add post quant (#4255)

configs/slim/README.md

@@ -4,10 +4,11 @@
 
 - [剪裁](prune)
 - [量化](quant)
+- [离线量化](post_quant)
 - [蒸馏](distill)
 - [联合策略](extensions)
 
-推荐您使用剪裁和蒸馏联合训练，或者使用剪裁和量化，进行检测模型压缩。 下面以YOLOv3为例，进行剪裁、蒸馏和量化实验。
+推荐您使用剪裁和蒸馏联合训练，或者使用剪裁、量化训练和离线量化，进行检测模型压缩。 下面以YOLOv3为例，进行剪裁、蒸馏和量化实验。
 
 ## 实验环境
 
@@ -20,7 +21,8 @@
 **PaddleDetection、 PaddlePaddle与PaddleSlim 版本关系:**
 |  PaddleDetection版本  | PaddlePaddle版本  | PaddleSlim版本 |  备注    |
 | :------------------: | :---------------: | :-------: |:---------------: |
-| release/2.1       |       >= 2.1.0       | 2.1      |  量化模型导出依赖最新Paddle develop分支，可在[PaddlePaddle每日版本](https://www.paddlepaddle.org.cn/documentation/docs/zh/install/Tables.html#whl-dev)中下载安装  |
+| release/2.3       |       >= 2.1       | 2.1      | 离线量化依赖Paddle 2.2及PaddleSlim 2.2 |
+| release/2.1 | 2.2    |       >= 2.1.0       | 2.1      |  量化模型导出依赖最新Paddle develop分支，可在[PaddlePaddle每日版本](https://www.paddlepaddle.org.cn/documentation/docs/zh/install/Tables.html#whl-dev)中下载安装  |
 | release/2.0       |       >= 2.0.1       | 2.0      | 量化依赖Paddle 2.1及PaddleSlim 2.1 |
 
 
@@ -145,6 +147,16 @@ python tools/export_model.py -c configs/{MODEL.yml} --slim_config configs/slim/{
 - 上述V100预测时延非量化模型均是使用TensorRT-FP32测试，量化模型均使用TensorRT-INT8测试，并且都包含NMS耗时。
 - SD855预测时延为使用PaddleLite部署，使用arm8架构并使用4线程(4 Threads)推理时延。
 
+### 离线量化
+需要准备val集，用来对离线量化模型进行校准，运行方式：
+```shell
+python tools/post_quant.py -c configs/{MODEL.yml} --slim_config configs/slim/post_quant/{SLIM_CONFIG.yml}
+```
+例如：
+```shell
+python3.7 tools/post_quant.py -c configs/ppyolo/ppyolo_mbv3_large_coco.yml --slim_config=configs/slim/post_quant/ppyolo_mbv3_large_ptq.yml
+```
+
 ### 蒸馏
 
 #### COCO上benchmark

configs/slim/post_quant/ppyolo_mbv3_large_ptq.yml

+weights: https://paddledet.bj.bcebos.com/models/ppyolo_mbv3_large_coco.pdparams
+slim: PTQ
+
+PTQ:
+  ptq_config: {
+      'activation_quantizer': 'HistQuantizer',
+      'upsample_bins': 127,
+      'hist_percent': 0.999}
+  quant_batch_num: 10
+  fuse: True

ppdet/engine/export_utils.py

@@ -20,6 +20,7 @@ import os
 import yaml
 from collections import OrderedDict
 
+import paddle
 from ppdet.data.source.category import get_categories
 
 from ppdet.utils.logger import setup_logger
@@ -50,6 +51,24 @@ KEYPOINT_ARCH = ['HigherHRNet', 'TopDownHRNet']
 MOT_ARCH = ['DeepSORT', 'JDE', 'FairMOT']
 
 
+def _prune_input_spec(input_spec, program, targets):
+    # try to prune static program to figure out pruned input spec
+    # so we perform following operations in static mode
+    paddle.enable_static()
+    pruned_input_spec = [{}]
+    program = program.clone()
+    program = program._prune(targets=targets)
+    global_block = program.global_block()
+    for name, spec in input_spec[0].items():
+        try:
+            v = global_block.var(name)
+            pruned_input_spec[0][name] = spec
+        except Exception:
+            pass
+    paddle.disable_static()
+    return pruned_input_spec
+
+
 def _parse_reader(reader_cfg, dataset_cfg, metric, arch, image_shape):
     preprocess_list = []
 
