* fix demo

* add unittest

* update unittest

* optimize readme

* update

* add eval.yaml

* fix train config

* fix

* Update Some Demos & Add Infer for ImageNet Demo

* Update Some Demos & Add Infer for ImageNet Demo

* Update Some Demos & Add Infer for ImageNet Demo

* Update Some Demos & Add Infer for ImageNet Demo

* fix save_inference_model

* Fix config and wrapper of dataloader

* Fix wrapper of dataloader
1. Skip wrapper when input is format like [{'x': data1, 'y': data2}]
2. Fix scripts in nlp demo of ACT

1. Make config of ACT support path, set, list and tuple
2. Refine the implementation of loading config
3. Add wrapper for dataloader to return data in dict format

* support voc dataset in act

* fix readme

* fix find op itself in transformer prune

* update

* optimization yaml

* update yaml

* update yaml

* fix unittest

* fix unittest

* replace eval dataloader with train dataloader if eval_dataloader is None

* update

* add YOLOv5s ACT demo

* fix comment

* fix docs

1. fix temp work dir not found error
2. move temp work dir to final directory formated by pid and datetime
3. remove temp work dir of ACT when job is done
4. remove temp work dir of quant_post_hpo when job is done

* solve prediction problems.

* fix docs.

* Update README.md

* Update README.md

* Update README.md

* [auto-compression-asp] support fleet training and asp+distillation in auto compression