[Fix]Dcff Deploy Revision

configs/pruning/mmcls/dcff/dcff_compact_resnet_8xb32_in1k.py

@@ -2,4 +2,7 @@
 
 # model settings
 model = _base_.model
+# Avoid pruning_ratio check in mutator
+model['fix_subnet'] = 'configs/pruning/mmcls/dcff/fix_subnet.yaml'
+model['target_pruning_ratio'] = None
 model['is_deployed'] = True

configs/pruning/mmcls/dcff/dcff_resnet_8xb32_in1k.py

@@ -75,7 +75,11 @@
         parse_cfg=dict(
             type='BackwardTracer',
             loss_calculator=dict(type='ImageClassifierPseudoLoss'))),
+    fix_subnet=None,
+    data_preprocessor=None,
     target_pruning_ratio=target_pruning_ratio,
     step_freq=1,
     linear_schedule=False,
     is_deployed=False)
+
+val_cfg = dict(_delete_=True, type='mmrazor.ItePruneValLoop')

configs/pruning/mmdet/dcff/dcff_compact_faster_rcnn_resnet50_8xb4_coco.py

@@ -2,4 +2,8 @@
 
 # model settings
 model = _base_.model
+model = _base_.model
+# Avoid pruning_ratio check in mutator
+model['fix_subnet'] = 'configs/pruning/mmdet/dcff/fix_subnet.yaml'
+model['target_pruning_ratio'] = None
 model['is_deployed'] = True

configs/pruning/mmdet/dcff/dcff_faster_rcnn_resnet50_8xb4_coco.py

@@ -65,19 +65,14 @@
     _delete_=True)
 train_cfg = dict(max_epochs=120, val_interval=1)
 
-# !dataset config
-# ==========================================================================
-# data preprocessor
-
 model = dict(
     _scope_='mmrazor',
     type='DCFF',
     architecture=_base_.architecture,
     mutator_cfg=dict(
         type='DCFFChannelMutator',
         channel_unit_cfg=dict(
-            type='DCFFChannelUnit',
-            units='configs/pruning/mmdet/dcff/resnet_det.json'),
+            type='DCFFChannelUnit', default_args=dict(choice_mode='ratio')),
         parse_cfg=dict(
             type='BackwardTracer',
             loss_calculator=dict(type='TwoStageDetectorPseudoLoss'))),
@@ -89,4 +84,4 @@
 model_wrapper = dict(
     type='mmcv.MMDistributedDataParallel', find_unused_parameters=True)
 
-val_cfg = dict(_delete_=True)
+val_cfg = dict(_delete_=True, type='mmrazor.ItePruneValLoop')

configs/pruning/mmpose/dcff/dcff_compact_topdown_heatmap_resnet50_coco.py

@@ -2,4 +2,8 @@
 
 # model settings
 model = _base_.model
+model = _base_.model
+# Avoid pruning_ratio check in mutator
+model['fix_subnet'] = 'configs/pruning/mmpose/dcff/fix_subnet.yaml'
+model['target_pruning_ratio'] = None
 model['is_deployed'] = True

configs/pruning/mmpose/dcff/dcff_topdown_heatmap_resnet50_coco.py

@@ -108,13 +108,11 @@
 model = dict(
     _scope_='mmrazor',
     type='DCFF',
-    architecture=dict(
-        cfg_path='mmcls::resnet/resnet50_8xb32_in1k.py', pretrained=False),
+    architecture=architecture,
     mutator_cfg=dict(
         type='DCFFChannelMutator',
         channel_unit_cfg=dict(
-            type='DCFFChannelUnit',
-            units='configs/pruning/mmpose/dcff/resnet_pose.json'),
+            type='DCFFChannelUnit', default_args=dict(choice_mode='ratio')),
         parse_cfg=dict(
             type='BackwardTracer',
             loss_calculator=dict(type='TopdownPoseEstimatorPseudoLoss'))),
@@ -125,7 +123,7 @@
 
 dataset_type = 'CocoDataset'
 data_mode = 'topdown'
-data_root = 'data/coco'
+data_root = 'data/coco/'
 
 file_client_args = dict(backend='disk')
 
