Semantic-Segmentation-Suite
Semantic-Segmentation-Suite copied to clipboard
GeorgeSeif
train DDSC failed
- Bug reports
Information
Please specify the following information when submitting an issue:
- What are your command line arguments?: python train.py --dataset Vehicle_512 --model DDSC
- Have you written any custom code?: no not at all. just using the fresh code.
- What have you done to try and solve this issue?: yeah, searched a lot to find the error. some people say the problem is with the cuda version. i tried different cuda versions, 8.0, 9.0,9.1, 9.2 none worked. Some also say the version of cudnn. I tried different cudnn versions, but no luck.
- TensorFlow version?: 1.11 cuda 9.2, cudnn 7.2.1
Describe the problem
Describe the problem clearly here. Be sure to convey here why it's a bug or a feature request. choosing the DDSC algorithm throws the following error:
2018-10-09 00:29:11.162019: F tensorflow/stream_executor/cuda/cuda_dnn.cc:430] could not convert BatchDescriptor {count: 1 feature_map_count: 1024 spatial: 0 0 value_min: 0.000000 value_max: 0.000000 layout: BatchDepthYX} to cudnn tensor descriptor: CUDNN_STATUS_BAD_PARAM
Aborted (core dumped)
Yup I know! Still trying to figure out what that bug means.... Thanks for opening up the issue though!
@GeorgeSeif I am also trying to figuring it out. as small side question, is the issue with InceptionV4 resolved? no matter which algorithm I choose, selecting InceptionV4 as backbone fails.
@GeorgeSeif I have tried to solve this error noticed this issue in Tensorflow https://github.com/tensorflow/tensorflow/issues/2107. Perhaps, we are fitting somewhere an empty array. I have tried with different input sizes and I just tried the last version of the github repo, but still the issue remains. With which input size you have trained DDSC? for me other networks can be trained but DDSC not.
Apologies for the late reply!
For DDSC I always had trouble, but left the code in there for reference. So I haven't fully trained it yet :)