Facenet: Error After Training

[ekarem]

I experience an error when training as classifier process is finished:

tensorflow/core/kernels/queue_base.cc:294] _0_input_producer/fraction_of_32_full/fraction_of_32_full: Skipping cancelled enqueue attempt with queue not closed
W tensorflow/core/kernels/queue_base.cc:302] _2_FIFOQueueV2: Skipping cancelled dequeue attempt with queue not closed
W tensorflow/core/kernels/queue_base.cc:302] _2_FIFOQueueV2: Skipping cancelled dequeue attempt with queue not closed
W tensorflow/core/kernels/queue_base.cc:302] _2_FIFOQueueV2: Skipping cancelled dequeue attempt with queue not closed
W tensorflow/core/kernels/queue_base.cc:302] _2_FIFOQueueV2: Skipping cancelled dequeue attempt with queue not closed

Is there anyone who has the error?

[davidsandberg]

Hi,
These are warnings that are harmless but a bit annoying. They come from closing the session (which was not done before) but I didn't find a way to make them disappear. I think you can safely ignore them for now.

[andyhx]

HI,I got the same issue,any way to solved it?the program stopped,the final message is something like this:

Epoch: [80][1999/2000]  Time 0.237  Loss 0.226  RegLoss 0.222
Epoch: [80][2000/2000]  Time 0.222  Loss 0.231  RegLoss 0.222
Saving variables
Variables saved in 0.28 seconds
W tensorflow/core/kernels/queue_base.cc:302] _2_FIFOQueueV2: Skipping cancelled dequeue attempt with queue not closed
W tensorflow/core/kernels/queue_base.cc:302] _2_FIFOQueueV2: Skipping cancelled dequeue attempt with queue not closed
W tensorflow/core/kernels/queue_base.cc:302] _2_FIFOQueueV2: Skipping cancelled dequeue attempt with queue not closed
W tensorflow/core/kernels/queue_base.cc:302] _2_FIFOQueueV2: Skipping cancelled dequeue attempt with queue not closed
W tensorflow/core/kernels/queue_base.cc:294] _0_input_producer/fraction_of_32_full/fraction_of_32_full: Skipping cancelled enqueue attempt with queue not closed

my running cmd is:
python src/train_softmax.py --logs_base_dir /home/public/logs/facenet/ --models_base_dir src/models/facenet/ --data_dir /home/public/Proj/20170603 --image_size 160 --model_def models.inception_resnet_v1 --learning_rate -0.0001 --max_nrof_epochs 80 --keep_probability 0.8 --random_crop --random_flip --learning_rate_schedule_file data/learning_rate_schedule_classifier_casia.txt --weight_decay 6e-5 --center_loss_factor 1e-2 --center_loss_alfa 0.7 --batch_size 10 --epoch_size 2000 --gpu_memory_fraction 0.21 --pretrained_model ~/git/facenet20170518/facenet/src/models/facenet/20170608-002050/model-20170608-002050.ckpt-160000
is there something to do with max_nrof_epochs settings?
thanks

[Bananajia]

@andyhx because of your paras that " --max_nrof_epochs 80" so your program ended up in epoch[80]

[andyhx]

@Bananajia ,thank your ,I will modify it to check that

[kay-max]

@andyhx did fixing the " --max_nrof_epochs 80" value solve the issue for you?