masahi commented on pull request #7513: URL: https://github.com/apache/tvm/pull/7513#issuecomment-785553307
I generally don't like a half baked solution, but if that allows more models to run on TVM, that doesn't sound too bad. If people agree with this, I think we can go with this @t-vi @siju-samuel @kevinthesun @codeislife99 @alexwong @apivovarov If you want to pursue this path, please revisit the `index_put` -> `scatter_nd` solution. This is what PyTorch ONNX converter does to convert `copy_`. Your solution probably doesn't work for dynamic shape inputs, but copying dynamic tensors should be conceptually no harder than for static case. ---------------------------------------------------------------- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: [email protected]
