fix bugs

configs/quantization/qat/lsq_resnet18_8xb32_in1k.py → configs/quantization/qat/lsq_resnet18_8xb16_cifar10.py

@@ -1,27 +1,22 @@
-_base_ = ['mmcls::resnet/resnet18_8xb32_in1k.py']
-
-train_cfg = dict(
-    _delete_=True,
-    type='mmrazor.QATEpochBasedLoop',
-    max_epochs=_base_.train_cfg.max_epochs)
+_base_ = ['mmcls::resnet/resnet18_8xb16_cifar10.py']
 
 resnet = _base_.model
-ckpt = 'https://download.openmmlab.com/mmclassification/v0/resnet/resnet18_8xb32_in1k_20210831-fbbb1da6.pth'  # noqa: E501
-resnet.init_cfg = dict(type='Pretrained', checkpoint=ckpt)
+pretrained_ckpt = 'https://download.openmmlab.com/mmclassification/v0/resnet/resnet18_b16x8_cifar10_20210528-bd6371c8.pth'  # noqa: E501
 
 model = dict(
     _delete_=True,
     _scope_='mmrazor',
     type='GeneralQuant',
-    # data_preprocessor = dict(
-    #     num_classes=1000,
-    #     # RGB format normalization parameters
-    #     mean=[123.675, 116.28, 103.53],
-    #     std=[58.395, 57.12, 57.375],
-    #     # convert image from BGR to RGB
-    #     to_rgb=True,
-    # ),
+    data_preprocessor=dict(
+        type='mmcls.ClsDataPreprocessor',
+        num_classes=10,
+        # RGB format normalization parameters
+        mean=[125.307, 122.961, 113.8575],
+        std=[51.5865, 50.847, 51.255],
+        # loaded images are already RGB format
+        to_rgb=False),
     architecture=resnet,
+    pretrained_ckpt=pretrained_ckpt,
     quantizer=dict(
         type='TensorRTQuantizer',
         skipped_methods=[
@@ -30,8 +25,8 @@
         ],
         qconfig=dict(
             qtype='affine',
-            w_observer=dict(type='mmrazor.MinMaxObserver'),
-            a_observer=dict(type='mmrazor.EMAMinMaxObserver'),
+            w_observer=dict(type='mmrazor.LSQObserver'),
+            a_observer=dict(type='mmrazor.LSQObserver'),
             w_fake_quant=dict(type='mmrazor.LearnableFakeQuantize'),
             a_fake_quant=dict(type='mmrazor.LearnableFakeQuantize'),
             w_qscheme=dict(
@@ -59,17 +54,17 @@
     begin=0,
     end=100)
 
-default_hooks = dict(
-    checkpoint=dict(
-        type='CheckpointHook',
-        interval=5,
-        max_keep_ckpts=3,
-        out_dir='/mnt/petrelfs/caoweihan.p/training_ckpt/quant'))
-
 model_wrapper_cfg = dict(
     type='mmrazor.GeneralQuantDDP',
     broadcast_buffers=False,
-    find_unused_parameters=False)
+    find_unused_parameters=True)
 
+# train, val, test setting
+train_cfg = dict(
+    _delete_=True,
+    type='mmrazor.QATEpochBasedLoop',
+    by_epoch=True,
+    max_epochs=100,
+    val_interval=1)
 val_cfg = dict(_delete_=True, type='mmrazor.QATValLoop')
 test_cfg = val_cfg

mmrazor/engine/hooks/__init__.py

@@ -1,7 +1,6 @@
 # Copyright (c) OpenMMLab. All rights reserved.
 from .dump_subnet_hook import DumpSubnetHook
 from .estimate_resources_hook import EstimateResourcesHook
+from .visualization_hook import RazorVisualizationHook
 
-# from .quant_hook import QuantitiveHook
-
-__all__ = ['DumpSubnetHook', 'EstimateResourcesHook']
+__all__ = ['DumpSubnetHook', 'EstimateResourcesHook', 'RazorVisualizationHook']

mmrazor/models/algorithms/__init__.py

@@ -3,7 +3,7 @@
 from .distill import (DAFLDataFreeDistillation, DataFreeDistillation,
                       FpnTeacherDistill, OverhaulFeatureDistillation,
                       SelfDistill, SingleTeacherDistill)
-from .nas import SPOS, AutoSlim, AutoSlimDDP, Darts, DartsDDP, Dsnas, DsnasDDP
+from .nas import DSNAS, DSNASDDP, SPOS, AutoSlim, AutoSlimDDP, Darts, DartsDDP
 from .pruning import SlimmableNetwork, SlimmableNetworkDDP
 from .pruning.ite_prune_algorithm import ItePruneAlgorithm
 from .quantization import GeneralQuant
@@ -13,5 +13,5 @@
     'SlimmableNetwork', 'SlimmableNetworkDDP', 'AutoSlim', 'AutoSlimDDP',
     'Darts', 'DartsDDP', 'SelfDistill', 'DataFreeDistillation',
     'DAFLDataFreeDistillation', 'OverhaulFeatureDistillation',
-    'ItePruneAlgorithm', 'Dsnas', 'DsnasDDP', 'GeneralQuant'
+    'ItePruneAlgorithm', 'DSNAS', 'DSNASDDP', 'GeneralQuant'
 ]

mmrazor/models/observers/minmax.py

@@ -47,8 +47,6 @@ def forward(self, x_orig):
         max_val = torch.max(self.max_val, max_val_cur)
         self.min_val = min_val
         self.max_val = max_val
-        # self.min_val.copy_(min_val)
-        # self.max_val.copy_(max_val)
 
         return x
 

mmrazor/models/quantizers/trt_quantizer.py

@@ -9,7 +9,7 @@ class TensorRTQuantizer(CustomQuantizer):
     """Quantizer for TensorRT backend."""
 
     def __init__(self,
-                 qconfig=DefalutQconfigs['default'],
+                 qconfig=DefalutQconfigs['tensorrt'],
                  is_qat=True,
                  skipped_methods=None,
                  prepare_custom_config_dict=None,

tests/test_models/test_task_modules/test_custom_tracer.py

@@ -1,17 +1,35 @@
 # Copyright (c) OpenMMLab. All rights reserved.
-class testCustomTracer:
+from unittest import TestCase
 
-    def test_init():
-        pass
+from mmrazor.models.task_modules import CustomTracer, UntracedMethodRegistry
+from mmrazor.testing import ConvBNReLU
+
+
+class testCustomTracer(TestCase):
+
+    def test_init(self):
+        tracer = CustomTracer()
+        assert tracer.skipped_methods.__len__() == 0
 
-    def test_trace():
+    def test_trace(self):
+        tracer = CustomTracer()
+        model = ConvBNReLU(3, 3, norm_cfg=dict(type='BN'))
+        graph = tracer.trace(model)  # noqa: F841
+
+    def test_auto_skip_call_module(self):
         pass
 
-    def test_auto_skip_call_module():
+    def test_auto_skip_call_method(self):
         pass
 
-    def test_auto_skip_call_method():
+    def test_configurable_skipped_methods(self):
         pass
 
-    def test_configurable_skipped_methods():
+
+class testUntracedMethodRgistry(TestCase):
+
+    def test_init(self):
+        self.assertEqual(len(UntracedMethodRegistry.method_dict), 0)
+
+    def test_add_method(self):
         pass

tools/tracer_demo.py

@@ -6,7 +6,7 @@
 from mmengine.config import Config
 from mmengine.registry import MODELS
 
-from mmrazor.models.task_modules.tracer import custom_symbolic_tracer
+from mmrazor.models.task_modules.tracer import custom_symbolic_trace
 
 cfg_path = 'configs/quantization/ptq/demo.py'
 _ADAROUND_SUPPORT_TYPE = (torch.nn.Conv2d, torch.nn.Linear)
@@ -73,7 +73,7 @@ def main():
     # load config
     cfg = Config.fromfile(cfg_path)
     model = MODELS.build(cfg.model)
-    symbolic_traced = custom_symbolic_tracer(
+    symbolic_traced = custom_symbolic_trace(
         model, concrete_args={'mode': 'tensor'})
     # block_slices = extract_blocks(symbolic_traced)
     block_slices = extract_layers(