When using the .Estimator estimate in tensorflow.contrib.learn, after training and forecasting, there are these files in modeldir :
- control point
- events.out.tfevents.1487956647
- events.out.tfevents.1487957016
- graph.pbtxt
- model.ckpt-101.data-00000-of-00001
- model.ckpt-101.index
- model.ckpt-101.meta li>
When a graph is complex or the number of variables is large, the graph.pbtxt file and event files can be very large. Is this a way not to write these files? Since only model control point files are needed to reload the model, their removal will not affect future estimation and forecasting.