PocketFlow

PocketFlow copied to clipboard

Describe the bug @jiaxiang-wu The compressed model with ChannelPrunedRmtLearner cannot be converted to .pd or .tflite file using tools/conversion/export_chn_pruned_tflite_model.py or tools/conversion/export_pb_tflite_models.py.

To Reproduce Steps to reproduce the behavior:

1. I have got the models_cpr_eval folder.
2. Then I used the tools/conversion/export_pb_tflite_models.py to convert the compressed model.
3. But I got this error.

Additional context And I read the usage and results of channel pruning for SSD model in PR #228. I think the respective compressed model by a different method, such as backbone-only (layers from VGG) or all layer (both layers from VGG and extra layers following the VGG), at the different prune ratio, should have the different model size.

Am I right?

if I am right, which file can show this phenomenon, the checkpoint file or the .pb or .tflite file converted from the checkpoint file?

Thanks a lot !

@jiaxiang-wu Can you take a look at this issue and provide some solutions?

你好，请问训练好了全精度的SSD后，要怎么用channel pruning自动压缩SSD？

@le8888e Please refer to #228

你好，请问训练好了全精度的SSD后，要怎么用channel pruning自动压缩SSD？

请问你的效果怎么样呀