archinetai / surgeon-pytorch Goto Github PK

Is it possible to save the results from the backbone and apply them on the heads of the all the other models. My goal was to try to save time by avoiding repeating the backbone part. Instead of running the 3 complete models (left), only run the backbone 1 time and switch only the heads for the 3 models (right), therefore not repeating executing the backbone every time in yolov5 model.

Thank you for the help!

Hi, I'm a student studying the deep learning compiler for DL frameworks.

I'm also making a program that handles a computational graph(CG) of pytorch by using the 'torch.fx' library, but always failed to derive a output result of node constituting a CG.

I wonder how you make it possible to read the output result of a node with the Extract class?

Is it related to the override function name of 'call_module'?

Thank you for reading my quetion.

Hi, I noticed that the current version does not support parallel models (at least those created using torch.nn.DataParallel) since the forward hook does not differentiate between the different copies of the model and a model wrapped with Inspect will just return the intermediate features of the last copy of the parallelized model to run.

Are you planning on fixing this issue/supporting this use case?