datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
cptac-3Clustered_UP000005640
Human reference proteome dataset (UP000005640) is first clustered with 80% similarity with MMSeqs2's easy-cluster.
The cluster representatives within the FASTA file are then searched against uniref30_2302 database with ColabFold's colabfold_search function.
The final aligned a3m files placed in a tarball and compressed with zstd.
cupcake-cpt-1wsi-cptac-vqa-2026