refine the paddle place support using str (#28769)

7dd551e0 · wangchaochaohu · GitHub · 404c1676 · 7dd551e0 · 7dd551e0
15 changed file
--- a/paddle/fluid/operators/reader/create_double_buffer_reader_op.cc
+++ b/paddle/fluid/operators/reader/create_double_buffer_reader_op.cc
@@ -47,11 +47,12 @@ class CreateDoubleBufferReaderOp : public framework::OperatorBase {
    platform::Place place;
    if (place_str == "AUTO") {
      place = dev_place;
-    } else if (place_str == "CPU") {
+    } else if (place_str == "CPUPLACE") {
      place = platform::CPUPlace();
    } else {
+      place_str = place_str.substr(0, place_str.length() - 1);
      std::istringstream sin(place_str);
-      sin.seekg(std::string("CUDA:").size(), std::ios::beg);
+      sin.seekg(std::string("CUDAPLACE(").size(), std::ios::beg);
      size_t num;
      sin >> num;
      place = platform::CUDAPlace(static_cast<int>(num));
@@ -78,9 +79,9 @@ class CreateDoubleBufferReaderOpMaker : public DecoratedReaderMakerBase {
    std::unordered_set<std::string> enum_range;
    constexpr size_t kMaxCUDADevs = 128;
    for (size_t i = 0; i < kMaxCUDADevs; ++i) {
-      enum_range.insert(string::Sprintf("CUDA:%d", i));
+      enum_range.insert(string::Sprintf("CUDAPLACE(%d)", i));
    }
-    enum_range.insert("CPU");
+    enum_range.insert("CPUPLACE");
    enum_range.insert("AUTO");
    AddAttr<std::string>("place", "The double buffer place")
        .SetDefault("AUTO")

--- a/python/paddle/fluid/compiler.py
+++ b/python/paddle/fluid/compiler.py
@@ -18,8 +18,8 @@ import six
 import sys
 from .. import compat as cpt
 from . import framework
+from .framework import _get_paddle_place, _get_paddle_place_list
 from .framework import cuda_places, cpu_places, xpu_places
-
 from . import core

 __all__ = ['CompiledProgram', 'ExecutionStrategy', 'BuildStrategy']
@@ -202,7 +202,7 @@ class CompiledProgram(object):
                Tensors to other devices when it is first executed, the CompiledProgram
                specified by share_vars_from must be run before the current CompiledProgram.
                The default is None.
-            places(list(CUDAPlace)|list(CPUPlace)|None): This parameter specifies the device
+            places(list(CUDAPlace)|list(CPUPlace)|list(str)|None): This parameter specifies the device
                on which the model is running. If you want to run on GPU0 and GPU1, places are
                [fluid.CUDAPlace(0), fluid.CUDAPlace(1)]; if you want to run with 2 CPUs, places are
                [fluid.CPUPlace()] * 2. If the parameter is not set, i.e. the parameter is None,
@@ -213,7 +213,8 @@ class CompiledProgram(object):
                CPU number is obtained from the environment variable CPU_NUM. For example,
                export CPU_NUM=4, if the environment variable is not set, the executor will
                add the variable to the environment variable and set its value to 1.
-                The default is None.
+                The default is None. If ``places`` is the list of string, the string in the list
+                can be ``cpu``, ``gpu:x``, where ``x`` is the index of the GPUs. 

        Returns:
            CompiledProgram
@@ -282,7 +283,10 @@ class CompiledProgram(object):
        self._exec_strategy = exec_strategy
        self._loss_name = loss_name
        self._share_vars_from = share_vars_from
-        self._places = places
+        if isinstance(places, (list, tuple)):
+            self._places = _get_paddle_place_list(places)
+        else:
+            self._places = _get_paddle_place(places)

        if _has_backward_op(self._graph):
            assert self._loss_name is not None, "The loss name of CompiledProgram is None. The loss name should be set if CompiledProgram contains backward part."

