Change Doc

docs/en_US/Compressor/Overview.md

@@ -118,7 +118,8 @@ class YourPruner(nni.compressors.tf_compressor.TfPruner):
     def update_epoch(self, epoch_num, sess):
         pass
 
-    def step(self):
+    # note for pytorch version, there is no sess in input arguments
+    def step(self, sess):
         # can do some processing based on the model or weights binded
         # in the func bind_model
         pass
@@ -159,10 +160,18 @@ class YourPruner(nni.compressors.tf_compressor.TfQuantizer):
     def update_epoch(self, epoch_num, sess):
         pass
 
-    def step(self):
+    # note for pytorch version, there is no sess in input arguments
+    def step(self, sess):
         # can do some processing based on the model or weights binded
         # in the func bind_model
         pass
+
+    # you can also design your method
+    def your_method(self, your_input):
+        #your code
+
+    def bind_model(self, model):
+        #preprocess model
 ```
 
 __[TODO]__ Will add another member function `quantize_layer_output`, as some quantization algorithms also quantize layers' output.

docs/en_US/Compressor/Pruner.md

@@ -13,16 +13,21 @@ We first sort the weights in the specified layer by their absolute values. And t
 
 Tensorflow code
 ```
-pruner = nni.compressors.tf_compressor.LevelPruner([{'sparsity':0.8,'support_type': 'default'}])
+configure_list = [{'sparsity':0.8,'support_type': 'default'}]
+pruner = nni.compressors.tf_compressor.LevelPruner(configure_list)
 pruner(model_graph)
 ```
 
 Pytorch code
 ```
-pruner = nni.compressors.torch_compressor.LevelPruner([{'sparsity':0.8,'support_type': 'default'}])
+configure_list = [{'sparsity':0.8,'support_type': 'default'}]
+pruner = nni.compressors.torch_compressor.LevelPruner(configure_list)
 pruner(model)
 ```
 
+#### User configuration for LevelPruner
+* **sparsity:** This is to specify the sparsity operations to be compressed to
+
 ***
 <a name="AGPruner"></a>
 
@@ -41,13 +46,29 @@ First, you should import pruner and add mask to model.
 Tensorflow code
 ```
 from nni.compressors.tfCompressor import AGPruner
-pruner = AGPruner(initial_sparsity=0, final_sparsity=0.8, start_epoch=1, end_epoch=10, frequency=1)
+configure_list = [{
+                        'initial_sparsity': 0,
+                        'final_sparsity': 0.8,
+                        'start_epoch': 1,
+                        'end_epoch': 10,
+                        'frequency': 1,
+                        'support_type': 'default'
+                    }]
+pruner = AGPruner(configure_list)
 pruner(tf.get_default_graph())
 ```
 Pytorch code
 ```
 from nni.compressors.torchCompressor import AGPruner
-pruner = AGPruner(initial_sparsity=0, final_sparsity=0.8, start_epoch=1, end_epoch=10, frequency=1)
+configure_list = [{
+                        'initial_sparsity': 0,
+                        'final_sparsity': 0.8,
+                        'start_epoch': 1,
+                        'end_epoch': 10,
+                        'frequency': 1,
+                        'support_type': 'default'
+                    }]
+pruner = AGPruner(configure_list)
 pruner(model)
 ```
 
@@ -62,6 +83,14 @@ Pytorch code
 pruner.update_epoch(epoch)
 ```
 You can view example for more information
+
+#### User configuration for AGPruner
+* **initial_sparsity:** This is to specify the sparsity when compressor starts to  compress
+* **final_sparsity:** This is to specify the sparsity when compressor finishes to  compress
+* **start_epoch:** This is to specify the epoch number when compressor starts to  compress
+* **end_epoch:** This is to specify the epoch number when compressor finishes to  compress
+* **frequency:** This is to specify every *frequency* number epochs compressor compress once
+
 ***
 <a name="SensitivityPruner"></a>
 
