Colab question

[lycorn]

Based on my experience I don´t think that will really make a difference. I sometimes have P100/T4 sessions lasting the full 10 hours and K80/P4 being killed after a short while. Not that I use many K80 or P4s, tho...
I think the whole thing boils down to GPUs availability at a given time.

[Chuck]

I like the new feature of fetching only one assignment at a time.

On my assignments page, there are two Colab assignments which I don't recognize:

Colab_RPi 95766551
Colab_SSG 95766493

[Chuck]

Another one just showed up on my assignments page:

Colab_CVH 95766191

[chalsall]

Originally Posted by Chuck

Another one just showed up on my assignments page:

Yeah... Don't worry, I'm just passing you and RichD a few low assignments to expediate clearing out 95M.

[RichD]

No problem. Let's clear them out...

[kladner]

Happiness here with the single assignment per instance policy. Less cluttered, easier to track.

Hey Chris, I don't care exactly what's running. I just like keeping it going. Anything you want to push is fine with me.

Out of 4 notebooks I have running, 3 right now are factoring 74-75, and 1 75-76. Changes are interesting. 75-76 has been the main thing running now, it seems, so something different gives a moment of amusement.

[Runtime Error]

(Replying here rather than on the GPU72 Status thread since this exchange has become a colab discussion and not really about the status anymore.)

Originally Posted by chalsall

OK, version 0.423 is now in production. Does the one assignment at a time thing.

The one assignment rule is great! I've actually managed to finish some of the 81 bit jobs in recent days. However, my session durations are dwindling. I'm often seeing fewer than 3 hours per session, and logging in for 3 hours a day for a week to do one assignment isn't fun. I will likely switch to take 332M to 75 or 78 bits instead once I finish my in-progress 81 bit jobs, but I appreciate that you set it up for us!

[Uncwilly]

Originally Posted by Runtime Error

I'm often seeing fewer than 3 hours per session,

You and everyone else.

[mrk74]

Originally Posted by Uncwilly

You and everyone else.

Ditto here. I haven't run GPU in about 3 days and tried and only got a few hours out of it and had to go TPU and it said it was for usage limits. Really?? I can usually run DAYS without a usage limit problem.

[kladner]

Originally Posted by bayanne

I wonder whether we might get longer sessions if we opted to use K80 and P4s ...

Belatedly: Another question would be, "Would the presumed time increase offset and overcome the slowness of these card?" That is, an hour of a T4=~1.5 hours of P100. P4s and K80s, as chalsall put it, are not worth wasting free time on. He did acknowledge the irony, but I fully understand it. It is something of a sport to game the system.

[chalsall]

Originally Posted by Uncwilly

Except the initial phase of the healing, your cat should largely be ok. Depth perception will be an issue.

Thanks for the encouraging words. Indeed, Mars (aka "Marzipan" aka "Super Mars!") is doing fine. And has a new nickname...

Since he now looks somewhat Borg-ish, he's now also known as "2^3 of 3^2". Or "8 of 9" to his friends...