python - Efficiently serialize tensorflow graph / session to db? -
from can tell, tf.train.saver()
interface can take file-prefix input, , tensorflow handles serialization of model directly disk.
i'm interested in writing current state of tensorflow
model postgresql database, , have been looking serializing current state of tensorflow
object location in memory, can save serialized state postgresql bytea
column. how might go doing this?
my current workaround involves allowing tf.train.saver()
write disk @ tempfile
location, reading , serializing contents of files, re-writing file locations , allowing tf.train.saver()
read them load graph in @ inference time.
Comments
Post a Comment