Welcome to Flättlibrettli
Your metrics stay on this machine. All parsing and plotting happens locally in your browser.
Load a .plattli archive or pick the run folder directly — each run folder should contain a plattli/ directory with config.json, plattli.json, and one file per metric (<metric>.indices + <metric>.<dtype> or <metric>.json). The zipped form typically lives at metrics.plattli in the run folder. If you run the bundled Python server, hit Connect Remote to stream those files over HTTP instead.
Don't have anything ready yet? I gotchu fam, here's an example archive for you to play around with.
Brettli
Hover a subplot to see run values.
def configs(c, run_id): [...] return True
✓
✗
?
def metrics(m, c): [...] return m
✓
⚡︎
⚙︎
✗
?
Select runs:
⟳
What varies?
?
Select runs to see which config keys differ.