patch nas&dist docs (#761)

* update nas and distill docs

docs/en/api_en/search_space_en.md

@@ -47,8 +47,8 @@ For example, how to add a search space with resnet block. New search space can N
 
 ```python
 ### import necessary head file
-from .search_space_base import SearchSpaceBase
-from .search_space_registry import SEARCHSPACE
+from paddleslim.nas import SearchSpaceBase
+from paddleslim.nas import SEARCHSPACE
 import numpy as np
 
 ### use decorator SEARCHSPACE.register to register yourself search space to search space NameSpace

docs/en/quick_start/distillation_tutorial_en.md

@@ -25,6 +25,7 @@ This tutorial trains and verifies distillation model on the MNIST dataset. The i
 Select `ResNet50` as the teacher to perform distillation training on the students of the` MobileNet` architecture.
 
 ```python
+import paddleslim as slim
 model = slim.models.MobileNet()
 student_program = fluid.Program()
 student_startup = fluid.Program()

docs/en/quick_start/nas_tutorial_en.md

@@ -111,6 +111,7 @@ archs = sanas.next_archs()[0]
 ### 7.2 build program
 Get program according to the function in Step3 and model architecture from Step 7.1.
 ```python
+paddle.enable_static()
 exe, train_program, eval_program, inputs, avg_cost, acc_top1, acc_top5 = build_program(archs)
 ```
 

docs/zh_cn/api_cn/dygraph/ofa/convert_supernet_api.rst

@@ -61,7 +61,7 @@ PaddleSlim提供了三种方式构造超网络，下面分别介绍这三种方
 
   model = mobilenet_v1()
   sp_net_config = supernet(kernel_size=(3, 5, 7), expand_ratio=[1, 2, 4])
-  sp_model = Convert(sp_net_config).convert(self.model)
+  sp_model = Convert(sp_net_config).convert(model)
 
 方式二
 ------------------

docs/zh_cn/api_cn/dygraph/ofa/ofa_api.rst

@@ -54,6 +54,8 @@ DistillConfig
 .. code-block:: python
 
   from paddleslim.nas.ofa import DistillConfig
+  from paddle.vision.models import mobilenet_v1
+  teacher_model = mobilenet_v1()
   default_distill_config = {
       'lambda_distill': 0.01,
       'teacher_model': teacher_model,

docs/zh_cn/api_cn/static/dist/single_distiller_api.rst

@@ -26,9 +26,10 @@ merge
 **使用示例：**
 
 .. code-block:: python
-
+   import paddle
    import paddle.fluid as fluid
    import paddleslim.dist as dist
+   paddle.enable_static()
    student_program = fluid.Program()
    with fluid.program_guard(student_program):
        x = fluid.layers.data(name='x', shape=[1, 28, 28])
@@ -73,9 +74,10 @@ fsp_loss出自论文 `A Gift from Knowledge Distillation: Fast Optimization, Net
 **使用示例：**
 
 .. code-block:: python
-
+   import paddle
    import paddle.fluid as fluid
    import paddleslim.dist as dist
+   paddle.enable_static()
    student_program = fluid.Program()
    with fluid.program_guard(student_program):
        x = fluid.layers.data(name='x', shape=[1, 28, 28])
@@ -119,9 +121,10 @@ l2_loss
 **使用示例：**
 
 .. code-block:: python
-
+   import paddle
    import paddle.fluid as fluid
    import paddleslim.dist as dist
+   paddle.enable_static()
    student_program = fluid.Program()
    with fluid.program_guard(student_program):
        x = fluid.layers.data(name='x', shape=[1, 28, 28])
@@ -169,9 +172,10 @@ soft_label_loss出自论文 `Distilling the Knowledge in a Neural Network <https
 **使用示例：**
 
 .. code-block:: python
-
+   import paddle
    import paddle.fluid as fluid
    import paddleslim.dist as dist
+   paddle.enable_static()
    student_program = fluid.Program()
    with fluid.program_guard(student_program):
        x = fluid.layers.data(name='x', shape=[1, 28, 28])
@@ -215,9 +219,10 @@ loss
 **使用示例：**
 
 .. code-block:: python
-
+   import paddle
    import paddle.fluid as fluid
    import paddleslim.dist as dist
+   paddle.enable_static()
    student_program = fluid.Program()
    with fluid.program_guard(student_program):
        x = fluid.layers.data(name='x', shape=[1, 28, 28])

docs/zh_cn/api_cn/static/nas/nas_api.rst

@@ -49,7 +49,7 @@ SANAS（Simulated Annealing Neural Architecture Search）是基于模拟退火
    from paddleslim.nas import SANAS
    config = [('MobileNetV2Space')]
    paddle.enable_static()
-   sanas = SANAS(configs=config)
+   sanas = SANAS(configs=config, server_addr=（"", 8881))
 
 .. note::
 
@@ -88,7 +88,7 @@ SANAS（Simulated Annealing Neural Architecture Search）是基于模拟退火
       from paddleslim.nas import SANAS
       config = [('MobileNetV2Space')]
       paddle.enable_static()
-      sanas = SANAS(configs=config)
+      sanas = SANAS(configs=config, server_addr=（"", 8882))
       input = paddle.static.data(name='input', shape=[None, 3, 32, 32], dtype='float32')
       archs = sanas.next_archs()
       for arch in archs:
@@ -115,7 +115,7 @@ SANAS（Simulated Annealing Neural Architecture Search）是基于模拟退火
       from paddleslim.nas import SANAS
       config = [('MobileNetV2Space')]
       paddle.enable_static()
-      sanas = SANAS(configs=config)
+      sanas = SANAS(configs=config, server_addr=（"", 8883))
       archs = sanas.next_archs()
       
       ### 假设网络计算出来的score是1，实际代码中使用时需要返回真实score。
@@ -142,7 +142,7 @@ SANAS（Simulated Annealing Neural Architecture Search）是基于模拟退火
       from paddleslim.nas import SANAS
       config = [('MobileNetV2Space')]
       paddle.enable_static()
-      sanas = SANAS(configs=config)
+      sanas = SANAS(configs=config, server_addr=（"", 8884))
       input = paddle.static.data(name='input', shape=[None, 3, 32, 32], dtype='float32')
       tokens = ([0] * 25)
       archs = sanas.tokens2arch(tokens)[0]
@@ -163,7 +163,7 @@ SANAS（Simulated Annealing Neural Architecture Search）是基于模拟退火
       from paddleslim.nas import SANAS
       config = [('MobileNetV2Space')]
       paddle.enable_static()
-      sanas = SANAS(configs=config)
+      sanas = SANAS(configs=config, server_addr=（"", 8885))
       print(sanas.current_info())
 
 
@@ -233,7 +233,7 @@ RLNAS (Reinforcement Learning Neural Architecture Search）是基于强化学习
    config = [('MobileNetV2Space')]
 
    paddle.enable_static()
-   rlnas = RLNAS(key='lstm', configs=config)
+   rlnas = RLNAS(key='lstm', configs=config, server_addr=（"", 8886))
 
 
 .. py:method:: next_archs(obs=None)
@@ -255,7 +255,7 @@ RLNAS (Reinforcement Learning Neural Architecture Search）是基于强化学习
   from paddleslim.nas import RLNAS
   config = [('MobileNetV2Space')]
   paddle.enable_static()
-  rlnas = RLNAS(key='lstm', configs=config)
+  rlnas = RLNAS(key='lstm', configs=config, server_addr=（"", 8887))
   input = paddle.static.data(name='input', shape=[None, 3, 32, 32], dtype='float32')
   archs = rlnas.next_archs(1)[0]
   for arch in archs:
@@ -280,7 +280,7 @@ RLNAS (Reinforcement Learning Neural Architecture Search）是基于强化学习
   from paddleslim.nas import RLNAS
   config = [('MobileNetV2Space')]
   paddle.enable_static()
-  rlnas = RLNAS(key='lstm', configs=config)
+  rlnas = RLNAS(key='lstm', configs=config, server_addr=（"", 8888))
   rlnas.next_archs(1)
   rlnas.reward(1.0)
 
