1. Remove fluid
2. Remove unused codes
2.1 remove adabert
2.2 remove ddpg controller of NAS
2.3 remove paddleslim.models
2.4 Remove PC-DARTS

* support fc in pruning

* support matmul and matmul_v2 pruning

* support ffn pruning with gelu activations in the middle

* support reshape2, transpose and split in transformer block

* support pattern: qkv gemm -> batched gemm -> out linear

* prune all fc layer

* support fc pruning

* almost done with few hardcode

* remove hardcode

* fix UT

* fix UT

* avoid setting attributes for reshape in fc pruning

* fix UT

* fix UT

* add static masks for quant

* Update unstructured_pruner.rst

* Update unstructured_prune_api.rst

* add the UT

* fix typo

* API docs
Co-authored-by: Ngmm <38800877+mmglove@users.noreply.github.com>

* fix a bug in skip params func

* polish distill

* optimize unstructured pruner using GMP

* add uniittests for GMP pruner

* add training scripts for fleet and gmp; add conv1x1 pruning

* add conv1x1 pruning to dygraph

* unify hyper parameters across methods and scripts, static graph

* unify hyper parameters across methods and scripts, dynamic graph

* refine test cases

* fix singlecard training in static graph

* refine readmes

* update unstructured pruner api doc static

* update api docs dygraph

* fix bugs in api docs

* rename UnstructuredPrunerGMP to GMPUnstructuredPruner

* fix some format issues

* remove redundant methods/arguments

* description for gmp configs

* fix minor bugs

* fix test cases accordingly
Co-authored-by: Nwhs <wanghaoshuang@baidu.com>

* refine distillation

* up

* add test

* fix coverage

* fix unit test error

* update quant docs

* Update quantization_api.rst

* Update quantization_api.rst

* Update quantization_api.rst

* Update qat.py

与之前版本兼容：之前版本直接改动传入的model，所以这里inplace改为true

* add dygraph ptq and refine dygraph qat

* add unit test for ptq

* update according to comments

* add distill dirs

* fix unittest

* prune the optimizer in dygraph mode

* set opt when initializing pruner

* tests passed locally

* update Chinese docs and tutorials

* refine docs style

* avoid ambiguation in docs

Fix the error of save_quantized_model when weight_preprocess is PACT