humu789
humu789
Sorry for not replying in time because of missing the GitHub notice. Could you provide us with your config for analyzing better?
Sorry, mmaction2 is not supported in the current version. Welcome to contribute the feature to mmrazor, or waiting for the support in the furture.
1. We will release quantization in the second half of the year. 2. Our quantization will be more friendly to OpenMMLab's users, it will be applied easily to the upstream...
@tanghy2016 `AutoSlim: Our current StructurePruner does not support pruning this architecture. StructurePruner is not perfect enough to handle all the corner cases. We will appreciate it if you create a...
Please give us your log to analyze it better
Got it，we'll check it as soon.
@JennieDD The size of distilled ckpt is bigger than original ckpt is due to that both of teacher's and student's ckpts are saved in one, we will fix it as...
The useage is same as `mmdetection`'s. You can try `--eval--options` -> `--eval-options`.
@Veiiling Could you provide us with your config to analyze it better?
 @Yaolang003 In the paper， experiments tell us that pre-training on ImageNet can get a better result. Whether do your final question mean to distill your searcher model with DetNAS?