search

open-mmlab / mmrazor

OpenMMLab Model Compression Toolbox and Benchmark.
https://mmrazor.readthedocs.io/en/latest/
Apache License 2.0
1.47k stars 227 forks source link

What should I do to prune my own segmentation model with my own datasets? #583

Open HePengguang opened 1 year ago

HePengguang commented 1 year ago

Could you please give us an example how to prune own segmentation model from scratch? Thank you so much.

LKJacky commented 1 year ago

Please refer to link, where we provide several configs to prune classification and detection models

HePengguang commented 1 year ago

But

Please refer to link, where we provide several configs to prune classification and detection models

It is different with mmcls or mmdet. And I cannot use this command to get the target_pruning_ratio, it always occurred error like "When I ran the script pruning/get_channel_units.py, I got this error: TypeError: expected Tensor as element 0 in argument 0, but got tuple. #585". So could you please roll out a model pruning process for semantic segmentation tasks, especially with custom datasets? Thank you so much. image