--- a/python/paddle/fluid/contrib/slim/quantization/quant2_int8_mkldnn_pass.py
+++ b/python/paddle/fluid/contrib/slim/quantization/quant2_int8_mkldnn_pass.py
@@ -15,6 +15,7 @@
 import numpy as np
 from .... import core
 from ....framework import IrGraph
+from ....framework import _get_paddle_place

 __all__ = ['Quant2Int8MkldnnPass']

@@ -43,7 +44,7 @@ class Quant2Int8MkldnnPass(object):
                 _core=None,
                 _debug=False):
        self._scope = _scope
-        self._place = _place
+        self._place = _get_paddle_place(_place)
        self._core = _core
        self._debug = _debug
        self._fake_quantize_types = [

--- a/python/paddle/fluid/contrib/slim/quantization/quant_int8_mkldnn_pass.py
+++ b/python/paddle/fluid/contrib/slim/quantization/quant_int8_mkldnn_pass.py
@@ -16,6 +16,7 @@ import numpy as np
 from .... import core
 from ....framework import IrGraph
 from ....framework import IrNode
+from ....framework import _get_paddle_place

 __all__ = ['QuantInt8MkldnnPass']

@@ -40,7 +41,8 @@ class QuantInt8MkldnnPass(object):
        r"""
        Args:
            scope(fluid.Scope): scope is used to initialize the new parameters.
-            place(fluid.CPUPlace): place is used to initialize the new parameters.
+            place(fluid.CPUPlace|str): place is used to initialize the new parameters.
+            When it is string, it can be only 'cpu'.


        Examples:
@@ -60,7 +62,7 @@ class QuantInt8MkldnnPass(object):
        """

        self._scope = _scope
-        self._place = _place
+        self._place = _get_paddle_place(_place)

        self._quantize_type = [
            'fake_quantize_moving_average_abs_max',

--- a/python/paddle/fluid/contrib/slim/quantization/quantization_pass.py
+++ b/python/paddle/fluid/contrib/slim/quantization/quantization_pass.py
@@ -25,6 +25,7 @@ from ....framework import Program, program_guard, default_startup_program
 from ....data import data
 from ....layers import mean
 from ....executor import scope_guard
+from ....framework import _get_paddle_place

 __all__ = [
    'QuantizationTransformPass', 'QuantizationFreezePass', 'ConvertToInt8Pass',
@@ -246,8 +247,9 @@ class QuantizationTransformPass(object):
            scope(fluid.Scope): When activation use 'range_abs_max' as the quantize
                type, this pass will create some new parameters. The scope is used to
                initialize these new parameters.
-            place(fluid.CPUPlace|fluid.CUDAPlace): place is used to initialize new
-                parameters described above.
+            place(fluid.CPUPlace|fluid.CUDAPlace|str): place is used to initialize new
+                parameters described above. If it's string, It can be ``cpu``, and ``gpu:x``,
+                where ``x`` is the index of the GPUs. 
            weight_bits(int): quantization bit number for weights,
                the bias is not quantized.
            activation_bits(int): quantization bit number for activation.
@@ -315,7 +317,7 @@ class QuantizationTransformPass(object):
            transform_pass.apply(graph)
        """
        self._scope = scope
-        self._place = place
+        self._place = _get_paddle_place(place)
        self._weight_bits = weight_bits
        self._activation_bits = activation_bits
        self._skip_pattern = skip_pattern
@@ -1057,7 +1059,8 @@ class QuantizationFreezePass(object):

        Args:
            scope(fluid.Scope): scope is used to get the weight tensor values.
-            place(fluid.CPUPlace|fluid.CUDAPlace): place is used to restore the weight tensors.
+            place(fluid.CPUPlace|fluid.CUDAPlace|str): place is used to restore the weight tensors.
+                If it's string, It can be ``cpu``, and ``gpu:x``, where ``x`` is the index of the GPUs.
            weight_bits(int): quantization bit number for weights.
            activation_bits(int): quantization bit number for activation.
            weight_quantize_type(str): quantization type for weights, support 'abs_max' and 
@@ -1071,7 +1074,7 @@ class QuantizationFreezePass(object):
        assert place is not None, \
            'The place cannot be set None.'
        self._scope = scope
-        self._place = place
+        self._place = _get_paddle_place(place)
        self._weight_bits = weight_bits
        self._activation_bits = activation_bits
        self._weight_quantize_type = weight_quantize_type
@@ -1365,8 +1368,9 @@ class ConvertToInt8Pass(object):

