remove two context param

example/bert_clue/README.md

@@ -58,7 +58,6 @@ options:
     --epoch_size               epoch size: N, default is 1
     --device_num               number of used devices: N, default is 1
     --device_id                device id: N, default is 0
-    --enable_task_sink         enable task sink: "true" | "false", default is "true"
     --enable_loop_sink         enable loop sink: "true" | "false", default is "true"
     --enable_mem_reuse         enable memory reuse: "true" | "false", default is "true"
     --enable_save_ckpt         enable save checkpoint: "true" | "false", default is "true"

example/bert_clue/run_distribute_pretrain.sh

@@ -50,7 +50,6 @@ do
     --epoch_size=$EPOCH_SIZE \
     --device_id=$DEVICE_ID \
     --device_num=$RANK_SIZE \
-    --enable_task_sink="true" \
     --enable_loop_sink="true" \
     --enable_mem_reuse="true" \
     --enable_save_ckpt="true" \

example/bert_clue/run_pretrain.py

@@ -59,7 +59,6 @@ def run_pretrain():
     parser.add_argument("--epoch_size", type=int, default="1", help="Epoch size, default is 1.")
     parser.add_argument("--device_id", type=int, default=0, help="Device id, default is 0.")
     parser.add_argument("--device_num", type=int, default=1, help="Use device nums, default is 1.")
-    parser.add_argument("--enable_task_sink", type=str, default="true", help="Enable task sink, default is true.")
     parser.add_argument("--enable_loop_sink", type=str, default="true", help="Enable loop sink, default is true.")
     parser.add_argument("--enable_mem_reuse", type=str, default="true", help="Enable mem reuse, default is true.")
     parser.add_argument("--enable_save_ckpt", type=str, default="true", help="Enable save checkpoint, default is true.")
@@ -76,8 +75,7 @@ def run_pretrain():
 
     args_opt = parser.parse_args()
     context.set_context(mode=context.GRAPH_MODE, device_target="Ascend", device_id=args_opt.device_id)
-    context.set_context(enable_task_sink=(args_opt.enable_task_sink == "true"),
-                        enable_loop_sink=(args_opt.enable_loop_sink == "true"),
+    context.set_context(enable_loop_sink=(args_opt.enable_loop_sink == "true"),
                         enable_mem_reuse=(args_opt.enable_mem_reuse == "true"))
     context.set_context(reserve_class_name_in_scope=False)
 

example/bert_clue/run_standalone_pretrain.sh

@@ -29,7 +29,6 @@ python run_pretrain.py  \
     --distribute="false" \
     --epoch_size=$EPOCH_SIZE \
     --device_id=$DEVICE_ID \
-    --enable_task_sink="true" \
     --enable_loop_sink="true" \
     --enable_mem_reuse="true" \
     --enable_save_ckpt="true" \

example/googlenet_cifar10/train.py

@@ -70,7 +70,6 @@ if __name__ == '__main__':
 
     context.set_context(mode=context.GRAPH_MODE, device_target=args_opt.device_target)
     context.set_context(device_id=args_opt.device_id)
-    context.set_context(enable_task_sink=True)
     context.set_context(enable_loop_sink=True)
     context.set_context(enable_mem_reuse=True)
 

example/mobilenetv2_imagenet2012/eval.py

@@ -34,7 +34,6 @@ args_opt = parser.parse_args()
 device_id = int(os.getenv('DEVICE_ID'))
 
 context.set_context(mode=context.GRAPH_MODE, device_target="Ascend", device_id=device_id, save_graphs=False)
-context.set_context(enable_task_sink=True)
 context.set_context(enable_loop_sink=True)
 context.set_context(enable_mem_reuse=True)
 

example/mobilenetv2_imagenet2012/train.py

@@ -54,7 +54,6 @@ rank_size = int(os.getenv('RANK_SIZE'))
 run_distribute = rank_size > 1
 
