diff --git a/blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_09bcbbfb-0d48-42d8-a39c-e7d01300a646.safetensors b/blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_09bcbbfb-0d48-42d8-a39c-e7d01300a646.safetensors deleted file mode 100644 index a496963d4aeb8636c359764c90401c5d42eb9bb0..0000000000000000000000000000000000000000 --- a/blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_09bcbbfb-0d48-42d8-a39c-e7d01300a646.safetensors +++ /dev/null @@ -1,3 +0,0 @@ -version https://git-lfs.github.com/spec/v1 -oid sha256:6029b05a935f29c145a19ebe7928f10871387124bb9d56e72f78248da1beb106 -size 4936 diff --git a/blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_1196e756-48f7-4f94-81a3-94a68f170b59.safetensors b/blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_1196e756-48f7-4f94-81a3-94a68f170b59.safetensors deleted file mode 100644 index 73ff8924cd23ac7a9e8cafed4538323fa62a5aef..0000000000000000000000000000000000000000 --- a/blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_1196e756-48f7-4f94-81a3-94a68f170b59.safetensors +++ /dev/null @@ -1,3 +0,0 @@ -version https://git-lfs.github.com/spec/v1 -oid sha256:e9e2d9e187fb73fa76039d00a97177dcdb7be30798273d3adf8958a45317334c -size 4488 diff --git a/blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_16d8ee84-f85b-4192-a58a-2228a5d9641e.safetensors b/blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_16d8ee84-f85b-4192-a58a-2228a5d9641e.safetensors deleted file mode 100644 index 31f7ceff6cf25777e52f6d8a301ff5db5f9dd9ca..0000000000000000000000000000000000000000 --- a/blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_16d8ee84-f85b-4192-a58a-2228a5d9641e.safetensors +++ /dev/null @@ -1,3 +0,0 @@ -version https://git-lfs.github.com/spec/v1 -oid sha256:f2b05c877e6670fe876c4ea47ea81c575d1c87923a03770cadbeb728ed1eb5fd -size 110308 diff --git a/blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_1735d27c-3a9a-44b4-9fa1-e9a0e3703d38.safetensors b/blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_1735d27c-3a9a-44b4-9fa1-e9a0e3703d38.safetensors deleted file mode 100644 index 8dd32dc997fb83d95cf7fe5c9dbc0617510b9dbd..0000000000000000000000000000000000000000 --- a/blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_1735d27c-3a9a-44b4-9fa1-e9a0e3703d38.safetensors +++ /dev/null @@ -1,3 +0,0 @@ -version https://git-lfs.github.com/spec/v1 -oid sha256:e18b47c17007cebe099f3633e49667ad39ac63d1082560609e362b7864b1cdb5 -size 2596 diff --git a/blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_1a1f0089-58e8-4bb5-81a6-fbe229e3b845.safetensors b/blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_1a1f0089-58e8-4bb5-81a6-fbe229e3b845.safetensors deleted file mode 100644 index 03118a5d78eec05f3ac2deb269a8d194dead3f28..0000000000000000000000000000000000000000 --- a/blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_1a1f0089-58e8-4bb5-81a6-fbe229e3b845.safetensors +++ /dev/null @@ -1,3 +0,0 @@ -version https://git-lfs.github.com/spec/v1 -oid sha256:2a39285b41ea7dd5ad2cb6834025dc66ac31f51550af5c8f8deba71de84df516 -size 2852 diff --git a/blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_1bdf652e-366a-4024-9236-ce54682a7cf7.safetensors b/blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_1bdf652e-366a-4024-9236-ce54682a7cf7.safetensors deleted file mode 100644 index 03bd6610f1cbae42e8494337480076affe5cb384..0000000000000000000000000000000000000000 --- a/blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_1bdf652e-366a-4024-9236-ce54682a7cf7.safetensors +++ /dev/null @@ -1,3 +0,0 @@ -version https://git-lfs.github.com/spec/v1 -oid sha256:cd41886645e74768ae33d892ce5d4d2fff1169caa751d74ab3a28fb1a2a6934f -size 57064 diff --git a/blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_1e0c7af6-f077-42f6-b5a0-3cbb2e30c5c3.safetensors b/blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_1e0c7af6-f077-42f6-b5a0-3cbb2e30c5c3.safetensors deleted file mode 100644 index d36cd4bb35ad2f86e6acdfc07a2ca5ef11466d46..0000000000000000000000000000000000000000 --- a/blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_1e0c7af6-f077-42f6-b5a0-3cbb2e30c5c3.safetensors +++ /dev/null @@ -1,3 +0,0 @@ -version https://git-lfs.github.com/spec/v1 -oid sha256:7e9521f54dba6336ed2e881c2f2748f4da08bb4f0976a84657313aae4067c7d8 -size 1180 diff --git a/blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_2b180048-9203-476c-a5b4-255118f15059.safetensors b/blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_2b180048-9203-476c-a5b4-255118f15059.safetensors deleted file mode 100644 index 517e45db9e8b8ccc397cf32a254e985062356c9b..0000000000000000000000000000000000000000 --- a/blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_2b180048-9203-476c-a5b4-255118f15059.safetensors +++ /dev/null @@ -1,3 +0,0 @@ -version https://git-lfs.github.com/spec/v1 -oid sha256:4257c5e2a5027a1c15f12c481457a30c76c7d9402a380505f42547c46beb0965 -size 191020 diff --git a/blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_51393402-8fec-4b52-88ca-3cdb2270d293.safetensors b/blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_51393402-8fec-4b52-88ca-3cdb2270d293.safetensors deleted file mode 100644 index da90faa9c2bbe9e5b382a2e45725afc146bbcb0a..0000000000000000000000000000000000000000 --- a/blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_51393402-8fec-4b52-88ca-3cdb2270d293.safetensors +++ /dev/null @@ -1,3 +0,0 @@ -version https://git-lfs.github.com/spec/v1 -oid sha256:934a282339e721dbe2d8cf3e82a2c2ad84302ac682d8ca8b14585b2327887189 -size 107076 diff --git a/blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_580fa22c-dee6-484e-8847-eb301b21057f.safetensors b/blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_580fa22c-dee6-484e-8847-eb301b21057f.safetensors deleted file mode 100644 index 64b3add14b3f4d0d0d131b40ec68ce723a18103b..0000000000000000000000000000000000000000 --- a/blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_580fa22c-dee6-484e-8847-eb301b21057f.safetensors +++ /dev/null @@ -1,3 +0,0 @@ -version https://git-lfs.github.com/spec/v1 -oid sha256:43fe2d472b0737a910c0c9892ccd28ed00603c0046e4f6f6e34651785c347c61 -size 1332 diff --git a/blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_721ff730-4d9e-49d8-9fd4-7ba713b7cfdb.safetensors b/blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_721ff730-4d9e-49d8-9fd4-7ba713b7cfdb.safetensors deleted file mode 100644 index 346cfc2c0c34f8aaf36996ab462076f4bedcdc99..0000000000000000000000000000000000000000 --- a/blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_721ff730-4d9e-49d8-9fd4-7ba713b7cfdb.safetensors +++ /dev/null @@ -1,3 +0,0 @@ -version https://git-lfs.github.com/spec/v1 -oid sha256:5f8ee5403b5264d69131641160b12fe971d36ec890b157bc516f596d24498a85 -size 55624 diff --git a/blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_7832673f-b1c1-4185-b298-d510edbceac4.safetensors b/blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_7832673f-b1c1-4185-b298-d510edbceac4.safetensors deleted file mode 100644 index 21784c97d6aba4bbb55690d91240a22abd8ea2a2..0000000000000000000000000000000000000000 --- a/blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_7832673f-b1c1-4185-b298-d510edbceac4.safetensors +++ /dev/null @@ -1,3 +0,0 @@ -version https://git-lfs.github.com/spec/v1 -oid sha256:36c43aa363876040abb4f39bacbf22888b41fbfba7adffce4eca977192d5d018 -size 193740 diff --git a/blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_8c88a86f-1843-42eb-9251-8a0ef4b6a1ee.safetensors b/blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_8c88a86f-1843-42eb-9251-8a0ef4b6a1ee.safetensors deleted file mode 100644 index 901aea4b0082b1773b72ede41f16d848a4794b10..0000000000000000000000000000000000000000 --- a/blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_8c88a86f-1843-42eb-9251-8a0ef4b6a1ee.safetensors +++ /dev/null @@ -1,3 +0,0 @@ -version https://git-lfs.github.com/spec/v1 -oid sha256:d8cc1d8ff4bb54a8f319890ce73c7aad7aa4ed5006b3851dcf1406a0de1cd99f -size 109636 diff --git a/blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_b013fc3f-eca8-4c18-a5a1-7231c26eb231.safetensors b/blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_b013fc3f-eca8-4c18-a5a1-7231c26eb231.safetensors deleted file mode 100644 index edae6fff68fc009ea4a51fffeaafc6be04548f67..0000000000000000000000000000000000000000 --- a/blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_b013fc3f-eca8-4c18-a5a1-7231c26eb231.safetensors +++ /dev/null @@ -1,3 +0,0 @@ -version https://git-lfs.github.com/spec/v1 -oid sha256:4a7144502e3e38588142893176c0cc4dd6ffa75ac1fa73cf2bba07f2bb81fd8f -size 1100 diff --git a/blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_c81ef23a-da17-4cdd-88c2-327a4af240d0.safetensors b/blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_c81ef23a-da17-4cdd-88c2-327a4af240d0.safetensors deleted file mode 100644 index 8a3abcf0609123b2591a4aa00a69b959cb3aa8b6..0000000000000000000000000000000000000000 --- a/blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_c81ef23a-da17-4cdd-88c2-327a4af240d0.safetensors +++ /dev/null @@ -1,3 +0,0 @@ -version https://git-lfs.github.com/spec/v1 -oid sha256:c1d917c6a8a76887b2f80e4b8713d7941b46de7c9893f85e743e80ce3bb7a258 -size 186924 diff --git a/blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_ed7c988c-a1a9-4e62-b410-2e33b2bdc1d8.safetensors b/blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_ed7c988c-a1a9-4e62-b410-2e33b2bdc1d8.safetensors deleted file mode 100644 index 908a99df340f4215d82d2b83c72c7258139f3a8f..0000000000000000000000000000000000000000 --- a/blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_ed7c988c-a1a9-4e62-b410-2e33b2bdc1d8.safetensors +++ /dev/null @@ -1,3 +0,0 @@ -version https://git-lfs.github.com/spec/v1 -oid sha256:b21e4fd7434bebf29c986f13d729820469e9b699c4ccb9bcfe809b79d7f46dcc -size 2404 diff --git a/blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_ee6efcf9-bf19-432f-bcd3-33f1c9ec599e.safetensors b/blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_ee6efcf9-bf19-432f-bcd3-33f1c9ec599e.safetensors deleted file mode 100644 index ecbd9b1d4d19682b4de8d75d5097744b59e5b3e5..0000000000000000000000000000000000000000 --- a/blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_ee6efcf9-bf19-432f-bcd3-33f1c9ec599e.safetensors +++ /dev/null @@ -1,3 +0,0 @@ -version https://git-lfs.github.com/spec/v1 -oid sha256:451e0495ee71e25a3f23ae8a40a8e624b60d80e6de0cc68b29e7d3604312ef4b -size 5384 diff --git a/blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_ff1d78b9-0383-4845-94df-904c49112856.safetensors b/blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_ff1d78b9-0383-4845-94df-904c49112856.safetensors deleted file mode 100644 index dd235ba265fc2bb5a427096d1317c41689b2d7e5..0000000000000000000000000000000000000000 --- a/blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_ff1d78b9-0383-4845-94df-904c49112856.safetensors +++ /dev/null @@ -1,3 +0,0 @@ -version https://git-lfs.github.com/spec/v1 -oid sha256:c8ad66849d6ad3e32f02f1f02f86a1a6959fffc2c446772bf822db37d026a56d -size 56904 diff --git a/blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps1/gqa_paged_decode_h32_kv8_d64_ps1_011db24f-e331-41e4-89fd-dff3bb7345e5.safetensors b/blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps1/gqa_paged_decode_h32_kv8_d64_ps1_011db24f-e331-41e4-89fd-dff3bb7345e5.safetensors deleted file mode 100644 index a1cb22abc614ff61355273cdab16f71aaa4b9919..0000000000000000000000000000000000000000 --- a/blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps1/gqa_paged_decode_h32_kv8_d64_ps1_011db24f-e331-41e4-89fd-dff3bb7345e5.safetensors +++ /dev/null @@ -1,3 +0,0 @@ -version https://git-lfs.github.com/spec/v1 -oid sha256:a8515217c3a58a0b25f457145ee5f96cac1f263b5d3b6292b4198c1e82224d53 -size 176 diff --git a/blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps1/gqa_paged_decode_h32_kv8_d64_ps1_11155649-fd33-4b5c-80d8-05553c0cec0b.safetensors b/blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps1/gqa_paged_decode_h32_kv8_d64_ps1_11155649-fd33-4b5c-80d8-05553c0cec0b.safetensors deleted file mode 100644 index b9904d811535875060f2037352700a7b08c95553..0000000000000000000000000000000000000000 --- a/blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps1/gqa_paged_decode_h32_kv8_d64_ps1_11155649-fd33-4b5c-80d8-05553c0cec0b.safetensors +++ /dev/null @@ -1,3 +0,0 @@ -version https://git-lfs.github.com/spec/v1 -oid sha256:eb95af94a5630438bfff0db796fdb2e84cd29dbbabe153af11b09599ee535a11 -size 204 diff --git a/blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps1/gqa_paged_decode_h32_kv8_d64_ps1_3d044768-4601-4b70-8472-85fbc34a4504.safetensors b/blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps1/gqa_paged_decode_h32_kv8_d64_ps1_3d044768-4601-4b70-8472-85fbc34a4504.safetensors deleted file mode 100644 index 65fcc9aebe79dfbc96127ebd96bc49269cbd19d2..0000000000000000000000000000000000000000 --- a/blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps1/gqa_paged_decode_h32_kv8_d64_ps1_3d044768-4601-4b70-8472-85fbc34a4504.safetensors +++ /dev/null @@ -1,3 +0,0 @@ -version https://git-lfs.github.com/spec/v1 -oid sha256:c14d6c71f58ae5becc77146223456c973828b39d9d87e1cdc9d0d0f0478adc24 -size 152 diff --git a/blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps1/gqa_paged_decode_h32_kv8_d64_ps1_5d4bb19c-8f86-41c2-a3dc-ab974ba60bdf.safetensors b/blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps1/gqa_paged_decode_h32_kv8_d64_ps1_5d4bb19c-8f86-41c2-a3dc-ab974ba60bdf.safetensors deleted file mode 100644 index 67e8f3a52afe8beac1b20a1f479dedbd81afd3c9..0000000000000000000000000000000000000000 --- a/blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps1/gqa_paged_decode_h32_kv8_d64_ps1_5d4bb19c-8f86-41c2-a3dc-ab974ba60bdf.safetensors +++ /dev/null @@ -1,3 +0,0 @@ -version https://git-lfs.github.com/spec/v1 -oid sha256:999f3eb7ac27170d0ce513b86797d92dc2d32396bc99d5069dfefd0485b2d3a5 -size 376 diff --git a/blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps1/gqa_paged_decode_h32_kv8_d64_ps1_7810a1b5-7bcd-4825-b36f-80c9b4ec61c6.safetensors b/blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps1/gqa_paged_decode_h32_kv8_d64_ps1_7810a1b5-7bcd-4825-b36f-80c9b4ec61c6.safetensors deleted file mode 100644 index 67e8f3a52afe8beac1b20a1f479dedbd81afd3c9..0000000000000000000000000000000000000000 --- a/blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps1/gqa_paged_decode_h32_kv8_d64_ps1_7810a1b5-7bcd-4825-b36f-80c9b4ec61c6.safetensors +++ /dev/null @@ -1,3 +0,0 @@ -version https://git-lfs.github.com/spec/v1 -oid sha256:999f3eb7ac27170d0ce513b86797d92dc2d32396bc99d5069dfefd0485b2d3a5 -size 376 diff --git a/blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps1/gqa_paged_decode_h32_kv8_d64_ps1_8342a570-03f6-42ab-b7fb-bc1eadc8759e.safetensors b/blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps1/gqa_paged_decode_h32_kv8_d64_ps1_8342a570-03f6-42ab-b7fb-bc1eadc8759e.safetensors deleted file mode 100644 index a1cb22abc614ff61355273cdab16f71aaa4b9919..0000000000000000000000000000000000000000 --- a/blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps1/gqa_paged_decode_h32_kv8_d64_ps1_8342a570-03f6-42ab-b7fb-bc1eadc8759e.safetensors +++ /dev/null @@ -1,3 +0,0 @@ -version https://git-lfs.github.com/spec/v1 -oid sha256:a8515217c3a58a0b25f457145ee5f96cac1f263b5d3b6292b4198c1e82224d53 -size 176 diff --git a/blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps1/gqa_paged_decode_h32_kv8_d64_ps1_921b9fdc-3937-4873-b11e-7f4650daf0f8.safetensors b/blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps1/gqa_paged_decode_h32_kv8_d64_ps1_921b9fdc-3937-4873-b11e-7f4650daf0f8.safetensors deleted file mode 100644 index b9904d811535875060f2037352700a7b08c95553..0000000000000000000000000000000000000000 --- a/blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps1/gqa_paged_decode_h32_kv8_d64_ps1_921b9fdc-3937-4873-b11e-7f4650daf0f8.safetensors +++ /dev/null @@ -1,3 +0,0 @@ -version https://git-lfs.github.com/spec/v1 -oid sha256:eb95af94a5630438bfff0db796fdb2e84cd29dbbabe153af11b09599ee535a11 -size 204 diff --git a/blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps1/gqa_paged_decode_h32_kv8_d64_ps1_e97b000c-8677-4877-86bf-2a82b9cb77e4.safetensors b/blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps1/gqa_paged_decode_h32_kv8_d64_ps1_e97b000c-8677-4877-86bf-2a82b9cb77e4.safetensors deleted file mode 100644 index 65fcc9aebe79dfbc96127ebd96bc49269cbd19d2..0000000000000000000000000000000000000000 --- a/blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps1/gqa_paged_decode_h32_kv8_d64_ps1_e97b000c-8677-4877-86bf-2a82b9cb77e4.safetensors +++ /dev/null @@ -1,3 +0,0 @@ -version https://git-lfs.github.com/spec/v1 -oid sha256:c14d6c71f58ae5becc77146223456c973828b39d9d87e1cdc9d0d0f0478adc24 -size 152 diff --git a/blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps64/gqa_paged_decode_h32_kv8_d64_ps64_3a4d1bf7-c9fe-40b9-a50d-9780a14fabf7.safetensors b/blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps64/gqa_paged_decode_h32_kv8_d64_ps64_3a4d1bf7-c9fe-40b9-a50d-9780a14fabf7.safetensors deleted file mode 100644 index ecd98db58d224c66f5bc03a1fe41e969239ba00c..0000000000000000000000000000000000000000 --- a/blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps64/gqa_paged_decode_h32_kv8_d64_ps64_3a4d1bf7-c9fe-40b9-a50d-9780a14fabf7.safetensors +++ /dev/null @@ -1,3 +0,0 @@ -version https://git-lfs.github.com/spec/v1 -oid sha256:20b7fb9f4e0e49e333bcc822a9935ffae73114d487186b3477d7bb40dcc6e209 -size 452 diff --git a/blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps64/gqa_paged_decode_h32_kv8_d64_ps64_4b40f9c1-2054-4b1c-ab80-0873f2801f7e.safetensors b/blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps64/gqa_paged_decode_h32_kv8_d64_ps64_4b40f9c1-2054-4b1c-ab80-0873f2801f7e.safetensors deleted file mode 100644 index 41828f32d2e0124b2abc280234c772b4032680cc..0000000000000000000000000000000000000000 --- a/blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps64/gqa_paged_decode_h32_kv8_d64_ps64_4b40f9c1-2054-4b1c-ab80-0873f2801f7e.safetensors +++ /dev/null @@ -1,3 +0,0 @@ -version https://git-lfs.github.com/spec/v1 -oid sha256:200a7c2e17a5b4fbfe8377a308633562787ea2f7adf5b03fe4d9b83cb5b8b19b -size 252 diff --git a/blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps64/gqa_paged_decode_h32_kv8_d64_ps64_687cdb3b-d2e0-4215-887b-fbf56b8569e7.safetensors b/blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps64/gqa_paged_decode_h32_kv8_d64_ps64_687cdb3b-d2e0-4215-887b-fbf56b8569e7.safetensors deleted file mode 100644 index 4f2d0edf7a4dd1cfaf21ae7a27073e776af4178f..0000000000000000000000000000000000000000 --- a/blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps64/gqa_paged_decode_h32_kv8_d64_ps64_687cdb3b-d2e0-4215-887b-fbf56b8569e7.safetensors +++ /dev/null @@ -1,3 +0,0 @@ -version https://git-lfs.github.com/spec/v1 -oid sha256:82f843fe2684e6603fe9dd1f05e08b0a5608a2a9ed357848af9fb654b41cbf48 -size 280 diff --git a/blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps64/gqa_paged_decode_h32_kv8_d64_ps64_68af5fd6-8da8-4070-ad1b-503ba3a56b94.safetensors b/blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps64/gqa_paged_decode_h32_kv8_d64_ps64_68af5fd6-8da8-4070-ad1b-503ba3a56b94.safetensors deleted file mode 100644 index 4d892db0c1955e9b82a908ec4f046c06b324b27f..0000000000000000000000000000000000000000 --- a/blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps64/gqa_paged_decode_h32_kv8_d64_ps64_68af5fd6-8da8-4070-ad1b-503ba3a56b94.safetensors +++ /dev/null @@ -1,3 +0,0 @@ -version https://git-lfs.github.com/spec/v1 -oid sha256:6e03d77904f34966f6e2eaf6898fda2c89fbb820a29a1d021106aee849c13c62 -size 228 diff --git a/blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps64/gqa_paged_decode_h32_kv8_d64_ps64_82761919-e520-4a9d-b18d-ef5b07f25245.safetensors b/blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps64/gqa_paged_decode_h32_kv8_d64_ps64_82761919-e520-4a9d-b18d-ef5b07f25245.safetensors deleted file mode 100644 index 4d892db0c1955e9b82a908ec4f046c06b324b27f..0000000000000000000000000000000000000000 --- a/blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps64/gqa_paged_decode_h32_kv8_d64_ps64_82761919-e520-4a9d-b18d-ef5b07f25245.safetensors +++ /dev/null @@ -1,3 +0,0 @@ -version https://git-lfs.github.com/spec/v1 -oid sha256:6e03d77904f34966f6e2eaf6898fda2c89fbb820a29a1d021106aee849c13c62 -size 228 diff --git a/blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps64/gqa_paged_decode_h32_kv8_d64_ps64_9d298dcb-2965-4e58-8c82-ea1b1fa49ccd.safetensors b/blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps64/gqa_paged_decode_h32_kv8_d64_ps64_9d298dcb-2965-4e58-8c82-ea1b1fa49ccd.safetensors deleted file mode 100644 index 4f2d0edf7a4dd1cfaf21ae7a27073e776af4178f..0000000000000000000000000000000000000000 --- a/blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps64/gqa_paged_decode_h32_kv8_d64_ps64_9d298dcb-2965-4e58-8c82-ea1b1fa49ccd.safetensors +++ /dev/null @@ -1,3 +0,0 @@ -version https://git-lfs.github.com/spec/v1 -oid sha256:82f843fe2684e6603fe9dd1f05e08b0a5608a2a9ed357848af9fb654b41cbf48 -size 280 diff --git a/blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps64/gqa_paged_decode_h32_kv8_d64_ps64_ea8adb18-8557-4dd2-9191-c1a045c7416d.safetensors b/blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps64/gqa_paged_decode_h32_kv8_d64_ps64_ea8adb18-8557-4dd2-9191-c1a045c7416d.safetensors deleted file mode 100644 index ecd98db58d224c66f5bc03a1fe41e969239ba00c..0000000000000000000000000000000000000000 --- a/blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps64/gqa_paged_decode_h32_kv8_d64_ps64_ea8adb18-8557-4dd2-9191-c1a045c7416d.safetensors +++ /dev/null @@ -1,3 +0,0 @@ -version https://git-lfs.github.com/spec/v1 -oid sha256:20b7fb9f4e0e49e333bcc822a9935ffae73114d487186b3477d7bb40dcc6e209 -size 452 diff --git a/blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps64/gqa_paged_decode_h32_kv8_d64_ps64_f5ae694d-0c9c-4534-8ab2-b4cdd38d6a35.safetensors b/blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps64/gqa_paged_decode_h32_kv8_d64_ps64_f5ae694d-0c9c-4534-8ab2-b4cdd38d6a35.safetensors deleted file mode 100644 index 41828f32d2e0124b2abc280234c772b4032680cc..0000000000000000000000000000000000000000 --- a/blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps64/gqa_paged_decode_h32_kv8_d64_ps64_f5ae694d-0c9c-4534-8ab2-b4cdd38d6a35.safetensors +++ /dev/null @@ -1,3 +0,0 @@ -version https://git-lfs.github.com/spec/v1 -oid sha256:200a7c2e17a5b4fbfe8377a308633562787ea2f7adf5b03fe4d9b83cb5b8b19b -size 252 diff --git a/blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps1/gqa_paged_prefill_causal_h32_kv8_d64_ps1_0b2f34f7-531b-436a-8c39-3d13f486cb78.safetensors b/blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps1/gqa_paged_prefill_causal_h32_kv8_d64_ps1_0b2f34f7-531b-436a-8c39-3d13f486cb78.safetensors deleted file mode 100644 index 27db13a5860d847a1dee7db1bdda794e205cbe8d..0000000000000000000000000000000000000000 --- a/blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps1/gqa_paged_prefill_causal_h32_kv8_d64_ps1_0b2f34f7-531b-436a-8c39-3d13f486cb78.safetensors +++ /dev/null @@ -1,3 +0,0 @@ -version https://git-lfs.github.com/spec/v1 -oid sha256:aa222372eadfe234266dee021521c4169be7147416b48dd714484a31263f7a85 -size 9448 diff --git a/blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps1/gqa_paged_prefill_causal_h32_kv8_d64_ps1_25732cc4-7e33-44bc-9a26-8f5f1ebd1da5.safetensors b/blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps1/gqa_paged_prefill_causal_h32_kv8_d64_ps1_25732cc4-7e33-44bc-9a26-8f5f1ebd1da5.safetensors deleted file mode 100644 index 5becf6d6eba7d604248878e61c0234e64e6f810c..0000000000000000000000000000000000000000 --- a/blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps1/gqa_paged_prefill_causal_h32_kv8_d64_ps1_25732cc4-7e33-44bc-9a26-8f5f1ebd1da5.safetensors +++ /dev/null @@ -1,3 +0,0 @@ -version https://git-lfs.github.com/spec/v1 -oid sha256:002ddb743fa6bdb8eac7b45e8c5082909abea8cbdc897331a74b32077e7fd9e3 -size 232 diff --git a/blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps1/gqa_paged_prefill_causal_h32_kv8_d64_ps1_297c3b37-1681-41f6-a052-1f449af8c580.safetensors b/blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps1/gqa_paged_prefill_causal_h32_kv8_d64_ps1_297c3b37-1681-41f6-a052-1f449af8c580.safetensors deleted file mode 100644 index 27db13a5860d847a1dee7db1bdda794e205cbe8d..0000000000000000000000000000000000000000 --- a/blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps1/gqa_paged_prefill_causal_h32_kv8_d64_ps1_297c3b37-1681-41f6-a052-1f449af8c580.safetensors +++ /dev/null @@ -1,3 +0,0 @@ -version https://git-lfs.github.com/spec/v1 -oid sha256:aa222372eadfe234266dee021521c4169be7147416b48dd714484a31263f7a85 -size 9448 diff --git a/blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps1/gqa_paged_prefill_causal_h32_kv8_d64_ps1_46907c3e-0b63-41a2-9ebf-63382ffffb60.safetensors b/blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps1/gqa_paged_prefill_causal_h32_kv8_d64_ps1_46907c3e-0b63-41a2-9ebf-63382ffffb60.safetensors deleted file mode 100644 index e880600dd6ac829c7fde68cf43600c0d80bebf40..0000000000000000000000000000000000000000 --- a/blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps1/gqa_paged_prefill_causal_h32_kv8_d64_ps1_46907c3e-0b63-41a2-9ebf-63382ffffb60.safetensors +++ /dev/null @@ -1,3 +0,0 @@ -version https://git-lfs.github.com/spec/v1 -oid sha256:2920da977aa49d7dcbe1d1854115692410bdc12e874111f4ae9bc9b2c0f85615 -size 2536 diff --git a/blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps1/gqa_paged_prefill_causal_h32_kv8_d64_ps1_8f24a1a5-f13a-4902-a0ee-f457a5074862.safetensors b/blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps1/gqa_paged_prefill_causal_h32_kv8_d64_ps1_8f24a1a5-f13a-4902-a0ee-f457a5074862.safetensors deleted file mode 100644 index e880600dd6ac829c7fde68cf43600c0d80bebf40..0000000000000000000000000000000000000000 --- a/blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps1/gqa_paged_prefill_causal_h32_kv8_d64_ps1_8f24a1a5-f13a-4902-a0ee-f457a5074862.safetensors +++ /dev/null @@ -1,3 +0,0 @@ -version https://git-lfs.github.com/spec/v1 -oid sha256:2920da977aa49d7dcbe1d1854115692410bdc12e874111f4ae9bc9b2c0f85615 -size 2536 diff --git a/blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps1/gqa_paged_prefill_causal_h32_kv8_d64_ps1_940a5fb1-7b82-40ad-b1ca-153bd7c170d9.safetensors b/blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps1/gqa_paged_prefill_causal_h32_kv8_d64_ps1_940a5fb1-7b82-40ad-b1ca-153bd7c170d9.safetensors deleted file mode 100644 index cf7ffca9ca273fa4da7344da5c33dc00cc6b3014..0000000000000000000000000000000000000000 --- a/blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps1/gqa_paged_prefill_causal_h32_kv8_d64_ps1_940a5fb1-7b82-40ad-b1ca-153bd7c170d9.safetensors +++ /dev/null @@ -1,3 +0,0 @@ -version https://git-lfs.github.com/spec/v1 -oid sha256:c246663796dc6217bbcf5bac4276c6d18d98883cc460a280eba5cf1f0694afe5 -size 33000 diff --git a/blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps1/gqa_paged_prefill_causal_h32_kv8_d64_ps1_efb8c6d0-1c04-4804-a29c-b5339ccdbe3a.safetensors b/blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps1/gqa_paged_prefill_causal_h32_kv8_d64_ps1_efb8c6d0-1c04-4804-a29c-b5339ccdbe3a.safetensors deleted file mode 100644 index 5becf6d6eba7d604248878e61c0234e64e6f810c..0000000000000000000000000000000000000000 --- a/blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps1/gqa_paged_prefill_causal_h32_kv8_d64_ps1_efb8c6d0-1c04-4804-a29c-b5339ccdbe3a.safetensors +++ /dev/null @@ -1,3 +0,0 @@ -version https://git-lfs.github.com/spec/v1 -oid sha256:002ddb743fa6bdb8eac7b45e8c5082909abea8cbdc897331a74b32077e7fd9e3 -size 232 diff --git a/blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps1/gqa_paged_prefill_causal_h32_kv8_d64_ps1_f8641e85-8041-426f-9996-e189e4aa7004.safetensors b/blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps1/gqa_paged_prefill_causal_h32_kv8_d64_ps1_f8641e85-8041-426f-9996-e189e4aa7004.safetensors deleted file mode 100644 index cf7ffca9ca273fa4da7344da5c33dc00cc6b3014..0000000000000000000000000000000000000000 --- a/blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps1/gqa_paged_prefill_causal_h32_kv8_d64_ps1_f8641e85-8041-426f-9996-e189e4aa7004.safetensors +++ /dev/null @@ -1,3 +0,0 @@ -version https://git-lfs.github.com/spec/v1 -oid sha256:c246663796dc6217bbcf5bac4276c6d18d98883cc460a280eba5cf1f0694afe5 -size 33000 diff --git a/blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps64/gqa_paged_prefill_causal_h32_kv8_d64_ps64_58b5a85a-8410-46bc-bf0f-fbc41ad6ad84.safetensors b/blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps64/gqa_paged_prefill_causal_h32_kv8_d64_ps64_58b5a85a-8410-46bc-bf0f-fbc41ad6ad84.safetensors deleted file mode 100644 index 0bf09f95d940ee4ff73f35419db654d0b6f543b6..0000000000000000000000000000000000000000 --- a/blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps64/gqa_paged_prefill_causal_h32_kv8_d64_ps64_58b5a85a-8410-46bc-bf0f-fbc41ad6ad84.safetensors +++ /dev/null @@ -1,3 +0,0 @@ -version https://git-lfs.github.com/spec/v1 -oid sha256:3abb75babc648fcc7f2c9f18444efdbe4cce696871a785eed5ab492f7a896105 -size 2616 diff --git a/blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps64/gqa_paged_prefill_causal_h32_kv8_d64_ps64_59f2a01c-bbcd-44b9-aac1-10878d8ec2b7.safetensors b/blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps64/gqa_paged_prefill_causal_h32_kv8_d64_ps64_59f2a01c-bbcd-44b9-aac1-10878d8ec2b7.safetensors deleted file mode 100644 index 04509b2a6d3a1790a3a7228895a472370542e383..0000000000000000000000000000000000000000 --- a/blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps64/gqa_paged_prefill_causal_h32_kv8_d64_ps64_59f2a01c-bbcd-44b9-aac1-10878d8ec2b7.safetensors +++ /dev/null @@ -1,3 +0,0 @@ -version https://git-lfs.github.com/spec/v1 -oid sha256:3f619f5d0f543185cc030897c1b12b7e92607e3421a673378eef40306ed59e8c -size 33088 diff --git a/blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps64/gqa_paged_prefill_causal_h32_kv8_d64_ps64_83cca546-a68d-47ee-ac1d-1883ea9a8cd3.safetensors b/blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps64/gqa_paged_prefill_causal_h32_kv8_d64_ps64_83cca546-a68d-47ee-ac1d-1883ea9a8cd3.safetensors deleted file mode 100644 index ae9879c31d826c3d9072080da1eac7efb21ec266..0000000000000000000000000000000000000000 --- a/blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps64/gqa_paged_prefill_causal_h32_kv8_d64_ps64_83cca546-a68d-47ee-ac1d-1883ea9a8cd3.safetensors +++ /dev/null @@ -1,3 +0,0 @@ -version https://git-lfs.github.com/spec/v1 -oid sha256:dc858cb7e49395441de71343d41564802a64eace09007ab8586820d39c31f3aa -size 312 diff --git a/blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps64/gqa_paged_prefill_causal_h32_kv8_d64_ps64_90a5a745-f1f4-42d6-8d63-a3ac06d27326.safetensors b/blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps64/gqa_paged_prefill_causal_h32_kv8_d64_ps64_90a5a745-f1f4-42d6-8d63-a3ac06d27326.safetensors deleted file mode 100644 index d275b9088a0a75c8c0d48bcdea9971a64aadd689..0000000000000000000000000000000000000000 --- a/blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps64/gqa_paged_prefill_causal_h32_kv8_d64_ps64_90a5a745-f1f4-42d6-8d63-a3ac06d27326.safetensors +++ /dev/null @@ -1,3 +0,0 @@ -version https://git-lfs.github.com/spec/v1 -oid sha256:cc03d37fd7436360fed2ecea7136b1ee84a438e208adcd1fa968efa10678ce71 -size 9528 diff --git a/blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps64/gqa_paged_prefill_causal_h32_kv8_d64_ps64_93cd7fca-8cef-4d07-aa0b-af998636b6bf.safetensors b/blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps64/gqa_paged_prefill_causal_h32_kv8_d64_ps64_93cd7fca-8cef-4d07-aa0b-af998636b6bf.safetensors deleted file mode 100644 index d275b9088a0a75c8c0d48bcdea9971a64aadd689..0000000000000000000000000000000000000000 --- a/blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps64/gqa_paged_prefill_causal_h32_kv8_d64_ps64_93cd7fca-8cef-4d07-aa0b-af998636b6bf.safetensors +++ /dev/null @@ -1,3 +0,0 @@ -version https://git-lfs.github.com/spec/v1 -oid sha256:cc03d37fd7436360fed2ecea7136b1ee84a438e208adcd1fa968efa10678ce71 -size 9528 diff --git a/blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps64/gqa_paged_prefill_causal_h32_kv8_d64_ps64_a10cccb9-d773-4921-b85e-b9bc75a9ed39.safetensors b/blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps64/gqa_paged_prefill_causal_h32_kv8_d64_ps64_a10cccb9-d773-4921-b85e-b9bc75a9ed39.safetensors deleted file mode 100644 index ae9879c31d826c3d9072080da1eac7efb21ec266..0000000000000000000000000000000000000000 --- a/blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps64/gqa_paged_prefill_causal_h32_kv8_d64_ps64_a10cccb9-d773-4921-b85e-b9bc75a9ed39.safetensors +++ /dev/null @@ -1,3 +0,0 @@ -version https://git-lfs.github.com/spec/v1 -oid sha256:dc858cb7e49395441de71343d41564802a64eace09007ab8586820d39c31f3aa -size 312 diff --git