        Args:
            scope(fluid.Scope): scope is used to get the weight tensor values.
-            place(fluid.CPUPlace|fluid.CUDAPlace): place is used to restore the
-                8bits weight tensors.
+            place(fluid.CPUPlace|fluid.CUDAPlace|str): place is used to restore the
+                8bits weight tensors. If it's string, It can be ``cpu``, and ``gpu:x``,
+                where ``x`` is the index of the GPUs.
            quantizable_op_type(list[str]): This input param will be removed latter. The pass
                will process all quantized op, so it is not necessary to set the input param.
        """
@@ -1375,7 +1379,7 @@ class ConvertToInt8Pass(object):
        assert place is not None, \
            'The place cannot be set None.'
        self._scope = scope
-        self._place = place
+        self._place = _get_paddle_place(place)

    def apply(self, graph):
        """
@@ -1495,11 +1499,13 @@ class OutScaleForTrainingPass(object):

        Args:
            scope(fluid.Scope): The scope is used to initialize these new parameters.
-            place(fluid.CPUPlace|fluid.CUDAPlace): The place is used to initialize new parameters.
+            place(fluid.CPUPlace|fluid.CUDAPlace|str): The place is used to initialize new parameters.
+                If it's string, It can be ``cpu``, and ``gpu:x``, where ``x`` is the
+                index of the GPUs.
            moving_rate(float): The decay coefficient of moving average. The default value is 0.9.
        """
        self._scope = scope
-        self._place = place
+        self._place = _get_paddle_place(place)
        self._moving_rate = moving_rate
        self._is_test = None
        self._teller_set = _out_scale_op_list
@@ -1688,8 +1694,9 @@ class AddQuantDequantPass(object):

        Args:
            scope(fluid.Scope): The scope is used to initialize these new parameters.
-            place(fluid.CPUPlace|fluid.CUDAPlace): place is used to initialize new
-                parameters described above.
+            place(fluid.CPUPlace|fluid.CUDAPlace|str): place is used to initialize new
+                parameters described above. If ``place`` is string, it can be It can be ``cpu``
+                or ``gpu:x``, where ``x`` is the index of the GPUs.
            moving_rate(float, optional): the param for 'quant_dequant_moving_average_abs_max' 
                quantization. Default is 0.9.
            quant_bits(int, optional): quantization bit number for activation. Default is 8.
@@ -1705,7 +1712,7 @@ class AddQuantDequantPass(object):
                quantizable_op_type.
        """
        self._scope = scope
-        self._place = place
+        self._place = _get_paddle_place(place)
        self._moving_rate = moving_rate
        self._quant_bits = quant_bits
        self._is_test = None

--- a/python/paddle/fluid/dygraph/base.py
+++ b/python/paddle/fluid/dygraph/base.py
@@ -25,6 +25,7 @@ from .tracer import Tracer
 import logging
 from ..data_feeder import convert_dtype
 import warnings
+from ..framework import _get_paddle_place

 __all__ = [
    'no_grad', 'no_grad_', 'grad', 'guard', 'enable_dygraph', 'disable_dygraph',
@@ -128,8 +129,9 @@ def enable_dygraph(place=None):
    This API turn OFF static graph mode. You can turn ON static graph mode by `enable_static <./disable_dygraph_en.html>`_ .

