[Fix] Fix torch2.0 dcn/mdcn symbolic (#2695)

* fix

* fix lint

mmcv/ops/deform_conv.py

@@ -108,8 +108,10 @@ class DeformConv2dFunction(Function):
             return output
         ctx.save_for_backward(input, offset, weight)
 
-        output = input.new_empty(
-            DeformConv2dFunction._output_size(ctx, input, weight))
+        output = input.new_empty([
+            int(i)
+            for i in DeformConv2dFunction._output_size(ctx, input, weight)
+        ])
 
         ctx.bufs_ = [input.new_empty(0), input.new_empty(0)]  # columns, ones
 

mmcv/ops/modulated_deform_conv.py

@@ -136,8 +136,10 @@ class ModulatedDeformConv2dFunction(Function):
                 ctx, input, offset, mask, weight, bias)
             return output
         ctx.save_for_backward(input, offset, mask, weight, bias)
-        output = input.new_empty(
-            ModulatedDeformConv2dFunction._output_size(ctx, input, weight))
+        output = input.new_empty([
+            int(i) for i in ModulatedDeformConv2dFunction._output_size(
+                ctx, input, weight)
+        ])
         ctx._bufs = [input.new_empty(0), input.new_empty(0)]
         ext_module.modulated_deform_conv_forward(
             input,

tests/test_ops/test_onnx.py

@@ -3,7 +3,6 @@ import os
 
 import numpy as np
 import onnx
-import onnxruntime as rt
 import pytest
 import torch
 import torch.nn as nn
@@ -37,6 +36,7 @@ class WrapFunction(nn.Module):
 
 
 def test_roialign():
+    rt = pytest.importorskip('onnxruntime')
     try:
         from mmcv.ops import roi_align
     except (ImportError, ModuleNotFoundError):
@@ -106,6 +106,7 @@ def test_roialign():
 
 @pytest.mark.skipif(not torch.cuda.is_available(), reason='test requires GPU')
 def test_roipool():
+    rt = pytest.importorskip('onnxruntime')
     from mmcv.ops import roi_pool
 
     # roi pool config
@@ -204,7 +205,7 @@ def test_deform_conv():
     from mmcv.ops import DeformConv2dPack
     x = torch.randn(1, 2, 4, 4, device='cuda')
     _test_symbolic(
-        DeformConv2dPack(2, 4, 3, 1, 1).cuda(), x, 'MMCVDeformConv2d')
+        DeformConv2dPack(2, 4, 3, 1, 1).cuda(), (x, ), 'MMCVDeformConv2d')
 
 
 @pytest.mark.skipif(not torch.cuda.is_available(), reason='test requires GPU')