remove deploy test loop and enable edit SDK (#1083)

mmdeploy/codebase/base/runner.py

@@ -1,12 +1,10 @@
 # Copyright (c) OpenMMLab. All rights reserved.
-from typing import Dict, Optional, Sequence, Union
+from typing import Dict, Optional, Union
 
-import torch
 from mmengine.device import get_device
 from mmengine.logging import MMLogger
 from mmengine.model import BaseModel
-from mmengine.registry import LOOPS
+from mmengine.runner import Runner
-from mmengine.runner import Runner, TestLoop, autocast
 
 
 class DeployTestRunner(Runner):
@@ -64,38 +62,3 @@ class DeployTestRunner(Runner):
             log_file = self._log_file
 
         return super().build_logger(log_level, log_file, **kwargs)
-
-
-@LOOPS.register_module()
-class DeployTestLoop(TestLoop):
-    """Loop for test. To skip data_preprocessor for SDK.
-
-    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 testing. Defaults to
-            False.
-    """
-
-    @torch.no_grad()
-    def run_iter(self, idx, data_batch: Sequence[dict]) -> None:
-        """Iterate one mini-batch.
-
-        Args:
-            data_batch (Sequence[dict]): Batch of data from dataloader.
-        """
-        self.runner.call_hook(
-            'before_test_iter', batch_idx=idx, data_batch=data_batch)
-        # predictions should be sequence of BaseDataElement
-        with autocast(enabled=self.fp16):
-            # skip data_preprocessor to avoid Normalize and Padding for SDK
-            outputs = self.runner.model._run_forward(
-                data_batch, mode='predict')
-        self.evaluator.process(data_samples=outputs, data_batch=data_batch)
-        self.runner.call_hook(
-            'after_test_iter',
-            batch_idx=idx,
-            data_batch=data_batch,
-            outputs=outputs)

mmdeploy/codebase/mmcls/deploy/classification_model.py

@@ -94,6 +94,10 @@ class End2EndModel(BaseBackendModel):
 class SDKEnd2EndModel(End2EndModel):
     """SDK inference class, converts SDK output to mmcls format."""
 
+    def __init__(self, *arg, **kwargs):
+        kwargs['data_preprocessor'] = None
+        super().__init__(*arg, **kwargs)
+
     def forward(self,
                 inputs: Sequence[torch.Tensor],
                 data_samples: Optional[List[BaseDataElement]] = None,

mmdeploy/codebase/mmedit/deploy/super_resolution.py

@@ -307,9 +307,6 @@ class SuperResolution(BaseTask):
             'valid_ratio'
         ]
         preprocess = model_cfg.test_pipeline
-        for item in preprocess:
-            if 'Normalize' == item['type'] and 'std' in item:
-                item['std'] = [255, 255, 255]
 
         preprocess.insert(1, model_cfg.model.data_preprocessor)
         transforms = preprocess
@@ -322,6 +319,7 @@ class SuperResolution(BaseTask):
                 transform['type'] = 'ImageToTensor'
             if transform['type'] == 'EditDataPreprocessor':
                 transform['type'] = 'Normalize'
+                transform['to_rgb'] = transform.get('to_rgb', False)
             if transform['type'] == 'PackEditInputs':
                 meta_keys += transform[
                     'meta_keys'] if 'meta_keys' in transform else []

mmdeploy/codebase/mmedit/deploy/super_resolution_model.py

@@ -113,8 +113,11 @@ class End2EndModel(BaseBackendModel):
 class SDKEnd2EndModel(End2EndModel):
     """SDK inference class, converts SDK output to mmedit format."""
 
+    def __init__(self, *args, **kwargs):
+        super(SDKEnd2EndModel, self).__init__(*args, **kwargs)
+
     def forward(self,
-                lq: torch.Tensor,
+                inputs: torch.Tensor,
                 data_samples: Optional[List[BaseDataElement]] = None,
                 mode: str = 'predict',
                 *args,
@@ -135,8 +138,25 @@ class SDKEnd2EndModel(End2EndModel):
         Returns:
             list | dict: High resolution image or a evaluation results.
         """
-        output = self.wrapper.invoke(lq[0].contiguous().detach().cpu().numpy())
+        if hasattr(self.data_preprocessor, 'destructor'):
-        return [output]
+            inputs = self.data_preprocessor.destructor(
+                inputs.to(self.data_preprocessor.input_std.device))
+        outputs = []
+        for i in range(inputs.shape[0]):
+            output = self.wrapper.invoke(inputs[i].permute(
+                1, 2, 0).contiguous().detach().cpu().numpy())
+            outputs.append(
+                torch.from_numpy(output).permute(2, 0, 1).contiguous())
+        outputs = torch.stack(outputs, 0) / 255.
+        if hasattr(self.data_preprocessor, 'destructor'):
+            outputs = self.data_preprocessor.destructor(
+                outputs.to(self.data_preprocessor.outputs_std.device))
+
+        for i, sr_pred in enumerate(outputs):
+            pred = EditDataSample()
+            pred.set_data(dict(pred_img=PixelData(**dict(data=sr_pred))))
+            data_samples[i].set_data(dict(output=pred))
+        return data_samples
 
 
 def build_super_resolution_model(

mmdeploy/codebase/mmocr/deploy/text_detection_model.py

@@ -23,10 +23,8 @@ class End2EndModel(BaseBackendModel):
         backend_files (Sequence[str]): Paths to all required backend files(e.g.
             '.onnx' for ONNX Runtime, '.param' and '.bin' for ncnn).
         device (str): A string represents device type.
-        deploy_cfg (str | mmengine.Config): Deployment config file or loaded
+        deploy_cfg (mmengine.Config | None): Loaded Config object of MMDeploy.
-            Config object.
+        model_cfg (mmengine.Config | None): Loaded Config object of MMOCR.
-        model_cfg (str | mmengine.Config): Model config file or loaded Config
-            object.
     """
 
