AttributeError: 'str' object has no attribute 'rank'

[TonyTangYu]

I run python train.py --gpu 0 --batch-size 32 --lr 0.001 but I get the following problem.

Using mxnet as:
<module 'mxnet' from '/THL5/home/daodao/softwares/mxnet0.11/python/mxnet/init.pyc'>
Warning: using pre-installed version of mxnet may cause unexpected error...
(export MXNET_EXAMPLE_SSD_DISABLE_PRE_INSTALLED=1) to prevent loading pre-installed mxnet.
[22:12:49] src/io/iter_image_det_recordio.cc:280: ImageDetRecordIOParser: /THL5/home/daodao/ty_project/model/mxnet-ssd/data/train.rec, use 27 threads for decoding..
[22:12:50] src/io/iter_image_det_recordio.cc:333: ImageDetRecordIOParser: /THL5/home/daodao/ty_project/model/mxnet-ssd/data/train.rec, label padding width: 350
[22:12:50] src/engine/threaded_engine_perdevice.cc:253: [added by cxt] engine is created by EnginePerDevice
[22:12:50] src/io/iter_image_det_recordio.cc:280: ImageDetRecordIOParser: /THL5/home/daodao/ty_project/model/mxnet-ssd/data/val.rec, use 27 threads for decoding..
[22:12:51] src/io/iter_image_det_recordio.cc:333: ImageDetRecordIOParser: /THL5/home/daodao/ty_project/model/mxnet-ssd/data/val.rec, label padding width: 350
INFO:root:Start training with (gpu(0)) from pretrained model /THL5/home/daodao/ty_project/model/mxnet-ssd/model/vgg16_reduced
[22:12:52] src/nnvm/legacy_json_util.cc:190: Loading symbol saved by previous version v0.8.0. Attempting to upgrade...
[22:12:52] src/nnvm/legacy_json_util.cc:198: Symbol successfully upgraded!
INFO:root:Freezed parameters: [conv1_1_weight,conv1_1_bias,conv1_2_weight,conv1_2_bias,conv2_1_weight,conv2_1_bias,conv2_2_weight,conv2_2_bias]
[22:13:03] src/operator/././cudnn_algoreg-inl.h:112: Running performance tests to find the best convolution algorithm, this can take a while... (setting env variable MXNET_CUDNN_AUTOTUNE_DEFAULT to 0 to disable)
[22:13:40] src/operator/convolution.cu:119: This convolution is not supported by cudnn, MXNET convolution is applied.
ff
Traceback (most recent call last):
File "train.py", line 148, in
tensorboard=args.tensorboard)
File "/THL5/home/daodao/ty_project/model/mxnet-ssd/train/train_net.py", line 354, in train_net
monitor=monitor)
File "/THL5/home/daodao/softwares/mxnet0.11/python/mxnet/module/base_module.py", line 568, in fit
rank=kvstore.rank, total_iter=temp_count,
AttributeError: 'str' object has no attribute 'rank'

It seems that it results from the mxnet. There is no solution corresponding to this problem. I wanna solve this problem but don't know how. Could you please help me get over with it? Thank you!

[zhreshold]

please check mxnet version, make sure it's 1.1 or 1.2

[TonyTangYu]

please check mxnet version, make sure it's 1.1 or 1.2

Thank you for your response. I checked the version of mxnet. It's 0.11. What can I do to solve this problem?

[TonyTangYu]

@zhreshold In fact, I downloaded the source code of mxnet 0.11 and try to train vgg16_reduced but got the same error. Is it a problem from the base_module.py or resulting from the change of mxnet's API?

[zhreshold]

@TonyTangYu I mean mxnet version is 1.1 not 0.11

[TonyTangYu]

@zhreshold I know. Do you mean I should install mxnet 1.1 or 1.2? However I want to know how to solve this problem when it's 0.11. Could you do me a favour?

[zhreshold]

@TonyTangYu

pip install mxnet==1.2.0 --upgrade

[TonyTangYu]

@zhreshold Thank you for your response. But I install mxnet0.11 on the cloud server. It cannot connect the Internet and the pip command won't work on this occasion. Could you please give me another solution? Thank you.

[TonyTangYu]

@zhreshold I upgraded the version of MXNet. It works now. Thank you!