    Parameters:
-        place(paddle.CPUPlace|paddle.CUDAPlace, optional): Place to run dynamic graph. Default: None. Which means that the running place will be 
-            determined according to the way of paddle compilation. 
+        place(paddle.CPUPlace|paddle.CUDAPlace|str, optional): Place to run dynamic graph. Default: None. Which means that the running place will be 
+            determined according to the way of paddle compilation. If ``place`` is string, It can be ``cpu``, and ``gpu:x``, where ``x`` is the
+            index of the GPUs.

    return:
        None
@@ -149,7 +151,8 @@ def enable_dygraph(place=None):
    """
    global _functional_dygraph_context_manager
    if _functional_dygraph_context_manager is None:
-        _functional_dygraph_context_manager = guard(place=place)
+        _functional_dygraph_context_manager = guard(
+            place=_get_paddle_place(place))
        _functional_dygraph_context_manager.__enter__()

        # call disable_dygraph when Python exit
@@ -343,8 +346,10 @@ def guard(place=None):
    This context will create a dygraph context for dygraph to run, using python ``with`` statement.

    Parameters:
-        place(fluid.CPUPlace or fluid.CUDAPlace, optional): Place to execute dygraph. 
-            If None, the running place will be determined according to the way of paddle compilation. Default: None
+        place(fluid.CPUPlace| fluid.CUDAPlace|str, optional): Place to execute dygraph. 
+            If None, the running place will be determined according to the way of paddle compilation.
+            If ``place`` is string, It can be ``cpu``, ``gpu:x`` and ``xpu:x``, where ``x`` is the
+            index of the GPUs or XPUs. Default: None

    return:
        None
@@ -371,7 +376,7 @@ def guard(place=None):
    VarBase = core.VarBase

    if place is not None:
-        expected_place = place
+        expected_place = _get_paddle_place(place)
    else:
        expected_place = framework._current_expected_place()
    tracer._expected_place = expected_place

--- a/python/paddle/fluid/executor.py
+++ b/python/paddle/fluid/executor.py
@@ -480,11 +480,13 @@ class Executor(object):
    and single/multiple-CPU running.

    Args:
-        place(paddle.CPUPlace()|paddle.CUDAPlace(n)|None): This parameter represents
+        place(paddle.CPUPlace()|paddle.CUDAPlace(n)|str|None): This parameter represents
            which device the executor runs on. When this parameter is None, PaddlePaddle
            will set the default device according to its installation version. If Paddle
            is CPU version, the default device would be set to `CPUPlace()` . If Paddle is
            GPU version, the default device would be set to `CUDAPlace(0)` . Default is None.
+            If ``place`` is string, it can be ``cpu``, and ``gpu:x``, where ``x`` 
+            is the index of the GPUs.

    Returns:
        Executor
@@ -550,7 +552,7 @@ class Executor(object):
            expected_place = framework._current_expected_place()
            self.place = expected_place
        else:
-            self.place = place
+            self.place = framework._get_paddle_place(place)
        self.program_caches = dict()
        self.ctx_caches = dict()
        self.scope_caches = dict()

