Add mobilenet & se_resnext to supported models

[kuke]

Use flowers data set (102 classes) to train the model.

MobileNet:

-----------  Configuration Arguments -----------
a: 0.0
b: 1.0
batch_size: 10
expected_decimal: 5
fluid_model: extras/mobilenet.inference.model/
onnx_model: mobilenet.onnx
------------------------------------------------
Inference results for fluid model:
[array([[0.01249686, 0.00800683, 0.00538686, ..., 0.00885275, 0.01019887,
        0.00645085],
       [0.01233307, 0.00802851, 0.00547722, ..., 0.00911936, 0.01021425,
        0.00654219],
       [0.01261485, 0.00802507, 0.00543512, ..., 0.00885483, 0.01012875,
        0.00641232],
       ...,
       [0.01282006, 0.0080833 , 0.00547658, ..., 0.00893381, 0.01017397,
        0.00650065],
       [0.01262719, 0.00823347, 0.00558914, ..., 0.00892504, 0.01017115,
        0.00638958],
       [0.01250273, 0.00804726, 0.00548675, ..., 0.0089457 , 0.0100737 ,
        0.00649769]], dtype=float32)]


Inference results for ONNX model:
Outputs(_0=array([[0.01249686, 0.00800683, 0.00538686, ..., 0.00885275, 0.01019888,
        0.00645085],
       [0.01233307, 0.00802851, 0.00547722, ..., 0.00911936, 0.01021425,
        0.00654219],
       [0.01261485, 0.00802507, 0.00543512, ..., 0.00885483, 0.01012875,
        0.00641232],
       ...,
       [0.01282006, 0.0080833 , 0.00547659, ..., 0.00893382, 0.01017398,
        0.00650065],
       [0.0126272 , 0.00823346, 0.00558914, ..., 0.00892504, 0.01017114,
        0.00638958],
       [0.01250273, 0.00804726, 0.00548675, ..., 0.00894571, 0.0100737 ,
        0.00649769]], dtype=float32))


The exported model achieves 5-decimal precision.

SE_ResNeXt:

-----------  Configuration Arguments -----------
a: 0.0
b: 1.0
batch_size: 10
expected_decimal: 5
fluid_model: extras/re_resnext_flowers.inference.model/
onnx_model: re_resnext_flowers.onnx
------------------------------------------------
Inference results for fluid model:
[array([[0.00420744, 0.00678189, 0.0047285 , ..., 0.00482811, 0.00527949,
        0.00482003],
       [0.00420748, 0.00678191, 0.00472855, ..., 0.0048281 , 0.00527948,
        0.00482005],
       [0.00420748, 0.00678191, 0.00472855, ..., 0.0048281 , 0.00527948,
        0.00482005],
       ...,
       [0.00420747, 0.0067819 , 0.00472854, ..., 0.0048281 , 0.00527948,
        0.00482005],
       [0.00420746, 0.0067819 , 0.00472853, ..., 0.0048281 , 0.00527948,
        0.00482004],
       [0.00420746, 0.0067819 , 0.00472852, ..., 0.0048281 , 0.00527949,
        0.00482004]], dtype=float32)]


Inference results for ONNX model:
Outputs(_0=array([[0.00420744, 0.00678189, 0.0047285 , ..., 0.00482811, 0.00527949,
        0.00482003],
       [0.00420748, 0.00678191, 0.00472855, ..., 0.0048281 , 0.00527948,
        0.00482005],
       [0.00420748, 0.00678191, 0.00472855, ..., 0.0048281 , 0.00527948,
        0.00482005],
       ...,
       [0.00420747, 0.0067819 , 0.00472854, ..., 0.0048281 , 0.00527948,
        0.00482005],
       [0.00420746, 0.0067819 , 0.00472853, ..., 0.0048281 , 0.00527948,
        0.00482005],
       [0.00420746, 0.0067819 , 0.00472852, ..., 0.0048281 , 0.00527949,
        0.00482004]], dtype=float32))


The exported model achieves 5-decimal precision.