a/blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps64/gqa_paged_prefill_causal_h32_kv8_d64_ps64_b5f2086c-8df8-4ee3-9883-b2ba121c7b69.safetensors b/blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps64/gqa_paged_prefill_causal_h32_kv8_d64_ps64_b5f2086c-8df8-4ee3-9883-b2ba121c7b69.safetensors deleted file mode 100644 index 0bf09f95d940ee4ff73f35419db654d0b6f543b6..0000000000000000000000000000000000000000 --- a/blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps64/gqa_paged_prefill_causal_h32_kv8_d64_ps64_b5f2086c-8df8-4ee3-9883-b2ba121c7b69.safetensors +++ /dev/null @@ -1,3 +0,0 @@ -version https://git-lfs.github.com/spec/v1 -oid sha256:3abb75babc648fcc7f2c9f18444efdbe4cce696871a785eed5ab492f7a896105 -size 2616 diff --git a/blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps64/gqa_paged_prefill_causal_h32_kv8_d64_ps64_ea6b783f-fdb1-4be4-97fe-180de19df270.safetensors b/blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps64/gqa_paged_prefill_causal_h32_kv8_d64_ps64_ea6b783f-fdb1-4be4-97fe-180de19df270.safetensors deleted file mode 100644 index 04509b2a6d3a1790a3a7228895a472370542e383..0000000000000000000000000000000000000000 --- a/blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps64/gqa_paged_prefill_causal_h32_kv8_d64_ps64_ea6b783f-fdb1-4be4-97fe-180de19df270.safetensors +++ /dev/null @@ -1,3 +0,0 @@ -version https://git-lfs.github.com/spec/v1 -oid sha256:3f619f5d0f543185cc030897c1b12b7e92607e3421a673378eef40306ed59e8c -size 33088 diff --git a/blob/workloads/gqa_ragged/gqa_ragged_prefill_causal_h32_kv8_d64/gqa_ragged_prefill_causal_h32_kv8_d64_17e47b78-d996-4fc1-af3e-56bd9c789a82.safetensors b/blob/workloads/gqa_ragged/gqa_ragged_prefill_causal_h32_kv8_d64/gqa_ragged_prefill_causal_h32_kv8_d64_17e47b78-d996-4fc1-af3e-56bd9c789a82.safetensors deleted file mode 100644 index e941a0709b16374dfe6c644e23323e2396e797ef..0000000000000000000000000000000000000000 --- a/blob/workloads/gqa_ragged/gqa_ragged_prefill_causal_h32_kv8_d64/gqa_ragged_prefill_causal_h32_kv8_d64_17e47b78-d996-4fc1-af3e-56bd9c789a82.safetensors +++ /dev/null @@ -1,3 +0,0 @@ -version https://git-lfs.github.com/spec/v1 -oid sha256:8def9f0d61e82d65fc22324b0d7158c00d816c8f021a787f70977dc0d4436633 -size 152 diff --git a/blob/workloads/gqa_ragged/gqa_ragged_prefill_causal_h32_kv8_d64/gqa_ragged_prefill_causal_h32_kv8_d64_1bf692f3-4891-4382-b3a0-452aff438463.safetensors b/blob/workloads/gqa_ragged/gqa_ragged_prefill_causal_h32_kv8_d64/gqa_ragged_prefill_causal_h32_kv8_d64_1bf692f3-4891-4382-b3a0-452aff438463.safetensors deleted file mode 100644 index e941a0709b16374dfe6c644e23323e2396e797ef..0000000000000000000000000000000000000000 --- a/blob/workloads/gqa_ragged/gqa_ragged_prefill_causal_h32_kv8_d64/gqa_ragged_prefill_causal_h32_kv8_d64_1bf692f3-4891-4382-b3a0-452aff438463.safetensors +++ /dev/null @@ -1,3 +0,0 @@ -version https://git-lfs.github.com/spec/v1 -oid sha256:8def9f0d61e82d65fc22324b0d7158c00d816c8f021a787f70977dc0d4436633 -size 152 diff --git a/blob/workloads/gqa_ragged/gqa_ragged_prefill_causal_h32_kv8_d64/gqa_ragged_prefill_causal_h32_kv8_d64_319d7c83-25e5-47c3-8de0-bf65e9e80a35.safetensors b/blob/workloads/gqa_ragged/gqa_ragged_prefill_causal_h32_kv8_d64/gqa_ragged_prefill_causal_h32_kv8_d64_319d7c83-25e5-47c3-8de0-bf65e9e80a35.safetensors deleted file mode 100644 index 44f57e18a8d52979470b543c0a4791a6e6423d86..0000000000000000000000000000000000000000 --- a/blob/workloads/gqa_ragged/gqa_ragged_prefill_causal_h32_kv8_d64/gqa_ragged_prefill_causal_h32_kv8_d64_319d7c83-25e5-47c3-8de0-bf65e9e80a35.safetensors +++ /dev/null @@ -1,3 +0,0 @@ -version https://git-lfs.github.com/spec/v1 -oid sha256:c167058df494f728412bdb28109f022887672dabcd886f893751fa9e7f29dde9 -size 152 diff --git a/blob/workloads/gqa_ragged/gqa_ragged_prefill_causal_h32_kv8_d64/gqa_ragged_prefill_causal_h32_kv8_d64_7adaac59-352e-4574-a959-d3470a966090.safetensors b/blob/workloads/gqa_ragged/gqa_ragged_prefill_causal_h32_kv8_d64/gqa_ragged_prefill_causal_h32_kv8_d64_7adaac59-352e-4574-a959-d3470a966090.safetensors deleted file mode 100644 index 48c1252e8b9fd0bea795fac519b42e259b814c3f..0000000000000000000000000000000000000000 --- a/blob/workloads/gqa_ragged/gqa_ragged_prefill_causal_h32_kv8_d64/gqa_ragged_prefill_causal_h32_kv8_d64_7adaac59-352e-4574-a959-d3470a966090.safetensors +++ /dev/null @@ -1,3 +0,0 @@ -version https://git-lfs.github.com/spec/v1 -oid sha256:a238940a59605fc0db1fbe169143251014daa6f31b20d5163c1ba12f18ad12ae -size 584 diff --git a/blob/workloads/gqa_ragged/gqa_ragged_prefill_causal_h32_kv8_d64/gqa_ragged_prefill_causal_h32_kv8_d64_d22f94aa-fc21-4bdd-ac49-f36bae0caa5c.safetensors b/blob/workloads/gqa_ragged/gqa_ragged_prefill_causal_h32_kv8_d64/gqa_ragged_prefill_causal_h32_kv8_d64_d22f94aa-fc21-4bdd-ac49-f36bae0caa5c.safetensors deleted file mode 100644 index 8b45c98c701eb9cfdd544f8ae0010a712384993b..0000000000000000000000000000000000000000 --- a/blob/workloads/gqa_ragged/gqa_ragged_prefill_causal_h32_kv8_d64/gqa_ragged_prefill_causal_h32_kv8_d64_d22f94aa-fc21-4bdd-ac49-f36bae0caa5c.safetensors +++ /dev/null @@ -1,3 +0,0 @@ -version https://git-lfs.github.com/spec/v1 -oid sha256:57d9aed8cc57093d9d0f6da9e63538322342da7c2317d94ea9ada7dfab890a59 -size 248 diff --git a/blob/workloads/gqa_ragged/gqa_ragged_prefill_causal_h32_kv8_d64/gqa_ragged_prefill_causal_h32_kv8_d64_db7cafbd-5dc4-4a6b-a073-38b22ea8c0b4.safetensors b/blob/workloads/gqa_ragged/gqa_ragged_prefill_causal_h32_kv8_d64/gqa_ragged_prefill_causal_h32_kv8_d64_db7cafbd-5dc4-4a6b-a073-38b22ea8c0b4.safetensors deleted file mode 100644 index ea8348cbc0042e7d4a01c548ede5a01acc9e64bf..0000000000000000000000000000000000000000 --- a/blob/workloads/gqa_ragged/gqa_ragged_prefill_causal_h32_kv8_d64/gqa_ragged_prefill_causal_h32_kv8_d64_db7cafbd-5dc4-4a6b-a073-38b22ea8c0b4.safetensors +++ /dev/null @@ -1,3 +0,0 @@ -version https://git-lfs.github.com/spec/v1 -oid sha256:0866a4834efdb10cf2e977710ff7917495ae8cf8133d4774be2e3cbc5242986f -size 400 diff --git a/blob/workloads/gqa_ragged/gqa_ragged_prefill_causal_h32_kv8_d64/gqa_ragged_prefill_causal_h32_kv8_d64_eaf31898-d557-474a-8967-08912443e85e.safetensors b/blob/workloads/gqa_ragged/gqa_ragged_prefill_causal_h32_kv8_d64/gqa_ragged_prefill_causal_h32_kv8_d64_eaf31898-d557-474a-8967-08912443e85e.safetensors deleted file mode 100644 index 44f57e18a8d52979470b543c0a4791a6e6423d86..0000000000000000000000000000000000000000 --- a/blob/workloads/gqa_ragged/gqa_ragged_prefill_causal_h32_kv8_d64/gqa_ragged_prefill_causal_h32_kv8_d64_eaf31898-d557-474a-8967-08912443e85e.safetensors +++ /dev/null @@ -1,3 +0,0 @@ -version https://git-lfs.github.com/spec/v1 -oid sha256:c167058df494f728412bdb28109f022887672dabcd886f893751fa9e7f29dde9 -size 152 diff --git a/blob/workloads/gqa_ragged/gqa_ragged_prefill_causal_h32_kv8_d64/gqa_ragged_prefill_causal_h32_kv8_d64_f75c1eec-0a27-4e06-bfb6-d6238049271d.safetensors b/blob/workloads/gqa_ragged/gqa_ragged_prefill_causal_h32_kv8_d64/gqa_ragged_prefill_causal_h32_kv8_d64_f75c1eec-0a27-4e06-bfb6-d6238049271d.safetensors deleted file mode 100644 index d08f4ec4cbe15c3b57a688cf2cce2d5f1e28659f..0000000000000000000000000000000000000000 --- a/blob/workloads/gqa_ragged/gqa_ragged_prefill_causal_h32_kv8_d64/gqa_ragged_prefill_causal_h32_kv8_d64_f75c1eec-0a27-4e06-bfb6-d6238049271d.safetensors +++ /dev/null @@ -1,3 +0,0 @@ -version https://git-lfs.github.com/spec/v1 -oid sha256:f608c48d19f97477e82ab82b2ad5873c11932b3024a6ce71e0b314f3457c4b43 -size 216 diff --git a/blob/workloads/sampling/top_k_top_p_sampling_from_probs_v248320/top_k_top_p_sampling_from_probs_v248320_0d6d294f-ff2d-48a3-a241-1f0740709b61.safetensors b/blob/workloads/sampling/top_k_top_p_sampling_from_probs_v248320/top_k_top_p_sampling_from_probs_v248320_0d6d294f-ff2d-48a3-a241-1f0740709b61.safetensors deleted file mode 100644 index f7753b22a4b1f8a806f2c652d01cf71a6608c8aa..0000000000000000000000000000000000000000 --- a/blob/workloads/sampling/top_k_top_p_sampling_from_probs_v248320/top_k_top_p_sampling_from_probs_v248320_0d6d294f-ff2d-48a3-a241-1f0740709b61.safetensors +++ /dev/null @@ -1,3 +0,0 @@ -version https://git-lfs.github.com/spec/v1 -oid sha256:ada8dabcfb5408cac8710c99d87419fdc11697c8d92aa15eae7a21ddb4eaf77f -size 100 diff --git a/blob/workloads/sampling/top_k_top_p_sampling_from_probs_v248320/top_k_top_p_sampling_from_probs_v248320_1050f50f-88d7-4f23-ba26-dd654440e6ed.safetensors b/blob/workloads/sampling/top_k_top_p_sampling_from_probs_v248320/top_k_top_p_sampling_from_probs_v248320_1050f50f-88d7-4f23-ba26-dd654440e6ed.safetensors deleted file mode 100644 index 486e9b46cf5cf49ef38b637f843e91813be357e7..0000000000000000000000000000000000000000 --- a/blob/workloads/sampling/top_k_top_p_sampling_from_probs_v248320/top_k_top_p_sampling_from_probs_v248320_1050f50f-88d7-4f23-ba26-dd654440e6ed.safetensors +++ /dev/null @@ -1,3 +0,0 @@ -version https://git-lfs.github.com/spec/v1 -oid sha256:9a3396781034d4c875370af4e64b03129778711d504876df64f4181d284194a9 -size 164 diff --git a/blob/workloads/sampling/top_k_top_p_sampling_from_probs_v248320/top_k_top_p_sampling_from_probs_v248320_11ba194f-96e1-485e-bdeb-55aac8f944db.safetensors b/blob/workloads/sampling/top_k_top_p_sampling_from_probs_v248320/top_k_top_p_sampling_from_probs_v248320_11ba194f-96e1-485e-bdeb-55aac8f944db.safetensors deleted file mode 100644 index 72966931c3d17476f1555f46b8214b1249b209f1..0000000000000000000000000000000000000000 --- a/blob/workloads/sampling/top_k_top_p_sampling_from_probs_v248320/top_k_top_p_sampling_from_probs_v248320_11ba194f-96e1-485e-bdeb-55aac8f944db.safetensors +++ /dev/null @@ -1,3 +0,0 @@ -version https://git-lfs.github.com/spec/v1 -oid sha256:8e08e283bcd4f46213d57cc14084f8fe4be6b863693f49f47590f8f8b52516b1 -size 172 diff --git a/blob/workloads/sampling/top_k_top_p_sampling_from_probs_v248320/top_k_top_p_sampling_from_probs_v248320_1aa3d7a2-e19b-4898-b468-1e89f9438b1f.safetensors b/blob/workloads/sampling/top_k_top_p_sampling_from_probs_v248320/top_k_top_p_sampling_from_probs_v248320_1aa3d7a2-e19b-4898-b468-1e89f9438b1f.safetensors deleted file mode 100644 index e96b6939fb49ce1253b29c185f5469e0987c53b6..0000000000000000000000000000000000000000 --- a/blob/workloads/sampling/top_k_top_p_sampling_from_probs_v248320/top_k_top_p_sampling_from_probs_v248320_1aa3d7a2-e19b-4898-b468-1e89f9438b1f.safetensors +++ /dev/null @@ -1,3 +0,0 @@ -version https://git-lfs.github.com/spec/v1 -oid sha256:dd1780894b3fb22797b9bf7576db6360c3a275445d1ebd36b465e3181cfbe2aa -size 80 diff --git a/blob/workloads/sampling/top_k_top_p_sampling_from_probs_v248320/top_k_top_p_sampling_from_probs_v248320_21b36673-7c3f-4d58-9e9d-688e7498557a.safetensors b/blob/workloads/sampling/top_k_top_p_sampling_from_probs_v248320/top_k_top_p_sampling_from_probs_v248320_21b36673-7c3f-4d58-9e9d-688e7498557a.safetensors deleted file mode 100644 index f1e7f690d841cb138e5acc2052e37e185ff56c91..0000000000000000000000000000000000000000 --- a/blob/workloads/sampling/top_k_top_p_sampling_from_probs_v248320/top_k_top_p_sampling_from_probs_v248320_21b36673-7c3f-4d58-9e9d-688e7498557a.safetensors +++ /dev/null @@ -1,3 +0,0 @@ -version https://git-lfs.github.com/spec/v1 -oid sha256:46f1052c3aa1276ca75f6e7c5b86361fb5d49e3b6f1d5d0859e15d4356e684b8 -size 196 diff --git a/blob/workloads/sampling/top_k_top_p_sampling_from_probs_v248320/top_k_top_p_sampling_from_probs_v248320_2413e30c-d007-4adc-b0ec-dfed05df6504.safetensors b/blob/workloads/sampling/top_k_top_p_sampling_from_probs_v248320/top_k_top_p_sampling_from_probs_v248320_2413e30c-d007-4adc-b0ec-dfed05df6504.safetensors deleted file mode 100644 index 4738093860b3932bdaaa464acba72c94935733a3..0000000000000000000000000000000000000000 --- a/blob/workloads/sampling/top_k_top_p_sampling_from_probs_v248320/top_k_top_p_sampling_from_probs_v248320_2413e30c-d007-4adc-b0ec-dfed05df6504.safetensors +++ /dev/null @@ -1,3 +0,0 @@ -version https://git-lfs.github.com/spec/v1 -oid sha256:fbfb7bccfa27808a275f92f939fee0c18ff6c1b14d9efd53b98f0e1bef971339 -size 472 diff --git a/blob/workloads/sampling/top_k_top_p_sampling_from_probs_v248320/top_k_top_p_sampling_from_probs_v248320_2ee765e1-0c33-4b92-a6bd-84c4ea1df0f2.safetensors b/blob/workloads/sampling/top_k_top_p_sampling_from_probs_v248320/top_k_top_p_sampling_from_probs_v248320_2ee765e1-0c33-4b92-a6bd-84c4ea1df0f2.safetensors deleted file mode 100644 index 303dde4478ca3cd12aceacbefce28cbc1b87b488..0000000000000000000000000000000000000000 --- a/blob/workloads/sampling/top_k_top_p_sampling_from_probs_v248320/top_k_top_p_sampling_from_probs_v248320_2ee765e1-0c33-4b92-a6bd-84c4ea1df0f2.safetensors +++ /dev/null @@ -1,3 +0,0 @@ -version https://git-lfs.github.com/spec/v1 -oid sha256:8acee33d50ac06e6c55e7be3988c3a160025cdab030b2fa2958111ba9cb5ab25 -size 464 diff --git a/blob/workloads/sampling/top_k_top_p_sampling_from_probs_v248320/top_k_top_p_sampling_from_probs_v248320_37e90fb8-4450-451b-8bff-5915d50e445c.safetensors b/blob/workloads/sampling/top_k_top_p_sampling_from_probs_v248320/top_k_top_p_sampling_from_probs_v248320_37e90fb8-4450-451b-8bff-5915d50e445c.safetensors deleted file mode 100644 index 77d939e9decb9a7dfa6e8483df42b77b764d75f9..0000000000000000000000000000000000000000 --- a/blob/workloads/sampling/top_k_top_p_sampling_from_probs_v248320/top_k_top_p_sampling_from_probs_v248320_37e90fb8-4450-451b-8bff-5915d50e445c.safetensors +++ /dev/null @@ -1,3 +0,0 @@ -version https://git-lfs.github.com/spec/v1 -oid sha256:138d756d2b50fa41097a9938755f2e911fd3dc1318e6d1e944b19f5be6254093 -size 188 diff --git a/blob/workloads/sampling/top_k_top_p_sampling_from_probs_v248320/top_k_top_p_sampling_from_probs_v248320_415daee6-51bb-4248-bc39-ffc9d51f4a59.safetensors b/blob/workloads/sampling/top_k_top_p_sampling_from_probs_v248320/top_k_top_p_sampling_from_probs_v248320_415daee6-51bb-4248-bc39-ffc9d51f4a59.safetensors deleted file mode 100644 index 4252b757cb8cbd878a2b1ae126eca90615b41ae6..0000000000000000000000000000000000000000 --- a/blob/workloads/sampling/top_k_top_p_sampling_from_probs_v248320/top_k_top_p_sampling_from_probs_v248320_415daee6-51bb-4248-bc39-ffc9d51f4a59.safetensors +++ /dev/null @@ -1,3 +0,0 @@ -version https://git-lfs.github.com/spec/v1 -oid sha256:b06e4b99c83f028213cbe19441807e028b2a9443e48033de7b49e8998f37398f -size 96 diff --git a/blob/workloads/sampling/top_k_top_p_sampling_from_probs_v248320/top_k_top_p_sampling_from_probs_v248320_59cb7944-0b44-429f-97ba-31d124e6c720.safetensors b/blob/workloads/sampling/top_k_top_p_sampling_from_probs_v248320/top_k_top_p_sampling_from_probs_v248320_59cb7944-0b44-429f-97ba-31d124e6c720.safetensors deleted file mode 100644 index 3793736659df5249a93cd1fa61c2d50187c93000..0000000000000000000000000000000000000000 --- a/blob/workloads/sampling/top_k_top_p_sampling_from_probs_v248320/top_k_top_p_sampling_from_probs_v248320_59cb7944-0b44-429f-97ba-31d124e6c720.safetensors +++ /dev/null @@ -1,3 +0,0 @@ -version https://git-lfs.github.com/spec/v1 -oid sha256:391e86baff088558bfab120d1903150de91bb1f5011c9dee0fcc7cf135d6df01 -size 468 diff --git a/blob/workloads/sampling/top_k_top_p_sampling_from_probs_v248320/top_k_top_p_sampling_from_probs_v248320_5b83f0e1-4d9e-436a-b6e1-874ecdbcaf1c.safetensors b/blob/workloads/sampling/top_k_top_p_sampling_from_probs_v248320/top_k_top_p_sampling_from_probs_v248320_5b83f0e1-4d9e-436a-b6e1-874ecdbcaf1c.safetensors deleted file mode 100644 index fb9e9106993b16dc0fc5891b3b8412c1df10bfa5..0000000000000000000000000000000000000000 --- a/blob/workloads/sampling/top_k_top_p_sampling_from_probs_v248320/top_k_top_p_sampling_from_probs_v248320_5b83f0e1-4d9e-436a-b6e1-874ecdbcaf1c.safetensors +++ /dev/null @@ -1,3 +0,0 @@ -version https://git-lfs.github.com/spec/v1 -oid sha256:e0c8e5869fe8acbfce261cf65cf4f0191edae35535d910fb3f5fbf56071d7fed -size 236 diff --git a/blob/workloads/sampling/top_k_top_p_sampling_from_probs_v248320/top_k_top_p_sampling_from_probs_v248320_6218b0d1-0ee3-44db-b8e7-604bc90c8335.safetensors b/blob/workloads/sampling/top_k_top_p_sampling_from_probs_v248320/top_k_top_p_sampling_from_probs_v248320_6218b0d1-0ee3-44db-b8e7-604bc90c8335.safetensors deleted file mode 100644 index f673a3ec92ecf7813b561a28347cab5a99b4b49a..0000000000000000000000000000000000000000 --- a/blob/workloads/sampling/top_k_top_p_sampling_from_probs_v248320/top_k_top_p_sampling_from_probs_v248320_6218b0d1-0ee3-44db-b8e7-604bc90c8335.safetensors +++ /dev/null @@ -1,3 +0,0 @@ -version https://git-lfs.github.com/spec/v1 -oid sha256:a50efa38967d3df0f776a7872d3c704bef6de50ab789ad04496e556bd2fc90d7 -size 184 diff --git