merge

Signed-off-by: zehao-intel <[email protected]>

Author: zehao-intel

.azure-pipelines/scripts/codeScan/pyspelling/inc_dict.txt

@@ -2397,15 +2397,25 @@ grappler
 amsgrad
 qoperator
 apis
+<<<<<<< HEAD:.azure-pipelines/scripts/codeScan/pyspelling/inc_dict.txt
 PostTrainingQuantConfig
 dgpu
 CPz
 PostTrainingQuantConfig
 dgpu
+=======
+AccuracyCriterion
+TuningCriterion
+CPz
+>>>>>>> 93a5017ddc484b28ed0169574a7127b184385b34:.azure-pipelines/scripts/codeScan/pyspelling/lpot_dict.txt
 Nsh
 UmK
 fe
 vmware
+<<<<<<< HEAD:.azure-pipelines/scripts/codeScan/pyspelling/inc_dict.txt
+=======
+PythonLauncher
+>>>>>>> 93a5017ddc484b28ed0169574a7127b184385b34:.azure-pipelines/scripts/codeScan/pyspelling/lpot_dict.txt
 keepachangelog
 vscode
 IntelNeuralCompressor

examples/tensorflow/nlp/bert_base_mrpc/quantization/ptq/README.md

@@ -87,6 +87,12 @@ This is a tutorial of how to enable bert model with Intel® Neural Compressor.
 2. User specifies fp32 *model*, calibration dataset *q_dataloader* and a custom *eval_func* which encapsulates the evaluation dataset and metric by itself.
 
 For bert, we applied the first one as we already have write dataset and metric for bert mrpc task. 
+<<<<<<< HEAD
+
+### Quantization Config
+The Quantization Config class has default parameters setting for running on Intel CPUs. If running this example on Intel GPUs, the 'backend' parameter should be set to 'itex' and the 'device' parameter should be set to 'gpu'.
+=======
+>>>>>>> 93a5017ddc484b28ed0169574a7127b184385b34
 
 ### Quantization Config
 The Quantization Config class has default parameters setting for running on Intel CPUs. If running this example on Intel GPUs, the 'backend' parameter should be set to 'itex' and the 'device' parameter should be set to 'gpu'.
@@ -100,6 +106,17 @@ config = PostTrainingQuantConfig(
     ...
     )
 ```
+<<<<<<< HEAD
+config = PostTrainingQuantConfig(
+    device="gpu",
+    backend="itex",
+    inputs=["input_file", "batch_size"],
+    outputs=["loss/Softmax:0", "IteratorGetNext:3"],
+    ...
+    )
+```
+=======
+>>>>>>> 93a5017ddc484b28ed0169574a7127b184385b34
 Here we set the input tensor and output tensors name into *inputs* and *outputs* args. In this case we calibrate and quantize the model, and use our calibration dataloader initialized from a 'Dataset' object.
 
 ### Code update
@@ -135,5 +152,5 @@ After prepare step is done, we add tune and benchmark code to generate quantized
         accuracy = evaluate(model.graph_def)
         print('Batch size = %d' % FLAGS.eval_batch_size)
         print("Accuracy: %.5f" % accuracy)
-```
+
 The Intel® Neural Compressor quantization.fit() function will return a best quantized model under time constraint.

examples/tensorflow/nlp/distilbert_base/quantization/ptq/README.md

@@ -170,7 +170,8 @@ After prepare step is done, we add the code for quantization tuning to generate
     from neural_compressor.benchmark import fit
     from neural_compressor.config import BenchmarkConfig
     if ARGS.mode == 'performance':
-        fit(model, conf=BenchmarkConfig(), b_func=eval_func)
+        conf = BenchmarkConfig(cores_per_instance=28, num_of_instance=1)
+        fit(graph, conf, b_func=self.eval_func)
     elif ARGS.mode == 'accuracy':
         self.eval_func(graph)
 ```