program is blocked when training about gazeml HOT 8 CLOSED

I have solved this. I changed with tf.Session() as session in elg_train.py to session=tf.Session() and fixed this error.

from gazeml.

Hi Yang. The training code allows for resumption of training, and so won't continue training if a model has exceeded (or matched) the number of steps expected based on what's specified at https://github.com/swook/GazeML/blob/master/src/elg_train.py#L101 and the current size of the training set. As the mentioned model was trained for a sufficient no. of steps it therefore does not continue and simply exits.

from gazeml.

@swook Hi, I am tring to train from the scratch, and i am pretty sure it not exceeds the expected steps, but i am still getting the same error. How could i slove it?
If i want to fine tune using your pre-trained model, how could i do it? Should the input image size and batch size compatiable with your original pre-trained model?
Thanks in advance.

from gazeml.

Hi @ningscapr ,
Could you share the line of code you edited? I tried to edited the code, but still get the same problem.
Thanks

from gazeml.

@keishatsai
In elg_train.py I changed

2019-01-17 14:11:57.398307: W tensorflow/core/kernels/queue_base.cc:277]_0_UnityEyes/random_shuffle_queue: Skipping cancelled enqueue attempt with queue not closed
2019-01-17 14:11:57.398371: W tensorflow/core/kernels/queue_base.cc:277]_0_UnityEyes/random_shuffle_queue: Skipping cancelled enqueue attempt with queue not closed
2019-01-17 14:11:57.398380: W tensorflow/core/kernels/queue_base.cc:277]_0_UnityEyes/random_shuffle_queue: Skipping cancelled enqueue attempt with queue not closed
2019-01-17 14:11:57.398386: W tensorflow/core/kernels/queue_base.cc:277]_0_UnityEyes/random_shuffle_queue: Skipping cancelled enqueue attempt with queue not closed
2019-01-17 14:11:57.398390: W tensorflow/core/kernels/queue_base.cc:277]_0_UnityEyes/random_shuffle_queue: Skipping cancelled enqueue attempt with queue not closed

from gazeml.

@ningscapr
Thanks for the help.
By the way, have you eventually managed to fine-tune the model? And how did you achieve that?
I also wonder to use some of my own data to fine tune, but don't know what kind of annotation do I need for training.

from gazeml.

@keishatsai It seems you can't use original mdoel to finetune, you have to train your own data from scratch.

from gazeml.

@ningscapr ,
Hmm.....could you explain why do you think that I have to do it from scratch?
Thank you.

from gazeml.