@@ -76,15 +105,15 @@ You can prune weight step by step and reach one target sparsity by SensitivityPr
 Tensorflow code
 ```
 from nni.compressors.tfCompressor import SensitivityPruner
-
-pruner = SensitivityPruner(sparsity = 0.8)
+configure_list = [{'sparsity':0.8,'support_type': 'default'}]
+pruner = SensitivityPruner(configure_list)
 pruner(tf.get_default_graph())
 ```
 Pytorch code
 ```
 from nni.compressors.torchCompressor import SensitivityPruner
-
-pruner = SensitivityPruner(sparsity = 0.8)
+configure_list = [{'sparsity':0.8,'support_type': 'default'}]
+pruner = SensitivityPruner(configure_list)
 pruner(model)
 ```
 Like AGPruner, you should update mask information every epoch by adding code below
@@ -98,4 +127,8 @@ Pytorch code
 pruner.update_epoch(epoch)
 ```
 You can view example for more information
+
+#### User configuration for SensitivityPruner
+* **sparsity:** This is to specify the sparsity operations to be compressed to
+
 ***

docs/en_US/Compressor/Quantizer.md

@@ -5,7 +5,7 @@ Quantizer on NNI Compressor
 
 ## NaiveQuantizer
 
-We provide NaiveQuantizer to quantizer weight to default 8 bits, you can use it to test quantize algorithm.
+We provide NaiveQuantizer to quantizer weight to default 8 bits, you can use it to test quantize algorithm without any configure.
 
 ### Usage
 tensorflow
@@ -16,6 +16,7 @@ pytorch
 ```
 nni.compressors.torch_compressor.NaiveQuantizer()(model)
 ```
+
 ***
 <a name="QATquantizer"></a>
 
@@ -34,18 +35,24 @@ You can quantize your model to 8 bits with the code below before your training c
 Tensorflow code
 ```
 from nni.compressors.tfCompressor import QATquantizer
-quantizer = QATquantizer(q_bits = 8)
+configure_list = [{'q_bits':8, 'support_type':'default'}]
+quantizer = QATquantizer(configure_list)
 quantizer(tf.get_default_graph())
 ```
 Pytorch code
 ```
 from nni.compressors.torchCompressor import QATquantizer
-quantizer = QATquantizer(q_bits = 8)
+configure_list = [{'q_bits':8, 'support_type':'default'}]
+quantizer = QATquantizer(configure_list)
 quantizer(model)
 ```
 
 You can view example for more information
 
+#### User configuration for QATquantizer
+* **q_bits:** This is to specify the q_bits operations to be quantized to
+
+
 ***
 <a name="DoReFaQuantizer"></a>
 
@@ -58,14 +65,19 @@ To implement DoReFaQuantizer, you can add code below before your training code
 Tensorflow code
 ```
 from nni.compressors.tfCompressor import DoReFaQuantizer
-quantizer = DoReFaQuantizer(q_bits = 8)
+configure_list = [{'q_bits':8, 'support_type':'default'}]
+quantizer = DoReFaQuantizer(configure_list)
 quantizer(tf.get_default_graph())
 ```
 Pytorch code
 ```
 from nni.compressors.torchCompressor import DoReFaQuantizer
-quantizer = DoReFaQuantizer(q_bits = 8)
+configure_list = [{'q_bits':8, 'support_type':'default'}]
+quantizer = DoReFaQuantizer(configure_list)
 quantizer(model)
 ```
 
 You can view example for more information
+
+#### User configuration for QATquantizer
+* **q_bits:** This is to specify the q_bits operations to be quantized to

src/sdk/pynni/nni/compressors/tf_compressor/_nnimc_tf.py

@@ -32,7 +32,7 @@ def compress(self, model):
         """
         assert self._bound_model is None, "Each NNI compressor instance can only compress one model"
         self._bound_model = model
-        self.preprocess_model(model)
+        self.bind_model(model)
 
     def compress_default_graph(self):
         """
@@ -42,7 +42,7 @@ def compress_default_graph(self):
         self.compress(tf.get_default_graph())
 
 
-    def preprocess_model(self, model):
+    def bind_model(self, model):
         """
         This method is called when a model is bound to the compressor.
         Users can optionally overload this method to do model-specific initialization.

src/sdk/pynni/nni/compressors/torch_compressor/_nnimc_torch.py

@@ -31,10 +31,10 @@ def compress(self, model):
         """
         assert self._bound_model is None, "Each NNI compressor instance can only compress one model"
         self._bound_model = model
-        self.preprocess_model(model)
+        self.bind_model(model)
 
 
-    def preprocess_model(self, model):
+    def bind_model(self, model):
         """
         This method is called when a model is bound to the compressor.
         Users can optionally overload this method to do model-specific initialization.