45.6 GB

Ctrl+K

2 contributors

gemma4_e4b_int4_rawmetal raw-Metal int4 pack on the native QAT grid — tokenizer, generation config and stop ids inside; manifest carries sizes and sha256s

6eb9514 verified about 1 month ago

ctx131072
e4b cap bundle: README with provenance, KV arithmetic, digests about 1 month ago
gpu-pipelined
prune: remove superseded/non-functional artifacts about 2 months ago
ios-frontend
Add PLE gather_raw table sidecar about 2 months ago
levered2
levered2/ — gather-first embedding, int4 PLE, fp16 head about 1 month ago
raw-metal
gemma4_e4b_int4_rawmetal raw-Metal int4 pack on the native QAT grid — tokenizer, generation config and stop ids inside; manifest carries sizes and sha256s about 1 month ago
ring
Add ring gemma4_e4b_qat_decode_int4lin_tbl_pf64_ring_c16384 about 2 months ago
ring3
Add ring3 gemma4_e4b_qat_decode_int4lin_tbl_pf64_ring_c4096_gqafull about 2 months ago
stable
Add stable gemma4_e4b_qat_decode_int4lin_tbl_pf64_stable_c16384 about 2 months ago
v4
v4: head-sidecar E4B — fp16 head as a bound input beside the PLE pair about 1 month ago
v5
v5: the verify entrypoint — full logits + activations for the assistant drafter about 1 month ago
w4a8
Add w4a8/gemma4_e4b_qat_decode_int4lin_a8_tbl_pf64 about 2 months ago
.gitattributes

6.32 kB
E2B raw-Metal int4 pack — our exact int4 recipe, bi dropped, softcapped head — plus the KV-sharing assistant pack; manifest carries sizes and sha256s about 1 month ago
README.md

23.8 kB
card: levered2/ — gather-first embedding, int4 PLE, fp16 head about 1 month ago