high-res spectrum comparison

[changhoonhahn]

@stephjuneau pointed out that some EDR spectra have high-resolution spectra from Gemini/GHOST.

These quasars could be nice comparisons for the EDR catalog paper. @SGontcho could you track down the GHOST spectra and their targetids in the EDR?

[SGontcho]

The GHOST Spectrum has TARGETID=39627775732093815.

[stephjuneau]

Yes, it's a BGS-AGN target in addition to being a Ly-a QSO :-)

[SGontcho]

Copy/pasting here some tutorial-like instructions I gave to @stephjuneau

Note that TARGETID=39627775732093815 will be in the /global/cfs/projectdirs/desi/users/chahah/spender_qso/DESIedr.qso_highz_*.pkl production.

You’ll notice there’s a couple libraries specific to SpenderQ needed, below is some instructions to have the proper setup at NERSC:

Installing SpenderQ @NERSC

conda create -y -n spenderq-env python pytorch accelerate h5py torchvision torchaudio pytorch-cuda=12.1 -c pytorch -c nvidia
conda activate spenderq-env
git clone https://github.com/changhoonhahn/SpenderQ.git
cd spender
pip install -e .

To resolve conflict between MKL and PyTorch

conda install mkl=2024.0.0

Then also add the following packages

conda install scipy numba

Jupyter Notebook at NERSC

Set your custom environment:

conda install ipykernel
python -m ipykernel install --user --name spenderq-env

=========

This notebook from the SpenderQ GitHub exemplifies how to grab data and manipulate them (in order to create some output files that are compatible with software from the LyA WG, but it's fine to follow along until cell 25 "Grab values from SpenderQ files" -- it shows how to open SpenderQ output and manipulate it): https://github.com/galactic-ai/SpenderQ/blob/main/bin/london/SpenderQ_to_Picca.ipynb

[changhoonhahn]

@stephjuneau Does the following look like the matching spectrum?

If so, it's in batch 15 of SpenderQ outputs. The corresponding SpenderQ reconstructions are: edr.lfsc.rsnr.sa1p1sb0p8.i3.[MODEL_NUM]_15.recons.npy where MODEL_NUM is 0-4.

[stephjuneau]

Yes, this looks like the one! Thanks @changhoonhahn