I think it would be preferable to add DCNv2 instead of original DCN at this point for both pytorch and onnx. https://github.com/CharlesShang/DCNv2

Yes, you are right. I want to convert yolact_plus to onnx but onnx don't support DCNv2.

import DCNv2
torch.onnx.export(net, batch, "yolact_plus.onnx",operator_export_type=DCNv2)

Error

Multiple GPUs detected! Turning off JIT.
Config not specified. Parsed yolact_plus_base_config from the file name.

Loading model... Done.
/home/kiwi/Document/yolact/DCNv2/dcn_v2.py:31: TracerWarning: Converting a tensor to a Python integer might cause the trace to be incorrect. We can't record the data flow of Python values, so this value will be treated as a constant in the future. This means that the trace might not generalize to other inputs!
ctx.deformable_groups)
/home/kiwi/Document/yolact/yolact.py:221: TracerWarning: Converting a tensor to a Python boolean might cause the trace to be incorrect. We can't record the data flow of Python values, so this value will be treated as a constant in the future. This means that the trace might not generalize to other inputs!
if self.last_img_size != (cfg._tmp_img_w, cfg._tmp_img_h):
Traceback (most recent call last):
File "eval.py", line 1123, in
evaluate(net, dataset)
File "eval.py", line 895, in evaluate
evalimage(net, args.image)
File "eval.py", line 609, in evalimage
torch.onnx.export(net, batch, "yolact_plus.onnx",operator_export_type=DCNv2)
File "/home/kiwi/miniconda3/envs/detectron2/lib/python3.6/site-packages/torch/onnx/init.py", line 148, in export
strip_doc_string, dynamic_axes, keep_initializers_as_inputs)
File "/home/kiwi/miniconda3/envs/detectron2/lib/python3.6/site-packages/torch/onnx/utils.py", line 66, in export
dynamic_axes=dynamic_axes, keep_initializers_as_inputs=keep_initializers_as_inputs)
File "/home/kiwi/miniconda3/envs/detectron2/lib/python3.6/site-packages/torch/onnx/utils.py", line 416, in _export
fixed_batch_size=fixed_batch_size)
File "/home/kiwi/miniconda3/envs/detectron2/lib/python3.6/site-packages/torch/onnx/utils.py", line 296, in _model_to_graph
fixed_batch_size=fixed_batch_size, params_dict=params_dict)
File "/home/kiwi/miniconda3/envs/detectron2/lib/python3.6/site-packages/torch/onnx/utils.py", line 135, in _optimize_graph
graph = torch._C._jit_pass_onnx(graph, operator_export_type)
TypeError: _jit_pass_onnx(): incompatible function arguments. The following argument types are supported:
1. (arg0: torch::jit::Graph, arg1: torch._C._onnx.OperatorExportTypes) -> torch::jit::Graph

Really need onnx to support DCN (deformal conv) operator.... #2485

Description

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions