training error

[wwyzby]

hello! when I try to use learning->training->denoising->main.py to train the net, I find it will report an error when running the "self.pred_depth_list = self.umgf_model(self.input_guide, self.input_target,self.niters)" in "self.build_model()" during the intial process. As the input shape is set as "shape=[None, None, None, self.inputI_chn]", it will make the "N = box_filter_op(tf.ones((1, 1, x_shape[2], x_shape[3]), dtype=x.dtype), r)" failed for the "x_shape[2]" shouldn't be nonetype. Could you plz help me out? thanks a lot!

[shizenglin]

Hi, I just run the code, and I didn't have your error. x_shape = tf.shape(x), which dynamically gets the tensor shape during running, refer to https://pgaleone.eu/tensorflow/2018/07/28/understanding-tensorflow-tensors-shape-static-dynamic/
Could you provide your complete error log?

[wwyzby]

Thank you for your reply! The problem is probably caused by a wrong vesion of tensorflow. it fails using tf-cpu,but everything will be ok using tf-gpu.