mmpose
mmpose copied to clipboard
How to export the model to torchscript?
How to export the model to torchscript?
Hi we haven't provide tools yet. Please request the feature here https://github.com/open-mmlab/mmpose/issues/9 to help us prioritize.
Can the network export the onnx including decode ? like HeatmapParser function
requesting module to the ~~DJ~~ programming team as well, keep up the good work guys you rock
any update on this?
onnx conversion works, but torchscript conversion fails due to custom sub-class of pytorch
I know im a little late on this thread but i was having a similar issue to @alicera for exporting a module using onnx. The model only returns the heat map and not the decoded keypoints. Is this still an issue? If not, can someone point me to an onnx example?
I also want to know, how to convert mmpose models to onnx. Does anyone have an example code?
@Monalsingh @rtaylo45 Now we have MMDeploy to support model export and deployment for MMPose. An example is provided in MMDeploy documents.
Would like to have Tracing support for mmPose models. I have been able to successfully use 'torch2torchscript' under mmdeploy.apis to trace mmSegmentation Models. However, using the same on mmPose (tried configs under dekr and associative+embedding) with the following mmdeploy config: '\mmdeploy\configs\mmpose\pose-detection_torchscript.py' would throw the following error:
File "mmdetection\mmpose\mmpose\datasets\pipelines\shared_transform.py", line 176, in call meta[key_tgt] = results[key_src]
KeyError: 'flip_index'
Added this to the feature request page. Posting here to see if there are any possible workarounds.
Now,Can export the model to torchscript?