@@ -307,7 +307,7 @@ RLNAS (Reinforcement Learning Neural Architecture Search）是基于强化学习
   from paddleslim.nas import RLNAS
   config = [('MobileNetV2Space')]
   paddle.enable_static()
-  rlnas = RLNAS(key='lstm', configs=config)
+  rlnas = RLNAS(key='lstm', configs=config, server_addr=（"", 8889))
   archs = rlnas.final_archs(1)
   print(archs)
 
@@ -330,7 +330,7 @@ RLNAS (Reinforcement Learning Neural Architecture Search）是基于强化学习
   from paddleslim.nas import RLNAS
   config = [('MobileNetV2Space')]
   paddle.enable_static()
-  rlnas = RLNAS(key='lstm', configs=config)
+  rlnas = RLNAS(key='lstm', configs=config, server_addr=（"", 8891))
   input = paddle.static.data(name='input', shape=[None, 3, 32, 32], dtype='float32')
   tokens = ([0] * 25)
   archs = rlnas.tokens2arch(tokens)[0]

docs/zh_cn/quick_start/static/distillation_tutorial.md

@@ -28,6 +28,7 @@ paddle.enable_static()
 选择`ResNet50`作为teacher对`MobileNet`结构的student进行蒸馏训练。
 
 ```python
+import paddleslim as slim
 model = slim.models.MobileNet()
 student_program = paddle.static.Program()
 student_startup = paddle.static.Program()

docs/zh_cn/tutorials/nas/dygraph/nas_ofa.md

@@ -29,38 +29,38 @@ OFA的基本流程分为以下步骤：
    训练配置默认根据论文中PS的训练模式进行配置，可进行配置的参数和含义可以参考: [RunConfig](https://paddleslim.readthedocs.io/zh_CN/latest/api_cn/dygraph/ofa/ofa_api.html)
 
 ```python
-  from paddleslim.nas.ofa import RunConfig
-  default_run_config = {
-      'train_batch_size': 256,
-      'n_epochs': [[1], [2, 3], [4, 5]],
-      'init_learning_rate': [[0.001], [0.003, 0.001], [0.003, 0.001]],
-      'dynamic_batch_size': [1, 1, 1],
-      'total_images': 1281167,
-      'elastic_depth': (2, 5, 8)
-  }
-  run_config = RunConfig(**default_run_config)
+from paddleslim.nas.ofa import RunConfig
+default_run_config = {
+    'train_batch_size': 256,
+    'n_epochs': [[1], [2, 3], [4, 5]],
+    'init_learning_rate': [[0.001], [0.003, 0.001], [0.003, 0.001]],
+    'dynamic_batch_size': [1, 1, 1],
+    'total_images': 1281167,
+    'elastic_depth': (2, 5, 8)
+}
+run_config = RunConfig(**default_run_config)
 ```
 
 ### 3. 蒸馏配置
   为OFA训练过程添加蒸馏配置，可进行配置的参数和含义可以参考: [DistillConfig](https://paddleslim.readthedocs.io/zh_CN/latest/api_cn/dygraph/ofa/ofa_api.html#distillconfig)
 
 ```python
-  from paddle.vision.models import mobilenet_v1
-  from paddleslim.nas.ofa import DistillConfig
-  teacher_model = mobilenet_v1()
-
-  default_distill_config = {
-      'teacher_model': teacher_model
-  }
-  distill_config = DistillConfig(**default_distill_config)
+from paddle.vision.models import mobilenet_v1
+from paddleslim.nas.ofa import DistillConfig
+teacher_model = mobilenet_v1()
+
+default_distill_config = {
+    'teacher_model': teacher_model
+}
+distill_config = DistillConfig(**default_distill_config)
 ```
 
 ### 4. 传入模型和相应配置
   用OFA封装模型、训练配置和蒸馏配置。配置完模型和正常模型训练流程相同。如果添加了蒸馏，则OFA封装后的模型会比原始模型多返回一组教师网络的输出。
 ```python
-  from paddleslim.nas.ofa import OFA
+from paddleslim.nas.ofa import OFA
 
-  ofa_model = OFA(model, run_config=run_config, distill_config=distill_config)
+ofa_model = OFA(model, run_config=run_config, distill_config=distill_config)
 ```
 
 ## 实验效果