robotzheng

THEANO_FLAGS=optimizer=fast_compile,device=gpu,floatX=float32 python nndial.py -config config/tracker.cfg -mode train also： including informable tracker loss ... including informable tracker loss ... including informable tracker loss ... including requestable tracker loss ... including requestable...

I use centos 7.5 K40

start work： number of parameters : 1103292 number of training parameters : 1096842 start network training ... Finishing 25 dialog in epoch 1 thanks to shawnwun

me too. T2T: 1.6.2 TF: 1.7.0 Pyhton: 3.5.3 [output_gen.txt](https://github.com/tensorflow/tensor2tensor/files/2148219/output_gen.txt)

Thanks for @huyvd7 . I have trained a GLR model, I will try your method.

hi @huyvd7 . Do you try resnet architecture for your deepGLR, add more direct connections?

I have trained four GLR models, but, when I train a DGLR model using them, learning rate is 1e-6, the loss is only dropped to several thousand, the test result...

[10_r.zip](https://github.com/huyvd7/pytorch-deepglr/files/4447742/10_r.zip) Thanks.

Hi @huyvd7, have you test this picture, maybe robust is not so easy, any clue for future?

Thanks for your answer, I will make more expriments.