I'm writing a simple model for Dist-Keras.
If I change some parameter of the Neural Network (for example the size of some layer), I get errors like:
ValueError: Cannot feed value of shape (16, 128) for Tensor u'Placeholder_1:0', which has shape '(15, 128)'
In fact, I changed a layer size from 16 to 15, but it seems like the system has memory of the old size and prompts this error.
I found the problem: there was an old dist-kers instance stuck on port 5000, and it contained the old version of the model. So I wasn't able to update my code.
I'm writing a simple model for Dist-Keras. If I change some parameter of the Neural Network (for example the size of some layer), I get errors like:
ValueError: Cannot feed value of shape (16, 128) for Tensor u'Placeholder_1:0', which has shape '(15, 128)'
In fact, I changed a layer size from 16 to 15, but it seems like the system has memory of the old size and prompts this error.