question for Martinos MNE-python users

✉️ Mailing List Archive (read-only)
1

Hello,
I've used the Martinos network version of ipython before with no
problems to run some stuff in MNE-python.

Now when I try to start it up by typing 'ipython' after setting the path
exactly as per 'Getting Started" instructions, it does not start for
many minutes (20+) and when it finally starts, it is extremely slow
(compared to previous usage or running a local version on a Mac laptop).

The problem is not with the machine, which has gobs of free RAM (8GB+)
and 4 CPUs that are mostly idle. Last time I quit ipython, it was using
>5GB of RAM, which seems excessive.

What am I missing here? Should I install a local version, e.g., Canopy?

Thanks,
Kestas

2

Hi Kestas,

It is likely related to the gpfs server problems mentioned on the
batch-users mailing list. Have you tried contacting help@?

D

3

Thanks, Dan, I will now.

4

Hi Kestas,

It is definitely due to the gpfs problems as Dan mentioned. The problem
is that when starting python many small files are accessed, which can be
slow. I assume you are using EPD? It is a bit faster if you use
anaconda, i.e., add this to your path:

/usr/pubsw/packages/python/anaconda/bin

and get an academic license from here:

https://store.continuum.io/cshop/academicanaconda

anaconda is faster to start than EPD because many files are included as
zip files, which reduces the number of file system operations when
starting python.

Hope this helps,

Martin

5

The 'gpfs problem' is resolved by rebooting, as we mentioned on batch-users.

I have instructed Kestas to reboot his computer.

-- Sam Mehl (http://help.nmr.mgh.harvard.edu)