Details about the Metrics Computation

[Nan-S]

Can you provide the details about the metrics computation?
I found the code in the following run file:

few-shot-diffusion-models/script/compute_fid.sh

Line 117 in f9e86f7

python metrics/metrics.py /scratch/gigi/fsddpm/cifar100_ddpm_sigma/sampling-conditional-out-distro-2022-04-11-01-22-01-001184/full_samples_conditional_10000x32x32x3_out-distro_5_transfer_minimagenet.npz /scratch/gigi/fsddpm/cifar100_ddpm_sigma/sampling-conditional-out-distro-2022-04-11-01-22-01-001184/full_samples_conditional_10000x32x32x3_out-distro_5_transfer_minimagenet.npz

Looks like the two '.npz' files are all full_samples_conditional_10000x32x32x3_in-distro_5.npz, how you selected the reference images in the new classes? Or can you provide the reference data?