a/blob/workloads/sampling/top_k_top_p_sampling_from_probs_v248320/top_k_top_p_sampling_from_probs_v248320_64a90d77-55d4-4b07-8493-baa18e016cd6.safetensors b/blob/workloads/sampling/top_k_top_p_sampling_from_probs_v248320/top_k_top_p_sampling_from_probs_v248320_64a90d77-55d4-4b07-8493-baa18e016cd6.safetensors deleted file mode 100644 index d2868f77319c3af1a217086c0aa86144680a8c3b..0000000000000000000000000000000000000000 --- a/blob/workloads/sampling/top_k_top_p_sampling_from_probs_v248320/top_k_top_p_sampling_from_probs_v248320_64a90d77-55d4-4b07-8493-baa18e016cd6.safetensors +++ /dev/null @@ -1,3 +0,0 @@ -version https://git-lfs.github.com/spec/v1 -oid sha256:90f129e519d3495085fb428e526905a4ad8ef0c5dde4c529ba30ccc303b3f8de -size 84 diff --git a/blob/workloads/sampling/top_k_top_p_sampling_from_probs_v248320/top_k_top_p_sampling_from_probs_v248320_975b78d0-9973-4ac9-9152-e100f8b6c2ba.safetensors b/blob/workloads/sampling/top_k_top_p_sampling_from_probs_v248320/top_k_top_p_sampling_from_probs_v248320_975b78d0-9973-4ac9-9152-e100f8b6c2ba.safetensors deleted file mode 100644 index cee4df922b652e0145b294326a96373d9d535edb..0000000000000000000000000000000000000000 --- a/blob/workloads/sampling/top_k_top_p_sampling_from_probs_v248320/top_k_top_p_sampling_from_probs_v248320_975b78d0-9973-4ac9-9152-e100f8b6c2ba.safetensors +++ /dev/null @@ -1,3 +0,0 @@ -version https://git-lfs.github.com/spec/v1 -oid sha256:0e9c3b9ad7f9c6769740da68cc662c610c41c0ff2098a131a176812ccac9ee38 -size 168 diff --git a/blob/workloads/sampling/top_k_top_p_sampling_from_probs_v248320/top_k_top_p_sampling_from_probs_v248320_98c33e79-e32a-4d9f-99e0-45c800dd883a.safetensors b/blob/workloads/sampling/top_k_top_p_sampling_from_probs_v248320/top_k_top_p_sampling_from_probs_v248320_98c33e79-e32a-4d9f-99e0-45c800dd883a.safetensors deleted file mode 100644 index a5c231f718aac0eebe5a9362b9da240104bac58b..0000000000000000000000000000000000000000 --- a/blob/workloads/sampling/top_k_top_p_sampling_from_probs_v248320/top_k_top_p_sampling_from_probs_v248320_98c33e79-e32a-4d9f-99e0-45c800dd883a.safetensors +++ /dev/null @@ -1,3 +0,0 @@ -version https://git-lfs.github.com/spec/v1 -oid sha256:42090f9967425d85f88d0396d92e5be2cc6eaa7dcbaa21f78f4bb2e3943cdb53 -size 460 diff --git a/blob/workloads/sampling/top_k_top_p_sampling_from_probs_v248320/top_k_top_p_sampling_from_probs_v248320_9def7b85-9198-4acd-a3d7-e79c608321e4.safetensors b/blob/workloads/sampling/top_k_top_p_sampling_from_probs_v248320/top_k_top_p_sampling_from_probs_v248320_9def7b85-9198-4acd-a3d7-e79c608321e4.safetensors deleted file mode 100644 index 4070c6ee25a881b8102c6cc86cd812f626019914..0000000000000000000000000000000000000000 --- a/blob/workloads/sampling/top_k_top_p_sampling_from_probs_v248320/top_k_top_p_sampling_from_probs_v248320_9def7b85-9198-4acd-a3d7-e79c608321e4.safetensors +++ /dev/null @@ -1,3 +0,0 @@ -version https://git-lfs.github.com/spec/v1 -oid sha256:08bc5de624a82c678edcaf95b064b67f5553537fe52df39261aefe5a3af09335 -size 452 diff --git a/blob/workloads/sampling/top_k_top_p_sampling_from_probs_v248320/top_k_top_p_sampling_from_probs_v248320_a2b286ac-256d-4c6c-a5b0-6076203fbefd.safetensors b/blob/workloads/sampling/top_k_top_p_sampling_from_probs_v248320/top_k_top_p_sampling_from_probs_v248320_a2b286ac-256d-4c6c-a5b0-6076203fbefd.safetensors deleted file mode 100644 index e5973777c2474d0c56d96706583a18cac53f5bdc..0000000000000000000000000000000000000000 --- a/blob/workloads/sampling/top_k_top_p_sampling_from_probs_v248320/top_k_top_p_sampling_from_probs_v248320_a2b286ac-256d-4c6c-a5b0-6076203fbefd.safetensors +++ /dev/null @@ -1,3 +0,0 @@ -version https://git-lfs.github.com/spec/v1 -oid sha256:be7b02d8ea3397384a25310757afbba640b0a4dfa3d503247470a8903599789f -size 456 diff --git a/blob/workloads/sampling/top_k_top_p_sampling_from_probs_v248320/top_k_top_p_sampling_from_probs_v248320_b11edc96-72c7-44fd-aec9-d53cf809382b.safetensors b/blob/workloads/sampling/top_k_top_p_sampling_from_probs_v248320/top_k_top_p_sampling_from_probs_v248320_b11edc96-72c7-44fd-aec9-d53cf809382b.safetensors deleted file mode 100644 index 869152ae7abc034a8a24d95f2f801c539936382f..0000000000000000000000000000000000000000 --- a/blob/workloads/sampling/top_k_top_p_sampling_from_probs_v248320/top_k_top_p_sampling_from_probs_v248320_b11edc96-72c7-44fd-aec9-d53cf809382b.safetensors +++ /dev/null @@ -1,3 +0,0 @@ -version https://git-lfs.github.com/spec/v1 -oid sha256:31537ba5119d440f16941eca0e35d68dbfa02b729ae3bfac50da41e0a3e7597c -size 104 diff --git a/blob/workloads/sampling/top_k_top_p_sampling_from_probs_v248320/top_k_top_p_sampling_from_probs_v248320_ce8e6cbc-010e-49bc-9b67-8523231e2e65.safetensors b/blob/workloads/sampling/top_k_top_p_sampling_from_probs_v248320/top_k_top_p_sampling_from_probs_v248320_ce8e6cbc-010e-49bc-9b67-8523231e2e65.safetensors deleted file mode 100644 index dd0a79e89389891960929cedef87b09f2cf9c3cd..0000000000000000000000000000000000000000 --- a/blob/workloads/sampling/top_k_top_p_sampling_from_probs_v248320/top_k_top_p_sampling_from_probs_v248320_ce8e6cbc-010e-49bc-9b67-8523231e2e65.safetensors +++ /dev/null @@ -1,3 +0,0 @@ -version https://git-lfs.github.com/spec/v1 -oid sha256:150b09e458e840e51398f33578b59b4b2e9f4c465a5d3e49f8a8f18954f7acb7 -size 212 diff --git a/blob/workloads/sampling/top_k_top_p_sampling_from_probs_v248320/top_k_top_p_sampling_from_probs_v248320_d6db6ce7-70e4-4775-859a-dfd6be2b6e53.safetensors b/blob/workloads/sampling/top_k_top_p_sampling_from_probs_v248320/top_k_top_p_sampling_from_probs_v248320_d6db6ce7-70e4-4775-859a-dfd6be2b6e53.safetensors deleted file mode 100644 index 41e6f9e72178ab18787a7b7a3f2a91797502f603..0000000000000000000000000000000000000000 --- a/blob/workloads/sampling/top_k_top_p_sampling_from_probs_v248320/top_k_top_p_sampling_from_probs_v248320_d6db6ce7-70e4-4775-859a-dfd6be2b6e53.safetensors +++ /dev/null @@ -1,3 +0,0 @@ -version https://git-lfs.github.com/spec/v1 -oid sha256:698936fe3f50b7b8ddea5557ade851bede834d336b6e48127b4204e3f7dc4091 -size 148 diff --git a/definitions/gdn/gdn_decode_qk8_v16_d128_k_last.json b/definitions/gdn/gdn_decode_qk8_v16_d128_k_last.json index 7151030d4dcbdf54099b8118a99802dbd9eca5a8..5c6fe74a1ff838664535db3ee6d172a24f39fca0 100644 --- a/definitions/gdn/gdn_decode_qk8_v16_d128_k_last.json +++ b/definitions/gdn/gdn_decode_qk8_v16_d128_k_last.json @@ -6,7 +6,6 @@ "stage:decode", "status:verified", "model:qwen3-next", - "model:qwen3.5-35b-a3b", "layout:k-last", "fi_api:flashinfer.gdn.gated_delta_rule_decode", "tp:2" diff --git a/definitions/gemm/gemm_n16384_k2048.json b/definitions/gemm/gemm_n16384_k2048.json deleted file mode 100644 index 5c8a19685ea5996fdfc7b68153a92bc82ccf6f8c..0000000000000000000000000000000000000000 --- a/definitions/gemm/gemm_n16384_k2048.json +++ /dev/null @@ -1,48 +0,0 @@ -{ - "name": "gemm_n16384_k2048", - "description": "General matrix multiply (GEMM) C = A @ B.T. Captured from Llama-3.2-1B mlp.gate_up_proj (fused gate+up: 2 * 8192 = 16384).", - "op_type": "gemm", - "tags": [ - "status:unverified", - "model:llama-3.2-1b" - ], - "axes": { - "M": { - "type": "var" - }, - "N": { - "type": "const", - "value": 16384 - }, - "K": { - "type": "const", - "value": 2048 - } - }, - "inputs": { - "A": { - "shape": [ - "M", - "K" - ], - "dtype": "float16" - }, - "B": { - "shape": [ - "N", - "K" - ], - "dtype": "float16" - } - }, - "outputs": { - "C": { - "shape": [ - "M", - "N" - ], - "dtype": "float16" - } - }, - "reference": "import torch\n\ndef run(A, B):\n C = torch.matmul(A, B.T)\n return C" -} diff --git a/definitions/gemm/gemm_n2048_k2048.json b/definitions/gemm/gemm_n2048_k2048.json deleted file mode 100644 index 0ab7af20171629d4593c4f8f974d3ef5b403b57d..0000000000000000000000000000000000000000 --- a/definitions/gemm/gemm_n2048_k2048.json +++ /dev/null @@ -1,48 +0,0 @@ -{ - "name": "gemm_n2048_k2048", - "description": "General matrix multiply (GEMM) C = A @ B.T. Captured from Llama-3.2-1B attn.o_proj.", - "op_type": "gemm", - "tags": [ - "status:unverified", - "model:llama-3.2-1b" - ], - "axes": { - "M": { - "type": "var" - }, - "N": { - "type": "const", - "value": 2048 - }, - "K": { - "type": "const", - "value": 2048 - } - }, - "inputs": { - "A": { - "shape": [ - "M", - "K" - ], - "dtype": "float16" - }, - "B": { - "shape": [ - "N", - "K" - ], - "dtype": "float16" - } - }, - "outputs": { - "C": { - "shape": [ - "M", - "N" - ], - "dtype": "float16" - } - }, - "reference": "import torch\n\ndef run(A, B):\n C = torch.matmul(A, B.T)\n return C" -} diff --git a/definitions/gemm/gemm_n2048_k8192.json b/definitions/gemm/gemm_n2048_k8192.json deleted file mode 100644 index 7149bca35d2c622a8053cc2b6b3ba07a9fa9171d..0000000000000000000000000000000000000000 --- a/definitions/gemm/gemm_n2048_k8192.json +++ /dev/null @@ -1,48 +0,0 @@ -{ - "name": "gemm_n2048_k8192", - "description": "General matrix multiply (GEMM) C = A @ B.T. Captured from Llama-3.2-1B mlp.down_proj.", - "op_type": "gemm", - "tags": [ - "status:unverified", - "model:llama-3.2-1b" - ], - "axes": { - "M": { - "type": "var" - }, - "N": { - "type": "const", - "value": 2048 - }, - "K": { - "type": "const", - "value": 8192 - } - }, - "inputs": { - "A": { - "shape": [ - "M", - "K" - ], - "dtype": "float16" - }, - "B": { - "shape": [ - "N", - "K" - ], - "dtype": "float16" - } - }, - "outputs": { - "C": { - "shape": [ - "M", - "N" - ], - "dtype": "float16" - } - }, - "reference": "import torch\n\ndef run(A, B):\n C = torch.matmul(A, B.T)\n return C" -} diff --git a/definitions/gemm/gemm_n3072_k2048.json b/definitions/gemm/gemm_n3072_k2048.json deleted file mode 100644 index ac32d0fca7a79321bc98d943bede9417974a4c3b..0000000000000000000000000000000000000000 --- a/definitions/gemm/gemm_n3072_k2048.json +++ /dev/null @@ -1,48 +0,0 @@ -{ - "name": "gemm_n3072_k2048", - "description": "General matrix multiply (GEMM) C = A @ B.T. Captured from Llama-3.2-1B attn.qkv_proj (fused q+k+v: 32*64 + 8*64 + 8*64 = 3072).", - "op_type": "gemm", - "tags": [ - "status:unverified", - "model:llama-3.2-1b" - ], - "axes": { - "M": { - "type": "var" - }, - "N": { - "type": "const", - "value": 3072 - }, - "K": { - "type": "const", - "value": 2048 - } - }, - "inputs": { - "A": { - "shape": [ - "M", - "K" - ], - "dtype": "float16" - }, - "B": { - "shape": [ - "N", - "K" - ], - "dtype": "float16" - } - }, - "outputs": { - "C": { - "shape": [ - "M", - "N" - ], - "dtype": "float16" - } - }, - "reference": "import torch\n\ndef run(A, B):\n C = torch.matmul(A, B.T)\n return C" -} diff --git a/definitions/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps1.json b/definitions/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps1.json deleted file mode 100644 index b79c02acabba7c897c959a1e45a3b0db4a9e78bc..0000000000000000000000000000000000000000 --- a/definitions/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps1.json +++ /dev/null @@ -1,111 +0,0 @@ -{ - "name": "gqa_paged_decode_h32_kv8_d64_ps1", - "description": "Batched Grouped Query Attention decode with a paged KV cache. Captured from Llama-3.2-1B.", - "op_type": "gqa_paged", - "tags": [ - "stage:decode", - "status:unverified", - "model:llama-3.2-1b", - "fi_api:flashinfer.decode.BatchDecodeWithPagedKVCacheWrapper", - "tp:1" - ], - "axes": { - "batch_size": { - "type": "var", - "description": "Total number of query tokens." - }, - "num_qo_heads": { - "type": "const", - "value": 32 - }, - "num_kv_heads": { - "type": "const", - "value": 8 - }, - "head_dim": { - "type": "const", - "value": 64 - }, - "num_pages": { - "type": "var" - }, - "page_size": { - "type": "const", - "value": 1 - }, - "len_indptr": { - "type": "var", - "description": "Length of kv_indptr array." - }, - "num_kv_indices": { - "type": "var", - "description": "Total number of KV page indices." - } - }, - "constraints": [ - "len_indptr == batch_size + 1", - "num_kv_indices == kv_indptr[-1].item()" - ], - "inputs": { - "q": { - "shape": [ - "batch_size", - "num_qo_heads", - "head_dim" - ], - "dtype": "bfloat16" - }, - "k_cache": { - "shape": [ - "num_pages", - "page_size", - "num_kv_heads", - "head_dim" - ], - "dtype": "bfloat16" - }, - "v_cache": { - "shape": [ - "num_pages", - "page_size", - "num_kv_heads", - "head_dim" - ], - "dtype": "bfloat16" - }, - "kv_indptr": { - "shape": [ - "len_indptr" - ], - "dtype": "int32" - }, - "kv_indices": { - "shape": [ - "num_kv_indices" - ], - "dtype": "int32" - }, - "sm_scale": { - "shape": null, - "dtype": "float32" - } - }, - "outputs": { - "output": { - "shape": [ - "batch_size", - "num_qo_heads", - "head_dim" - ], - "dtype": "bfloat16" - }, - "lse": { - "shape": [ - "batch_size", - "num_qo_heads" - ], - "dtype": "float32" - } - }, - "reference": "import torch\nimport math\n\n\n@torch.no_grad()\ndef run(q, k_cache, v_cache, kv_indptr, kv_indices, sm_scale):\n batch_size, num_qo_heads, head_dim = q.shape\n _, page_size, num_kv_heads, _ = k_cache.shape\n\n assert num_qo_heads == 32\n assert num_kv_heads == 8\n assert head_dim == 64\n assert page_size == 1\n\n assert kv_indptr.shape[0] == batch_size + 1\n assert kv_indices.shape[0] == kv_indptr[-1].item()\n\n device = q.device\n output = torch.zeros(\n (batch_size, num_qo_heads, head_dim), dtype=torch.bfloat16, device=device\n )\n lse = torch.full(\n (batch_size, num_qo_heads), -float(\"inf\"), dtype=torch.float32, device=device\n )\n\n gqa_ratio = num_qo_heads // num_kv_heads\n k_flat = k_cache.squeeze(1).to(torch.float32)\n v_flat = v_cache.squeeze(1).to(torch.float32)\n q_f32 = q.to(torch.float32)\n\n for b in range(batch_size):\n ps = int(kv_indptr[b].item())\n pe = int(kv_indptr[b + 1].item())\n if ps >= pe:\n output[b].zero_()\n continue\n\n idx = kv_indices[ps:pe].to(torch.long)\n k = k_flat[idx].permute(1, 0, 2).repeat_interleave(gqa_ratio, dim=0)\n v = v_flat[idx].permute(1, 0, 2).repeat_interleave(gqa_ratio, dim=0)\n q_b = q_f32[b].unsqueeze(1)\n\n logits = torch.bmm(q_b, k.transpose(1, 2)).squeeze(1) * sm_scale\n lse[b] = torch.logsumexp(logits, dim=-1) / math.log(2.0)\n attn = torch.softmax(logits, dim=-1)\n output[b] = torch.bmm(attn.unsqueeze(1), v).squeeze(1).to(torch.bfloat16)\n\n return output, lse" -} diff --git a/definitions/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps64.json b/definitions/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps64.json deleted file mode 100644 index 865a5502ffd2ec952e92185f5be4ef5eeb481a60..0000000000000000000000000000000000000000 --- a/definitions/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps64.json +++ /dev/null @@ -1,122 +0,0 @@ -{ - "name": "gqa_paged_decode_h32_kv8_d64_ps64", - "description": "Batched Grouped Query Attention decode with a paged KV cache (page_size=64). Captured from Llama-3.2-1B.", - "op_type": "gqa_paged", - "tags": [ - "stage:decode", - "status:unverified", - "model:llama-3.2-1b", - "fi_api:flashinfer.decode.BatchDecodeWithPagedKVCacheWrapper", - "tp:1" - ], - "axes": { - "batch_size": { - "type": "var", - "description": "Total number of query tokens." - }, - "num_qo_heads": { - "type": "const", - "value": 32, - "description": "Number of query/output attention heads." - }, - "num_kv_heads": { - "type": "const", - "value": 8, - "description": "Number of key-value attention heads." - }, - "head_dim": { - "type": "const", - "value": 64, - "description": "Dimension of each attention head." - }, - "num_pages": { - "type": "var", - "description": "Total number of allocated pages in the KV cache." - }, - "page_size": { - "type": "const", - "value": 64, - "description": "Number of tokens stored per page." - }, - "len_indptr": { - "type": "var", - "description": "Length of kv_indptr array." - }, - "num_kv_indices": { - "type": "var", - "description": "Total number of KV page indices." - } - }, - "constraints": [ - "len_indptr == batch_size + 1", - "num_kv_indices == kv_indptr[-1].item()" - ], - "inputs": { - "q": { - "shape": [ - "batch_size", - "num_qo_heads", - "head_dim" - ], - "dtype": "bfloat16" - }, - "k_cache": { - "shape": [ - "num_pages", - "page_size", - "num_kv_heads", - "head_dim" - ], - "dtype": "bfloat16" - }, - "v_cache": { - "shape": [ - "num_pages", - "page_size", - "num_kv_heads", - "head_dim" - ], - "dtype": "bfloat16" - }, - "kv_indptr": { - "shape": [ - "len_indptr" - ], - "dtype": "int32" - }, - "kv_indices": { - "shape": [ - "num_kv_indices" - ], - "dtype": "int32" - }, - "kv_last_page_len": { - "shape": [ - "batch_size" - ], - "dtype": "int32" - }, - "sm_scale": { - "shape": null, - "dtype": "float32" - } - }, - "outputs": { - "output": { - "shape": [ - "batch_size", - "num_qo_heads", - "head_dim" - ], - "dtype": "bfloat16" - }, - "lse": { - "shape": [ - "batch_size", - "num_qo_heads" - ], - "dtype": "float32" - } - }, - "reference": "import torch\nimport math\n\n\n@torch.no_grad()\ndef run(q, k_cache, v_cache, kv_indptr, kv_indices, kv_last_page_len, sm_scale):\n batch_size, num_qo_heads, head_dim = q.shape\n _, page_size, num_kv_heads, _ = k_cache.shape\n\n assert num_qo_heads == 32\n assert num_kv_heads == 8\n assert head_dim == 64\n assert page_size == 64\n\n device = q.device\n output = torch.zeros(\n (batch_size, num_qo_heads, head_dim), dtype=torch.bfloat16, device=device\n )\n lse = torch.full(\n (batch_size, num_qo_heads), -float(\"inf\"), dtype=torch.float32, device=device\n )\n\n gqa_ratio = num_qo_heads // num_kv_heads\n k_cache_f32 = k_cache.to(torch.float32)\n v_cache_f32 = v_cache.to(torch.float32)\n q_f32 = q.to(torch.float32)\n\n for b in range(batch_size):\n ps = int(kv_indptr[b].item())\n pe = int(kv_indptr[b + 1].item())\n last_len = int(kv_last_page_len[b].item())\n if ps >= pe:\n output[b].zero_()\n continue\n\n page_ids = kv_indices[ps:pe].to(torch.long)\n num_full_pages = len(page_ids) - 1\n\n if num_full_pages > 0:\n k_full = k_cache_f32[page_ids[:num_full_pages]].reshape(-1, num_kv_heads, head_dim)\n v_full = v_cache_f32[page_ids[:num_full_pages]].reshape(-1, num_kv_heads, head_dim)\n else:\n k_full = torch.empty(0, num_kv_heads, head_dim, device=device)\n v_full = torch.empty(0, num_kv_heads, head_dim, device=device)\n k_tokens = torch.cat([k_full, k_cache_f32[page_ids[-1], :last_len]], dim=0)\n v_tokens = torch.cat([v_full, v_cache_f32[page_ids[-1], :last_len]], dim=0)\n\n k = k_tokens.permute(1, 0, 2).repeat_interleave(gqa_ratio, dim=0)\n v = v_tokens.permute(1, 0, 2).repeat_interleave(gqa_ratio, dim=0)\n q_b = q_f32[b].unsqueeze(1)\n\n logits = torch.bmm(q_b, k.transpose(1, 2)).squeeze(1) * sm_scale\n lse[b] = torch.logsumexp(logits, dim=-1) / math.log(2.0)\n attn = torch.softmax(logits, dim=-1)\n output[b] = torch.bmm(attn.unsqueeze(1), v).squeeze(1).to(torch.bfloat16)\n\n return output, lse" -} diff --git a/definitions/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps1.json b/definitions/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps1.json deleted file mode 100644 index 81bf834cb302274b90206b4c6400803b96230667..0000000000000000000000000000000000000000 --- a/definitions/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps1.json +++ /dev/null @@ -1,122 +0,0 @@ -{ - "name": "gqa_paged_prefill_causal_h32_kv8_d64_ps1", - "description": "Batched Grouped Query Attention prefill with a paged KV cache. Causal mask is applied. Captured from Llama-3.2-1B during incremental prefill.", - "op_type": "gqa_paged", - "tags": [ - "stage:prefill", - "status:unverified", - "model:llama-3.2-1b", - "fi_api:flashinfer.prefill.BatchPrefillWithPagedKVCacheWrapper", - "tp:1" - ], - "axes": { - "num_qo_heads": { - "type": "const", - "value": 32 - }, - "num_kv_heads": { - "type": "const", - "value": 8 - }, - "head_dim": { - "type": "const", - "value": 64 - }, - "page_size": { - "type": "const", - "value": 1 - }, - "len_indptr": { - "type": "var", - "description": "Length of indptr arrays. Should be the same for qo_indptr and kv_indptr (batch_size + 1)." - }, - "total_q": { - "type": "var", - "description": "Total number of query tokens." - }, - "num_kv_indices": { - "type": "var", - "description": "Total number of KV page indices." - }, - "num_pages": { - "type": "var" - } - }, - "constraints": [ - "total_q == qo_indptr[-1].item()", - "num_kv_indices == kv_indptr[-1].item()" - ], - "inputs": { - "q": { - "shape": [ - "total_q", - "num_qo_heads", - "head_dim" - ], - "dtype": "bfloat16" - }, - "k_cache": { - "shape": [ - "num_pages", - "page_size", - "num_kv_heads", - "head_dim" - ], - "dtype": "bfloat16" - }, - "v_cache": { - "shape": [ - "num_pages", - "page_size", - "num_kv_heads", - "head_dim" - ], - "dtype": "bfloat16" - }, - "qo_indptr": { - "shape": [ - "len_indptr" - ], - "dtype": "int32", - "description": "Query offsets for each sequence." - }, - "kv_indptr": { - "shape": [ - "len_indptr" - ], - "dtype": "int32", - "description": "KV page offsets for each sequence." - }, - "kv_indices": { - "shape": [ - "num_kv_indices" - ], - "dtype": "int32", - "description": "Page IDs for KV cache lookups." - }, - "sm_scale": { - "shape": null, - "dtype": "float32", - "description": "Softmax scale. Default is (1/sqrt(head_dim))." - } - }, - "outputs": { - "output": { - "shape": [ - "total_q", - "num_qo_heads", - "head_dim" - ], - "dtype": "bfloat16" - }, - "lse": { - "shape": [ - "total_q", - "num_qo_heads" - ], - "dtype": "float32", - "description": "The 2-based log-sum-exp of attention logits." - } - }, - "reference": "import torch\nimport math\n\nCHUNK_Q = 512 # chunk query tokens to bound peak memory for large prefills\n\n\n@torch.no_grad()\ndef run(q, k_cache, v_cache, qo_indptr, kv_indptr, kv_indices, sm_scale):\n total_q, num_qo_heads, head_dim = q.shape\n num_pages, page_size, num_kv_heads, _ = k_cache.shape\n batch_size = int(qo_indptr.shape[0]) - 1\n\n assert num_qo_heads == 32\n assert num_kv_heads == 8\n assert head_dim == 64\n assert page_size == 1\n\n device = q.device\n output = torch.zeros((total_q, num_qo_heads, head_dim), dtype=torch.bfloat16, device=device)\n lse = torch.full((total_q, num_qo_heads), -float(\"inf\"), dtype=torch.float32, device=device)\n\n gqa_ratio = num_qo_heads // num_kv_heads\n q_f32 = q.to(torch.float32)\n k_flat = k_cache.squeeze(1).to(torch.float32)\n v_flat = v_cache.squeeze(1).to(torch.float32)\n\n for b in range(batch_size):\n qs = int(qo_indptr[b].item())\n qe = int(qo_indptr[b + 1].item())\n kvs = int(kv_indptr[b].item())\n kve = int(kv_indptr[b + 1].item())\n if qs >= qe or kvs >= kve:\n continue\n\n page_ids = kv_indices[kvs:kve].to(torch.long)\n k = k_flat[page_ids]\n v = v_flat[page_ids]\n num_kv = k.shape[0]\n num_q = qe - qs\n delta = num_kv - num_q\n\n k_exp = k.permute(1, 0, 2).repeat_interleave(gqa_ratio, dim=0)\n v_exp = v.permute(1, 0, 2).repeat_interleave(gqa_ratio, dim=0)\n kv_pos = torch.arange(num_kv, device=device)\n\n for chunk_start in range(0, num_q, CHUNK_Q):\n chunk_end = min(chunk_start + CHUNK_Q, num_q)\n q_chunk = q_f32[qs + chunk_start:qs + chunk_end]\n\n logits = torch.einsum(\"qhd,hkd->hqk\", q_chunk, k_exp) * sm_scale\n\n q_pos = torch.arange(chunk_start, chunk_end, device=device).unsqueeze(1)\n mask = kv_pos.unsqueeze(0) > q_pos + delta\n logits.masked_fill_(mask.unsqueeze(0), float(\"-inf\"))\n\n lse[qs + chunk_start:qs + chunk_end] = (\n torch.logsumexp(logits, dim=-1) / math.log(2.0)\n ).permute(1, 0)\n\n attn = torch.softmax(logits, dim=-1)\n output[qs + chunk_start:qs + chunk_end] = torch.einsum(\n \"hqk,hkd->qhd\", attn, v_exp\n ).to(torch.bfloat16)\n\n return output, lse" -} diff --git a/definitions/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps64.json b/definitions/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps64.json deleted file mode 100644 index 79b3eed0f9c55397580f95075ded188f93e9ec35..0000000000000000000000000000000000000000 --- a/definitions/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps64.json +++ /dev/null @@ -1,132 +0,0 @@ -{ - "name": "gqa_paged_prefill_causal_h32_kv8_d64_ps64", - "description": "Batched Grouped Query Attention prefill with a paged KV cache (page_size=64). Causal mask is applied. Captured from Llama-3.2-1B during incremental prefill.", - "op_type": "gqa_paged", - "tags": [ - "stage:prefill", - "status:unverified", - "model:llama-3.2-1b", - "fi_api:flashinfer.prefill.BatchPrefillWithPagedKVCacheWrapper", - "tp:1" - ], - "axes": { - "num_qo_heads": { - "type": "const", - "value": 32, - "description": "Number of query/output attention heads." - }, - "num_kv_heads": { - "type": "const", - "value": 8, - "description": "Number of key-value attention heads." - }, - "head_dim": { - "type": "const", - "value": 64, - "description": "Dimension of each attention head." - }, - "page_size": { - "type": "const", - "value": 64, - "description": "Number of tokens stored per page." - }, - "len_indptr": { - "type": "var", - "description": "Length of indptr arrays. Should be the same for qo_indptr and kv_indptr (batch_size + 1)." - }, - "total_q": { - "type": "var", - "description": "Total number of query tokens." - }, - "num_kv_indices": { - "type": "var", - "description": "Total number of KV page indices." - }, - "num_pages": { - "type": "var", - "description": "Total number of allocated pages in the KV cache." - }, - "batch_size": { - "type": "var", - "description": "Number of sequences in the batch." - } - }, - "constraints": [ - "total_q == qo_indptr[-1].item()", - "num_kv_indices == kv_indptr[-1].item()" - ], - "inputs": { - "q": { - "shape": [ - "total_q", - "num_qo_heads", - "head_dim" - ], - "dtype": "bfloat16" - }, - "k_cache": { - "shape": [ - "num_pages", - "page_size", - "num_kv_heads", - "head_dim" - ], - "dtype": "bfloat16" - }, - "v_cache": { - "shape": [ - "num_pages", - "page_size", - "num_kv_heads", - "head_dim" - ], - "dtype": "bfloat16" - }, - "qo_indptr": { - "shape": [ - "len_indptr" - ], - "dtype": "int32" - }, - "kv_indptr": { - "shape": [ - "len_indptr" - ], - "dtype": "int32" - }, - "kv_indices": { - "shape": [ - "num_kv_indices" - ], - "dtype": "int32" - }, - "kv_last_page_len": { - "shape": [ - "batch_size" - ], - "dtype": "int32" - }, - "sm_scale": { - "shape": null, - "dtype": "float32" - } - }, - "outputs": { - "output": { - "shape": [ - "total_q", - "num_qo_heads", - "head_dim" - ], - "dtype": "bfloat16" - }, - "lse": { - "shape": [ - "total_q", - "num_qo_heads" - ], - "dtype": "float32" - } - }, - "reference": "import torch\nimport math\n\nCHUNK_Q = 512\n\n\n@torch.no_grad()\ndef run(q, k_cache, v_cache, qo_indptr, kv_indptr, kv_indices, kv_last_page_len, sm_scale):\n total_q, num_qo_heads, head_dim = q.shape\n num_pages, page_size, num_kv_heads, _ = k_cache.shape\n batch_size = int(qo_indptr.shape[0]) - 1\n\n assert num_qo_heads == 32\n assert num_kv_heads == 8\n assert head_dim == 64\n assert page_size == 64\n\n device = q.device\n output = torch.zeros((total_q, num_qo_heads, head_dim), dtype=torch.bfloat16, device=device)\n lse = torch.full((total_q, num_qo_heads), -float(\"inf\"), dtype=torch.float32, device=device)\n\n gqa_ratio = num_qo_heads // num_kv_heads\n q_f32 = q.to(torch.float32)\n k_cache_f32 = k_cache.to(torch.float32)\n v_cache_f32 = v_cache.to(torch.float32)\n\n for b in range(batch_size):\n qs = int(qo_indptr[b].item())\n qe = int(qo_indptr[b + 1].item())\n kvs = int(kv_indptr[b].item())\n kve = int(kv_indptr[b + 1].item())\n last_len = int(kv_last_page_len[b].item())\n if qs >= qe or kvs >= kve:\n continue\n\n page_ids = kv_indices[kvs:kve].to(torch.long)\n num_full_pages = len(page_ids) - 1\n\n if num_full_pages > 0:\n k_full = k_cache_f32[page_ids[:num_full_pages]].reshape(-1, num_kv_heads, head_dim)\n v_full = v_cache_f32[page_ids[:num_full_pages]].reshape(-1, num_kv_heads, head_dim)\n else:\n k_full = torch.empty(0, num_kv_heads, head_dim, device=device)\n v_full = torch.empty(0, num_kv_heads, head_dim, device=device)\n k_tokens = torch.cat([k_full, k_cache_f32[page_ids[-1], :last_len]], dim=0)\n v_tokens = torch.cat([v_full, v_cache_f32[page_ids[-1], :last_len]], dim=0)\n\n num_kv = k_tokens.shape[0]\n num_q = qe - qs\n delta = num_kv - num_q\n\n k_exp = k_tokens.permute(1, 0, 2).repeat_interleave(gqa_ratio, dim=0)\n v_exp = v_tokens.permute(1, 0, 2).repeat_interleave(gqa_ratio, dim=0)\n kv_pos = torch.arange(num_kv, device=device)\n\n for chunk_start in range(0, num_q, CHUNK_Q):\n chunk_end = min(chunk_start + CHUNK_Q, num_q)\n q_chunk = q_f32[qs + chunk_start:qs + chunk_end]\n\n logits = torch.einsum(\"qhd,hkd->hqk\", q_chunk, k_exp) * sm_scale\n\n q_pos = torch.arange(chunk_start, chunk_end, device=device).unsqueeze(1)\n mask = kv_pos.unsqueeze(0) > q_pos + delta\n logits.masked_fill_(mask.unsqueeze(0), float(\"-inf\"))\n\n lse[qs + chunk_start:qs + chunk_end] = (\n torch.logsumexp(logits, dim=-1) / math.log(2.0)\n ).permute(1, 0)\n\n attn = torch.softmax(logits, dim=-1)\n output[qs + chunk_start:qs + chunk_end] = torch.einsum(\n \"hqk,hkd->qhd\", attn, v_exp\n ).to(torch.bfloat16)\n\n return output, lse" -} diff --git a/definitions/gqa_ragged/gqa_ragged_prefill_causal_h32_kv8_d64.json b/definitions/gqa_ragged/gqa_ragged_prefill_causal_h32_kv8_d64.json deleted file mode 100644 index 7465f1a3ed2f85f4020a84311e552be42bf742b8..0000000000000000000000000000000000000000 --- a/definitions/gqa_ragged/gqa_ragged_prefill_causal_h32_kv8_d64.json +++ /dev/null @@ -1,102 +0,0 @@ -{ - "name": "gqa_ragged_prefill_causal_h32_kv8_d64", - "description": "Batched Grouped Query Attention prefill with ragged (variable-length) inputs. Causal mask is applied. Captured from Llama-3.2-1B during total prefill.", - "op_type": "gqa_ragged", - "tags": [ - "stage:prefill", - "status:unverified", - "model:llama-3.2-1b", - "fi_api:flashinfer.prefill.BatchPrefillWithRaggedKVCacheWrapper", - "tp:1" - ], - "axes": { - "num_qo_heads": { - "type": "const", - "value": 32 - }, - "num_kv_heads": { - "type": "const", - "value": 8 - }, - "head_dim": { - "type": "const", - "value": 64 - }, - "len_indptr": { - "type": "var", - "description": "Length of indptr arrays. Should be the same for qo_indptr and kv_indptr (batch_size + 1)." - }, - "total_q": { - "type": "var", - "description": "Total number of query tokens." - }, - "total_kv": { - "type": "var", - "description": "Total key-value tokens across all sequences." - } - }, - "constraints": [ - "total_q == qo_indptr[-1].item()", - "total_kv == kv_indptr[-1].item()" - ], - "inputs": { - "q": { - "shape": [ - "total_q", - "num_qo_heads", - "head_dim" - ], - "dtype": "bfloat16" - }, - "k": { - "shape": [ - "total_kv", - "num_kv_heads", - "head_dim" - ], - "dtype": "bfloat16" - }, - "v": { - "shape": [ - "total_kv", - "num_kv_heads", - "head_dim" - ], - "dtype": "bfloat16" - }, - "qo_indptr": { - "shape": [ - "len_indptr" - ], - "dtype": "int32" - }, - "kv_indptr": { - "shape": [ - "len_indptr" - ], - "dtype": "int32" - }, - "sm_scale": { - "shape": null, - "dtype": "float32" - } - }, - "outputs": { - "output": { - "shape": [ - "total_q", - "num_qo_heads", - "head_dim" - ], - "dtype": "bfloat16" - }, - "lse": { - "shape": [ - "total_q", - "num_qo_heads" - ], - "dtype": "float32" - } - }, - "reference": "import torch\nimport math\n\n\n@torch.no_grad()\ndef run(q, k, v, qo_indptr, kv_indptr, sm_scale):\n total_q, num_qo_heads, head_dim = q.shape\n total_kv, num_kv_heads, _ = k.shape\n len_indptr = qo_indptr.shape[0]\n\n assert num_qo_heads == 32\n assert num_kv_heads == 8\n assert head_dim == 64\n\n assert total_q == qo_indptr[-1].item()\n assert total_kv == kv_indptr[-1].item()\n\n device = q.device\n\n output = torch.zeros(\n (total_q, num_qo_heads, head_dim), dtype=torch.bfloat16, device=device\n )\n lse = torch.full(\n (total_q, num_qo_heads), -float(\"inf\"), dtype=torch.float32, device=device\n )\n\n gqa_ratio = num_qo_heads // num_kv_heads\n\n q_f32 = q.to(torch.float32)\n k_f32 = k.to(torch.float32)\n v_f32 = v.to(torch.float32)\n\n for b in range(len_indptr - 1):\n q_start = int(qo_indptr[b].item())\n q_end = int(qo_indptr[b + 1].item())\n\n kv_start = int(kv_indptr[b].item())\n kv_end = int(kv_indptr[b + 1].item())\n\n if q_start >= q_end or kv_start >= kv_end:\n continue\n\n q_batch = q_f32[q_start:q_end]\n k_batch = k_f32[kv_start:kv_end]\n v_batch = v_f32[kv_start:kv_end]\n\n num_q_tokens = q_batch.shape[0]\n num_kv_tokens = k_batch.shape[0]\n delta = num_kv_tokens - num_q_tokens\n\n k_expanded = k_batch.repeat_interleave(gqa_ratio, dim=1)\n v_expanded = v_batch.repeat_interleave(gqa_ratio, dim=1)\n\n logits = torch.einsum('qhd,khd->qhk', q_batch, k_expanded) * sm_scale\n\n q_positions = torch.arange(num_q_tokens, device=device)\n kv_positions = torch.arange(num_kv_tokens, device=device)\n\n causal_mask = kv_positions[None, :] < (q_positions[:, None] + 1 + delta)\n logits = logits.masked_fill(~causal_mask[:, None, :], float('-inf'))\n\n lse_batch = torch.logsumexp(logits, dim=-1) / math.log(2.0)\n lse[q_start:q_end] = lse_batch\n\n attn_weights = torch.softmax(logits, dim=-1)\n output_batch = torch.einsum('qhk,khd->qhd', attn_weights, v_expanded)\n output[q_start:q_end] = output_batch.to(torch.bfloat16)\n\n return output, lse" -} diff --git a/definitions/mla_paged/mla_paged_decode_h8_ckv512_kpe64_ps1.json b/definitions/mla_paged/mla_paged_decode_h8_ckv512_kpe64_ps1.json index 3e5e92a4f5866ada74f57db2c9987fb1662ea116..6a60bb19f508fca007d141f451f90525152782f2 100644 --- a/definitions/mla_paged/mla_paged_decode_h8_ckv512_kpe64_ps1.json +++ b/definitions/mla_paged/mla_paged_decode_h8_ckv512_kpe64_ps1.json @@ -1,12 +1,11 @@ { "name": "mla_paged_decode_h8_ckv512_kpe64_ps1", - "description": "Batched Multi-head Latent Attention decode with a paged KV cache. Captured from Kimi K2 / Kimi K2.5 with tensor parallel size 8 (64/8=8 query heads). The Kimi K2.5 text backbone (text_config.model_type=kimi_k2, DeepseekV3ForCausalLM) shares the same MLA shape as Kimi K2: kv_lora_rank=512, qk_rope_head_dim=64, qk_nope_head_dim=128, v_head_dim=128, num_attention_heads=64 \u2192 h=8 at TP=8.", + "description": "Batched Multi-head Latent Attention decode with a paged KV cache. Captured from Kimi K2 with tensor parallel size 8 (64/8=8 query heads).", "op_type": "mla_paged", "tags": [ "stage:decode", "status:verified", "model:kimi-k2", - "model:kimi-k2.5", "fi_api:flashinfer.mla.BatchMLAPagedAttentionWrapper", "tp:8" ], diff --git a/definitions/mla_paged/mla_paged_prefill_causal_h8_ckv512_kpe64_ps1.json b/definitions/mla_paged/mla_paged_prefill_causal_h8_ckv512_kpe64_ps1.json index 1686a66eaf17b15a5f70c1de522ac707a48d2db2..9fdfbce76f1c1f30f6624e0aeb47e5191f659e5f 100644 --- a/definitions/mla_paged/mla_paged_prefill_causal_h8_ckv512_kpe64_ps1.json +++ b/definitions/mla_paged/mla_paged_prefill_causal_h8_ckv512_kpe64_ps1.json @@ -1,12 +1,11 @@ { "name": "mla_paged_prefill_causal_h8_ckv512_kpe64_ps1", - "description": "Batched Multi-head Latent Attention prefill with a paged KV cache. Causal mask is applied. Captured from Kimi K2 / Kimi K2.5 during incremental prefill with tensor parallel size 8 (64/8=8 query heads). Kimi K2.5 shares this shape via its DeepseekV3ForCausalLM text backbone.", + "description": "Batched Multi-head Latent Attention prefill with a paged KV cache. Causal mask is applied. Captured from Kimi K2 during incremental prefill with tensor parallel size 8 (64/8=8 query heads).", "op_type": "mla_paged", "tags": [ "stage:prefill", "status:verified", "model:kimi-k2", - "model:kimi-k2.5", "fi_api:flashinfer.mla.BatchMLAPagedAttentionWrapper", "tp:8" ], diff --git a/definitions/mla_ragged/mla_ragged_prefill_causal_h8_qk192_vo128.json b/definitions/mla_ragged/mla_ragged_prefill_causal_h8_qk192_vo128.json index 1dcaefe83accd8fb621c21356ab33f3a0088e4a4..bf028de7704a80d93074177f95358aed3d45db6c 100644 --- a/definitions/mla_ragged/mla_ragged_prefill_causal_h8_qk192_vo128.json +++ b/definitions/mla_ragged/mla_ragged_prefill_causal_h8_qk192_vo128.json @@ -1,12 +1,11 @@ { "name": "mla_ragged_prefill_causal_h8_qk192_vo128", - "description": "Batched Multi-head Latent Attention prefill with ragged (variable-length) inputs. Uses the absorbed MLA formulation with combined QK dimension (qk_nope=128 + qk_rope=64 = 192) and value output dimension 128. Causal mask is applied. Captured from Kimi K2 / Kimi K2.5 during total prefill (no prefix cache) with tensor parallel size 8 (64/8=8 query heads). Kimi K2.5 shares this shape via its DeepseekV3ForCausalLM text backbone.", + "description": "Batched Multi-head Latent Attention prefill with ragged (variable-length) inputs. Uses the absorbed MLA formulation with combined QK dimension (qk_nope=128 + qk_rope=64 = 192) and value output dimension 128. Causal mask is applied. Captured from Kimi K2 during total prefill (no prefix cache) with tensor parallel size 8 (64/8=8 query heads).", "op_type": "mla_ragged", "tags": [ "stage:prefill", "status:verified", "model:kimi-k2", - "model:kimi-k2.5", "fi_api:flashinfer.prefill.BatchPrefillWithRaggedKVCacheWrapper", "tp:8" ], diff --git a/definitions/sampling/top_k_sampling_from_probs_v163840.json b/definitions/sampling/top_k_sampling_from_probs_v163840.json deleted file mode 100644 index cfa50ddda247e7cbffa858fc452ce094b39d783b..0000000000000000000000000000000000000000 --- a/definitions/sampling/top_k_sampling_from_probs_v163840.json +++ /dev/null @@ -1,48 +0,0 @@ -{ - "name": "top_k_sampling_from_probs_v163840", - "op_type": "sampling", - "description": "Top-k sampling from probabilities with vocab_size=163840. Keeps only the k highest probability tokens, renormalizes, then samples from the filtered distribution. Captured from Kimi K2.5 (moonshotai/Kimi-K2.5, text_config.vocab_size=163840).", - "tags": [ - "status:reference", - "model:kimi-k2.5", - "fi_api:flashinfer.sampling.top_k_sampling_from_probs" - ], - "axes": { - "batch_size": { - "type": "var", - "description": "Number of sequences to sample from" - }, - "vocab_size": { - "type": "const", - "value": 163840, - "description": "Size of the vocabulary for Kimi K2.5 (text_config.vocab_size)." - } - }, - "inputs": { - "probs": { - "shape": [ - "batch_size", - "vocab_size" - ], - "dtype": "float32", - "description": "Probability distributions (after softmax)" - }, - "top_k": { - "shape": [ - "batch_size" - ], - "dtype": "int32", - "description": "Number of top tokens to consider for sampling per sequence" - } - }, - "outputs": { - "samples": { - "shape": [ - "batch_size" - ], - "dtype": "int64", - "description": "Sampled token indices" - } - }, - "reference": "import torch\n\n@torch.no_grad()\ndef run(probs, top_k):\n batch_size, vocab_size = probs.shape\n device = probs.device\n\n # Check constants\n assert vocab_size == 163840\n\n probs = probs.to(torch.float32)\n samples = torch.empty(batch_size, dtype=torch.int64, device=device)\n\n for i in range(batch_size):\n row = probs[i]\n k = int(top_k[i].item())\n\n if 0 < k < vocab_size:\n idx_sorted = torch.argsort(row, descending=True)\n keep_idx = idx_sorted[:k]\n\n filtered = torch.zeros_like(row)\n filtered[keep_idx] = row[keep_idx]\n\n row = filtered / filtered.sum()\n\n samples[i] = torch.multinomial(row, 1, replacement=True).squeeze(0)\n\n return samples\n" -} diff --git a/definitions/sampling/top_k_top_p_sampling_from_probs_v163840.json b/definitions/sampling/top_k_top_p_sampling_from_probs_v163840.json deleted file mode 100644 index 50f96cd1be41cc6d9caf2cff1ba274d16ba01b5e..0000000000000000000000000000000000000000 --- a/definitions/sampling/top_k_top_p_sampling_from_probs_v163840.json +++ /dev/null @@ -1,55 +0,0 @@ -{ - "name": "top_k_top_p_sampling_from_probs_v163840", - "op_type": "sampling", - "description": "Top-k top-p (nucleus) sampling from probabilities with vocab_size=163840. Filters probabilities using top-k and top-p constraints, then samples from the filtered distribution. Captured from Kimi K2.5 (moonshotai/Kimi-K2.5, text_config.vocab_size=163840).", - "tags": [ - "status:reference", - "model:kimi-k2.5", - "fi_api:flashinfer.sampling.top_k_top_p_sampling_from_probs" - ], - "axes": { - "batch_size": { - "type": "var", - "description": "Number of sequences to sample from" - }, - "vocab_size": { - "type": "const", - "value": 163840, - "description": "Size of the vocabulary for Kimi K2.5 (text_config.vocab_size)." - } - }, - "inputs": { - "probs": { - "shape": [ - "batch_size", - "vocab_size" - ], - "dtype": "float32", - "description": "Probability distributions (after softmax)" - }, - "top_k": { - "shape": [ - "batch_size" - ], - "dtype": "int32", - "description": "Number of top tokens to consider for sampling per sequence" - }, - "top_p": { - "shape": [ - "batch_size" - ], - "dtype": "float32", - "description": "Cumulative probability threshold for nucleus sampling per sequence" - } - }, - "outputs": { - "samples": { - "shape": [ - "batch_size" - ], - "dtype": "int64", - "description": "Sampled token indices" - } - }, - "reference": "import torch\n\n@torch.no_grad()\ndef run(probs, top_k, top_p):\n batch_size, vocab_size = probs.shape\n device = probs.device\n\n # Check constants\n assert vocab_size == 163840\n\n probs = probs.to(torch.float32)\n samples = torch.empty(batch_size, dtype=torch.int64, device=device)\n\n for i in range(batch_size):\n row = probs[i]\n k = int(top_k[i].item())\n p = float(top_p[i].item())\n\n # Apply top-k filtering\n if 0 < k < vocab_size:\n idx_sorted = torch.argsort(row, descending=True)\n keep_idx_k = idx_sorted[:k]\n filtered_k = torch.zeros_like(row)\n filtered_k[keep_idx_k] = row[keep_idx_k]\n row = filtered_k / filtered_k.sum()\n\n # Then apply top-p filtering\n if p <= 0.0:\n samples[i] = torch.argmax(row).to(torch.int64)\n continue\n\n if p < 1.0:\n vals, idx = torch.sort(row, descending=True)\n cdf = torch.cumsum(vals, dim=0)\n\n to_remove = cdf > p\n if vocab_size > 1:\n to_remove[1:] = to_remove[:-1].clone()\n to_remove[0] = False\n\n keep_idx_p = idx[~to_remove]\n filtered_p = torch.zeros_like(row)\n filtered_p[keep_idx_p] = row[keep_idx_p]\n row = filtered_p / filtered_p.sum()\n\n samples[i] = torch.multinomial(row, 1, replacement=True).squeeze(0)\n\n return samples\n" -} diff --git a/definitions/sampling/top_k_top_p_sampling_from_probs_v248320.json b/definitions/sampling/top_k_top_p_sampling_from_probs_v248320.json deleted file mode 100644 index 19d0b19a48e48e2a535184c7150f0249230f7e44..0000000000000000000000000000000000000000 --- a/definitions/sampling/top_k_top_p_sampling_from_probs_v248320.json +++ /dev/null @@ -1,54 +0,0 @@ -{ - "name": "top_k_top_p_sampling_from_probs_v248320", - "op_type": "sampling", - "description": "Top-k top-p (nucleus) sampling from probabilities with vocab_size=248320. Filters probabilities using top-k and top-p constraints, then samples from the filtered distribution. Captured from Qwen3.5-35B-A3B.", - "tags": [ - "status:verified", - "model:qwen3.5-35b-a3b", - "fi_api:flashinfer.sampling.top_k_top_p_sampling_from_probs" - ], - "axes": { - "batch_size": { - "type": "var", - "description": "Number of sequences to sample from" - }, - "vocab_size": { - "type": "const", - "value": 248320, - "description": "Size of the vocabulary for Qwen3.5" - } - }, - "inputs": { - "probs": { - "shape": [ - "batch_size", - "vocab_size" - ], - "dtype": "float32", - "description": "Probability distributions (after softmax)" - }, - "top_k": { - "shape": [ - "batch_size" - ], - "dtype": "int32", - "description": "Number of top tokens to consider for sampling per sequence" - }, - "top_p": { - "shape": [ - "batch_size" - ], - "dtype": "float32", - "description": "Cumulative probability threshold for nucleus sampling per sequence" - } - }, - "outputs": { - "samples": { - "shape": [ - "batch_size" - ], - "dtype": "int64", - "description": "Sampled token indices" - } - } -} diff --git a/definitions/sampling/top_p_sampling_from_probs_v163840.json b/definitions/sampling/top_p_sampling_from_probs_v163840.json deleted file mode 100644 index f7b70b0a3c76bf764c89a159024be661012c38ce..0000000000000000000000000000000000000000 --- a/definitions/sampling/top_p_sampling_from_probs_v163840.json +++ /dev/null @@ -1,48 +0,0 @@ -{ - "name": "top_p_sampling_from_probs_v163840", - "op_type": "sampling", - "description": "Top-p (nucleus) sampling from probabilities with vocab_size=163840. Filters probabilities using cumulative probability threshold, then samples from the filtered distribution. Captured from Kimi K2.5 (moonshotai/Kimi-K2.5, text_config.vocab_size=163840).", - "tags": [ - "status:reference", - "model:kimi-k2.5", - "fi_api:flashinfer.sampling.top_p_sampling_from_probs" - ], - "axes": { - "batch_size": { - "type": "var", - "description": "Number of sequences to sample from" - }, - "vocab_size": { - "type": "const", - "value": 163840, - "description": "Size of the vocabulary for Kimi K2.5 (text_config.vocab_size)." - } - }, - "inputs": { - "probs": { - "shape": [ - "batch_size", - "vocab_size" - ], - "dtype": "float32", - "description": "Probability distributions (after softmax)" - }, - "top_p": { - "shape": [ - "batch_size" - ], - "dtype": "float32", - "description": "Cumulative probability threshold for nucleus sampling per sequence" - } - }, - "outputs": { - "samples": { - "shape": [ - "batch_size" - ], - "dtype": "int64", - "description": "Sampled token indices" - } - }, - "reference": "import torch\n\n@torch.no_grad()\ndef run(probs, top_p):\n batch_size, vocab_size = probs.shape\n device = probs.device\n\n # Check constants\n assert vocab_size == 163840\n\n probs = probs.to(torch.float32)\n out = torch.empty(batch_size, dtype=torch.int64, device=device)\n\n for i in range(batch_size):\n row = probs[i]\n p = float(top_p[i].item())\n\n if p <= 0.0:\n out[i] = torch.argmax(row).to(torch.int64)\n continue\n\n if p < 1.0:\n vals, idx = torch.sort(row, descending=True)\n cdf = torch.cumsum(vals, dim=0)\n\n to_remove = cdf > p\n to_remove[1:] = to_remove[:-1].clone()\n to_remove[0] = False\n keep = ~to_remove\n keep_idx = idx[keep]\n\n filtered = torch.zeros_like(row)\n filtered[keep_idx] = row[keep_idx]\n row = filtered / filtered.sum()\n\n out[i] = torch.multinomial(row, 1, replacement=True).squeeze(0)\n\n return out" -} diff --git a/skills/add-flashinfer-solution/SKILL.md b/skills/add-flashinfer-solution/SKILL.md deleted file mode 100644 index c796b40b112bb4f67763e078e9aae00a9153b54e..0000000000000000000000000000000000000000 --- a/skills/add-flashinfer-solution/SKILL.md +++ /dev/null @@ -1,821 +0,0 @@ ---- -name: add-flashinfer-solution -description: End-to-end workflow for adding a new Solution to the flashinfer-trace dataset, running its benchmark with flashinfer-bench, and visualizing traces in the web UI / public leaderboard. Use when implementing a new attention / GEMM / MoE / RMSNorm / sampling kernel as a Solution against an existing Definition, integrating a third-party kernel library (Flash Attention, FLA, SGLang, TRT-LLM, vLLM, cuDNN, etc.), or onboarding new contributors to the benchmark workflow. ---- - -# Add a New Solution → Run Benchmark → Visualize Traces - -End-to-end guide. Read sections in order on first use; jump to specific sections via the index after that. - -## Index - -1. [When to use this skill](#1-when-to-use-this-skill) -2. [Background: flashinfer-bench vs flashinfer-trace](#2-background-flashinfer-bench-vs-flashinfer-trace) -3. [Prerequisites](#3-prerequisites) -4. [Workflow overview (7 steps)](#4-workflow-overview) -5. [Step 1 — Pick a Definition + audit existing solutions](#step-1--pick-a-definition--audit-existing-solutions) -6. [Step 2 — Pull workload LFS data](#step-2--pull-workload-lfs-data) -7. [Step 3 — Write the wrapper (`main.py`)](#step-3--write-the-wrapper-mainpy) -8. [Step 4 — Write the Solution JSON](#step-4--write-the-solution-json) -9. [Step 5 — Run benchmark with `flashinfer-bench run`](#step-5--run-benchmark-with-flashinfer-bench-run) -10. [Step 6 — Inspect generated traces](#step-6--inspect-generated-traces) -11. [Step 7 — Visualize](#step-7--visualize) -12. [Common gotchas](#common-gotchas) -13. [Reference & templates](#reference--templates) - ---- - -## 1. When to use this skill - -Trigger this skill when: -- You are implementing a **new kernel** as a Solution for an existing Definition in `flashinfer-trace` -- You are wrapping a **third-party library** (Flash Attention, Flash Linear Attention, SGLang vendored kernel, TRT-LLM, vLLM, cuDNN, xFormers, …) so it can run inside `flashinfer-bench` -- You need to **run the benchmark** and produce traces (latency / speedup / correctness) for a Solution -- You want to **visualize results** in the local Next.js web UI or upload to https://bench.flashinfer.ai - -Do NOT use this skill for: -- Adding a new Definition (use the official `extract-kernel-definitions` skill) -- Collecting Workload tensor data from a SGLang inference run (use `collect-workloads`) -- Onboarding a brand-new model (use `onboard-model`) - -## 2. Background: flashinfer-bench vs flashinfer-trace - -Two repos, two roles: - -| Repo | Role | Contains | -|---|---|---| -| **flashinfer-bench** (https://github.com/flashinfer-ai/flashinfer-bench) | benchmark **codebase / framework** | Python package `flashinfer_bench/` + CLI (`flashinfer-bench run`, `flashinfer-bench report …`) + Next.js web UI under `web/` | -| **flashinfer-trace** (https://huggingface.co/datasets/flashinfer-ai/flashinfer-trace) | benchmark **dataset** | `definitions/`, `solutions/`, `workloads/`, `blob/workloads/` (LFS), `traces/` | - -Data model: - -``` -Definition ──── what kernel + parameter space (axes, inputs, outputs, reference impl) -Workload ──── concrete tensor inputs for one Definition instance (UUID-keyed) -Solution ──── one implementation of a Definition (Python / Triton / CUDA / …) -Trace ──── result of running a Solution on a Workload (latency, speedup, status) -``` - -Adding a new Solution = create the JSON + `main.py` under `solutions////`, then run `flashinfer-bench run` to evaluate it against existing Workloads, which writes Traces under `traces///.jsonl`. - -## 3. Prerequisites - -### 3.1 Hardware -- Linux x86_64 + NVIDIA GPU (H100 PCIe / SXM, B200, L40, A100 ... see Definition `target_hardware` to check compatibility) - -### 3.2 Disk space -- **Important**: home quota is typically tight (~5 GB). Place all heavy dirs on a scratch path: - ``` - HOME (small): - /home//kernel_arena/ - flashinfer-bench → symlink → SCRATCH - flashinfer-trace → symlink → SCRATCH - scripts/ → symlink → SCRATCH - results/ → symlink → SCRATCH - - SCRATCH (~700 GB): - /home/scratch./kernel_arena/ - flashinfer-bench/ # Codebase clone - flashinfer-trace/ # Dataset clone (LFS data lives here) - scripts/ # Your wrapper scripts, run scripts - results/ # Run logs - ``` -- LFS data per Workload: ~500 KB–2 MB; pulling 50–100 Workloads is typically <100 MB - -### 3.3 Environment — recommended: pre-built `.sqsh` image - -If your team has a pre-built `flashinfer-bench-runner.sqsh` on shared -scratch (see Section 3.5 for how to build one), use it directly. Zero pip -install per invocation, ~5 min saved per `crun` job. This is the -recommended path for collaborators consuming the image. - -```bash -crun -q 'gpu.chip=gh100 and cpu.arch=x86_64' --gpus=1 -C \ - -img /home/scratch./containers/flashinfer-bench-runner.sqsh \ - -r /tmp .sh -``` - -FYI: /home/scratch.yuny_wwfo/containers/flashinfer-bench-runner.sqsh is the one I built - -Inside `.sh`, **no pip install needed** — everything is baked -in. Jump straight to the dataset: - -```bash -cd /home/scratch./kernel_arena/flashinfer-trace -flashinfer-bench run --local . --definitions --solutions ... -``` - -### 3.4 Environment — fallback: `crun -img nvcr.io/...` + per-run pip install - -If a pre-built `.sqsh` is not available (first contributor on a new -cluster; one-off Solution authoring; you can't or don't want to build an -image), the fallback is the path used to verify the three reference -Solutions before the sqsh existed. Zero NGC API key setup, zero docker -daemon, zero custom image — `crun` pulls the NGC base on the compute node -(which has cluster-configured NGC credentials), and the user script -installs pip dependencies into `/tmp/pip-pkgs` inside that container. - -Per-run cost: ~5 min pip install at start of every `crun` invocation. The -torch 2.5 → 2.12 upgrade triggered by `fla-core` is the dominant time. - -```bash -crun -q 'gpu.chip=gh100 and cpu.arch=x86_64' --gpus=1 -C \ - -img nvcr.io/nvidia/pytorch:24.10-py3 \ - -r /tmp .sh -``` - -Inside `.sh`: - -```bash -PIP_TARGET=/tmp/pip-pkgs -mkdir -p "$PIP_TARGET" -export PYTHONPATH="$PIP_TARGET:$PYTHONPATH" -export PATH="$PIP_TARGET/bin:$PATH" - -# Match the stack used to verify the three reference Solutions: -# `flashinfer-python==0.6.9` is pinned for CuTe DSL ABI; everything else -# lets pip pick the latest compatible (which is what the reference Solutions -# were verified against — flashinfer-bench 0.1.2, fla-core 0.5.0, -# nvidia-cutlass-dsl 4.5.0, cuda-python 13.2.0, torch upgraded to 2.12). -pip install --target "$PIP_TARGET" --no-cache-dir \ - flashinfer-python==0.6.9 \ - flashinfer-bench \ - flash-linear-attention \ - "nvidia-cutlass-dsl[cu13]" \ - cuda-python \ - safetensors huggingface-hub -``` - -**Why `flashinfer-python` is pinned but the rest are not**: the CuTe DSL -GDN kernel inside `flashinfer-python==0.6.9` has tight ABI coupling with -the cutlass-dsl + cuda-python versions; pinning the latest of those + a -known-good `flashinfer-python` is what the three reference Solutions were -verified with. Other packages float to latest deliberately so the script -keeps working as PyPI publishes minor revisions. - -For wrappers that need additional libraries: -- **FA3**: add a source-build step (`cd /tmp && git clone https://github.com/Dao-AILab/flash-attention.git && cd flash-attention/hopper && python setup.py install`); ~15-25 min one-time compile -- **SGLang vendored kernel**: vendor the kernel file into your Solution's `sources/` — no pip install required - -### 3.5 Building a `.sqsh` image (one-time, ~30 min) - -The pre-built `.sqsh` is created once on a GPU compute node (frontend -cannot run `enroot start` because of VS Code SSH session cgroup namespace -constraints) and dropped at a shared scratch path for the team to consume. - -**Prerequisites**: -- NGC API key in `~/.config/enroot/.credentials` (one-time): - ```bash - mkdir -p ~/.config/enroot && chmod 700 ~/.config/enroot - cat > ~/.config/enroot/.credentials < - machine authn.nvidia.com login \$oauthtoken password - EOF - chmod 600 ~/.config/enroot/.credentials - ``` - Generate the key at https://ngc.nvidia.com → Account → API Keys. - -- Scratch path with ~50 GB free (NGC base sqsh 18 GB + final sqsh 22 GB + - enroot cache 10 GB). - -**Build script** (drop at `/home/scratch./.fbench-build/build_on_gpu.sh`): - -```bash -#!/bin/bash -set -e -export ENROOT_CACHE_PATH=/home/scratch./.enroot-cache -export ENROOT_DATA_PATH=/home/scratch./.enroot-data -mkdir -p $ENROOT_CACHE_PATH $ENROOT_DATA_PATH \ - /home/scratch./containers -cd /home/scratch./containers - -echo "=== HOST: $(hostname) @ $(date) ===" - -# Step 1: pull NGC base → sqsh (~10 min, first time only; subsequent builds reuse) -BASE_SQSH=$(ls *pytorch+24.10-py3*.sqsh 2>/dev/null | head -1) -if [ -z "$BASE_SQSH" ]; then - enroot import 'docker://nvcr.io#nvidia/pytorch:24.10-py3' - BASE_SQSH=$(ls *pytorch+24.10-py3*.sqsh | head -1) -fi -echo "Base: $BASE_SQSH" - -# Step 2: create instance + install pip layer -enroot remove -f fbench-build 2>/dev/null || true -enroot create --name fbench-build "$BASE_SQSH" - -enroot start --rw --root fbench-build bash -c ' - set -e - # The NGC base image ships a broken /root/.local/bin/pip whose shebang - # points to /bin/python3.11, which does not exist. Use `python -m pip` - # to invoke the container python directly. - rm -rf /root/.local - apt-get update && apt-get install -y --no-install-recommends git-lfs - python -m pip install --no-cache-dir \ - flashinfer-python==0.6.9 \ - flashinfer-bench \ - flash-linear-attention \ - "nvidia-cutlass-dsl[cu13]" \ - cuda-python safetensors huggingface-hub - python -c " -import torch; print(\"torch:\", torch.__version__) -import flashinfer_bench; print(\"flashinfer-bench OK\") -import fla; print(\"fla OK\") -" -' - -# Step 3: export final sqsh (~5 min mksquashfs compression) -rm -f flashinfer-bench-runner.sqsh -enroot export --output flashinfer-bench-runner.sqsh fbench-build -enroot remove -f fbench-build -chmod 644 flashinfer-bench-runner.sqsh -ls -lh flashinfer-bench-runner.sqsh -echo "=== DONE @ $(date) ===" -``` - -**Submit on a GPU compute node** (not frontend — frontend's VS Code SSH -cgroup namespace breaks `enroot start`): - -```bash -srun -p 'a100-80gb-pcie@cr+mp/h12sswnt/1gpu-16cpu-128gb' \ - --gres=gpu:1 -t 0:45:00 \ - bash /home/scratch./.fbench-build/build_on_gpu.sh -``` - -A100 1-GPU partition is fine — the build doesn't use the GPU, it just -needs a host shell outside the front-end VS Code session cgroup. - -Total wall time: ~25-30 min on first build, ~20 min on rebuilds (Step 1 -base sqsh cached). - -After the run, the final `flashinfer-bench-runner.sqsh` (~22 GB) is at -`/home/scratch./containers/flashinfer-bench-runner.sqsh`. Move it to -team-shared scratch and tell collaborators to use it via Section 3.3. - -### 3.6 Clone the dataset -```bash -cd /home/scratch./kernel_arena -git clone https://huggingface.co/datasets/flashinfer-ai/flashinfer-trace -cd flashinfer-trace -git-lfs install --local -# Don't pull all LFS yet — pull only the Workloads you need (Step 2) -``` - -### 3.7 (Optional) Clone the framework codebase for the local web UI -```bash -git clone https://github.com/flashinfer-ai/flashinfer-bench.git -``` - -## 4. Workflow overview - -``` -Step 1 ── Pick a Definition + audit existing solutions -Step 2 ── Pull workload LFS data (just for that Definition) -Step 3 ── Write wrapper (main.py) -Step 4 ── Write Solution JSON -Step 5 ── Run benchmark with `flashinfer-bench run` -Step 6 ── Inspect generated traces -Step 7 ── Visualize (web UI / public site) -``` - -Estimated time for an experienced contributor: -- First time end-to-end: **half a day to a full day** (env setup dominates) -- Subsequent solutions on same env: **1–2 hours per Solution** - ---- - -## Step 1 — Pick a Definition + audit existing solutions - -### 1.1 Find candidate Definition - -Browse `flashinfer-trace/definitions//` and pick a JSON file. The filename matches the Definition `name`. - -```bash -ls /home/scratch./kernel_arena/flashinfer-trace/definitions/ -# dsa_paged gdn gemm gqa_paged gqa_ragged mamba_ssu -# mla_paged mla_ragged moe rmsnorm rope sampling -ls /home/scratch./kernel_arena/flashinfer-trace/definitions/gqa_paged/ -# gqa_paged_decode_h32_kv8_d128_ps1.json -# gqa_paged_decode_h32_kv8_d128_ps64.json -# ... -``` - -### 1.2 Read the Definition JSON - -Field semantics → see [`reference/definition_schema.md`](reference/definition_schema.md). Key fields to confirm before writing the wrapper: - -| Field | What to confirm | -|---|---| -| `axes` | Which dims are `const` (compile-time) vs `var` (runtime). Wrapper signature must accept all `var` axes | -| `inputs` | Tensor names + dtypes + shapes. Wrapper signature must match exactly | -| `outputs` | Whether output is allocated by wrapper (return) or pre-allocated (`destination_passing_style: true`) | -| `tags` | `fi_api:` tells you which FlashInfer API the Definition was modeled after — your Solution may want to follow the same contract | -| `reference.code` | The PyTorch reference implementation. Read it to understand semantics (sm_scale, masking, LSE base, etc.) | - -### 1.3 Audit existing solutions for that Definition - -```bash -DEF=mla_paged_decode_h16_ckv512_kpe64_ps1 -OP=mla_paged -TRACE=/home/scratch./kernel_arena/flashinfer-trace -for author in baseline claude-opus-4-1-20250805 gemini-2.5-pro gpt-5-2025-08-07 gpt-o3; do - p="$TRACE/solutions/$author/$OP/$DEF" - if [ -d "$p" ]; then - echo "$author:" - ls "$p" - fi -done -``` - -Why: read existing Solutions' `main.py` to understand: -- How they map `axes` / `inputs` to function arguments -- How they handle paged KV layout (flat indices vs 2D page table) -- How they convert LSE base, sm_scale, dtype -- Whether they pre-allocate output (`destination_passing_style`) - -The `baseline/` solution (FlashInfer wrapper) is the ground-truth reference — match its function signature. - ---- - -## Step 2 — Pull workload LFS data - -Workload tensors live in `blob/workloads///_.safetensors`, stored as Git-LFS. Pull selectively to save disk: - -### 2.1 Pull all workloads for one Definition -```bash -cd /home/scratch./kernel_arena/flashinfer-trace -DEF=mla_paged_decode_h16_ckv512_kpe64_ps1 -OP=mla_paged -git-lfs pull --include="blob/workloads/$OP/$DEF/*" -``` - -### 2.2 Pull only N workloads (for quick iteration on a tight quota) - -Each Workload UUID is one entry in `workloads//.jsonl`. Use head to take the first N: - -```bash -WL_FILE=workloads/$OP/$DEF.jsonl -UUIDS=$(head -5 "$WL_FILE" | python3 -c "import json,sys; [print(json.loads(l)['uuid']) for l in sys.stdin]") -INCLUDES="" -for u in $UUIDS; do - INCLUDES="${INCLUDES}${INCLUDES:+,}blob/workloads/$OP/$DEF/${DEF}_${u}.safetensors" -done -git-lfs pull --include="$INCLUDES" -``` - -### 2.3 Temp-trim the workloads jsonl to N entries (so the runner only sees those N) -```bash -WL_FILE=workloads/$OP/$DEF.jsonl -cp "$WL_FILE" "$WL_FILE.bak" -head -5 "$WL_FILE.bak" > "$WL_FILE" -# After your run, restore: -# mv "$WL_FILE.bak" "$WL_FILE" -``` - -Tip: a `bash trap` cleanup line is the safest way to ensure the original jsonl is restored even on script failure. - -### 2.4 Verify workloads exist -```bash -find blob/workloads/$OP/$DEF -name "*.safetensors" -size +1k | head -5 -``` - ---- - -## Step 3 — Write the wrapper (`main.py`) - -The wrapper is a single Python file at: -``` -solutions////main.py -``` - -### 3.1 Wrapper signature contract - -Function signature must match the Definition's `inputs` exactly: -- Each `inputs[].name` becomes a positional or keyword argument -- Each `axes[].name` (the `var` ones) is also passed -- Tensor dtypes / shapes match `inputs[].dtype` / `inputs[].shape` - -Default entry point is `main.py::run`. Function returns the output tensor(s) in the order declared by `outputs[]`. - -If `spec.destination_passing_style: true`, output tensors are pre-allocated and passed in — modify them in place, then return None or a status flag. - -### 3.2 Three wrapper patterns (templates) - -Pick the template closest to your situation: - -| Pattern | When to use | Template | -|---|---|---| -| **Dense baseline** | Wrapping a generic dense impl (PyTorch SDPA / cuDNN frontend) for reference | [`templates/dense_baseline_main.py`](templates/dense_baseline_main.py) | -| **External Python lib** | Wrapping a pip-installable third-party lib (`fla-core`, `flash-attn`, `xformers`, `triton.ops` …) | [`templates/linear_attention_main.py`](templates/linear_attention_main.py) | -| **Vendored kernel** | Bringing in a Triton / CUDA kernel file from another project (SGLang / vLLM / private) without depending on the upstream package | [`templates/vendored_kernel_main.py`](templates/vendored_kernel_main.py) | - -### 3.3 Writing checklist - -- [ ] Function signature matches Definition `inputs` + `var` axes -- [ ] Output shape / dtype matches Definition `outputs` -- [ ] LSE (if produced) uses **base-2** convention (FlashInfer convention; convert from natural-log via `* (1.0 / math.log(2.0))`) -- [ ] `sm_scale` is taken from the input parameter, NOT recomputed as `1/sqrt(d)` -- [ ] Paged KV cache is read using the supplied `kv_indices` / `kv_indptr` flat layout (not assumed dense) -- [ ] GQA: kv heads broadcasted to query heads if the underlying lib expects MHA (`repeat_interleave(H_q // H_kv, dim=-2)`) -- [ ] Persistent state (e.g. recurrent kernels) returns the new state in the order Definition declares -- [ ] No global state mutation across batches (avoid using cached buffers keyed only on shape — also key on device + dtype) - -See [`reference/wrapper_gotchas.md`](reference/wrapper_gotchas.md) for a full list of common pitfalls. - ---- - -## Step 4 — Write the Solution JSON - -Create `solutions////.json`: - -```json -{ - "name": "my_solution_v1", - "definition": "mla_paged_decode_h16_ckv512_kpe64_ps1", - "author": "", - "spec": { - "language": "python", - "target_hardware": ["NVIDIA H100", "NVIDIA B200"], - "entry_point": "main.py::run", - "dependencies": [], - "destination_passing_style": false - }, - "sources": [ - {"path": "main.py", "content": ""} - ] -} -``` - -### 4.1 Field semantics - -Full schema → [`reference/solution_schema.md`](reference/solution_schema.md). Quick reference: - -| Field | Notes | -|---|---| -| `name` | Solution identifier. Convention: `__`, e.g. `fa3_gqa_paged_decode_v1`, `sglang_mla_decode_v1` | -| `definition` | MUST exactly match the Definition `name` | -| `author` | Subdir name under `solutions/`. Pick a stable identifier for your team / lab (e.g. `acme-research`) | -| `spec.language` | `python` (most common) / `triton` / `cuda` / `tilelang` / `tvm_ffi` | -| `spec.target_hardware` | List of strings; framework checks current GPU is in this list before running | -| `spec.entry_point` | `::`; default is `main.py::run` | -| `spec.dependencies` | pip-installable packages required at runtime (e.g. `["flash-attn>=3.0.0", "fla-core"]`); leave `[]` if everything is vendored | -| `spec.destination_passing_style` | `true` if outputs are pre-allocated and passed in; `false` (default) if the entry function returns the outputs | -| `sources` | List of `{path, content}` dicts; include `main.py` plus any vendored `.py` / `.cu` / `.triton` files | - -### 4.2 Multi-file Solution - -For vendored kernels or multi-module wrappers, list every file in `sources`: - -```json -"sources": [ - {"path": "main.py", "content": "..."}, - {"path": "vendored_kernel.py", "content": "..."}, - {"path": "kernel.cu", "content": "..."} -] -``` - -Files are extracted to a temp dir at run time; relative imports work as expected. - -### 4.3 Tooling - -`flashinfer-bench solution build .