logboard
Monitor and Compare Logs on Browser/Terminal.
Description
Inspired by tensorboard, grip and notable, all of which serve light-weight GUI by
- only using static files (e.g., markdown files without DB);
- single command (e.g.,
tensorboard --logdir logs/
andgrip README.md
).
tensorboard
?
Why not I also use tensorboard
in addition to this tool.
But currently tensorboard
doesn't support comparing different configurations
for each log (e.g., git-hash of the code, learning rate, training strategy).
logboard
is a kind of extra plugin to tensorboard
(but you need to run in different terminal, unfortunately).
I expect this kind of feature will be included in tensorboard
in the future.
Installation
pip install logboard
Usage
logboard --logdir logs/
)
Browser ($ cd examples
$ cat logs/20190310_093252.724597/args
{
"loglevel": "info",
"gpu": 0,
"seed": 0,
"class_ids": [
1
],
"lr": 0.001,
"timestamp": "2019-03-10T09:32:52.724597",
"out": "/home/wkentaro/logboard/examples/logs/20190310_093252.724597",
"hostname": "computer1",
"githash": "b48ce48"
}
# similar to tensorboard --logdir logs/
$ logboard --logdir logs/ --filter out timestamp loglevel gpu seed 'lr .*' '.*main/loss.*(max)' '.*loss_.*'
logtable --logdir logs/
)
Terminal ($ cd examples
$ logtable --logdir logs --filter out timestamp loglevel gpu seed 'lr .*' '.*main/loss.*(max)' '.*loss_.*'
* Log directory: logs
โโโโโโคโโโโโโโโโโโโโโโโโโโโโโโโโคโโโโโโโโโโคโโโโโโโโโโโโโโคโโโโโโโโโโโโโโโโโคโโโโโโโโโโโโโโโโโโโโคโโโโโโโโโโโโโโคโโโโโโโโโโโโคโโโโโโโโโโโโโคโโโโโโโโคโโโโโโโโโโโโโโโคโโโโโโโโโโโโโโโโ
โ โ log_dir โ epoch โ iteration โ elapsed_time โ updated_at โ class_ids โ githash โ hostname โ lr โ main/ โ validation/ โ
โ โ โ โ โ โ โ โ โ โ โ loss (min) โ main/ โ
โ โ โ โ โ โ โ โ โ โ โ โ loss (min) โ
โโโโโโชโโโโโโโโโโโโโโโโโโโโโโโโโชโโโโโโโโโโชโโโโโโโโโโโโโโชโโโโโโโโโโโโโโโโโชโโโโโโโโโโโโโโโโโโโโชโโโโโโโโโโโโโโชโโโโโโโโโโโโชโโโโโโโโโโโโโชโโโโโโโโชโโโโโโโโโโโโโโโชโโโโโโโโโโโโโโโโก
โ 0 โ 20190310_093252.724597 โ 1 โ 1740 โ 1:47:02 โ 88 days, 14:24:22 โ [1] โ b48ce48 โ computer1 โ 0.001 โ 0.0088 โ 0.18 โ
โ โ โ โ โ โ โ โ โ โ โ (1, 1580) โ (0, 880) โ
โโโโโโผโโโโโโโโโโโโโโโโโโโโโโโโโผโโโโโโโโโโผโโโโโโโโโโโโโโผโโโโโโโโโโโโโโโโโผโโโโโโโโโโโโโโโโโโโโผโโโโโโโโโโโโโโผโโโโโโโโโโโโผโโโโโโโโโโโโโผโโโโโโโโผโโโโโโโโโโโโโโโผโโโโโโโโโโโโโโโโค
โ 1 โ 20190310_093829.691289 โ 1 โ 1720 โ 1:45:37 โ 88 days, 14:24:22 โ [1] โ f766b97 โ computer2 โ 0.001 โ 0.012 โ 0.19 โ
โ โ โ โ โ โ โ โ โ โ โ (1, 1620) โ (0, 440) โ
โโโโโโงโโโโโโโโโโโโโโโโโโโโโโโโโงโโโโโโโโโโงโโโโโโโโโโโโโโงโโโโโโโโโโโโโโโโโงโโโโโโโโโโโโโโโโโโโโงโโโโโโโโโโโโโโงโโโโโโโโโโโโงโโโโโโโโโโโโโงโโโโโโโโงโโโโโโโโโโโโโโโงโโโโโโโโโโโโโโโโ