I am trying to train darkflow tiny-yolo on my custom dataset. The problem I am facing is that after setting up tensorflow gpu several times and starting the training using:
flow --model cfg/tiny-yolo-obj.cfg --labels labels.txt --train --dataset dataset/images/ --annotation dataset/annotations/ --batch 1 --gpu 0.9
the model does not train. removing the --gpu and running on cpu does work fine however.
I have tried multiple configurations for tf and cuda.
Configs tried:
tensorflow = 1.12.0
cuda = 9.2
cudnn = 7
The error on this was the Check failed: stream->parent()->GetConvolveAlgorithms(&algorithms)
The current config that I am trying is
tensorflow = 1.4.0
cuda = 8.0
cudnn = 6.0
There is no explicit error for this but the program crashes with a message Killed
Can someone please mention the config that they were able to train on, or maybe help with the errors I am facing.
I am using ubuntu 16.04 with python 3.5 and RTX 2080
Hi,
I am trying to train darkflow tiny-yolo on my custom dataset. The problem I am facing is that after setting up tensorflow gpu several times and starting the training using:
flow --model cfg/tiny-yolo-obj.cfg --labels labels.txt --train --dataset dataset/images/ --annotation dataset/annotations/ --batch 1 --gpu 0.9the model does not train. removing the --gpu and running on cpu does work fine however. I have tried multiple configurations for tf and cuda. Configs tried:tensorflow = 1.12.0 cuda = 9.2 cudnn = 7 The error on this was the
Check failed: stream->parent()->GetConvolveAlgorithms(&algorithms)The current config that I am trying is tensorflow = 1.4.0 cuda = 8.0 cudnn = 6.0 There is no explicit error for this but the program crashes with a message
KilledCan someone please mention the config that they were able to train on, or maybe help with the errors I am facing. I am using ubuntu 16.04 with python 3.5 and RTX 2080