TensorFlow: Where is the trained model stored and how to access it?
Where and how does TensorFlow save its models (graphs)? Is it inside files checkpoint
?
I know that in Caffe, the model is stored in a file .caffemodel
that is used to deploy (test) the model after training it.
I just prepared the model in this tutorial and now has the following files in my directory /tmp/mnist_convnet_model
:
Which file (s) contains the actual model?
For example, if I want to classify a single image using this trained model, which file is used for classification?
source to share
Yes, the model is within different files checkpoint
, although the file is *.ckpt
now split into data
and index
files. I'm not sure what is in the index file, but the files *.data*
contain weight and offset values. The files *.meta
contain the graph: variables, data structures, transformations, etc.
Here is a good tutorial on grooming and feeding.
source to share