@@ -186,3 +184,5 @@
     type='mmpose.CocoMetric',
     ann_file=data_root + 'annotations/person_keypoints_val2017.json')
 test_evaluator = val_evaluator
+
+val_cfg = dict(_delete_=True, type='mmrazor.ItePruneValLoop')

configs/pruning/mmseg/dcff/dcff_compact_pointrend_resnet50_8xb2_cityscapes.py

@@ -2,4 +2,7 @@
 
 # model settings
 model = _base_.model
+# Avoid pruning_ratio check in mutator
+model['fix_subnet'] = 'configs/pruning/mmseg/dcff/fix_subnet.yaml'
+model['target_pruning_ratio'] = None
 model['is_deployed'] = True

configs/pruning/mmseg/dcff/dcff_pointrend_resnet50_8xb2_cityscapes.py

@@ -80,13 +80,11 @@
 model = dict(
     _scope_='mmrazor',
     type='DCFF',
-    architecture=dict(
-        cfg_path='mmcls::resnet/resnet50_8xb32_in1k.py', pretrained=False),
+    architecture=_base_.architecture,
     mutator_cfg=dict(
         type='DCFFChannelMutator',
         channel_unit_cfg=dict(
-            type='DCFFChannelUnit',
-            units='configs/pruning/mmseg/dcff/resnet_seg.json'),
+            type='DCFFChannelUnit', default_args=dict(choice_mode='ratio')),
         parse_cfg=dict(
             type='BackwardTracer',
             loss_calculator=dict(type='CascadeEncoderDecoderPseudoLoss'))),
@@ -97,3 +95,5 @@
 
 model_wrapper = dict(
     type='mmcv.MMDistributedDataParallel', find_unused_parameters=True)
+
+val_cfg = dict(_delete_=True, type='mmrazor.ItePruneValLoop')

mmrazor/engine/__init__.py

@@ -3,13 +3,14 @@
 from .optimizers import SeparateOptimWrapperConstructor
 from .runner import (AutoSlimValLoop, DartsEpochBasedTrainLoop,
                      DartsIterBasedTrainLoop, EvolutionSearchLoop,
-                     GreedySamplerTrainLoop, SelfDistillValLoop,
-                     SingleTeacherDistillValLoop, SlimmableValLoop)
+                     GreedySamplerTrainLoop, ItePruneValLoop,
+                     SelfDistillValLoop, SingleTeacherDistillValLoop,
+                     SlimmableValLoop)
 
 __all__ = [
     'SeparateOptimWrapperConstructor', 'DumpSubnetHook',
     'SingleTeacherDistillValLoop', 'DartsEpochBasedTrainLoop',
     'DartsIterBasedTrainLoop', 'SlimmableValLoop', 'EvolutionSearchLoop',
     'GreedySamplerTrainLoop', 'AutoSlimValLoop', 'EstimateResourcesHook',
-    'SelfDistillValLoop'
+    'SelfDistillValLoop', 'ItePruneValLoop'
 ]

mmrazor/engine/runner/__init__.py

@@ -3,11 +3,13 @@
 from .darts_loop import DartsEpochBasedTrainLoop, DartsIterBasedTrainLoop
 from .distill_val_loop import SelfDistillValLoop, SingleTeacherDistillValLoop
 from .evolution_search_loop import EvolutionSearchLoop
+from .iteprune_val_loop import ItePruneValLoop
 from .slimmable_val_loop import SlimmableValLoop
 from .subnet_sampler_loop import GreedySamplerTrainLoop
 
