Morphology classification dataset

[Eason-nuosen]

I'm trying to do your downstream tasks, but the data of 'images_npix152_0%02d000000_0%02d000000.h5' from legacy survey is too large for me. Could you support the file after crossing match? Thanks very much!

[lsarra]

Hi!

Do you mean this dataset? (56Gb)

AstroCLIP/astroclip/data/dataset.py

Line 35 in e6c23e0

"joint": "https://users.flatironinstitute.org/~flanusse/astroclip_desi.1.1.5.h5",

Or are you looking for only the embeddings obtained after going through astroclip (i.e. OBJECT_ID, EMBEDDING)? This is here

• https://github.com/lsarra/astrolit/blob/5d9c9fceba78d2c0be7f5c36e45cf97940a2e583/astrolit.py#L83C8-L83C72

[Eason-nuosen]

Thanks! But What I mean is about the downstream task (morphology classification). https://github.com/PolymathicAI/AstroCLIP/tree/main/downstream_tasks/morphology_classification
In cross_match.py, you mentioned that you used the full DESI-LS survey. It is too large. So I will really appreciate if you could support the result file after cross_match .