Remove some redundant classes for quantization, benchmark and mixed precision (#840)

Signed-off-by: Cheng, Penghui <[email protected]>

Author: PenghuiCheng

docs/source/benchmark.md

@@ -49,7 +49,7 @@ And please make sure `cores_per_instance * num_of_instance` must be less than CP
 from neural_compressor.config import BenchmarkConfig
 from neural_compressor.benchmark import fit
 conf = BenchmarkConfig(warmup=10, iteration=100, cores_per_instance=4, num_of_instance=7)
-fit(model='./int8.pb', config=conf, b_dataloader=eval_dataloader)
+fit(model='./int8.pb', conf=conf, b_dataloader=eval_dataloader)
 ```
 
 ## Examples

docs/source/mixed_precision.md

@@ -42,8 +42,8 @@ Supported precisions for mix precision include bf16 and fp16. If users want to g
 from neural_compressor import mix_precision
 from neural_compressor.config import MixedPrecisionConfig
 
-conf = MixedPrecisionConfig(precision='bf16')
-converted_model = mix_precision.fit(model, config=conf)
+conf = MixedPrecisionConfig() # default precision is bf16
+converted_model = mix_precision.fit(model, conf=conf)
 converted_model.save('./path/to/save/')
 ```
 
@@ -56,8 +56,8 @@ from neural_compressor.config import MixedPrecisionConfig
 conf = MixedPrecisionConfig(
         backend='onnxrt_cuda_ep',
         device='gpu',
-        precision='fp16')
-converted_model = mix_precision.fit(model, config=conf)
+        precisions='fp16')
+converted_model = mix_precision.fit(model, conf=conf)
 converted_model.save('./path/to/save/')
 ```
 

docs/source/pruning.md

@@ -301,7 +301,8 @@ The following section exemplifies how to use hooks in user pass-in training func
      [**Experimental option** ]Modify model and optimizer.
 
     ```python
-        from neural_compressor.training import prepare_pruning, WeightPruningConfig 
+        from neural_compressor import WeightPruningConfig
+        from neural_compressor.experimental.compression import prepare_pruning
         config = WeightPruningConfig(configs)
         prepare_pruning(config, model, optimizer) # modify model and optimizer
         for epoch in range(num_train_epochs):

examples/helloworld/tf_example5/README.md

@@ -52,7 +52,7 @@ python test.py --benchmark --dataset_location=/path/to/imagenet/
 ```python
     from neural_compressor.benchmark import fit
     conf = BenchmarkConfig(iteration=100, cores_per_instance=4, num_of_instance=1)
-    fit(model='./int8.pb', config=conf, b_dataloader=eval_dataloader)
+    fit(model='./int8.pb', conf=conf, b_dataloader=eval_dataloader)
 
 ```
 

examples/keras/image_recognition/inception_resnet_v2/quantization/ptq/main.py

@@ -119,7 +119,7 @@ def main(_):
     if FLAGS.tune:
         from neural_compressor.quantization import fit
         from neural_compressor.config import PostTrainingQuantConfig
-        from neural_compressor.utils.utility import set_random_seed
+        from neural_compressor import set_random_seed
         set_random_seed(9527)
         config = PostTrainingQuantConfig(backend='itex', 
             calibration_sampling_size=[20, 150])
@@ -139,7 +139,7 @@ def main(_):
             fit(FLAGS.input_model, conf, b_func=evaluate)
         else:
             from neural_compressor.model.model import Model
-            accuracy = evaluate(Model(FLAGS.input_model, backend='keras').model)
+            accuracy = evaluate(Model(FLAGS.input_model, backend='itex').model)
             logger.info('Batch size = %d' % FLAGS.batch_size)
             logger.info("Accuracy: %.5f" % accuracy)
 

examples/keras/image_recognition/inception_v3/quantization/ptq/main.py

@@ -119,7 +119,7 @@ def main(_):
     if FLAGS.tune:
         from neural_compressor.quantization import fit
         from neural_compressor.config import PostTrainingQuantConfig
-        from neural_compressor.utils.utility import set_random_seed
+        from neural_compressor import set_random_seed
         set_random_seed(9527)
         config = PostTrainingQuantConfig(backend='itex', 
             calibration_sampling_size=[50, 100])
@@ -139,9 +139,10 @@ def main(_):
