Some important feedforward models to be supported at first stage

[kuke]

We are going to support the conversion of models as follows at first stage. They all can be found in our models bank, some maybe only need to be verified after #19 merged, and some lack of necessary operators.

• recognize_digits ( Support conversion of recognize_digits models #19 @kuke)
• ResNet50 (Enhance batch_norm_op & rename output args sharing names with inputs #30 Add vgg16 & resnet50 to supported models #32 @kuke)
• VGG16 (Enhance batch_norm_op & rename output args sharing names with inputs #30 Add vgg16 & resnet50 to supported models #32 @kuke)
• MobileNet (Enhance batch_norm_op & rename output args sharing names with inputs #30 Add mobilenet & se_resnext to supported models #33 @kuke)
• SE-ResNeXt (Enhance batch_norm_op & rename output args sharing names with inputs #30 Add mobilenet & se_resnext to supported models #33 @kuke)
• Inception-v4 (first add the Fluid config in models/fluid/image_classification) (#900 #49 @kuke)

Even all the models above are supported, only a small subset of operators are used and verified. So for the rest operators, we may need another task:

• Design the unit test framework to test every operator's conversion (Add unit test framework for operators' conversion #29 @kuke)

Hints:

• Choose one task every time and assign yourself when you are really ready for working on it;
• Please append the model to the supported models section in README like Support conversion of recognize_digits models #19 after its conversion being validated.

[pkuyym]

@kuke I think we can collect all the ops needed to be converted required by the listed models.
It's better to do Design the unit test framework to test every operator's conversion first and we can implement converter for fluid ops parallel. When all operator converter finished, I think it would be easy to validate the models.

[sidgoyal78]

Yeah, I agree with @pkuyym . I think it might be better to do design the unit test framework for a variety of operators.

[kuke]

I also agree with that unit test is very important, so I assign it to myself. But if you are interested in verifying the correctness of models' conversion, you can work on one now. From my experience, a runnable model will help you test the related operator's conversion and the unit test is not necessary.

[varunarora]

I'll take READMEs/installation, etc. Thanks @kuke and @pkuyym for your work on the models.

[kuke]

@varunarora Great! It is the time to have a friendly README and installation. I am going to continue to work on models' conversion. I think rest of us can pick up a model or general operators' conversion. You see we have a lot of operators waiting to be implemented the conversion.