save checkpoint

[kbrajwani]

hey thanks for great work. i am training this model on icdar 2015 dataset. i saw you save checkpoint when loss will be in 0.3 to 0.4 and epoch % 4 ==0 or loss is 0.3. it took me 83 epoch to reach below 0.4 loss. so i want to ask you is this normal time also what is a minimum loss that you get on icdar 15 dataset on which epoch. 2021-01-10 15:31:26 INFO train.py: [83/600], train_loss: 0.3949, time: 390.9523, lr: 0.0001

I have also tried https://github.com/whai362/PSENet repo where i am always getting loss of 1.000 so if you know then can you tell me what is a problem in this.