Colab question

[kriesel]

Originally Posted by kladner

Give me a little while to take care of stuff here, and I will write it up.

It will be interesting to see how it would differ from the content of https://www.mersenneforum.org/showthread.php?t=24839

[kladner]

Originally Posted by kriesel

It will be interesting to see how it would differ from the content of https://www.mersenneforum.org/showthread.php?t=24839

At a glance there are two things. I'm running in Windows 10, and my operations are entirely manual. There is no scripting. I used to get along with batch files for some things, but this is totally primitive hands-on.

[S485122]

Originally Posted by kladner

...
Readers: Remember this helpful hint regarding who's really paying:

If it is free, it is not the product.

You are.

We just don't know what the payoff for them is.

I can't agree with that statement : it may be free in the sense that you don't pay directly. But not only are we (or rather our data) the product but in the end we pay for it. All that datamining rage is to flood us with ads. The thing everybody forgets is that the consumers pay for the whole advertisement industry since it is part of almost anything we buy.

I would add, straggling even more of topic, that (almost ?) everybody would prefer to have no adds (uBlockOrigin or Adblock anyone ?) no lengthy interruption of television or radio programs, no huge billboards defiguring the public space... I could go on, but will stop here ;-)

Jacob

[Nick]

Originally Posted by S485122

All that datamining rage is to flood us with ads.

Not only that - in the long term, it will be used for many other things too, such as price discrimination.

[Aramis Wyler]

Originally Posted by LOBES

How do you run two separate colabs using the same Google account? As soon as I try to start up a second one when logged into the same account, it immediately connects to the other hosted runtime. I don't see how to run more than one colab window using the same Google account.

You can only have one notebook per session - the sessions are linked to notebooks because it saves the results of your session in the data portion of your notebook, and so you can't just run it 4 times for 4 sessions. I have 9 notebooks - 4 are gpu721-4, 4 are gpuOwl1-4, and 1 is ATH, the one that can download and compile a new gpuowl binary and supporting files from the git repository (based on ATH's posts).


I have 1 account, and it is paid. I run 4 notebooks at a time, and I start them once per day leaving the 4 tabs open in my browser. The gpuowl ones:

1. Mounts the google drive.
2. Copies the libstd files to the remote computer.
3. Echos my worktodo.txt file (I like to see it).
4. chmod my scripts and binaries to 777.
5. Runs primenet.py (just a one-off), to upload all of the results and proofs from yesterday and fetch new work.
6. Starts gpuOwl.

And that's all, I run 4 at a time (one notebook per tab) and have to get on once a day to start them. I don't run any free accounts, and they complete PRPs in the 105M range in 17 hours and 35 minutes. So most days primenet.py uploads one result and proof (on each notebook) but some days some notebooks give 2.

[kladner]

Originally Posted by S485122

I can't agree with that statement : it may be free in the sense that you don't pay directly. But not only are we (or rather our data) the product but in the end we pay for it. All that datamining rage is to flood us with ads. The thing everybody forgets is that the consumers pay for the whole advertisement industry since it is part of almost anything we buy.

I would add, straggling even more of topic, that (almost ?) everybody would prefer to have no adds (uBlockOrigin or Adblock anyone ?) no lengthy interruption of television or radio programs, no huge billboards defiguring the public space... I could go on, but will stop here ;-)

Jacob

The word "Free" should have had "scare quotes." Its use is ironic. Sorry for launching a digression.

[Ensigm]

Originally Posted by kriesel

But as of 2020-02-10, there's this, which should streamline things considerably: https://twitter.com/GoogleColab/stat...29213560610818
(Thanks for the notification, kracker)

Thanks for the post. This really saves me lots of hassle.


Also that the 'Mount Drive' button is now represented by an icon (without the text).

[Aramis Wyler]

Hey, that's nice! Got my notebook down to 6 lines and I don't have to deal with clicking through the other tab and doing the copy paste.

[butera]

Is there a way to set up a notebook to trial factor, but use google drive to set up a worktodo file if I want to manually TF some exponents?

[chalsall]

Originally Posted by butera

Is there a way to set up a notebook to trial factor, but use google drive to set up a worktodo file if I want to manually TF some exponents?

Sure... Take a look at the sixth post in the Google Co-Educational Notebook? thread which started this whole adventure.

[Aramis Wyler]

That notebook is pre-google_drive_mount_button, so you'd probably want to ignore the first 6 lines of that and use the mount button instead.

And since you can execute scripts from the notebook, I find it convenient to put the upload/fetch script in a shared drive folder folder and call it from each notebook, giving each notebook it's own folder to run executables from and worktodo/results/config/uploads/etc.

Here's my owl one:

%cd '/content/drive/My Drive/owl1'
!chmod 755 ../bin/*
!cp ../bin/libstdc* /usr/lib/x86_64-linux-gnu/
!../bin/primenet.py -u staffen -p YeahRight --dirs ./ -w PRP -t 0
!cat worktodo.txt
!../bin/gpuowl -user staffen -maxAlloc 15360 -cpu colab1 -log 2000000