With MLtraq, upstreaming results is as easy as persisting the experiment to a new session. You can use any SQL database supported by SQLAlchemy. No need to add more complexity with more dedicated running services.
Upstreaming experiments
| from mltraq import create_session
from mltraq.utils.fs import tmpdir_ctx
with tmpdir_ctx():
# Creating a session to a local MLtraq db
local = create_session("sqlite:///local.db")
# Working on a new experiment ...
local.create_experiment("iris").persist()
# Upstreaming results to a (simulated) remote db ...
remote = create_session("sqlite:///remote.db")
remote.persist_experiment(
local.load_experiment("iris"), if_exists="replace"
)
|