 __all__ = [
     'SingleTeacherDistillValLoop', 'DartsEpochBasedTrainLoop',
     'DartsIterBasedTrainLoop', 'SlimmableValLoop', 'EvolutionSearchLoop',
-    'GreedySamplerTrainLoop', 'AutoSlimValLoop', 'SelfDistillValLoop'
+    'GreedySamplerTrainLoop', 'AutoSlimValLoop', 'SelfDistillValLoop',
+    'ItePruneValLoop'
 ]

mmrazor/engine/runner/iteprune_val_loop.py

@@ -0,0 +1,46 @@
+# Copyright (c) OpenMMLab. All rights reserved.
+import os.path as osp
+
+from mmengine import fileio
+from mmengine.runner import ValLoop
+
+from mmrazor.registry import LOOPS
+from mmrazor.structures import export_fix_subnet
+
+
+@LOOPS.register_module()
+class ItePruneValLoop(ValLoop):
+    """Pruning loop for validation. Export fixed subnet configs.
+
+    Args:
+        runner (Runner): A reference of runner.
+        dataloader (Dataloader or dict): A dataloader object or a dict to
+            build a dataloader.
+        evaluator (Evaluator or dict or list): Used for computing metrics.
+        fp16 (bool): Whether to enable fp16 validation. Defaults to
+            False.
+    """
+
+    def run(self):
+        """Launch validation."""
+        self.runner.call_hook('before_val')
+        self.runner.call_hook('before_val_epoch')
+        self.runner.model.eval()
+        for idx, data_batch in enumerate(self.dataloader):
+            self.run_iter(idx, data_batch)
+
+        # compute metrics
+        metrics = self.evaluator.evaluate(len(self.dataloader.dataset))
+        self._save_fix_subnet()
+        self.runner.call_hook('after_val_epoch', metrics=metrics)
+        self.runner.call_hook('after_val')
+        return metrics
+
+    def _save_fix_subnet(self):
+        """Save model subnet config."""
+        fix_subnet = export_fix_subnet(self.model)
+        save_name = 'fix_subnet.yaml'
+        fileio.dump(fix_subnet, osp.join(self.runner.work_dir, save_name))
+        self.runner.logger.info(
+            'export finished and '
+            f'{save_name} saved in {self.runner.work_dir}.')

mmrazor/models/algorithms/pruning/dcff.py

@@ -8,10 +8,9 @@
 from mmengine.model import BaseModel
 from mmengine.structures import BaseDataElement
 
-from mmrazor.models.mutables import BaseMutable
 from mmrazor.models.mutators import DCFFChannelMutator
 from mmrazor.registry import MODELS
-from mmrazor.structures.subnet.fix_subnet import _dynamic_to_static
+from mmrazor.utils import ValidFixMutable
 from .ite_prune_algorithm import ItePruneAlgorithm, ItePruneConfigManager
 
 LossResults = Dict[str, torch.Tensor]
@@ -30,8 +29,8 @@ class DCFF(ItePruneAlgorithm):
     Args:
         architecture (Union[BaseModel, Dict]): The model to be pruned.
         mutator_cfg (Union[Dict, ChannelMutator], optional): The config
-            of a mutator. Defaults to dict( type='ChannelMutator',
-            channel_unit_cfg=dict( type='SequentialMutableChannelUnit')).
+            of a mutator. Defaults to dict( type='DCFFChannelMutator',
+            channel_unit_cfg=dict( type='DCFFChannelUnit')).
         data_preprocessor (Optional[Union[Dict, nn.Module]], optional):
             Defaults to None.
         target_pruning_ratio (dict, optional): The prune-target. The template
@@ -56,6 +55,7 @@ def __init__(self,
                  mutator_cfg: Union[Dict, DCFFChannelMutator] = dict(
                      type=' DCFFChannelMutator',
                      channel_unit_cfg=dict(type='DCFFChannelUnit')),
+                 fix_subnet: Optional[ValidFixMutable] = None,
                  data_preprocessor: Optional[Union[Dict, nn.Module]] = None,
                  target_pruning_ratio: Optional[Dict[str, float]] = None,
                  step_freq=1,
@@ -64,27 +64,9 @@ def __init__(self,
                  linear_schedule=False,
                  is_deployed=False) -> None:
         # invalid param prune_times, reset after message_hub get [max_epoch]
-        super().__init__(architecture, mutator_cfg, data_preprocessor,
-                         target_pruning_ratio, step_freq, prune_times,
-                         init_cfg, linear_schedule)
-        self.is_deployed = is_deployed
-        if (self.is_deployed):
-            # To static ops for loaded pruned network.
-            self._deploy()
-
-    def _fix_archtecture(self):
-        for module in self.architecture.modules():
-            if isinstance(module, BaseMutable):
-                if not module.is_fixed:
-                    module.fix_chosen(None)
-
-    def _deploy(self):
-        config = self.prune_config_manager.prune_at(self._iter)
-        self.mutator.set_choices(config)
-        self.mutator.fix_channel_mutables()
-        self._fix_archtecture()
-        _dynamic_to_static(self.architecture)
-        self.is_deployed = True
+        super().__init__(architecture, mutator_cfg, fix_subnet,
+                         data_preprocessor, target_pruning_ratio, step_freq,
+                         prune_times, init_cfg, linear_schedule, is_deployed)
 
