TypeError: unorderable types: NoneType() <= int()

[qjqjqjj]

Hi all,

While I was training my own data set, all is fine for 6 hours but this error suddenly appear.

Epoch: [90][1000/1000] Time 0.516 Loss 0.198 Xent 0.019 RegLoss 0.179 Accuracy 0.989 Lr 0.00050 Cl 0.915
Running forward pass on validation set

Validation Epoch: 90 Time 0.000 Loss nan Xent nan Accuracy nan
Saving variables
Variables saved in 0.11 seconds
Saving statistics
Traceback (most recent call last):
File "src/train_softmax.py", line 580, in
main(parse_arguments(sys.argv[1:]))
File "src/train_softmax.py", line 234, in main
prelogits, prelogits_center_loss, args.random_rotate, args.random_crop, args.random_flip, prelogits_norm, args.prelogits_hist_max, args.use_fixed_image_standardization)
File "src/train_softmax.py", line 308, in train
if lr<=0:
TypeError: unorderable types: NoneType() <= int()

Does anyone know what this error is? I used the following parameters:
python3 src/train_softmax.py
--logs_base_dir ~/logs/facenet/
--models_base_dir ~/models/facenet/
--data_dir ~/facenet/done/
--image_size 160
--model_def models.inception_resnet_v1
--optimizer ADAM --learning_rate -1
--max_nrof_epochs 150
--keep_probability 0.8
--random_crop
--random_flip
--use_fixed_image_standardization
--learning_rate_schedule_file data/learning_rate_schedule_classifier_casia.txt
--weight_decay 5e-4
--embedding_size 512
--lfw_distance_metric 1
--lfw_use_flipped_images
--lfw_subtract_mean
--validation_set_split_ratio 0.05
--validate_every_n_epochs 5
--prelogits_norm_loss_factor 5e-4

[qjqjqjj]

I found out the cause to this as in learning_rate_schedule_classifier_casia.txt if the epoch is grater than 90 then it will stop. But can I check why there isn't any .pb file being genterated.

[rgsousa88]

I had this same error when using learning_rate_schedule_classifier_vggface2.txt as learing_rate_schedule_file and 500 epochs. I've searched in that file and facenet.py code and I've found that when epoch number is grater than 275, a None object is returnd by the function get_learning_rate_from_file. So, I've solved this by adding some lines to verify the return from this function when it is called in train_softmax.py script.

About .pb model file, this one is generated by freeze_graph.py script. This train script generates metagraph and checkpoint files. So, when running freeze_graph script, you have to pass the directory having these two files and a the output path where .pb should be saved.

[hsm4703]

I found out the cause to this as in learning_rate_schedule_classifier_casia.txt if the epoch is grater than 90 then it will stop. But can I check why there isn't any .pb file being genterated.
how to make .pb file can u tell me thanks

[hsm4703]

回溯（最近一次調用最後一次）：
文件“src / train_softmax.py”，第580行，在
main（parse_arguments（sys.argv [1：]））
文件“src / train_softmax.py”，第234行，在主要的
prelogits中， prelogits_center_loss，args.random_rotate，args.random_crop，args.random_flip，prelogits_norm，args.prelogits_hist_max，args.use_fixed_image_standardization）
文件“SRC / train_softmax.py”，線306，在列車
LR = facenet.get_learning_rate_from_file（learning_rate_schedule_file，歷元）
文件“ C：\ facenet-master \ src \ facenet.py“，第295行，在get_learning_rate_from_file中
e = int（par [0]）
ValueError：對於帶有基數為10的int（）的無效文字：''
e = int(par[0]) this way how can i fix
i change code and use your code but i have this problem can u fix it ?