Bytedance_ICME2019_challenge_baseline icon indicating copy to clipboard operation
Bytedance_ICME2019_challenge_baseline copied to clipboard

Published 20 hours ago • verified publisher icon shenweichen

finish task loss cannot converge

Open ponpontio opened this issue 6 years ago • 1 comments

i trained the model with two epochs. the loss is like:

19622340/19622340 [==============================] - 417s 21us/step - loss: 0.5772 - finish_loss: 0.5248 - like_loss: 0.0292

the finish task seems not converging.

ponpontio avatar Feb 19 '19 04:02 ponpontio

@newtiopon Hi,I think you can check if it is overfitted when training 2 or more epochs

shenweichen avatar Feb 19 '19 04:02 shenweichen