I'm using Centos 6.6 on a GPU server. I installed mxnet following the instructions here. Everything goes right until I tried to test the installation. I run the following commands:
python
>> import mxnet
Then I got errors saying
Check failed: e == fmap_.at(alias) (0x10fc620 vs. 0x103adf0) Entry add_n already registered under different entry
The full error message is as follows:
[16:38:16] /home/mypath/software/try_mxnet/mxnet/dmlc-core/include/dmlc/./logging.h:300: [16:38:16] /home/mypath/software/try_mxnet/mxnet/dmlc-core/include/dmlc/registry.h:66: Check failed: e == fmap_.at(alias) (0x10fc620 vs. 0x103adf0) Entry add_n already registered under different entry
Stack trace returned 4 entries:
[bt] (0) /home/mypath/anaconda2/lib/python2.7/site-packages/mxnet-0.9.4-py2.7.egg/mxnet/libmxnet.so(ZN4dmlc8RegistryIN4nnvm2OpEE8AddAliasERKSsS5+0x79b) [0x7fa4c820635b]
[bt] (1) /home/mypath/anaconda2/lib/python2.7/site-packages/mxnet-0.9.4-py2.7.egg/mxnet/libmxnet.so(_ZN4nnvm2Op9add_aliasERKSs+0x1f) [0x7fa4c820493f]
[bt] (2) /home/mypath/anaconda2/lib/python2.7/site-packages/mxnet-0.9.4-py2.7.egg/mxnet/libmxnet.so(+0xb3cea9) [0x7fa4c7073ea9]
[bt] (3) /home/mypath/anaconda2/lib/python2.7/site-packages/mxnet-0.9.4-py2.7.egg/mxnet/libmxnet.so(+0x1cff226) [0x7fa4c8236226]
terminate called after throwing an instance of 'dmlc::Error'
what(): [16:38:16] /home/mypath/software/try_mxnet/mxnet/dmlc-core/include/dmlc/registry.h:66: Check failed: e == fmap_.at(alias) (0x10fc620 vs. 0x103adf0) Entry add_n already registered under different entry
Stack trace returned 4 entries:
[bt] (0) /home/mypath/anaconda2/lib/python2.7/site-packages/mxnet-0.9.4-py2.7.egg/mxnet/libmxnet.so(ZN4dmlc8RegistryIN4nnvm2OpEE8AddAliasERKSsS5+0x79b) [0x7fa4c820635b]
[bt] (1) /home/mypath/anaconda2/lib/python2.7/site-packages/mxnet-0.9.4-py2.7.egg/mxnet/libmxnet.so(_ZN4nnvm2Op9add_aliasERKSs+0x1f) [0x7fa4c820493f]
[bt] (2) /home/mypath/anaconda2/lib/python2.7/site-packages/mxnet-0.9.4-py2.7.egg/mxnet/libmxnet.so(+0xb3cea9) [0x7fa4c7073ea9]
[bt] (3) /home/mypath/anaconda2/lib/python2.7/site-packages/mxnet-0.9.4-py2.7.egg/mxnet/libmxnet.so(+0x1cff226) [0x7fa4c8236226]
Aborted
I have no idea what's wrong with my installation. The server has 2 x Nvidia Tesla K80 and 128 GB memory, it should be powerful enough. I hope there is someone who can give me some help. Thank you very much!!!
As answered by lynguyen, a fix has been made here: github.com/dmlc/mxnet/issues/5255
I have tested that solution and it works for me. I'm using gcc-4.8.5 and maybe it is the version of gcc that leads to this problem. Thanks!