feat: onnx auto bn for kneron quantization

tools/pytorch2onnx_kneron.py

@@ -2,6 +2,7 @@
 # Original: tools/pytorch2onnx.py, modified by Kneron
 import argparse
 
+import os
 import onnx
 import mmcv
 import numpy as np
@@ -19,6 +20,7 @@ from mmseg.apis.inference import LoadImage
 from mmseg.datasets.pipelines import Compose
 from mmseg.models import build_segmentor
 
+from optimizer_scripts.tools import other
 from optimizer_scripts.pytorch_exported_onnx_preprocess import (
     torch_exported_onnx_flow,
 )
@@ -26,6 +28,19 @@ from optimizer_scripts.pytorch_exported_onnx_preprocess import (
 torch.manual_seed(3)
 
 
+def _parse_normalize_cfg(test_pipeline):
+    transforms = None
+    for pipeline in test_pipeline:
+        if 'transforms' in pipeline:
+            transforms = pipeline['transforms']
+            break
+    assert transforms is not None, 'Failed to find `transforms`'
+    norm_config_li = [_ for _ in transforms if _['type'] == 'Normalize']
+    assert len(norm_config_li) == 1, '`norm_config` should only have one'
+    norm_config = norm_config_li[0]
+    return norm_config
+
+
 def _convert_batchnorm(module):
     module_output = module
     if isinstance(module, torch.nn.SyncBatchNorm):
@@ -80,6 +95,7 @@ def _prepare_input_img(img_path,
 
 def pytorch2onnx(model,
                  img,
+                 norm_cfg=None,
                  opset_version=11,
                  show=False,
                  output_file='tmp.onnx',
@@ -130,6 +146,7 @@ def pytorch2onnx(model,
         m.ir_version = 6
     m = torch_exported_onnx_flow(m, disable_fuse_bn=False)
     onnx.save(m, output_file)
+    print(f'{output_file} optimized by KNERON successfully.')
 
     if verify:
         onnx_model = onnx.load(output_file)
@@ -190,6 +207,40 @@ def pytorch2onnx(model,
             err_msg='The outputs are different between Pytorch and ONNX')
         print('The outputs are same between Pytorch and ONNX')
 
+    if norm_cfg is not None:
+        mean = norm_cfg['mean']
+        std = norm_cfg['std']
+        # TODO: figure out should we add the lines below?
+        '''
+        if all(_ == 128. for _ in mean) and all(_ == 256. for _ in std):
+            print("normalization config perfectly matches kneron "
+                  "quantization requirement; not prepending batch norm.")
+            return
+        print("normalization config does not match kneron "
+              "quantization requirement; prepending batch norm...")
+        '''
+        i_n = m.graph.input[0]
+        if (
+            i_n.type.tensor_type.shape.dim[1].dim_value != len(mean)
+            or i_n.type.tensor_type.shape.dim[1].dim_value != len(std)
+        ):
+            raise ValueError(
+                f"--pixel-bias-value ({mean}) and --pixel-scale-value "
+                f"({std}) should be same as input dimension: "
+                f"{i_n.type.tensor_type.shape.dim[1].dim_value}"
+            )
+        norm_bn_bias = [-1 * cm / cs + 128. / cs for cm, cs in zip(mean, std)]
+        norm_bn_scale = [1 / cs for cs in std]
+        other.add_bias_scale_bn_after(
+            m.graph, i_n.name, norm_bn_bias, norm_bn_scale
+        )
+        m = other.polish_model(m)
+        bn_outf = os.path.splitext(output_file)[0] + "_kneron_optimized.onnx"
+        onnx.save(m, bn_outf)
+        print(f"ONNX for quantization saved to {bn_outf}")
+
+    return
+
 
 def parse_args():
     parser = argparse.ArgumentParser(description='Convert MMSeg to ONNX')
@@ -276,10 +327,12 @@ if __name__ == '__main__':
     else:
         img = _demo_mm_inputs(input_shape)
 
+    norm_cfg = _parse_normalize_cfg(cfg.test_pipeline)
     # convert model to onnx file
     pytorch2onnx(
         segmentor,
         img,
+        norm_cfg=norm_cfg,
         opset_version=args.opset_version,
         show=args.show,
         output_file=args.output_file,