[TOPI] Improve get_valid_count and nms performance for CUDA (#5339)

* get_valid_count updated to have correct results

* speedup nms

* update nms

* revert back nms

* recover one test for get_valid_count

python/tvm/relay/frontend/mxnet.py

@@ -853,7 +853,6 @@ def _mx_smooth_l1(inputs, attrs):
 
 def _mx_deformable_convolution(inputs, attrs):
     new_attrs = {}
-    assert attrs.get_bool("no_bias")
     new_attrs["kernel_size"] = attrs.get_int_tuple("kernel")
     new_attrs["strides"] = attrs.get_int_tuple("stride")
     new_attrs["padding"] = attrs.get_int_tuple("pad")

tests/python/relay/test_op_level5.py

@@ -225,6 +225,9 @@ def test_get_valid_counts():
             intrp = relay.create_executor("debug", ctx=ctx, target=target)
             out = intrp.evaluate(func)(np_data)
             tvm.testing.assert_allclose(out[0].asnumpy(), np_out1, rtol=1e-3, atol=1e-04)
+            # get_valid_count for cuda doesn't do data rearrangement
+            if target == 'cuda':
+                return
             tvm.testing.assert_allclose(out[1].asnumpy(), np_out2, rtol=1e-3, atol=1e-04)
 
     verify_get_valid_counts((1, 2500, 6), 0, 0, 1)

topi/python/topi/nn/deformable_conv2d.py

@@ -106,7 +106,7 @@ def deformable_conv2d_nchw(data, offset, kernel, strides, padding, dilation, def
                                     (kh * kernel_w + kw) * 2, y, x],
                              x * stride_w - pad_left + kw * dilation_w +
                              offset[n, c // ic_per_dgroup * (kernel_w*kernel_h*2) +
-                                    (kh * kernel_w + kw) * 2 + 1, y, x]))
+                                    (kh * kernel_w + kw) * 2 + 1, y, x]), tag="data_deform")
     return te.compute(
         (batch, out_channel, out_height, out_width),
         lambda n, f, y, x: te.sum(

topi/tests/python/test_topi_vision.py

@@ -106,8 +106,8 @@ def verify_get_valid_counts(dshape, score_threshold, id_index, score_index):
     """ Skip this test as it is intermittent
         see https://github.com/apache/incubator-tvm/pull/4901#issuecomment-595040094
     for device in ['llvm', 'cuda', 'opencl']:
-        # Disable opencl test for now
-        if device != "llvm" and device != "cuda":
+        # Disable gpu test for now
+        if device != "llvm":
             continue
         check_device(device)
     """