@@ -97,7 +116,7 @@ def _dump_infer_config(config, path, image_shape, model):
     arch_state = False
     from ppdet.core.config.yaml_helpers import setup_orderdict
     setup_orderdict()
-    use_dynamic_shape = True if image_shape[1] == -1 else False
+    use_dynamic_shape = True if image_shape[2] == -1 else False
     infer_cfg = OrderedDict({
         'mode': 'fluid',
         'draw_threshold': 0.5,
@@ -141,7 +160,7 @@ def _dump_infer_config(config, path, image_shape, model):
         dataset_cfg = config['TestDataset']
 
     infer_cfg['Preprocess'], infer_cfg['label_list'] = _parse_reader(
-        reader_cfg, dataset_cfg, config['metric'], label_arch, image_shape)
+        reader_cfg, dataset_cfg, config['metric'], label_arch, image_shape[1:])
 
     yaml.dump(infer_cfg, open(path, 'w'))
     logger.info("Export inference config file to {}".format(os.path.join(path)))

ppdet/engine/trainer.py

@@ -41,7 +41,7 @@ import ppdet.utils.stats as stats
 from ppdet.utils import profiler
 
 from .callbacks import Callback, ComposeCallback, LogPrinter, Checkpointer, WiferFaceEval, VisualDLWriter
-from .export_utils import _dump_infer_config
+from .export_utils import _dump_infer_config, _prune_input_spec
 
 from ppdet.utils.logger import setup_logger
 logger = setup_logger('ppdet.engine')
@@ -541,12 +541,7 @@ class Trainer(object):
         name, ext = os.path.splitext(image_name)
         return os.path.join(output_dir, "{}".format(name)) + ext
 
-    def export(self, output_dir='output_inference'):
-        self.model.eval()
-        model_name = os.path.splitext(os.path.split(self.cfg.filename)[-1])[0]
-        save_dir = os.path.join(output_dir, model_name)
-        if not os.path.exists(save_dir):
-            os.makedirs(save_dir)
+    def _get_infer_cfg_and_input_spec(self, save_dir, prune_input=True):
         image_shape = None
         if self.cfg.architecture in MOT_ARCH:
             test_reader_name = 'TestMOTReader'
@@ -555,9 +550,11 @@ class Trainer(object):
         if 'inputs_def' in self.cfg[test_reader_name]:
             inputs_def = self.cfg[test_reader_name]['inputs_def']
             image_shape = inputs_def.get('image_shape', None)
-        # set image_shape=[3, -1, -1] as default
+        # set image_shape=[None, 3, -1, -1] as default
         if image_shape is None:
-            image_shape = [3, -1, -1]
+            image_shape = [None, 3, -1, -1]
+        if len(image_shape) == 3:
+            image_shape = [None] + image_shape
 
         if hasattr(self.model, 'deploy'):
             self.model.deploy = True
@@ -574,7 +571,7 @@ class Trainer(object):
 
         input_spec = [{
             "image": InputSpec(
-                shape=[None] + image_shape, name='image'),
+                shape=image_shape, name='image'),
             "im_shape": InputSpec(
                 shape=[None, 2], name='im_shape'),
             "scale_factor": InputSpec(
@@ -585,13 +582,29 @@ class Trainer(object):
                 "crops": InputSpec(
                     shape=[None, 3, 192, 64], name='crops')
             })
+        if prune_input:
+            static_model = paddle.jit.to_static(
+                self.model, input_spec=input_spec)
+            # NOTE: dy2st do not pruned program, but jit.save will prune program
+            # input spec, prune input spec here and save with pruned input spec
+            pruned_input_spec = _prune_input_spec(
+                input_spec, static_model.forward.main_program,
+                static_model.forward.outputs)
+        else:
+            static_model = None
+            pruned_input_spec = input_spec
+
+        return static_model, pruned_input_spec
+
+    def export(self, output_dir='output_inference'):
+        self.model.eval()
+        model_name = os.path.splitext(os.path.split(self.cfg.filename)[-1])[0]
+        save_dir = os.path.join(output_dir, model_name)
+        if not os.path.exists(save_dir):
+            os.makedirs(save_dir)
 
-        static_model = paddle.jit.to_static(self.model, input_spec=input_spec)
-        # NOTE: dy2st do not pruned program, but jit.save will prune program
-        # input spec, prune input spec here and save with pruned input spec
-        pruned_input_spec = self._prune_input_spec(
-            input_spec, static_model.forward.main_program,
-            static_model.forward.outputs)
+        static_model, pruned_input_spec = self._get_infer_cfg_and_input_spec(
+            save_dir)
 
         # dy2st and save model
         if 'slim' not in self.cfg or self.cfg['slim_type'] != 'QAT':
@@ -606,22 +619,26 @@ class Trainer(object):
                 input_spec=pruned_input_spec)
         logger.info("Export model and saved in {}".format(save_dir))
 
-    def _prune_input_spec(self, input_spec, program, targets):
-        # try to prune static program to figure out pruned input spec
-        # so we perform following operations in static mode
-        paddle.enable_static()
-        pruned_input_spec = [{}]
-        program = program.clone()
-        program = program._prune(targets=targets)
-        global_block = program.global_block()
-        for name, spec in input_spec[0].items():
-            try:
-                v = global_block.var(name)
-                pruned_input_spec[0][name] = spec
-            except Exception:
-                pass
-        paddle.disable_static()
-        return pruned_input_spec
+    def post_quant(self, output_dir='output_inference'):
+        model_name = os.path.splitext(os.path.split(self.cfg.filename)[-1])[0]
+        save_dir = os.path.join(output_dir, model_name)
+        if not os.path.exists(save_dir):
+            os.makedirs(save_dir)
+
+        for idx, data in enumerate(self.loader):
+            self.model(data)
+            if idx == int(self.cfg.get('quant_batch_num', 10)):
+                break
+
+        # TODO: support prune input_spec
+        _, pruned_input_spec = self._get_infer_cfg_and_input_spec(
+            save_dir, prune_input=False)
+
+        self.cfg.slim.save_quantized_model(
+            self.model,
+            os.path.join(save_dir, 'model'),
+            input_spec=pruned_input_spec)
+        logger.info("Export Post-Quant model and saved in {}".format(save_dir))
 
