Request to package frame reading software for pip
Details
- Package name: gwpy and frameCPP
- Version:
Description
Here is a request we got from a person named Federico Muciaccia. I would appreciate guidance in how to reply to this person. Any notes on the availability (or not) of frame reading via pypi would be most helpful. Thank you!! -jonah
Dear GWOSC support team,
I have a problem trying to read GWF frames files using Python.
Im trying to use GWpy, but it is not able to read GWF frames files out of the box.
In the GWpy documentation webpage
https://gwpy.github.io/docs/stable/timeseries/io/
under the section about reading GWF files, there is mention to some (required) additional dependencies: LDAStools.frameCPP or framel or lalframe. I have problems with all of them...
1) LDAStools.frameCPP
The "LDAStools.frameCPP" link is behind password. It thus seams not available to the open pubblic.
https://ldas-jobs.ligo.caltech.edu/~emaros/share/doc/ldas-tools/framecpp/html/
2) framel
The "framel" link points to the "Fr" LIGO Git repository.
http://lappweb.in2p3.fr/virgo/FrameL/
Inside that repository, there isn't an already-packed executable.
The software cannot be installed with Python Pip (which should be the default way to go) because no package seems present in any Pip database.
There is, however, an already-packed version for Anaconda (in the conda-forge repository).
The problem is that many researchers do prefer to use regular Pip packages instead of Conda ones or the official distribution ones (eg: the ones provided by Ubuntu or Debian). This is because Conda packages are slowly updated and distribution packages have even slower updates.
Especially in Machine Learning (which is my field of research) this can lead to a lack of new features or even incompatibilities, especially (in my previous experience) with Nvidia CUDA, which are mandatory for GPU acceleration of the computation.
So, Pip package is absent, and I have to install the whole Anaconda (10+ GB) in order to try to be able to use the "framel" package inside GWpy to be able to read the .gwf files.
But a Conda-installed "framel" works by default with a Conda-installed "gwpy", not a Pip-installed one.
So, to avoid the 10+ GB bargain on my computer, I tried using a Google Colab notebook (which is very utilized in the Machine Learning community), installed a mini version of Anaconda, and tried to Conda-install the "gwpy" and "framel" packages. But, as somehow expected, there is a problem with the GWpy package in Conda during the installation, because it has some conflicts with the Conda-installed CUDA.
So it seems there is no straightforward way to use "framel".
3) lalframe
The "lalframe" link points to
http://software.ligo.org/docs/lalsuite/lalframe/
and it simply returns a "404 Not Found" error.
...
In conclusion, I think the simple solution to all those problems could be to provide a standard Pip package that can be installed together with the Pip-installed GWpy (or, even better, make "framel" a default dependence of it).
Can you please make it?
Thanks in advance
Federico Muciaccia (Virgo-Rome data analysis group)
Python versions
Please select the Python versions appropriate for this package (deselect those that are not required):
-
3.9
Request type
Please select all of the following that apply:
-
this is a request to package frame reading software via pip -
this is a new package -
this is a backwards-compatible update -
this is a backwards-incompatible update [API/ABI changes]