     def __init__(
@@ -34,19 +32,18 @@ class End2EndModel(BaseBackendModel):
         backend: Backend,
         backend_files: Sequence[str],
         device: str,
-        deploy_cfg: Union[str, mmengine.Config] = None,
+        deploy_cfg: Optional[mmengine.Config] = None,
-        model_cfg: Union[str, mmengine.Config] = None,
+        model_cfg: Optional[mmengine.Config] = None,
         **kwargs,
     ):
-        super(End2EndModel, self).__init__(deploy_cfg=deploy_cfg)
+        super(End2EndModel, self).__init__(
-        model_cfg, deploy_cfg = load_config(model_cfg, deploy_cfg)
+            deploy_cfg=deploy_cfg,
+            data_preprocessor=model_cfg.model.data_preprocessor)
         self.deploy_cfg = deploy_cfg
         self.show_score = False
 
         from mmocr.registry import MODELS
         self.det_head = MODELS.build(model_cfg.model.det_head)
-        self.data_preprocessor = MODELS.build(
-            model_cfg.model.data_preprocessor)
         self._init_wrapper(
             backend=backend,
             backend_files=backend_files,
@@ -125,6 +122,10 @@ class End2EndModel(BaseBackendModel):
 class SDKEnd2EndModel(End2EndModel):
     """SDK inference class, converts SDK output to mmocr format."""
 
+    def __init__(self, *args, **kwargs):
+        kwargs['model_cfg'].model.data_preprocessor = None
+        super(SDKEnd2EndModel, self).__init__(*args, **kwargs)
+
     def forward(self,
                 inputs: torch.Tensor,
                 data_samples: Optional[List[BaseDataElement]] = None,

mmdeploy/codebase/mmocr/deploy/text_recognition_model.py

@@ -1,5 +1,5 @@
 # Copyright (c) OpenMMLab. All rights reserved.
-from typing import Sequence, Union
+from typing import Optional, Sequence, Union
 
 import mmengine
 import torch
@@ -23,10 +23,8 @@ class End2EndModel(BaseBackendModel):
         backend_files (Sequence[str]): Paths to all required backend files(e.g.
             '.onnx' for ONNX Runtime, '.param' and '.bin' for ncnn).
         device (str): A string represents device type.
-        deploy_cfg (str | mmengine.Config): Deployment config file or loaded
+        deploy_cfg (mmengine.Config | None): Loaded Config object of MMDeploy.
-            Config object.
+        model_cfg (mmengine.Config | None): Loaded Config object of MMOCR.
-        model_cfg (str | mmengine.Config): Model config file or loaded Config
-            object.
     """
 
     def __init__(
@@ -34,10 +32,12 @@ class End2EndModel(BaseBackendModel):
         backend: Backend,
         backend_files: Sequence[str],
         device: str,
-        deploy_cfg: Union[str, mmengine.Config] = None,
+        deploy_cfg: Optional[mmengine.Config] = None,
-        model_cfg: Union[str, mmengine.Config] = None,
+        model_cfg: Optional[mmengine.Config] = None,
     ):
-        super(End2EndModel, self).__init__(deploy_cfg=deploy_cfg)
+        super(End2EndModel, self).__init__(
+            deploy_cfg=deploy_cfg,
+            data_preprocessor=model_cfg.model.data_preprocessor)
         model_cfg, deploy_cfg = load_config(model_cfg, deploy_cfg)
         self.deploy_cfg = deploy_cfg
         self.show_score = False
@@ -53,8 +53,6 @@ class End2EndModel(BaseBackendModel):
         if decoder.get('dictionary', None) is None:
             decoder.update(dictionary=self.dictionary)
         self.decoder = MODELS.build(decoder)
-        self.data_preprocessor = MODELS.build(
-            model_cfg.model.data_preprocessor)
         self._init_wrapper(
             backend=backend, backend_files=backend_files, device=device)
 
@@ -113,6 +111,10 @@ class End2EndModel(BaseBackendModel):
 class SDKEnd2EndModel(End2EndModel):
     """SDK inference class, converts SDK output to mmocr format."""
 
+    def __init__(self, *args, **kwargs):
+        kwargs['model_cfg'].model.data_preprocessor = None
+        super(SDKEnd2EndModel, self).__init__(*args, **kwargs)
+
     def forward(self, inputs: Sequence[torch.Tensor],
                 data_samples: RecSampleList, *args, **kwargs):
         """Run forward inference.

tools/test.py

@@ -6,8 +6,7 @@ from copy import deepcopy
 from mmengine import DictAction
 
 from mmdeploy.apis import build_task_processor
-from mmdeploy.utils.config_utils import get_backend, load_config
+from mmdeploy.utils.config_utils import load_config
-from mmdeploy.utils.constants import Backend
 from mmdeploy.utils.timer import TimeCounter
 
 
@@ -89,8 +88,6 @@ def main():
 
     # load deploy_cfg
     deploy_cfg, model_cfg = load_config(deploy_cfg_path, model_cfg_path)
-    if get_backend(deploy_cfg) == Backend.SDK:
-        model_cfg.test_cfg.type = 'DeployTestLoop'
 
     # work_dir is determined in this priority: CLI > segment in file > filename
     if args.work_dir is not None: