Default Branch

develop 4975bd0c8e · [rocm-libraries] ROCm/rocm-libraries#9232 (commit 2d25e39) · Updated 2026-07-10 12:22:23 +00:00

ZIP  TAR.GZ

Branches

yewang12/flash-attn-yiding-ck 82ee798809 · Fix nsplits to min(8, nc) for chunk-level dyn_naive load balancing under NCCL overlap · Updated 2026-06-02 19:58:12 +00:00	3454 4048		ZIP  TAR.GZ
jukorhon/unified-attention-ck 64d3e00077 · Merge remote-tracking branch 'origin/dlejeune/ua-swa-v2' into jukorhon/unified-attention-ck · Updated 2026-06-01 12:45:43 +00:00	400 194		ZIP  TAR.GZ
dlejeune/ua-sink 3f930e18f5 · UA sink: add instances for fp8 · Updated 2026-05-29 13:56:23 +00:00	400 199		ZIP  TAR.GZ
dlejeune/ua-swa-v2 2b4020af0d · Merge branch 'jukorhon/unified-attention-ck' into dlejeune/ua-swa-v2 · Updated 2026-05-28 09:18:57 +00:00	400 190		ZIP  TAR.GZ
new_develop 58e2ab1fc7 · [rocm-libraries] ROCm/rocm-libraries#6761 (commit d19f6f1) · Updated 2026-05-27 18:55:15 +00:00	114 0	Included	ZIP  TAR.GZ
develop-archive-2026-05-27 c73c50a96e · [rocm-libraries] ROCm/rocm-libraries#7714 (commit 13ae6d6) · Updated 2026-05-27 18:11:21 +00:00	400 237		ZIP  TAR.GZ
AITERKER-112-temp 28814b4cfc · optimize batch_prefill pipeline based on fwd one · Updated 2026-05-27 15:59:40 +00:00	400 218		ZIP  TAR.GZ
samaario/ua-tune 92a7a354e1 · WIP · Updated 2026-05-27 13:32:47 +00:00	400 168		ZIP  TAR.GZ
jukorhon/ua-blocktables-oob-fix badc807025 · CK-UA: enable Tier-2 LDS page-table cache on decode + fix split-KV bulk-load OOB · Updated 2026-05-26 08:21:10 +00:00	400 178		ZIP  TAR.GZ
thomas/experiment 2a2992a4f0 · [rocm-libraries] ROCm/rocm-libraries#4871 (commit 7d4c040) · Updated 2026-05-22 18:39:52 +00:00	400 232		ZIP  TAR.GZ
matthias.gfx11_ck 15b4a580dc · [CK] AIESW-32282: bake bf16 truncate-to-bf16 conversion as the only behavior; drop *Truncate variants and DequantPolicyFor trait · Updated 2026-05-21 10:22:38 +00:00	400 221		ZIP  TAR.GZ
aghamari/cross-layer-5d-kv-cache 84354a0232 · ck_tile: drop customer-specific qualifier from cross-layer KV cache enum doc · Updated 2026-05-21 10:14:12 +00:00	400 216		ZIP  TAR.GZ
aiterker-112-fp8-per-token-per-head-prefill ee3ada6e4a · [AITERKER-112] PER_TOKEN_HEAD: support page_size < kN0 via cross-page dequant · Updated 2026-05-20 14:21:12 +00:00	400 216		ZIP  TAR.GZ
mohsen-backup ee3ada6e4a · [AITERKER-112] PER_TOKEN_HEAD: support page_size < kN0 via cross-page dequant · Updated 2026-05-20 14:21:12 +00:00	400 216		ZIP  TAR.GZ
jukorhon/ua-multipagetile-fix-fp8-rebased 589fe55d48 · CK-UA: lazy per-issue SRD rebase for the int32-overflow K/V load path · Updated 2026-05-20 10:24:12 +00:00	400 176		ZIP  TAR.GZ
ginolu/sparge_attention fb75da2467 · sparse_attn: wire -mask and -attention_sink (block-map prune + attn mask) · Updated 2026-05-20 03:22:00 +00:00	400 244		ZIP  TAR.GZ
jukorhon/ua-multipagetile-fix-fp8-thd 4aff2fa016 · CK-UA: fix no-mask multi-Q-block path — was reading too-short K prefix · Updated 2026-05-19 14:29:48 +00:00	400 177		ZIP  TAR.GZ
jukorhon/ua-multipagetile-fix-fp8 06e1a70e7a · CK-UA: constexpr page_size (Tier 3) — prefill_d64 fp8 -15.8%, prefill_d128 fp8 -6.3% · Updated 2026-05-19 12:46:39 +00:00	400 175		ZIP  TAR.GZ
users/msaffari-amd/ck/dispatcher_test_in_aiter 789a7f9c18 · Import composablekernel from rocm-libraries develop branch (d5c2ef6929) · Updated 2026-05-18 13:56:37 +00:00	3454 1		ZIP  TAR.GZ
jukorhon/ua-multipagetile-fix c0e985d075 · CK-UA: document why per-issue SRD-rebase path was tried and dropped · Updated 2026-05-15 10:18:39 +00:00	400 169		ZIP  TAR.GZ