], dtype='int64')
             hidden = data
-            for _ in six.moves.xrange(num):
+            for _ in range(num):
                 hidden = fluid.layers.fc(hidden, size=128, act='relu')
             loss = fluid.layers.cross_entropy(input=hidden, label=label)
             loss = paddle.mean(loss)

python/paddle/fluid/tests/unittests/mkldnn/test_activation_bf16_mkldnn_op.py

@@ -12,7 +12,6 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-import six
 import abc
 import unittest
 import numpy as np
@@ -24,8 +23,7 @@ from paddle.fluid.tests.unittests.test_gelu_op import gelu
 
 
 @OpTestTool.skip_if_not_cpu_bf16()
-@six.add_metaclass(abc.ABCMeta)
-class MKLDNNBF16ActivationOp(object):
+class MKLDNNBF16ActivationOp(metaclass=abc.ABCMeta):
 
     @abc.abstractmethod
     def config(self):

python/paddle/fluid/tests/unittests/mlu/collective_allgather_api.py

@@ -20,7 +20,6 @@ import signal
 import time
 import socket
 from contextlib import closing
-from six import string_types
 import math
 import paddle
 import paddle.fluid as fluid