Fix parrots compatibility issues (#4143)

* fix parrots compatibility

* add comments

mmdet/core/export/pytorch2onnx.py

@@ -5,11 +5,6 @@ import numpy as np
 import torch
 from mmcv.runner import load_checkpoint
 
-try:
-    from mmcv.onnx.symbolic import register_extra_symbolics
-except ModuleNotFoundError:
-    raise NotImplementedError('please update mmcv to version>=v1.0.4')
-
 
 def generate_inputs_and_wrap_model(config_path, checkpoint_path, input_config):
     """Prepare sample input and wrap model for ONNX export.
@@ -51,6 +46,12 @@ def generate_inputs_and_wrap_model(config_path, checkpoint_path, input_config):
     # pytorch has some bug in pytorch1.3, we have to fix it
     # by replacing these existing op
     opset_version = 11
+    # put the import within the function thus it will not cause import error
+    # when not using this function
+    try:
+        from mmcv.onnx.symbolic import register_extra_symbolics
+    except ModuleNotFoundError:
+        raise NotImplementedError('please update mmcv to version>=v1.0.4')
     register_extra_symbolics(opset_version)
 
     return model, tensor_data

mmdet/datasets/samplers/group_sampler.py

@@ -103,8 +103,11 @@ class DistributedGroupSampler(Sampler):
             if size > 0:
                 indice = np.where(self.flag == i)[0]
                 assert len(indice) == size
-                indice = indice[list(torch.randperm(int(size),
-                                                    generator=g))].tolist()
+                # add .numpy() to avoid bug when selecting indice in parrots.
+                # TODO: check whether torch.randperm() can be replaced by
+                # numpy.random.permutation().
+                indice = indice[list(
+                    torch.randperm(int(size), generator=g).numpy())].tolist()
                 extra = int(
                     math.ceil(
                         size * 1.0 / self.samples_per_gpu / self.num_replicas)