af3 for academic use

chai1: add chai-lab's structure token-pair weights (diffuser/chai1_structure_token_pair/{weights,bias}, diffuser/chai1_structure_bond/weights) - the second halves of input_projs.TOKEN_PAIR and bond_loss_input_proj, which chai-lab's diffusion module reads as its pair input. Every existing parameter byte-identical; JAX loads them unused. Needed by LocalFold's native CUDA port, which reads these blobs.

Upload esmfold2/esmfold2_lm600m.fp16.bin.zst with huggingface_hub

Upload esmfold2/esmfold2_lm300m.fp16.bin.zst with huggingface_hub

Upload esmfold2/esmfold2_lm600m.int8.bin.zst with huggingface_hub

Upload esmfold2/esmfold2_lm600m.bin.zst with huggingface_hub

Upload esmfold2/esmfold2_lm300m.int8.bin.zst with huggingface_hub

Upload esmfold2/esmfold2_lm300m.bin.zst with huggingface_hub

int8 rosettafold3/rosettafold3.int8.bin.zst: per-block scales, and the parameters the old blob was missing

int8 protenix/protenix2.int8.bin.zst: per-block scales, and the parameters the old blob was missing

int8 protenix/protenix1.int8.bin.zst: per-block scales, and the parameters the old blob was missing

int8 openfold3/openfold3.int8.bin.zst: per-block scales, and the parameters the old blob was missing

int8 opendde/opendde.int8.bin.zst: per-block scales, and the parameters the old blob was missing

int8 openfold3/openbind0.int8.bin.zst: per-block scales, and the parameters the old blob was missing

int8 intellifold2/intellifold2.int8.bin.zst: per-block scales, and the parameters the old blob was missing

int8 esmfold2/esmfold2_lm600m.int8.bin.zst: per-block scales, and the parameters the old blob was missing

int8 esmfold2/esmfold2_lm300m.int8.bin.zst: per-block scales, and the parameters the old blob was missing

int8 esmfold2/esmfold2_fast.int8.bin.zst: per-block scales, and the parameters the old blob was missing

int8 esmfold2/esmfold2.int8.bin.zst: per-block scales, and the parameters the old blob was missing

int8 chai1/chai1.int8.bin.zst: per-block scales, and the parameters the old blob was missing

int8 boltz2/boltz2.int8.bin.zst: per-block scales, and the parameters the old blob was missing

Upload rosettafold3/rosettafold3.fp16.bin.zst with huggingface_hub

Upload protenix/protenix2.fp16.bin.zst with huggingface_hub

Upload protenix/protenix1.fp16.bin.zst with huggingface_hub

Upload openfold3/openfold3.fp16.bin.zst with huggingface_hub

Upload opendde/opendde.int8.bin.zst with huggingface_hub

Upload opendde/opendde.fp16.bin.zst with huggingface_hub

Upload opendde/opendde.bin.zst with huggingface_hub

Upload openfold3/openbind0.fp16.bin.zst with huggingface_hub

Upload intellifold2/intellifold2.int8.bin.zst with huggingface_hub

Upload intellifold2/intellifold2.fp16.bin.zst with huggingface_hub

Upload intellifold2/intellifold2.bin.zst with huggingface_hub

Upload esmfold2/esmfold2_lm600m.fp16.bin.zst with huggingface_hub

Upload esmfold2/esmfold2_lm300m.fp16.bin.zst with huggingface_hub

Upload esmfold2/esmfold2_fast.int8.bin.zst with huggingface_hub

Upload esmfold2/esmfold2_fast.fp16.bin.zst with huggingface_hub

Upload esmfold2/esmfold2_fast.bin.zst with huggingface_hub

Upload esmfold2/esmfold2.int8.bin.zst with huggingface_hub

Upload esmfold2/esmfold2.fp16.bin.zst with huggingface_hub

Upload esmfold2/esmfold2.bin.zst with huggingface_hub

Upload chai1/chai1.int8.bin.zst with huggingface_hub

Upload chai1/chai1.fp16.bin.zst with huggingface_hub

Upload boltz2/boltz2.int8.bin.zst with huggingface_hub

Upload boltz2/boltz2.fp16.bin.zst with huggingface_hub

Upload boltz2/boltz2.bin.zst with huggingface_hub

Upload esmfold2/esmfold2_lm300m.int8.bin.zst with huggingface_hub

Upload esmfold2/esmfold2_lm600m.int8.bin.zst with huggingface_hub

Upload esmfold2/esmfold2_exp_fast_cutoff2025.int8.bin.zst with huggingface_hub

Upload esmfold2/esmfold2_exp_cutoff2025.int8.bin.zst with huggingface_hub

Upload esmfold2/esmfold2_exp_fast.int8.bin.zst with huggingface_hub