     def _calc_temperature(self, cur_num: int, max_num: int):
         """Calculate temperature param."""

mmrazor/models/algorithms/pruning/ite_prune_algorithm.py

@@ -10,6 +10,7 @@
 from mmrazor.models.mutables import MutableChannelUnit
 from mmrazor.models.mutators import ChannelMutator
 from mmrazor.registry import MODELS
+from mmrazor.utils import ValidFixMutable
 from ..base import BaseAlgorithm
 
 LossResults = Dict[str, torch.Tensor]
@@ -97,6 +98,8 @@ class ItePruneAlgorithm(BaseAlgorithm):
         mutator_cfg (Union[Dict, ChannelMutator], optional): The config
             of a mutator. Defaults to dict( type='ChannelMutator',
             channel_unit_cfg=dict( type='SequentialMutableChannelUnit')).
+        fix_subnet (str | dict | :obj:`FixSubnet`): The path of yaml file or
+            loaded dict or built :obj:`FixSubnet`. Defaults to None.
         data_preprocessor (Optional[Union[Dict, nn.Module]], optional):
             Defaults to None.
         target_pruning_ratio (dict, optional): The prune-target. The template
@@ -110,6 +113,8 @@ class ItePruneAlgorithm(BaseAlgorithm):
             Defaults to None.
         linear_schedule (bool, optional): flag to set linear ratio schedule.
             Defaults to True.
+        is_deployed (bool, optional): flag to set deployed algorithm.
+            Defaults to False.
     """
 
     def __init__(self,
@@ -118,12 +123,14 @@ def __init__(self,
                      type='ChannelMutator',
                      channel_unit_cfg=dict(
                          type='SequentialMutableChannelUnit')),
+                 fix_subnet: Optional[ValidFixMutable] = None,
                  data_preprocessor: Optional[Union[Dict, nn.Module]] = None,
                  target_pruning_ratio: Optional[Dict[str, float]] = None,
-                 step_freq=-1,
-                 prune_times=-1,
+                 step_freq=1,
+                 prune_times=1,
                  init_cfg: Optional[Dict] = None,
-                 linear_schedule=True) -> None:
+                 linear_schedule=True,
+                 is_deployed=False) -> None:
 
         super().__init__(architecture, data_preprocessor, init_cfg)
 
@@ -132,10 +139,17 @@ def __init__(self,
         self.step_freq = step_freq
         self.prune_times = prune_times
         self.linear_schedule = linear_schedule
+        self.is_deployed = is_deployed
 
-        # mutator
-        self.mutator: ChannelMutator = MODELS.build(mutator_cfg)
-        self.mutator.prepare_from_supernet(self.architecture)
+        if self.is_deployed:
+            assert fix_subnet is not None
+            # Avoid circular import
+            from mmrazor.structures import load_fix_subnet
+            load_fix_subnet(self.architecture, fix_subnet)
+        else:
+            # init mutator
+            self.mutator: ChannelMutator = MODELS.build(mutator_cfg)
+            self.mutator.prepare_from_supernet(self.architecture)
 
     def group_target_pruning_ratio(
         self, target: Dict[str, float],