x and y must have the same dtype, got tf.float32 != tf.resource

[Chuong98CC]

Version: Tensorflow 1.12 Stable. When I run the code, it appears the error:

x and y must have the same dtype, got tf.float32 != tf.resource in "AdaBound.py", line 132, in _resource_apply_dense final_lr = self._final_lr * lr_t / base_lr_t

[Jasonnor]

+1 same issue here, also tf 1.12

[fastalgo]

+1 same issue here, tf 1.13.0-rc2

[fastalgo]

Dear author, which tf version did you use? I just want to reproduce your results.

[taki0112]

I tested in tf 1.8. I will fix it as soon as possible

[joeyearsley]

I managed to fix by removing all .handle calls in that function using tf 1.10

[fastalgo]

The current version works for me. tf 1.12.0 on v100 GPUs

[taki0112]

fix it by @huyu398 I merged it