Vaishaal
Vaishaal
Yeah the original hash was computed during commonpool download and then passed around to the dc1b parquet, maybe there was a bug in that data processing pipeline... ________________________________ From: pfischer-nvidia...
hello we are looking into this right now! So the download code that computes the hash is right here: https://github.com/rom1504/img2dataset/blob/main/img2dataset/downloader.py#L203-L318C6 We are checking our internal pool to see what the...
yeah I believe so, my money is on 2. What we can do is just look at our internal copy of one of the hash mismatched images and see if...
Which dataset would you like to access? I can create a public url for you.
Ah you need to use ZCA pre-processing as mentioned in the paper. 76 seems correct for standard pre_processing. Here is a link for ZCA pre-processed cifar-10: https://imagenetv2public.s3.us-west-2.amazonaws.com/cifar_zca.npz
Ah I am pretty sure the code only works on V100s or TitanV right now. I am working on an implementation based on Jax that supports more GPU types but...
Noted! What was the exception, that should not happen. I can try to fix it.