fix quantization docs (#760)

* modified README and demo

* Update qat.rst

* Update quantization_api.rst

* Update quant_post_static_tutorial_en.md

* Update quant_aware_tutorial_en.md

* Update quant_aware_training_tutorial.md

* Update quant_aware_tutorial.md

* Update dygraph_quant_post_tutorial.md

* Update quant_post_static_tutorial_en.md

* Update quant_aware_tutorial_en.md

* Update quant_aware_training_tutorial.md

docs/en/quick_start/quant_aware_tutorial_en.md

@@ -10,13 +10,14 @@ This tutorial shows how to do training-aware quantization using [API](https://pa
 6. Save model after quantization
 
 ## 1. Necessary imports
-PaddleSlim depends on Paddle1.7. Please make true that you have installed Paddle correctly. Then do the necessary imports:
+Please make true that you have installed Paddle correctly. Then do the necessary imports:
 
 ```python
 import paddle
 import paddle.fluid as fluid
 import paddleslim as slim
 import numpy as np
+paddle.enable_static()
 ```
 
 ## 2. Model architecture
@@ -56,7 +57,7 @@ Define functions to train and test model. We only need call the functions when f
 def train(prog):
     iter = 0
     for data in train_reader():
-        acc1, acc5, loss = exe.run(prog, feed=train_feeder.feed(data), fetch_list=outputs)
+        acc1, acc5, loss, out = exe.run(prog, feed=train_feeder.feed(data), fetch_list=outputs)
         if iter % 100 == 0:
             print('train iter={}, top1={}, top5={}, loss={}'.format(iter, acc1.mean(), acc5.mean(), loss.mean()))
         iter += 1
@@ -65,7 +66,7 @@ def test(prog):
     iter = 0
     res = [[], []]
     for data in train_reader():
-        acc1, acc5, loss = exe.run(prog, feed=train_feeder.feed(data), fetch_list=outputs)
+        acc1, acc5, loss, out = exe.run(prog, feed=train_feeder.feed(data), fetch_list=outputs)
         if iter % 100 == 0:
             print('test iter={}, top1={}, top5={}, loss={}'.format(iter, acc1.mean(), acc5.mean(), loss.mean()))
         res[0].append(acc1.mean())

docs/en/quick_start/quant_post_static_tutorial_en.md

@@ -16,6 +16,7 @@ import paddle
 import paddle.fluid as fluid
 import paddleslim as slim
 import numpy as np
+paddle.enable_static()
 ```
 ## 2. Model architecture
 
@@ -39,10 +40,10 @@ To speed up training process, we select MNIST dataset to train image classificat
 
