santi-pdp / segan

Speech Enhancement Generative Adversarial Network in TensorFlow
MIT License
809 stars 281 forks source link

En error with Gradient disappearance #70

Open marvin-nj opened 5 years ago

marvin-nj commented 5 years ago

Hello , Thank you for sharing ! the Gradient of my d_fk_loss disappear when i train my chinese experiment in the 22 epochs ,all config is initial. image

so, what is wrong with it ?