An issue I have there is that you should provide the links to the piece of the script, everywhere you mention the computer name (same as you do in the [URL="https://www.gpu72.com/account/computers/"]computers[/URL] page, I click the link and see the code for that specific notebook, with the assignment key and everything). So, the link should be provided everywhere you display the notebook name, with it. The reason is that when we completely delete a "notebook" the code is gone, and then we have to navigate gpu72 site till we find the link to the code, to copy/paste it. Our memory is very good but we regularly change computers here around, we don't remember where we put this good memory.... Now, well, do not ask us why we need to completely delete a notebook, the reason may be related to the fact the notebooks have some memory (wild ass guess here) and they will continue to allocate us the k80, unless we delete it and start from scratch, when we may get a T4 (we tried this successfully few times, but no idea if it was just luck).

Hello! Where I can download the latest version CudaLucas compiled for Linux/colab? On [url]https://sourceforge.net/projects/cudalucas/files/2.06/[/url] I found only Windows version.

[QUOTE=LaurV;527134]Now, well, do not ask us why we need to completely delete a notebook, the reason may be related to the fact the notebooks have some memory (wild ass guess here) and they will continue to allocate us the k80, unless we delete it and start from scratch, when we may get a T4 (we tried this successfully few times, but no idea if it was just luck).[/QUOTE]

I haven't had a T4 since the first time I used this. I'm going to try making a new notebook and see if that helps.

[QUOTE=chalsall;527124]Please note that if you're going to manually feed these checkpoint records into an instance for completion, mfaktc is a little finicky about the format of the M\d+.ckp files.

Specifically, single string; no CR/LF. When doing this kind of thing by hand, I had to use the "truncate" command to generate legal files after a "vi" or "echo".[/QUOTE]

This is good. But for me it is easier to download from the Colab notebook the checkpoint file and the worktodo.txt from the /home/gpu72 directory and move them into my local files.

It takes 3 hours 15 min on the K80 for me to complete one TF, so you get three completed and a partial out of each 12 hour session.

[QUOTE=Chuck;527162]I haven't had a T4 since the first time I used this. I'm going to try making a new notebook and see if that helps.[/QUOTE]
[QUOTE=Chuck;527163]It takes 3 hours 15 min on the K80 for me to complete one TF, so you get three completed and a partial out of each 12 hour session.[/QUOTE]It has been a bit since I have seen a T4. But, I have been getting more than 30 hours a week. It seems that there maybe a 12 in 24 restriction. I am using a corporate gmail login. So, I have been having it run during work hours. I have tried closing down the session, exiting my browser, reopening the browser, and spinning up a new session. It does not seem to get me more run time.

Getting error in new notebook
 

[CODE]20191002_151022: Installing needed packages (1/3)
20191002_151026: Installing needed packages (2/3)
20191002_151027: Installing needed packages (3/3)
20191002_151028: Fetching initial work...
20191002_151033: Running GPU type Tesla K80

20191002_151033: running a simple selftest...
20191002_151038: Selftest statistics
20191002_151038: number of tests 107
20191002_151038: successfull tests 107
20191002_151038: selftest PASSED!
20191002_151038: ERROR: get_next_assignment(): no valid assignment found in "worktodo.txt"
20191002_151038: Bootstrap finished. Exiting.[/CODE]

Says it is fetching work, but empty worktodo file

[QUOTE=Chuck;527175]Says it is fetching work, but empty worktodo file[/QUOTE]

Hmmm... What instance is this? (Name would be fine; please don't give me your AKey(s) publicly.)

There have been *no* code deltas for the last two days wrt assignment code-paths.

When you try re-running it immediately, what happens?

[QUOTE=chalsall;527176]Hmmm... What instance is this? (Name would be fine; please don't give me your AKey(s) publicly.)

There have been *no* code deltas for the last two days wrt assignment code-paths.

When you try re-running it immediately, what happens?[/QUOTE]

Sorry I already deleted the notebooks. I did try rerunning it before I deleted it, and got the same result. I mounted the Google drive, and there was an empty worktodo.txt file

This was Colab and I have only one version in my Computers list.

I see there are no available 69M exponents for TF. Could this be the problem?

[QUOTE=Chuck;527181]I see there are no available 69M exponents for TF. Could this be the problem?[/QUOTE]

Dooohhh!!! Now that's an example of a profoundly stupid programmer error!!!

Please give it another go...

Working on 65M
 

Working....

I thought something might be amiss when I checked my assignments earlier and found that Colab was not keeping three exponents in reserve.