Checkpoints in TensorFlow save the parameters of a model but lack a description of the computations involved. The SavedModel format addresses this by containing both the parameter values and a serialized version of the computation, making it usable across various programming languages. The guide provides instructions on saving and restoring TensorFlow models, including writing checkpoints and using high-level APIs to manage model variables. It emphasizes that tracking parameters can be efficiently done by attaching them to Python object attributes.
Checkpoints capture the exact value of all parameters (tf.Variable objects) used by a model but do not contain any description of the computation defined by the model.
The SavedModel format includes a serialized description of the computation defined by the model in addition to the parameter values, making models independent of the source code.
Collection
[
|
...
]