--- a/python/paddle/fluid/framework.py
+++ b/python/paddle/fluid/framework.py
@@ -5805,3 +5805,64 @@ def get_flags(flags):
    else:
        raise TypeError('Flags in get_flags should be a list, tuple or string.')
    return flags_value
+
+
+def _get_paddle_place(place):
+    "convert the string to paddle Place"
+    if place is None:
+        return place
+    if isinstance(place, (core.Place, core.XPUPlace, core.CPUPlace,
+                          core.CUDAPinnedPlace, core.CUDAPlace)):
+        return place
+
+    if not isinstance(place, str):
+        raise ValueError(
+            "place only support string which is 'Place' and so on.")
+
+    place = place.lower()
+    if (place == "cpu"):
+        return core.CPUPlace()
+    if (place == "device"):
+        return core.Place()
+
+    avaliable_gpu_place = re.match(r'gpu:\d+', place)
+    if place == "gpu_pinned" or place == "gpu" or avaliable_gpu_place:
+        if not core.is_compiled_with_cuda():
+            raise ValueError(
+                "The device should not be {}, since PaddlePaddle is " \
+                "not compiled with CUDA".format(avaliable_gpu_place))
+        if place == "gpu_pinned":
+            return core.CUDAPinnedPlace()
+        elif place == "gpu":
+            return core.CUDAPlace(0)
+        else:
+            place_info_list = place.split(':', 1)
+            device_id = place_info_list[1]
+            device_id = int(device_id)
+            return core.CUDAPlace(device_id)
+    avaliable_xpu_place = re.match(r'xpu:\d+', place)
+    if avaliable_xpu_place:
+        if not core.is_compiled_with_xpu():
+            raise ValueError(
+                "The device should not be {}, since PaddlePaddle is " \
+                "not compiled with XPU".format(avaliable_xpu_place))
+        place_info_list = place.split(':', 1)
+        device_id = place_info_list[1]
+        device_id = int(device_id)
+        return core.XPUPlace(device_id)
+    raise ValueError(
+        "paddle support CPUPlace, CUDAPlace,CUDAPinnedPlace and XPUPlace, Please check your Place Input"
+    )
+
+
+def _get_paddle_place_list(places):
+
+    if not isinstance(places, (list, tuple)):
+        raise TypeError("places must to be List or Tuple")
+
+    ret = []
+    for p in places:
+        p = _get_paddle_place(p)
+        ret.append(p)
+
+    return ret
--- a/python/paddle/fluid/generator.py
+++ b/python/paddle/fluid/generator.py
@@ -14,6 +14,7 @@
 """This is definition of generator class, which is for managing the state of the algorithm that produces pseudo random numbers."""

 from . import core
+from .framework import _get_paddle_place

 __all__ = ['Generator']

@@ -26,14 +27,15 @@ class Generator(core.Generator):
        Create a generator object which manages the random number generation. ( Experimental Feature )

        Parameters:
-            place(CPUPlace|CUDAPinnedPlace|CUDAPlace, optional): The place to allocate Tensor. Can be  
-                CPUPlace, CUDAPinnedPlace, CUDAPlace. Default: None, means global place.
+            place(CPUPlace|CUDAPinnedPlace|CUDAPlace|str,optional): The place to allocate Tensor. Can be  
+                CPUPlace, CUDAPinnedPlace, CUDAPlace. Default: None, means global place. If ``place`` is
+                string, it can be ``cpu`` and ``gpu:x``, where ``x`` is the index of the GPUs.

        Returns:
            Generator: A generator object.

        """
-        self.place = place
+        self.place = _get_paddle_place(place)
        if not place:
            place = core.CPUPlace()
        if isinstance(place, core.CPUPlace):

--- a/python/paddle/fluid/layers/io.py
+++ b/python/paddle/fluid/layers/io.py
@@ -32,6 +32,7 @@ from ..unique_name import generate as unique_name
 import logging
 from ..data_feeder import check_dtype, check_type
 from paddle.fluid.framework import static_only
+from ..framework import _get_paddle_place

 __all__ = [
    'data', 'read_file', 'double_buffer', 'py_reader',
@@ -842,7 +843,8 @@ def double_buffer(reader, place=None, name=None):

    Args:
        reader (Variable): The Reader Variable need to be wrapped.
-        place (Place, optional): The place of target data, such as CPU, GPU, and if use GPU, it's necessary to point out which card is involved. Default is the sample place of executor perform.
+        place (Place|str, optional): The place of target data, such as CPU, GPU, and if use GPU, it's necessary to point out which card is involved. Default is the sample place of executor perform.
+            if ``place`` is string, It can be ``cpu``, ``gpu:x``, where ``x`` is the ndex of the GPUs. 
        name (str, optional): Variable name. Normally there is no need for user to set this property. For more information, please refer to :ref:`api_guide_Name`. Default is None. 

