dcrnn_train.py : Found Inf or NaN global norm: Tensor had NaN values

liyaguang / DCRNN

Implementation of Diffusion Convolutional Recurrent Neural Network in Tensorflow

MIT License

1.21k stars 400 forks source link

dcrnn_train.py : Found Inf or NaN global norm: Tensor had NaN values #35

Closed Cancerman-Mylo closed 4 years ago

Cancerman-Mylo commented 5 years ago

hello , I get an InvalidArgumentError when running dcrnn_train.py, it shows: when I google it, someone said use smaller batch size but it doesn't work. what do you think the error happens ?

liyaguang commented 5 years ago

I have not come this error. Are you using a new dataset? Can you try to comment the line, i.e., grads, _ = tf.clip_by_global_norm(grads, max_grad_norm) which is not essential to the training.