tensorflow.python.framework.errors_impl.InvalidArgumentError: 2 root error(s) found.

错误日志：

tensorflow.python.framework.errors_impl.InvalidArgumentError: 2 root error(s) found.
  (0) Invalid argument: You must feed a value for placeholder tensor 'input_9' with dtype float and shape [?,256]
         [[{{node input_9}}]]
         [[training_4/Adam/gradients/gradients/model_3/batch_normalization_16/cond_grad/StatelessIf/then/_2762/gradients/FusedBatchNormV3_grad/FusedBatchNormGradV3/_4657]]
  (1) Invalid argument: You must feed a value for placeholder tensor 'input_9' with dtype float and shape [?,256]
         [[{{node input_9}}]]
0 successful operations.
0 derived errors ignored.

原因：GPU显存不足

解决方案：

from tensorflow.compat.v1 import ConfigProto 
from tensorflow.compat.v1 import InteractiveSession  
config = ConfigProto() 
config.gpu_options.allow_growth = True 
session = InteractiveSession(config=config)

动态分配GPU显存