 context.set_context(mode=context.GRAPH_MODE, device_target="Ascend", device_id=device_id, save_graphs=False)
-context.set_context(enable_task_sink=True)
 context.set_context(enable_loop_sink=True)
 context.set_context(enable_mem_reuse=True)
 

example/resnet101_imagenet2012/eval.py

@@ -46,7 +46,6 @@ args_opt = parser.parse_args()
 device_id = int(os.getenv('DEVICE_ID'))
 
 context.set_context(mode=context.GRAPH_MODE, device_target="Ascend", save_graphs=False, device_id=device_id)
-context.set_context(enable_task_sink=True)
 context.set_context(enable_loop_sink=True)
 context.set_context(enable_mem_reuse=True)
 

example/resnet101_imagenet2012/train.py

@@ -49,7 +49,6 @@ args_opt = parser.parse_args()
 device_id = int(os.getenv('DEVICE_ID'))
 
 context.set_context(mode=context.GRAPH_MODE, device_target="Ascend", save_graphs=False, device_id=device_id)
-context.set_context(enable_task_sink=True)
 context.set_context(enable_loop_sink=True)
 context.set_context(enable_mem_reuse=True)
 

example/resnet50_cifar10/eval.py

@@ -39,7 +39,7 @@ args_opt = parser.parse_args()
 device_id = int(os.getenv('DEVICE_ID'))
 
 context.set_context(mode=context.GRAPH_MODE, device_target="Ascend", save_graphs=False)
-context.set_context(enable_task_sink=True, device_id=device_id)
+context.set_context(device_id=device_id)
 context.set_context(enable_loop_sink=True)
 context.set_context(enable_mem_reuse=True)
 

example/resnet50_cifar10/train.py

@@ -42,7 +42,7 @@ args_opt = parser.parse_args()
 device_id = int(os.getenv('DEVICE_ID'))
 
 context.set_context(mode=context.GRAPH_MODE, device_target="Ascend", save_graphs=False)
-context.set_context(enable_task_sink=True, device_id=device_id)
+context.set_context(device_id=device_id)
 context.set_context(enable_loop_sink=True)
 context.set_context(enable_mem_reuse=True)
 

example/ssd_coco2017/eval.py

@@ -71,7 +71,7 @@ if __name__ == '__main__':
     args_opt = parser.parse_args()
 
     context.set_context(mode=context.GRAPH_MODE, device_target="Ascend", device_id=args_opt.device_id)
-    context.set_context(enable_task_sink=True, enable_loop_sink=True, enable_mem_reuse=True)
+    context.set_context(enable_loop_sink=True, enable_mem_reuse=True)
 
     config = ConfigSSD()
     prefix = "ssd_eval.mindrecord"

example/ssd_coco2017/train.py

@@ -93,7 +93,7 @@ def main():
     args_opt = parser.parse_args()
 
     context.set_context(mode=context.GRAPH_MODE, device_target="Ascend", device_id=args_opt.device_id)
-    context.set_context(enable_task_sink=True, enable_loop_sink=True, enable_mem_reuse=True)
+    context.set_context(enable_loop_sink=True, enable_mem_reuse=True)
 
     if args_opt.distribute:
         device_num = args_opt.device_num

example/vgg16_cifar10/train.py

@@ -64,7 +64,6 @@ if __name__ == '__main__':
 
     context.set_context(mode=context.GRAPH_MODE, device_target=args_opt.device_target)
     context.set_context(device_id=args_opt.device_id)
-    context.set_context(enable_task_sink=True)
     context.set_context(enable_loop_sink=True)
     context.set_context(enable_mem_reuse=True)
 

example/yolov3_coco2017/eval.py

@@ -82,7 +82,7 @@ if __name__ == '__main__':
     args_opt = parser.parse_args()
 
     context.set_context(mode=context.GRAPH_MODE, device_target="Ascend", device_id=args_opt.device_id)
-    context.set_context(enable_task_sink=True, enable_loop_sink=True, enable_mem_reuse=True)
+    context.set_context(enable_loop_sink=True, enable_mem_reuse=True)
 
