Keras-retinanet: Error "ValueError: generator already executing" during training

[mohammad-saber]

Hi,

I face an error when I train the network. I use following command in Terminal:

python train.py --batch-size 20 --multi-gpu 1 --epochs 20 --steps 200 --snapshot-path /snapshots --workers 4  pascal /Dataset

I am using my own dataset for training.
Environment:
Ubuntu 16
tensorflow 1.14.0
keras 2.2.4

Expected behaviour:
training RetinaNet model without error

Actual behaviour:
In the first epoch, training stops with following error:
ValueError: generator already executing

Note:
When I use --epochs 2 --steps 2, I don't face any error. When I increase the number of epochs or steps, the above-mentioned error occurs.

[mak4alex]

@mohammad-saber use --workers 1 to use only 1 generator

[mohammad-saber]

I tried --workers 1. Now, it works without error. Thanks.