             fit(FLAGS.input_model, conf, b_func=evaluate)
         else:
             from neural_compressor.model.model import Model
-            accuracy = evaluate(Model(FLAGS.input_model, backend='keras').model)
+            accuracy = evaluate(Model(FLAGS.input_model, backend='itex').model)
             logger.info('Batch size = %d' % FLAGS.batch_size)
             logger.info("Accuracy: %.5f" % accuracy)
 
+
 if __name__ == "__main__":
     tf.compat.v1.app.run()

examples/keras/image_recognition/mobilenet_v2/quantization/ptq/main.py

@@ -118,7 +118,7 @@ def main(_):
     if FLAGS.tune:
         from neural_compressor.quantization import fit
         from neural_compressor.config import PostTrainingQuantConfig
-        from neural_compressor.utils.utility import set_random_seed
+        from neural_compressor import set_random_seed
         set_random_seed(9527)
         config = PostTrainingQuantConfig(backend='itex', 
             calibration_sampling_size=[50, 100])
@@ -138,7 +138,7 @@ def main(_):
             fit(FLAGS.input_model, conf, b_func=evaluate)
         else:
             from neural_compressor.model.model import Model
-            accuracy = evaluate(Model(FLAGS.input_model, backend='keras').model)
+            accuracy = evaluate(Model(FLAGS.input_model, backend='itex').model)
             logger.info('Batch size = %d' % FLAGS.batch_size)
             logger.info("Accuracy: %.5f" % accuracy)
 

examples/keras/image_recognition/resnet101/quantization/ptq/main.py

@@ -125,7 +125,7 @@ def main(_):
     if FLAGS.tune:
         from neural_compressor.quantization import fit
         from neural_compressor.config import PostTrainingQuantConfig
-        from neural_compressor.utils.utility import set_random_seed
+        from neural_compressor import set_random_seed
         set_random_seed(9524)
         config = PostTrainingQuantConfig(backend='itex', 
             calibration_sampling_size=[10, 15])
@@ -145,7 +145,7 @@ def main(_):
             fit(FLAGS.input_model, conf, b_func=evaluate)
         else:
             from neural_compressor.model.model import Model
-            accuracy = evaluate(Model(FLAGS.input_model, backend='keras').model)
+            accuracy = evaluate(Model(FLAGS.input_model, backend='itex').model)
             logger.info('Batch size = %d' % FLAGS.batch_size)
             logger.info("Accuracy: %.5f" % accuracy)
 

examples/keras/image_recognition/resnet50/quantization/ptq/main.py

@@ -110,7 +110,7 @@ def eval_func(dataloader, metric):
     return acc
 
 def main(_):
-    from neural_compressor.utils import set_random_seed
+    from neural_compressor import set_random_seed
     set_random_seed(9527)
     if FLAGS.tune:
         from neural_compressor import quantization
@@ -130,7 +130,7 @@ def main(_):
             fit(FLAGS.input_model, conf, b_func=evaluate)
         else:
             from neural_compressor.model import Model
-            model = Model(FLAGS.input_model, backend='keras').model
+            model = Model(FLAGS.input_model, backend='itex').model
             accuracy = evaluate(model)
             print('Batch size = %d' % FLAGS.batch_size)
             print("Accuracy: %.5f" % accuracy)

examples/keras/image_recognition/resnet50_fashion/quantization/ptq/main.py

@@ -100,7 +100,7 @@ def eval_func(data_loader, metric):
     return acc
 
 def main(_):
-    from neural_compressor.utils import set_random_seed
+    from neural_compressor import set_random_seed
     set_random_seed(9527)
     if FLAGS.tune:
         from neural_compressor import quantization
@@ -128,7 +128,7 @@ def main(_):
             fit(FLAGS.input_model, conf, b_func=evaluate)
         else:
             from neural_compressor.model import Model
-            model = Model(FLAGS.input_model, backend='keras').model
+            model = Model(FLAGS.input_model, backend='itex').model
             accuracy = evaluate(model)
             print('Batch size = %d' % FLAGS.batch_size)
             print("Accuracy: %.5f" % accuracy)