    Returns:
@@ -861,7 +863,8 @@ def double_buffer(reader, place=None, name=None):
    """
    attrs = dict()
    if place is not None:
-        attrs['place'] = str(place).upper()
+        attrs['place'] = str(_get_paddle_place(place)).upper()
+
    return __create_unshared_decorated_reader__(
        'create_double_buffer_reader', reader, attrs, name=name)


--- a/python/paddle/fluid/reader.py
+++ b/python/paddle/fluid/reader.py
@@ -27,6 +27,7 @@ from .dataloader.dataloader_iter import _DataLoaderIterSingleProcess, _DataLoade
 from .dataloader.batch_sampler import _InfiniteIterableSampler
 from .layers.io import monkey_patch_reader_methods, _copy_reader_var_, double_buffer
 from .unique_name import UniqueNameGenerator
+from .framework import _get_paddle_place, _get_paddle_place_list
 import logging
 import warnings

@@ -186,10 +187,12 @@ class DataLoader(object):
            The Tensors should be created by :code:`paddle.static.data()`.
            :attr:`feed_list` must be set if :attr:`return_list` is
            False. Default None.
-        places(list(Place)|tuple(Place)|optional): a list of Place,
+        places(list(Place)|tuple(Place)|list(str)|optional): a list of Place,
            to put data onto, :attr:`places` can be None, if 
            :attr:`places` is None, default place(CPUPlace or CUDAPlace(0))
-            will be used. Default None.
+            will be used. Default None. If ``places`` is list of string,
+            the string in the list can be ``cpu``, ``gpu:x`` and ``gpu_pinned``,
+            where ``x`` is the index of the GPUs.
        return_list (bool): whether the return value on each device is 
            presented as a list. If :attr:`return_list=False`, the return
            value on each device would be a dict of str -> Tensor, where
@@ -335,6 +338,10 @@ class DataLoader(object):

        if places is None:
            places = _current_expected_place()
+        if isinstance(places, (list, tuple)):
+            places = _get_paddle_place_list(places)
+        else:
+            places = _get_paddle_place(places)
        self.places = _convert_places(places)

        assert num_workers >= 0, "num_workers should be a non-negative value"
@@ -752,8 +759,9 @@ class DataLoader(object):

        Args:
            dataset (InMemoryDataset|QueueDataset): the dataset object.
-            places (list(CUDAPlace)|list(CPUPlace)): places where the result 
-                data should be converted.   
+            places (list(CUDAPlace)|list(CPUPlace)|list(str)): places where the result 
+                data should be converted. If places is list of string, the string in the list 
+                can be ``cpu``, ``gpu:x`` and ``gpu_pinned``, where x is the index of the GPUs.   
            drop_last (bool): whether to drop the last batch whose sample 
                number is less than batch size. If drop_last = True, they
                would be dropped. If drop_last = False, they would be kept. 
@@ -1030,6 +1038,10 @@ class DygraphGeneratorLoader(DataLoaderBase):
                             drop_last=True,
                             places=None):
        assert batch_size > 0, "batch_size must be larger than 0"
+        if isinstance(places, (list, tuple)):
+            places = _get_paddle_place_list(places)
+        else:
+            places = _get_paddle_place(places)
        self.set_sample_list_generator(
            paddle.batch(
                reader, batch_size=batch_size, drop_last=drop_last),
@@ -1037,6 +1049,11 @@ class DygraphGeneratorLoader(DataLoaderBase):
        return self

    def set_sample_list_generator(self, reader, places=None):
+        if isinstance(places, (list, tuple)):
+            places = _get_paddle_place_list(places)
+        else:
+            places = _get_paddle_place(places)
+
        def __batch_reader_impl__():
            for batch in reader():
                slots = []
@@ -1052,6 +1069,10 @@ class DygraphGeneratorLoader(DataLoaderBase):
        return self

    def set_batch_generator(self, reader, places=None):
+        if isinstance(places, (list, tuple)):
+            places = _get_paddle_place_list(places)
+        else:
+            places = _get_paddle_place(places)
        self._batch_reader = reader
        if places is None:
            places = _current_expected_place()
@@ -1275,6 +1296,10 @@ class GeneratorLoader(DataLoaderBase):
                             drop_last=True,
                             places=None):
        assert batch_size > 0, "batch_size must be larger than 0"
+        if isinstance(places, (list, tuple)):
+            places = _get_paddle_place_list(places)
+        else:
+            places = _get_paddle_place(places)
        has_lod = False
        for f in self._feed_list:
            if f.lod_level != 0:
@@ -1297,6 +1322,10 @@ class GeneratorLoader(DataLoaderBase):
        return self

