eval common function that manage eval result and handle db transaction
actual eval functions for each neural networks
on prediction time we need to cache model and graph for faster response speed
we need to prepare for the situation that many users tries to execute train at a same time
which means we need job handler managing number of jobs can be run at once.
we need to make jupyter samples and java samples for pilot programs
data nodes and net nodes must be worked on all combinations (that's why we create feeder node)
need to replace all deprecated method to new function (you can check it on console)
we need services to see, modify and delete what we set on nodes (for now we only have insert functions)
batch version also need to be handled (need to build common functions first)
check nginx server and celery works fine on real server (starting with 172...)
seq2seq (performance, parms(hard coded like vector size), early stop function
we also need crud functions on the level of state (including all nodes on certain net and version)
new features has to be developed (autoencoder, ontology(sparql), gan, reinforcement.. etc)
hope we can fix all above problems in 2 weeks !
good luck guys