     # It will generate mindrecord file in args_opt.mindrecord_dir,
     # and the file name is yolo.mindrecord0, 1, ... file_num.

example/yolov3_coco2017/train.py

@@ -85,7 +85,7 @@ def main():
     args_opt = parser.parse_args()
 
     context.set_context(mode=context.GRAPH_MODE, device_target="Ascend", device_id=args_opt.device_id)
-    context.set_context(enable_task_sink=True, enable_loop_sink=True, enable_mem_reuse=True)
+    context.set_context(enable_loop_sink=True, enable_mem_reuse=True)
     if args_opt.distribute:
         device_num = args_opt.device_num
         context.reset_auto_parallel_context()

mindspore/ccsrc/pipeline/init.cc

@@ -115,12 +115,8 @@ PYBIND11_MODULE(_c_expression, m) {
     .def("set_device_id", &mindspore::MsContext::set_device_id, "Set device id.")
     .def("open_tsd", &mindspore::MsContext::OpenTsd, "Open tdt dataset client.")
     .def("close_tsd", &mindspore::MsContext::CloseTsd, "Close tdt dataset client.")
-    .def("set_task_sink_flag", &mindspore::MsContext::set_enable_task_sink, "Set enable task sink.")
-    .def("get_task_sink_flag", &mindspore::MsContext::enable_task_sink, "Get whether to enable task sink.")
     .def("get_save_graphs_flag", &mindspore::MsContext::save_graphs_flag, "Get whether to save graphs.")
     .def("set_save_graphs_flag", &mindspore::MsContext::set_save_graphs_flag, "Set whether to save graphs.")
-    .def("get_ir_fusion_flag", &mindspore::MsContext::ir_fusion_flag, "Get whether to enable ir fusion.")
-    .def("set_ir_fusion_flag", &mindspore::MsContext::set_ir_fusion_flag, "Set whether to enable ir fusion.")
     .def("get_auto_mixed_precision_flag", &mindspore::MsContext::auto_mixed_precision_flag,
          "Get whether to enable auto mixed precision.")
     .def("set_auto_mixed_precision_flag", &mindspore::MsContext::set_auto_mixed_precision_flag,

mindspore/ccsrc/utils/context/ms_context.h

@@ -62,7 +62,6 @@ class MsContext {
   bool enable_pynative_infer() const { return enable_pynative_infer_; }
   void set_enable_pynative_infer(bool enable_pynative_infer) { enable_pynative_infer_ = enable_pynative_infer; }
 
-  void set_enable_task_sink(bool enable_task_sink) { enable_task_sink_ = enable_task_sink; }
   bool enable_task_sink() const { return enable_task_sink_; }
 
   void set_precompile_only(bool precompile_only) { precompile_only_ = precompile_only; }
@@ -90,7 +89,6 @@ class MsContext {
   bool enable_hccl() const { return enable_hccl_; }
   bool PynativeInitGe();
 
-  void set_ir_fusion_flag(bool ir_fusion_flag) { ir_fusion_flag_ = ir_fusion_flag; }
   bool ir_fusion_flag() const { return ir_fusion_flag_; }
 
   void set_loop_sink_flag(bool loop_sink_flag) { enable_loop_sink_ = loop_sink_flag; }

mindspore/context.py

@@ -142,15 +142,6 @@ class _Context:
             raise ValueError("Context handle is none in context!!!")
         return value
 
-    # For Ascend task sink mode execution
-    @property
-    def enable_task_sink(self):
-        return self._context_handle.get_task_sink_flag()
-
-    @enable_task_sink.setter
-    def enable_task_sink(self, task_sink):
-        self._context_handle.set_task_sink_flag(task_sink)
-
     @property
     def mode(self):
         return self._context_handle.get_execution_mode()
@@ -224,14 +215,6 @@ class _Context:
         if not success:
             raise RuntimeError("Device id set failed!!!")
 
