Open Alex-Songs opened 2 years ago
hello @Alex-Songs, you could customize the replace function and pass it in by customized_replace_func
API
please refer Conv2d replacement and implement Conv1d: https://github.com/microsoft/nni/blob/7811307c240ff3e87166d48c8d012da426852838/nni/compression/pytorch/speedup/compress_modules.py#L443
It's welcome to contribute back to NNI after you implement it.😆
RuntimeError: Has not supported replacing module with type: Conv1d, you could report an issue at https://github.com/microsoft/nni. If you know how to replace Conv1d, you could implement module replacement by passing in
customized_replace_func
toModelSpeedup
. You are welcome to contribute back to nni as native support if you have implemented the replacement function, so that more users can benefit from your contributions.