Can't show all the data points that have been tracked!

[ArkiZh]

🐛 Bug

Can't show all the data points that have been tracked!

To reproduce

from aim import Run

import random

run = Run(experiment="aim_bug")

for i in range(1000):
    value = random.normalvariate(100, 10) if i%100==0 else 0
    run.track(value, name="value", step=i)

Expected behavior

Show all the data points, but now only some of them are sampled!

Environment

• Aim Version 3.25.0
• Python version 3.11
• pip version 24.2
• OS Mac

[mihran113]

Hey @ArkiZh! Thanks for reaching out. This is done by design considering performance limitations of rendering charts on browsers for providing smooth experience.
We're showing maximum 500 randomly uniform sampled data points and it's natural that some of them didn't make it.
Can you describe what has driven you to open this issue and also the use case where all the data points are essential in the charts?

[dmacd]

I'm attempting to use aim to debug and gain insight about some complicated internal dynamics of novel attention mechanisms I'm exploring. Not being able to see all the points collected even when zoomed in is a dealbreaker for research use cases. Shame, its otherwise a pretty great tool.

@mihran113 Can you point me in the direction of what needs to be hacked/tweaked to enable this?