-    @property
-    def enable_ir_fusion(self):
-        return self._context_handle.get_ir_fusion_flag()
-
-    @enable_ir_fusion.setter
-    def enable_ir_fusion(self, enable_ir_fusion):
-        self._context_handle.set_ir_fusion_flag(enable_ir_fusion)
-
     @property
     def enable_loop_sink(self):
         return self._context_handle.get_loop_sink_flag()
@@ -485,11 +468,9 @@ def reset_auto_parallel_context():
     _reset_auto_parallel_context()
 
 
-@args_type_check(mode=int, precompile_only=bool, device_target=str,
-                 device_id=int, enable_ir_fusion=bool, save_graphs=bool,
-                 enable_task_sink=bool, save_graphs_path=str, enable_loop_sink=bool,
-                 enable_mem_reuse=bool, save_ms_model=bool, save_ms_model_path=str,
-                 enable_auto_mixed_precision=bool, enable_dump=bool, save_dump_path=str,
+@args_type_check(mode=int, precompile_only=bool, device_target=str, device_id=int, save_graphs=bool,
+                 save_graphs_path=str, enable_loop_sink=bool, enable_mem_reuse=bool, save_ms_model=bool,
+                 save_ms_model_path=str, enable_auto_mixed_precision=bool, enable_dump=bool, save_dump_path=str,
                  enable_reduce_precision=bool, graph_memory_max_size=str,
                  variable_memory_max_size=str, enable_profiling=bool, profiling_options=str)
 def set_context(**kwargs):
@@ -517,10 +498,8 @@ def set_context(**kwargs):
         device_target (str): The target device to run, support "Ascend", "GPU", "CPU". Default: "Ascend".
         device_id (int): Id of target device, the value must be in [0, device_num_per_host-1],
                     while device_num_per_host should no more than 4096. Default: 0.
-        enable_ir_fusion (bool): Whether to enable ir fusion. Default: True.
         save_graphs (bool): Whether to save graphs. Default: False.
         enable_loop_sink (bool): Whether to enable loop sink. Default: True.
-        enable_task_sink (bool): Whether to enable task sink. Default: True.
         enable_mem_reuse (bool): Whether to enable memory reuse. Default: True.
         save_ms_model (bool): Whether to save lite model converted by graph. Default: False.
         save_ms_model_path (str): Path to save converted lite model. Default: "."
@@ -559,7 +538,6 @@ def set_context(**kwargs):
         >>> context.set_context(device_target="Ascend")
         >>> context.set_context(device_id=0)
         >>> context.set_context(save_graphs=True, save_graphs_path="./model.ms")
-        >>> context.set_context(enable_task_sink=True)
         >>> context.set_context(enable_mem_reuse=True)
         >>> context.set_context(enable_reduce_precision=True)
         >>> context.set_context(save_ms_model=True, save_ms_model_path=".")

tests/st/auto_parallel/onehot_model_parallel.py

@@ -33,9 +33,7 @@ def setup_module():
     global rank_id
     np.random.seed(0)
     context.set_context(mode=context.GRAPH_MODE, device_target="Ascend")
-    context.set_context(enable_task_sink=True,
-                        device_id=device_id)
-    context.set_context(enable_ir_fusion=True)
+    context.set_context(device_id=device_id)
     context.set_context(enable_loop_sink=False)
     distributedTool.init()
     device_num = distributedTool.get_group_size()
@@ -86,15 +84,15 @@ class DataGenerator():
         return data
 
     def input_data(self, shape):
-        data = (self.generate_data(shape)*2).astype(np.float32)
-        stra = [1]*len(shape)
+        data = (self.generate_data(shape) * 2).astype(np.float32)
+        stra = [1] * len(shape)
         stra[0] = device_num
         datas = self.get_parallel_blocks(data, stra)
         return Tensor(data), Tensor(datas[rank_id])
 