    def set_sample_list_generator(self, reader, places=None):
+        if isinstance(places, (list, tuple)):
+            places = _get_paddle_place_list(places)
+        else:
+            places = _get_paddle_place(places)
        with program_guard(Program(), Program()):
            feeder = DataFeeder(
                feed_list=self._feed_list, place=core.CPUPlace())
@@ -1310,6 +1339,10 @@ class GeneratorLoader(DataLoaderBase):
        return self

    def set_batch_generator(self, reader, places=None):
+        if isinstance(places, (list, tuple)):
+            places = _get_paddle_place_list(places)
+        else:
+            places = _get_paddle_place(places)
        self._tensor_reader = reader
        if self._iterable:
            assert places is not None, "Places cannot be None when DataLoader is iterable"
@@ -1784,6 +1817,10 @@ class DatasetLoader(DataLoaderBase):
                          DatasetBase), "dataset must be type of DatasetBase"
        assert not in_dygraph_mode(
        ), "DatasetLoader is not supported in dygraph mode yet"
+        if isinstance(places, (list, tuple)):
+            places = _get_paddle_place_list(places)
+        else:
+            places = _get_paddle_place(places)

        thread_num = len(places)


--- a/python/paddle/fluid/tests/unittests/test_py_reader_error_msg.py
+++ b/python/paddle/fluid/tests/unittests/test_py_reader_error_msg.py
@@ -15,6 +15,7 @@
 import paddle.fluid as fluid
 import unittest
 import numpy as np
+import paddle


 class TestPyReaderErrorMsg(unittest.TestCase):
@@ -35,5 +36,19 @@ class TestPyReaderErrorMsg(unittest.TestCase):
            ])


+class TestDoubleBufferAPI(unittest.TestCase):
+    def test_double_buffer(self):
+        paddle.enable_static()
+        if fluid.core.is_compiled_with_cuda():
+            reader = fluid.layers.py_reader(
+                capacity=64,
+                shapes=[(-1, 1, 28, 28), (-1, 1)],
+                dtypes=['float32', 'int64'],
+                use_double_buffer=False)
+            reader = fluid.layers.double_buffer(
+                reader, place=fluid.core.CUDAPlace(0))
+            image, label = fluid.layers.read_file(reader)
+
+
 if __name__ == '__main__':
    unittest.main()
--- a/python/paddle/fluid/tests/unittests/test_var_base.py
+++ b/python/paddle/fluid/tests/unittests/test_var_base.py
@@ -149,9 +149,12 @@ class TestVarBase(unittest.TestCase):
                    paddle.to_tensor([[1], [2, 3]], place=1)

        _test_place(core.CPUPlace())
+        _test_place("cpu")
        if core.is_compiled_with_cuda():
            _test_place(core.CUDAPinnedPlace())
+            _test_place("gpu_pinned")
            _test_place(core.CUDAPlace(0))
+            _test_place("gpu:0")

    def test_to_variable(self):
        with fluid.dygraph.guard():

--- a/python/paddle/hapi/model.py
+++ b/python/paddle/hapi/model.py
@@ -31,7 +31,7 @@ import paddle
 from paddle import fluid
 from paddle.fluid import core
 from paddle.fluid.framework import in_dygraph_mode, Variable, ParamBase, _current_expected_place
-from paddle.fluid.framework import in_dygraph_mode, Variable
+from paddle.fluid.framework import in_dygraph_mode, Variable, _get_paddle_place
 from paddle.fluid.framework import _current_expected_place as _get_device
 from paddle.fluid.executor import global_scope
 from paddle.fluid.io import is_belong_to_optimizer
@@ -167,6 +167,7 @@ def prepare_distributed_context(place=None):
        place = fluid.CUDAPlace(ParallelEnv().dev_id) if ParallelEnv().nranks > 1 \
            else fluid.CUDAPlace(0)