 ```python
 import paddle.dataset.mnist as reader
+paddle.enable_static()
 train_reader = paddle.fluid.io.batch(
         reader.train(), batch_size=128, drop_last=True)
 test_reader = paddle.fluid.io.batch(
-cs/en/quick_start/quant_aware_tutorial_en.md
         reader.train(), batch_size=128, drop_last=True)
 train_feeder = fluid.DataFeeder(inputs, fluid.CPUPlace())
 ```
@@ -55,7 +56,7 @@ Define functions to train and test model. We only need call the functions when f
 def train(prog):
     iter = 0
     for data in train_reader():
-        acc1, acc5, loss = exe.run(prog, feed=train_feeder.feed(data), fetch_list=outputs)
+        acc1, acc5, loss, out = exe.run(prog, feed=train_feeder.feed(data), fetch_list=outputs)
         if iter % 100 == 0:
             print('train', acc1.mean(), acc5.mean(), loss.mean())
         iter += 1
@@ -64,7 +65,7 @@ def test(prog, outputs=outputs):
     iter = 0
     res = [[], []]
     for data in train_reader():
-        acc1, acc5, loss = exe.run(prog, feed=train_feeder.feed(data), fetch_list=outputs)
+        acc1, acc5, loss, out = exe.run(prog, feed=train_feeder.feed(data), fetch_list=outputs)
         if iter % 100 == 0:
             print('test', acc1.mean(), acc5.mean(), loss.mean())
         res[0].append(acc1.mean())

docs/zh_cn/api_cn/dygraph/quanter/qat.rst

@@ -78,6 +78,7 @@ QAT
 
    .. code-block:: python
 
+      import paddle
       from paddle.vision.models import mobilenet_v1
       from paddleslim import QAT
       net = mobilenet_v1(pretrained=False) 
@@ -86,7 +87,7 @@ QAT
           'quantizable_layer_type': ['Conv2D', 'Linear'],
       }
       quanter = QAT(config=quant_config)
-      quanter.quantize(lenet)
+      quanter.quantize(net)
       paddle.summary(net, (1, 3, 224, 224))
    
    ..  
@@ -111,6 +112,7 @@ QAT
 
    .. code-block:: python
 
+      import paddle
       from paddle.vision.models import mobilenet_v1
       from paddleslim import QAT
       net = mobilenet_v1(pretrained=False) 
@@ -119,7 +121,7 @@ QAT
           'quantizable_layer_type': ['Conv2D', 'Linear'],
       }
       quanter = QAT(config=quant_config)
-      quanter.quantize(lenet)
+      quanter.quantize(net)
       paddle.summary(net, (1, 3, 224, 224))
 
       quanter.save_quantized_model(

docs/zh_cn/api_cn/static/quant/quantization_api.rst

@@ -78,10 +78,12 @@ quant_post_dynamic
 
 .. code-block:: python
 
+   import paddle
    import paddle.fluid as fluid
    import paddle.dataset.mnist as reader
    from paddleslim.quant import quant_post_dynamic
    
+   paddle.enable_static()
    quant_post_dynamic(
            model_dir='./model_path',
            save_model_dir='./save_path',
@@ -166,9 +168,11 @@ quant_post_static
 
 .. code-block:: python
 
+   import paddle
    import paddle.fluid as fluid
    import paddle.dataset.mnist as reader
    from paddleslim.quant import quant_post_static
+   paddle.enable_static()
    val_reader = reader.train()
    use_gpu = True
    place = fluid.CUDAPlace(0) if use_gpu else fluid.CPUPlace()
@@ -265,10 +269,11 @@ convert
 .. code-block:: python
 
    #encoding=utf8
+   import paddle
    import paddle.fluid as fluid
    import paddleslim.quant as quant
    
-   
+   paddle.enable_static()
    train_program = fluid.Program()
    
    with fluid.program_guard(train_program):
@@ -398,9 +403,11 @@ fluid.Program
 
 .. code-block:: python
 
+   import paddle
    import paddle.fluid as fluid
    import paddleslim.quant as quant
    
+   paddle.enable_static()
    train_program = fluid.Program()
    with fluid.program_guard(train_program):
        input_word = fluid.data(name="input_word", shape=[None, 1], dtype='int64')

docs/zh_cn/quick_start/dygraph/dygraph_quant_post_tutorial.md

@@ -20,6 +20,7 @@
 
 ```python
 import paddle
+import paddleslim
 import paddle.vision.models as models
 from paddle.static import InputSpec as Input
 from paddle.vision.datasets import Cifar10
@@ -61,7 +62,7 @@ model.evaluate(val_dataset, batch_size=256, verbose=1)
 
 训练完成后导出预测模型:
 ```python
-paddle.jit.save(net, "./fp32_inference_model", input_spec=[inputs])
+paddle.jit.save(net, "./fp32_inference_model", input_spec=inputs)
 ```
 
 
@@ -79,7 +80,7 @@ paddleslim.quant.quant_post_static(
         model_filename='fp32_inference_model.pdmodel',
         params_filename='fp32_inference_model.pdiparams',
         quantize_model_path='./quant_post_static_model',
-        sample_generator=train_dataset,
+        sample_generator=paddle.dataset.cifar.test10(),
         batch_nums=10)
 ```
 

docs/zh_cn/quick_start/static/quant_aware_tutorial.md

@@ -168,7 +168,7 @@ paddle.static.save_inference_model(
         feed_vars=[image],
         fetch_vars=target_vars,
         executor=exe,
-        program=float_prog)
+        program=quant_infer_program)
 ```
 
 根据业务场景，可以使用PaddleLite将该量化模型部署到移动端（ARM CPU），或者使用PaddleInference将该量化模型部署到服务器端（NV GPU和Intel CPU）。

docs/zh_cn/tutorials/quant/dygraph/quant_aware_training_tutorial.md

@@ -39,7 +39,10 @@ quant_config = {
 在确认好我们的量化配置以后，我们可以根据这个配置把我们定义好的一个普通模型转换为一个模拟量化模型。转换的方式也很简单:
 
 ```python
+import paddle
 import paddleslim
+from paddle.vision.models import mobilenet_v1
+net = mobilenet_v1()
 quanter = paddleslim.QAT(config=quant_config)
 quanter.quantize(net)
 ```
@@ -54,10 +57,11 @@ quanter.quantize(net)
 
 ```python
 import paddleslim
+save_path = './model'
 quanter.save_quantized_model(
-  model,
+  net,
   save_path,
-  input_spec=[paddle.static.InputSpec()])
+  input_spec=[paddle.static.InputSpec(shape=[None, 3, 224, 224], dtype='float32')])
 ```
 
 量化预测模型可以使用`netron`软件打开，进行可视化查看。该量化预测模型和普通FP32预测模型一样，可以使用PaddleLite和PaddleInference加载预测，具体请参考`推理部署`章节。