     def label_data(self, shape, classes):
-        data = (self.generate_data(shape)*(classes-1)).astype(np.int32)
-        stra = [1]*len(shape)
+        data = (self.generate_data(shape) * (classes - 1)).astype(np.int32)
+        stra = [1] * len(shape)
         stra[0] = device_num
         datas = self.get_parallel_blocks(data, stra)
         return Tensor(data), Tensor(datas[rank_id])

tests/st/auto_parallel/soft_entropy_loss_expand_parallel.py

@@ -37,7 +37,7 @@ device_id = int(os.getenv('DEVICE_ID'))
 rank_id = 0
 embed = 128
 classes = 32
-batch_size = 32*2
+batch_size = 32 * 2
 MatmulParamShape = (classes, embed)
 
 
@@ -46,9 +46,7 @@ def setup_module():
     global rank_id
     np.random.seed(0)
     context.set_context(mode=context.GRAPH_MODE, device_target="Ascend")
-    context.set_context(enable_task_sink=True,
-                        device_id=device_id)
-    context.set_context(enable_ir_fusion=True)
+    context.set_context(device_id=device_id)
     context.set_context(enable_loop_sink=False)
     distributedTool.init()
     rank_id = distributedTool.get_rank()
@@ -77,20 +75,20 @@ class DataGenerator():
     def generate_data(self, shape):
         size = np.cumprod(shape)[-1]
         num_range = min(size, 1000)
-        data = (np.arange(0, size) % num_range)/num_range
+        data = (np.arange(0, size) % num_range) / num_range
         data = np.reshape(data, shape)
         return data
 
     def input_data(self, shape):
-        data = (self.generate_data(shape)*0.1).astype(np.float32)
-        stra = [1]*len(shape)
+        data = (self.generate_data(shape) * 0.1).astype(np.float32)
+        stra = [1] * len(shape)
         stra[0] = device_num
         datas = self.get_parallel_blocks(data, stra)
         return Tensor(data), Tensor(datas[rank_id])
 
     def label_data(self, shape, embed):
-        data = (self.generate_data(shape)*(embed-1)).astype(np.int32)
-        stra = [1]*len(shape)
+        data = (self.generate_data(shape) * (embed - 1)).astype(np.int32)
+        stra = [1] * len(shape)
         stra[0] = device_num
         datas = self.get_parallel_blocks(data, stra)
         return Tensor(data), Tensor(datas[rank_id])
@@ -141,7 +139,7 @@ class SoftmaxCrossEntropyExpand(Cell):
     def __init__(self, sparse=False, stra_list=[]):
         super(SoftmaxCrossEntropyExpand, self).__init__()
         if len(stra_list) < 11:
-            stra_list = [None]*11
+            stra_list = [None] * 11
         self.exp = P.Exp()
         self.reduce_sum = P.ReduceSum(keep_dims=True).set_strategy(strategy=stra_list[1])
         self.onehot = P.OneHot().set_strategy(strategy=stra_list[2])

tests/st/auto_parallel/test_resnet50_expand_loss_2p.py

@@ -31,8 +31,7 @@ from mindspore.train.callback import Callback
 from mindspore.parallel import set_algo_parameters
 
 context.set_context(mode=context.GRAPH_MODE, device_target="Ascend")
-context.set_context(enable_task_sink=True, device_id=int(os.getenv('DEVICE_ID')))
-context.set_context(enable_ir_fusion=True)
+context.set_context(device_id=int(os.getenv('DEVICE_ID')))
 context.set_context(enable_loop_sink=False)
 init()
 context.set_auto_parallel_context(mirror_mean=True, parallel_mode=ParallelMode.AUTO_PARALLEL)
@@ -316,14 +315,14 @@ class DataGenerator():
 