+    place = _get_paddle_place(place)
    strategy = fluid.dygraph.parallel.ParallelStrategy()
    strategy.nranks = ParallelEnv().nranks
    strategy.local_rank = ParallelEnv().local_rank

--- a/python/paddle/tensor/creation.py
+++ b/python/paddle/tensor/creation.py
@@ -17,7 +17,7 @@ import numpy as np

 from ..fluid.framework import Variable
 from ..fluid.framework import unique_name
-from ..fluid.framework import _current_expected_place
+from ..fluid.framework import _current_expected_place, _get_paddle_place
 from ..fluid.framework import dygraph_only
 from ..fluid.initializer import Constant
 from ..fluid.layers import core
@@ -25,7 +25,6 @@ from ..fluid.layer_helper import LayerHelper
 from ..fluid.data_feeder import check_variable_and_dtype, check_type, check_dtype, convert_dtype
 from ..fluid.framework import convert_np_dtype_to_dtype_, in_dygraph_mode, _varbase_creator, device_guard, OpProtoHolder
 from paddle.common_ops_import import *
-
 # TODO: define functions to get create a tensor  
 from ..fluid.layers import linspace  #DEFINE_ALIAS
 import paddle
@@ -69,8 +68,9 @@ def to_tensor(data, dtype=None, place=None, stop_gradient=True):
            'float32' , 'float64' , 'int8' , 'int16' , 'int32' , 'int64' , 'uint8',
            'complex64' , 'complex128'. Default: None, infers dtype from ``data`` 
            except for python float number which gets dtype from ``get_default_type`` .
-        place(CPUPlace|CUDAPinnedPlace|CUDAPlace, optional): The place to allocate Tensor. Can be  
-            CPUPlace, CUDAPinnedPlace, CUDAPlace. Default: None, means global place.
+        place(CPUPlace|CUDAPinnedPlace|CUDAPlace|str, optional): The place to allocate Tensor. Can be  
+            CPUPlace, CUDAPinnedPlace, CUDAPlace. Default: None, means global place. If ``place`` is 
+            string, It can be ``cpu``, ``gpu:x`` and ``gpu_pinned``, where ``x`` is the index of the GPUs. 
        stop_gradient(bool, optional): Whether to block the gradient propagation of Autograd. Default: True.

    Returns:
@@ -80,7 +80,7 @@ def to_tensor(data, dtype=None, place=None, stop_gradient=True):
        TypeError: If the data type of ``data`` is not scalar, list, tuple, numpy.ndarray, paddle.Tensor
        ValueError: If ``data`` is tuple|list, it can't contain nested tuple|list with different lengths , such as: [[1, 2], [3, 4, 5]]
        TypeError: If ``dtype`` is not bool, float16, float32, float64, int8, int16, int32, int64, uint8, complex64, complex128
-        ValueError: If ``place`` is not paddle.CPUPlace, paddle.CUDAPinnedPlace, paddle.CUDAPlace
+        ValueError: If ``place`` is not paddle.CPUPlace, paddle.CUDAPinnedPlace, paddle.CUDAPlace or specified pattern string. 

    Examples:

@@ -118,10 +118,12 @@ def to_tensor(data, dtype=None, place=None, stop_gradient=True):
        #         [(3+2j), (4+0j)]])
    """

+    place = _get_paddle_place(place)
    if place is None:
        place = _current_expected_place()
-    elif not isinstance(place, (core.Place, core.CPUPlace, core.CUDAPinnedPlace,
-                                core.CUDAPlace)):
+    elif not isinstance(
+            place,
+        (core.Place, core.CPUPlace, core.CUDAPinnedPlace, core.CUDAPlace)):
        raise ValueError(
            "'place' must be any of paddle.Place, paddle.CPUPlace, paddle.CUDAPinnedPlace, paddle.CUDAPlace"
        )