nni
nni copied to clipboard
The accuracy and loss of ENAS algorithm are not convergent. Is there any appropriate parameter setting for the optimizer
Describe the issue: Tensorboard shows:The accuracy and loss of ENAS algorithm are not convergent. Is there any appropriate parameter setting for the optimizer. Environment:
- NNI version:2.10
- Training service (local|remote|pai|aml|etc):remote
- Client OS:Ubuntu
How to reproduce it?:
Lower learning rate might be helpful.