     def input_data(self, shape):
         data = (self.generate_data(shape)).astype(np.float32)
-        stra = [1]*len(shape)
+        stra = [1] * len(shape)
         stra[0] = device_num
         datas = self.get_parallel_blocks(data, stra)
         return Tensor(data), Tensor(datas[rank_id])
 
     def label_data(self, shape):
-        data = (self.generate_data(shape)*1000/np.prod(shape)).astype(np.int32)
-        stra = [1]*len(shape)
+        data = (self.generate_data(shape) * 1000 / np.prod(shape)).astype(np.int32)
+        stra = [1] * len(shape)
         stra[0] = device_num
         datas = self.get_parallel_blocks(data, stra)
         return Tensor(data), Tensor(datas[rank_id])
@@ -378,8 +377,8 @@ def test_train_feed(num_classes=8192):
     set_algo_parameters(elementwise_op_strategy_follow=True)
     parallel_callback = ModelCallback()
     dataGen = DataGenerator()
-    input_full, input_part = dataGen.input_data((32*2, 3, 224, 224))
-    label_full, label_part = dataGen.label_data((32*2,))
+    input_full, input_part = dataGen.input_data((32 * 2, 3, 224, 224))
+    label_full, label_part = dataGen.label_data((32 * 2,))
     dataset = Dataset(input_part, label_part)
     net = resnet50(num_classes)
     loss = SoftmaxCrossEntropyExpand(sparse=True)
@@ -398,8 +397,8 @@ def test_train_feed2(num_classes=1001):
     set_algo_parameters(elementwise_op_strategy_follow=True)
     parallel_callback = ModelCallback()
     dataGen = DataGenerator()
-    input_full, input_part = dataGen.input_data((32*2, 3, 224, 224))
-    label_full, label_part = dataGen.label_data((32*2,))
+    input_full, input_part = dataGen.input_data((32 * 2, 3, 224, 224))
+    label_full, label_part = dataGen.label_data((32 * 2,))
     dataset = Dataset(input_part, label_part)
     net = resnet50(num_classes)
     loss = SoftmaxCrossEntropyExpand(sparse=True)

tests/st/control/test_multigraph_sink.py

@@ -14,17 +14,14 @@
 # ============================================================================
 """ test_multigraph_sink """
 import pytest
-import numpy as np
-import mindspore.nn as nn
 import mindspore.context as context
 from mindspore.common.tensor import Tensor
 from mindspore.common import dtype as mstype
 from mindspore.common import ms_function
-from mindspore.ops import operations as P
 
 
 def setup_module(module):
-    context.set_context(mode = context.PYNATIVE_MODE, device_target = "Ascend")
+    context.set_context(mode=context.PYNATIVE_MODE, device_target="Ascend")
 
 
 c1 = Tensor([2], mstype.int32)
@@ -208,4 +205,3 @@ def test_while_in_while_in_while():
     output = while_in_while_in_while(c1, c2, c3)
     expect = Tensor([2534], mstype.int32)
     assert output == expect
-

tests/st/control/test_while.py

@@ -31,7 +31,6 @@ def t1_while(x, y, z):
 
 def test_net():
     context.set_context(mode=context.GRAPH_MODE, device_target="Ascend")
-    context.set_context(enable_task_sink=True)
     c1 = Tensor([2], mstype.int32)
     c2 = Tensor([14], mstype.int32)
     c3 = Tensor([1], mstype.int32)

tests/st/fusion/test_conv_bn1_fusion.py

@@ -21,7 +21,7 @@ from mindspore.common.initializer import initializer
 from mindspore.nn.loss import SoftmaxCrossEntropyWithLogits
 from mindspore.nn.optim import Momentum
 
-context.set_context(device_target="Ascend", enable_task_sink=True)
+context.set_context(device_target="Ascend")
 
 input_channel = 2048
 output_channel = 512