     def _flops(self, loader):
         self.model.eval()

ppdet/slim/__init__.py

@@ -48,6 +48,14 @@ def build_slim_model(cfg, slim_cfg, mode='train'):
             load_pretrain_weight(model, weights)
         cfg['model'] = model
         cfg['slim_type'] = cfg.slim
+    elif slim_load_cfg['slim'] == 'PTQ':
+        model = create(cfg.architecture)
+        load_config(slim_cfg)
+        load_pretrain_weight(model, cfg.weights)
+        slim = create(cfg.slim)
+        cfg['slim_type'] = cfg.slim
+        cfg['model'] = slim(model)
+        cfg['slim'] = slim
     else:
         load_config(slim_cfg)
         model = create(cfg.architecture)

ppdet/slim/quant.py

@@ -49,3 +49,36 @@ class QAT(object):
     def save_quantized_model(self, layer, path, input_spec=None, **config):
         self.quanter.save_quantized_model(
             model=layer, path=path, input_spec=input_spec, **config)
+
+
+@register
+@serializable
+class PTQ(object):
+    def __init__(self,
+                 ptq_config,
+                 quant_batch_num=10,
+                 output_dir='output_inference',
+                 fuse=True,
+                 fuse_list=None):
+        super(PTQ, self).__init__()
+        self.ptq_config = ptq_config
+        self.quant_batch_num = quant_batch_num
+        self.output_dir = output_dir
+        self.fuse = fuse
+        self.fuse_list = fuse_list
+
+    def __call__(self, model):
+        paddleslim = try_import('paddleslim')
+        self.ptq = paddleslim.PTQ(**self.ptq_config)
+        model.eval()
+        quant_model = self.ptq.quantize(
+            model, fuse=self.fuse, fuse_list=self.fuse_list)
+
+        return quant_model
+
+    def save_quantized_model(self,
+                             quant_model,
+                             quantize_model_path,
+                             input_spec=None):
+        self.ptq.save_quantized_model(quant_model, quantize_model_path,
+                                      input_spec)

tools/post_quant.py

+# Copyright (c) 2021 PaddlePaddle Authors. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+from __future__ import absolute_import
+from __future__ import division
+from __future__ import print_function
+
+import os
+import sys
+
+# add python path of PadleDetection to sys.path
+parent_path = os.path.abspath(os.path.join(__file__, *(['..'] * 2)))
+sys.path.insert(0, parent_path)
+
+# ignore warning log
+import warnings
+warnings.filterwarnings('ignore')
+
+import paddle
+
+from ppdet.core.workspace import load_config, merge_config
+from ppdet.utils.check import check_gpu, check_version, check_config
+from ppdet.utils.cli import ArgsParser
+from ppdet.engine import Trainer
+from ppdet.slim import build_slim_model
+
+from ppdet.utils.logger import setup_logger
+logger = setup_logger('post_quant')
+
+
+def parse_args():
+    parser = ArgsParser()
+    parser.add_argument(
+        "--output_dir",
+        type=str,
+        default="output_inference",
+        help="Directory for storing the output model files.")
+    parser.add_argument(
+        "--slim_config",
+        default=None,
+        type=str,
+        help="Configuration file of slim method.")
+    args = parser.parse_args()
+    return args
+
+
+def run(FLAGS, cfg):
+    # build detector
+    trainer = Trainer(cfg, mode='eval')
+
+    # load weights
+    if cfg.architecture in ['DeepSORT']:
+        if cfg.det_weights != 'None':
+            trainer.load_weights_sde(cfg.det_weights, cfg.reid_weights)
+        else:
+            trainer.load_weights_sde(None, cfg.reid_weights)
+    else:
+        trainer.load_weights(cfg.weights)
+
+    # post quant model
+    trainer.post_quant(FLAGS.output_dir)
+
+
+def main():
+    FLAGS = parse_args()
+    cfg = load_config(FLAGS.config)
+    # TODO: to be refined in the future
+    if 'norm_type' in cfg and cfg['norm_type'] == 'sync_bn':
+        FLAGS.opt['norm_type'] = 'bn'
+    merge_config(FLAGS.opt)
+
+    if FLAGS.slim_config:
+        cfg = build_slim_model(cfg, FLAGS.slim_config, mode='test')
+
+    # FIXME: Temporarily solve the priority problem of FLAGS.opt
+    merge_config(FLAGS.opt)
+    check_config(cfg)
+    check_gpu(cfg.use_gpu)
+    check_version()
+
+    run(FLAGS, cfg)
+
+
+if __name__ == '__main__':
+    main()