EmbedMask ONNX导出和TensorRT加速

---

• 2019.12.8:
  ONNX导出已经支持，代码在神力平台：
  http://manaai.cn
  TensorRT 加速 WIP

运行python3 demo/demo_embed_mask.py报错

Traceback (most recent call last):
File "demo/demo_embed_mask.py", line 124, in
main()
File "demo/demo_embed_mask.py", line 90, in main
min_image_size=args.min_image_size
File "./demo/predictor.py", line 119, in init
_ = checkpointer.load(cfg.MODEL.WEIGHT)
File "./fcos_core/utils/checkpoint.py", line 62, in load
self._load_model(checkpoint)
File "./fcos_core/utils/checkpoint.py", line 98, in _load_model
load_state_dict(self.model, checkpoint.pop("model"))
File "./fcos_core/utils/model_serialization.py", line 80, in load_state_dict
model.load_state_dict(model_state_dict)
File "/usr/local/lib/python3.6/dist-packages/torch/nn/modules/module.py", line 845, in load_state_dict
self.class.name, "\n\t".join(error_msgs)))
RuntimeError: Error(s) in loading state_dict for GeneralizedRCNN:
size mismatch for rpn.head.proposal_margin_pred.weight: copying a param with shape torch.Size([1, 256, 3, 3]) from checkpoint, the shape in current model is torch.Size([1, 4, 1, 1]).

@jinfagang 谢谢你们的工作，等待开源

我们现在在做centermask了。
embedmask似乎没有centermask效果好。

EmbedMask ONNX 支持c++的tensorRT加速么？