[RELAY][FRONEND] Frontend scalfolding

[tqchen]

This is an issue to coordinate frontend scaffolding for relay. To help us coordinate, I have tagged a list of folks who originated the initial scaffolding of nnvm part to initiate the move, please feel free to reply to this issue and coordinate the tasks.

• ONNX @zhreshold
• Keras @Huyuwei
• Tensorflow @srkreddy1238
• DarkNet @PariksheetPinjari909
• Caffe2
• CoreML
• NNVMv1 to relay
• MXNet

[srkreddy1238]

#2216 for tensorflow

[zhreshold]

#2302 for onnx

[Huyuwei]

#2336 for keras

[tqchen]

@Huyuwei can you also help to migrate the frontend tutorial into tutorials/frontend?

[Huyuwei]

#2479 migrates from_keras tutorial to tutorials/frontend

[makihiro]

#2507 for Caffe2

[tqchen]

We still have CoreML left, @kazum would you be interested in help on that?

[zhiics]

Is #2476 the CoreML frontend?

[tqchen]

Sorry I missed track of things, yes CoreML is already in, only Darknet left then :) Actually, I see why I missed the CoreML, because we have yet to update the tutorials in https://docs.tvm.ai/tutorials/index.html#compile-deep-learning-models

[kazum]

@tqchen @zhiics I created a CoreML tutorial instead, #2667 :)

[kazum]

@PariksheetPinjari909, do you have a time to implement a DarkNet support? If no, I'd be happy to take over the work from you :) What do you think of it?

[srkreddy1238]

I know @siju-samuel was working on this front end some time back.
@siju-samuel Can you update the status ?

[siju-samuel]

I was working on bringing yolov3 to Darknet relay frontend sometime back and then i got busy with some other work. In a week, i can finish the remaining and PR.
Darknet frontend depends on #1941 and #2275, i will work on this PRs and then i will upstream the frontend PR.