Zero Accuracy in Trainer

[init1mhd]

i have this config file :
number: '0123456789'
symbol: " !"$%&'()*+,-./:;<=>?@[]\_`{}|~©°—‘’“”"
lang_char: 'ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz'
experiment_name: 'en_filtered5'
train_data: 'all_data'
valid_data: 'all_data/'
manualSeed: 1111
workers: 6
batch_size: 32
num_iter: 100
valInterval: 4
saved_model: '' # saved_models/en_filtered/iter_300000.pth
FT: False
optim: adam # default is Adadelta
lr: 1.0
beta1: 0.9
rho: 0.95
eps: 1e-8
grad_clip: 5
select_data: '' # this is dataset folder in train_data
batch_ratio: '1'
total_data_usage_ratio: 1.0
batch_max_length: 20
imgH: 64
imgW: 600
rgb: False
contrast_adjust: False
sensitive: True
PAD: True
contrast_adjust: 0.0
data_filtering_off: False
Transformation: 'None'
FeatureExtraction: 'VGG'
SequenceModeling: 'BiLSTM'
Prediction: 'CTC'
num_fiducial: 20
input_channel: 1
output_channel: 256
hidden_size: 256
decode: 'greedy'
new_prediction: False
freeze_FeatureExtraction: False
freeze_SequenceModeling: False
but i have zero accuracy :

[100/100] Train loss: nan, Valid loss: nan, Elapsed_time: 00:02:50
Current_accuracy : 0.440, Current_norm_ED : 0.0000
Best_accuracy : 0.440, Best_norm_ED : 0.0119

Ground Truth | Prediction | Confidence Score & T/F

of | | nan False
Within | | nan False

---

Ground Truth | Prediction | Confidence Score & T/F

of | | nan False
Within | | nan False

why ???