Traceback (most recent call last):
File "./dff_deeplab/demo.py", line 312, in <module>
main()
File "./dff_deeplab/demo.py", line 136, in main
key_sym = sym_instance.get_key_test_symbol(config)
File "/home/daeyun/git/Accel/dff_deeplab/symbols/accel_18.py", line 133, in get_key_test_symbol
conv_feat = self.get_resnet_dcn(data)
File "/home/daeyun/git/Accel/dff_deeplab/symbols/resnet_v1_101_flownet_deeplab.py", line 1235, in get_resnet_dcn
res5a_branch2b = mx.contrib.symbol.DeformableConvolution(name='res5a_branch2b', data=res5a_branch2a_relu, offset=res5a_branch2b_offset,
AttributeError: 'module' object has no attribute 'DeformableConvolution'
When I search the codebase using grep -iR "deformable" mxnet, I see that there is an implementation in ./mxnet/3rdparty/tvm. Is there something I need to do when I build mxnet to expose this?
README.md refers to MXNet@(commit 62ecb60) version 0.9.3~0.9.5. But I had to use version 0.12.0 to run the code.
It doesn't seem like DeformableConvolution is accessible from
mx.contrib.symbol
. I get this error when I runpython ./dff_deeplab/demo.py --version 18 --interval 5 --num_ex 10
When I search the codebase using
grep -iR "deformable" mxnet
, I see that there is an implementation in./mxnet/3rdparty/tvm
. Is there something I need to do when I build mxnet to expose this?