I have the checkpoint of a pre-trained model. I want to feed a tensor to this model as the input and get the output features (tensor). I don't want to restore the model using saver.restore and run the session. Coz I'm in the training process and the only thing that I have is tensor (without value) and need tensor as the output. Is there any way to do that with checkpoints? As far as I know, we can do the same when we have .pb model, but not sure if we can extract features (in tensor) from .ckpt as well.
2.1m questions
2.1m answers
60 comments
57.0k users