` (in the solution dir) helps generate the JSON `sources` blob from on-disk files, but writing manually with `json.dumps` from a quick Python script also works. - ---- - -## Step 5 — Run benchmark with `flashinfer-bench run` - -### 5.1 Bare `flashinfer-bench run` command (inside the container) - -```bash -cd /home/scratch./kernel_arena/flashinfer-trace - -flashinfer-bench run \ - --local . \ - --definitions \ - --solutions [ ...] \ - --warmup-runs 5 \ - --iterations 20 \ - --num-trials 1 \ - --timeout 600 \ - --log-level INFO -``` - -### 5.1a Full `crun` invocation (computelab / SLURM cluster) — what to actually type - -On computelab and similar clusters, wrap the command in a shell script and -launch via `crun`. The script reproduces the exact path used to verify the -three reference Solutions: - -```bash -#!/bin/bash -# my_run.sh -set -e -LOGFILE=/home/yuny/kernel_arena/results/my_run_$(date +%Y%m%d_%H%M%S).log -mkdir -p $(dirname $LOGFILE) -exec > >(tee -a "$LOGFILE") 2>&1 - -echo "=== HOST: $(hostname) TIME: $(date) ===" - -# 1. Install dependency stack into ephemeral /tmp/pip-pkgs -PIP_TARGET=/tmp/pip-pkgs -mkdir -p $PIP_TARGET -pip install --target $PIP_TARGET --no-cache-dir \ - flashinfer-python==0.6.9 \ - flashinfer-bench \ - flash-linear-attention \ - "nvidia-cutlass-dsl[cu13]" \ - cuda-python \ - safetensors huggingface-hub 2>&1 | tail -3 -export PYTHONPATH=$PIP_TARGET:$PYTHONPATH -export PATH=$PIP_TARGET/bin:$PATH - -# 2. (Optional) Deploy your solution into the dataset before running -DATASET=/home/scratch.yuny_wwfo/kernel_arena/flashinfer-trace -DST=$DATASET/solutions/// -mkdir -p $DST -cp /home/yuny/kernel_arena/solutions//*.json $DST/ -cp /home/yuny/kernel_arena/solutions//main.py $DST/ - -# 3. Run the benchmark -cd $DATASET -flashinfer-bench run \ - --local . \ - --definitions \ - --solutions \ - --warmup-runs 5 \ - --iterations 20 \ - --num-trials 1 \ - --timeout 600 \ - --log-level INFO -``` - -Then submit: - -```bash -crun -q 'gpu.chip=gh100 and cpu.arch=x86_64' --gpus=1 -C \ - -img nvcr.io/nvidia/pytorch:24.10-py3 \ - -r /tmp /home/yuny/kernel_arena/scripts/my_run.sh -``` - -GPU chip query examples: -- `gpu.chip=gh100` → H100 (PCIe / SXM) -- `gpu.chip=gb200` → B200 -- `gpu.chip=ga100` → A100 - -**With a pre-built `.sqsh`** (Section 3.3): replace the `-img` URL with the -sqsh path, and **drop the pip-install block from the script** — deps are -already baked in: - -```bash -crun -q 'gpu.chip=gh100 and cpu.arch=x86_64' --gpus=1 -C \ - -img /home/scratch./containers/flashinfer-bench-runner.sqsh \ - -r /tmp /home/yuny/kernel_arena/scripts/my_run.sh -``` - -### 5.2 Key flags - -| Flag | Meaning | -|---|---| -| `--local ` | Point to a local trace dataset clone (not HF) | -| `--definitions` | Whitelist Definitions to run (space-separated; can also pass `--definitions all`) | -| `--solutions` | Whitelist Solutions to run; `baseline` (FlashInfer wrapper) typically runs as the comparison reference automatically | -| `--warmup-runs` | Forward passes before timing (default 10) | -| `--iterations` | Forward passes for timing each trial | -| `--num-trials` | Repeat trials, take median (default 3) | -| `--timeout` | Per-(solution × workload) wall time, seconds | -| `--log-level` | `DEBUG` / `INFO` / `WARNING` / `ERROR` | -| `--no-save-results` | Don't write traces (only print summary). Omit to write traces | - -### 5.3 What happens - -For each (Solution × Workload) pair the runner: -1. Loads input tensors from the workload `safetensors` -2. Builds a baseline output via Definition `reference.code` (PyTorch) -3. Calls the Solution's `entry_point` -4. Compares Solution output vs baseline (rtol=1e-2, atol=1e-2 by default → status PASSED / INCORRECT_NUMERICAL) -5. Times `iterations` warmup + measured passes → reports latency, ref_latency, speedup - -Result is a row in `traces///.jsonl`. - -### 5.4 Status enum - -| Status | Meaning | -|---|---| -| `PASSED` | Output within tolerance, latency measured | -| `INCORRECT_NUMERICAL` | rtol / atol exceeded | -| `RUNTIME_ERROR` | Exception during execution (true exception is hidden by worker isolation — see Common gotchas) | -| `TIMEOUT` | Exceeded `--timeout` | -| `SETUP_FAILED` | Failed to set up baseline / solution before measurement | - ---- - -## Step 6 — Inspect generated traces - -After `flashinfer-bench run`, check: - -```bash -TRACE=/home/scratch./kernel_arena/flashinfer-trace -find $TRACE/traces -name "*.jsonl" -newer $TRACE/run.log 2>/dev/null -# Or just list everything: -find $TRACE/traces -name "**.jsonl" -size +100c -``` - -### 6.1 Per-trace inspection -```bash -head -1 $TRACE/traces///.jsonl | python3 -m json.tool | head -40 -``` - -Key fields per trace row: -- `definition`, `solution`, `workload.uuid` -- `evaluation.status` (PASSED / RUNTIME_ERROR / …) -- `evaluation.environment.hardware` (e.g. `NVIDIA H100 PCIe`) -- `evaluation.performance.speedup_factor`, `latency_ms`, `reference_latency_ms` -- `evaluation.correctness.max_absolute_error`, `max_relative_error` - -### 6.2 Quick stats from CLI -```bash -flashinfer-bench report best --local . 2>&1 | head -40 -flashinfer-bench report summary --local . 2>&1 | head -40 -``` - -> Note: The `flashinfer-bench report visualize` CLI subcommand has a known issue in the current 0.6.x release where it treats Pydantic objects as dicts (`.get(...)`) and raises AttributeError. Use `report best` / `report summary` for CLI-level inspection, and the web UI (Step 7) for full visualization. - ---- - -## Step 7 — Visualize - -Two paths: - -### 7.1 Public site (read-only, official traces only) - -https://bench.flashinfer.ai - -This site is the production build of `flashinfer-bench/web/`. It shows the official `baseline` + `claude-…` / `gemini-…` / `gpt-…` author traces but NOT your local traces. It also has a `/viewer` page that accepts a single trace JSON as paste-in for inspection. - -### 7.2 Local web UI (sees your traces) - -The local Next.js dev server reads traces from the local `flashinfer-trace` clone: - -```bash -# Install pnpm (if not already; standalone binary is the simplest path) -curl -fsSL -o ~/bin/pnpm \ - https://github.com/pnpm/pnpm/releases/download/v9.15.0/pnpm-linuxstatic-x64 -chmod +x ~/bin/pnpm -export PATH=~/bin:$PATH - -# Configure pnpm to store packages on scratch (avoids home-quota issues) -mkdir -p /home/scratch./.pnpm-store -pnpm config set store-dir /home/scratch./.pnpm-store - -# Install web app dependencies -cd /home/scratch./kernel_arena/flashinfer-bench/web -pnpm install # ~2-3 minutes; produces ~1.9 GB node_modules - -# Point the data loader at your local dataset -export FIB_DATASET_PATH=/home/scratch./kernel_arena/flashinfer-trace -# Belt-and-suspenders: the data loader also looks at three fallback paths; -# the surest cover is to symlink one of the fallback locations: -ln -sfn $FIB_DATASET_PATH /tmp/flashinfer-trace -ln -sfn $FIB_DATASET_PATH /home/scratch./kernel_arena/flashinfer-bench/flashinfer_trace - -# Start the dev server (apps/web only; skip apps/docs to save resources) -cd apps/web -nohup pnpm dev > /tmp/dev.log 2>&1 & - -# Wait until ready (typically 60–90 s on a frontend-only machine) -until grep -q "Ready in" /tmp/dev.log; do sleep 2; done -echo "Open http://localhost:3000" -``` - -Notes on the web UI: -- Default port 3000 (web app), 3030 (docs) -- VS Code Remote-SSH auto-forwards port 3000 to local; otherwise set up an SSH tunnel -- **First page-load is slow** in dev mode — Next.js compiles each route on demand; `/` typically takes 200–400 s the first time, then is cached. Subsequent loads are seconds -- For a faster experience: `pnpm build && pnpm start` produces a production bundle (~10 min build, then sub-second navigation forever) - -### 7.3 Pages worth opening - -| Path | Contents | -|---|---| -| `http://localhost:3000` | Leaderboard / kernel list / model list (homepage) | -| `http://localhost:3000/kernels/` | Per-Definition page: Solutions table, fast_p curve over workloads, top-5 by speedup | -| `http://localhost:3000/models/` | Per-model coverage page (which Definitions belong to this model, status) | -| `http://localhost:3000/viewer` | Paste a single Trace JSON to inspect raw structure | - -### 7.4 Sharing traces with collaborators - -Traces are jsonl files in `traces///.jsonl`. To share: -- Copy the jsonl directly (one file per (author, op_type, def)) -- Open the public Viewer (https://bench.flashinfer.ai/viewer) and paste a single line of jsonl as a Trace JSON -- Or commit + push to a branch of `flashinfer-trace` and open a PR (then the public site updates after merge) - ---- - -## Common gotchas - -Curated from real run failures. Full list with diagnostics → [`reference/wrapper_gotchas.md`](reference/wrapper_gotchas.md). - -### Numerical / contract mismatches - -| Symptom | Likely cause | Fix | -|---|---|---| -| `INCORRECT_NUMERICAL` with `max_rel_error` ~ 1e0 | LSE base mismatch (natural log vs base-2) | Multiply natural-log LSE by `1 / math.log(2)` | -| `INCORRECT_NUMERICAL`, max_abs_error proportional to `1 / sqrt(d)` | sm_scale dropped, lib defaulted to `1 / sqrt(d)` instead of the supplied value | Pass `scale=sm_scale` (or `softmax_scale=`) explicitly | -| `INCORRECT_NUMERICAL` only on long-KV workloads | GQA expansion (`repeat_interleave`) mis-axis | Expand `H_kv → H_q` along the *head* axis only | -| Output values look like garbage | `transpose_state_layout=True` not set on FLA / k-first vs k-last layout | Match the Definition's KV layout flag | -| MLA wrapper passes correctness but speedup is < 1× | Kernel is fall-back path (e.g. FA3 ps=1 hits cp.async slow path) | Try ps=64 variant or different backend | - -### Runtime / framework - -| Symptom | Likely cause | Fix | -|---|---|---| -| `RUNTIME_ERROR` with no traceback shown | Worker subprocess swallowed the exception | Reproduce by `import` + calling the function directly outside `flashinfer-bench run` (skip the runner) | -| First Solution fails 3 times → all subsequent runs `SKIPPED` | `flashinfer-bench` skips Solutions with 3 consecutive failures | Fix the underlying issue or rerun targeting a different Solution; see PersistentRunner state caveats below | -| Second `flashinfer-bench run` re-uses stale baseline | PersistentRunner caches reference output keyed on (definition, workload); kill the persistent worker between runs | Pass `--runner=isolated` or restart the bench session | -| `ImportError` for vendored module despite being in `sources` | Path issue: framework extracts sources to a temp dir but `sys.path` excludes that dir | At top of `main.py`, do `sys.path.insert(0, os.path.dirname(__file__))` before importing vendored modules | - -### Environment / disk - -| Symptom | Likely cause | Fix | -|---|---|---| -| `git-lfs pull` fails or outputs only pointers | git-lfs not installed locally | `git-lfs install --local`, then re-pull | -| Disk-full mid-run | Home quota too small | Move repo + scripts + results to scratch (Section 3.3) | -| `ImportError: flashinfer_bench has no attribute apply` | Old pip cache with stale flashinfer-bench | `pip install --target /tmp/pip-pkgs --no-cache-dir flashinfer-bench` | - -### Hardware-specific - -| Symptom | Likely cause | Fix | -|---|---|---| -| `flashinfer.gdn` raises `'NoneType' object is not callable` (`run_pretranspose_decode is None`) on NGC PyTorch 24.10 | FlashInfer 0.6.x GDN uses CuTe DSL whose ABI doesn't match torch 2.11 in NGC 24.10 | Use FLA wrapper (Linear-Attention pattern) as the Solution; file an upstream issue if needed | -| FA3 wheel won't import (`undefined symbol`) | C++11 ABI mismatch between prebuilt wheel and PyTorch in container | Source-build FA3: `git clone Dao-AILab/flash-attention && cd hopper && python setup.py install` | -| Solution fails on B200 but works on H100 | `target_hardware` includes only H100 / wrapper hard-codes SM90 | Add `"NVIDIA B200"` to `spec.target_hardware`; verify cubin / kernel template covers SM100 | - ---- - -## Reference & templates - -### Reference docs (deep dives) - -- [`reference/definition_schema.md`](reference/definition_schema.md) — full Definition JSON schema with field semantics -- [`reference/solution_schema.md`](reference/solution_schema.md) — full Solution JSON schema -- [`reference/wrapper_gotchas.md`](reference/wrapper_gotchas.md) — extended troubleshooting list -- [`reference/visualization.md`](reference/visualization.md) — full web-UI / viewer setup notes including production build path - -### Wrapper templates (copy-paste starting points) - -- [`templates/dense_baseline_main.py`](templates/dense_baseline_main.py) — PyTorch SDPA-style dense baseline (paged GQA decode example) -- [`templates/dense_baseline_solution.json`](templates/dense_baseline_solution.json) — matching Solution JSON -- [`templates/linear_attention_main.py`](templates/linear_attention_main.py) — third-party Python lib wrapper (FLA-style; recurrent state, gated delta rule) -- [`templates/linear_attention_solution.json`](templates/linear_attention_solution.json) — matching Solution JSON -- [`templates/vendored_kernel_main.py`](templates/vendored_kernel_main.py) — vendored Triton kernel (SGLang-style; MLA decode with split-K + LSE reduction) -- [`templates/vendored_kernel_solution.json`](templates/vendored_kernel_solution.json) — matching Solution JSON - -Each template is heavily commented; the comments mark the lines you typically need to change for a new Definition. - ---- - -## Quick reference: end-to-end run script - -A complete script for running a single (Definition, Solution) pair from scratch: - -```bash -#!/bin/bash -set -e -USER_NAME=$(whoami) -SCRATCH=/home/scratch.${USER_NAME}_* -TRACE_ROOT=$SCRATCH/kernel_arena/flashinfer-trace -DEF=mla_paged_decode_h16_ckv512_kpe64_ps1 # ← Edit -OP=mla_paged # ← Edit -SOL=my_solution_v1 # ← Edit -AUTHOR=acme-research # ← Edit - -export PIP_TARGET=/tmp/pip-pkgs -mkdir -p "$PIP_TARGET" -export PYTHONPATH="$PIP_TARGET:$PYTHONPATH" -export PATH="$PIP_TARGET/bin:$PATH" -pip install --target "$PIP_TARGET" --no-cache-dir flashinfer-bench 2>&1 | tail -2 - -cd "$TRACE_ROOT" -git-lfs install --local -git-lfs pull --include="blob/workloads/$OP/$DEF/*" - -# (Solution files at solutions/$AUTHOR/$OP/$DEF/ are assumed already created) - -flashinfer-bench run \ - --local . \ - --definitions "$DEF" \ - --solutions "$SOL" \ - --warmup-runs 5 \ - --iterations 20 \ - --num-trials 1 \ - --timeout 600 \ - --log-level INFO - -# Inspect -find traces -name "*${DEF}*.jsonl" -size +100c | xargs wc -l -``` - -Save as `run_single_solution.sh`, set the four `← Edit` variables, and run. - ---- - -**END OF SKILL** diff --git a/skills/add-flashinfer-solution/reference/definition_schema.md b/skills/add-flashinfer-solution/reference/definition_schema.md deleted file mode 100644 index 2040ae42d652b56cd2e51acd456bd61a6473e39d..0000000000000000000000000000000000000000 --- a/skills/add-flashinfer-solution/reference/definition_schema.md +++ /dev/null @@ -1,198 +0,0 @@ -# Definition JSON Schema (full reference) - -Each Definition JSON file lives at `definitions//.json` in the `flashinfer-trace` repo. It declares a kernel's parameter space and the reference (PyTorch) implementation used for correctness checking. - -## Top-level schema - -```json -{ - "name": "", - "description": "", - "axes": { ... }, - "inputs": [ ... ], - "outputs": [ ... ], - "constraints": [ ... ], - "tags": [ ... ], - "reference": { ... } -} -``` - -| Field | Type | Required | Notes | -|---|---|---|---| -| `name` | string | yes | MUST equal filename stem (no `.json`) | -| `description` | string | yes | One-line natural language description: what the op does, source model, deployment context | -| `axes` | object | yes | Map of axis_name → axis spec | -| `inputs` | list of objects | yes | Input tensor declarations | -| `outputs` | list of objects | yes | Output tensor declarations | -| `constraints` | list of strings | optional | Inter-axis constraints (natural language or symbolic) | -| `tags` | list of strings | yes | Metadata; see Tag taxonomy below | -| `reference` | object | yes | PyTorch reference impl; sees `code` field | - -## `axes` field - -```json -"axes": { - "batch_size": {"type": "var", "description": "Number of sequences"}, - "num_qo_heads": {"type": "const", "value": 16, "description": "Query/output heads"}, - "num_kv_heads": {"type": "const", "value": 1, "description": "Grouped KV heads"}, - "head_dim": {"type": "const", "value": 128}, - "page_size": {"type": "const", "value": 64}, - "num_pages": {"type": "var"}, - "kv_seqlen": {"type": "var", "description": "KV cache length per sequence"}, - ... -} -``` - -Per-axis sub-schema: - -| Sub-field | Type | When | Notes | -|---|---|---|---| -| `type` | `"const"` or `"var"` | always | `const` = compile-time fixed (e.g. head_dim, sm_arch), captured in the def name; `var` = workload-time variable (batch, seq_len) | -| `value` | int / float / string | when `type=="const"` | The fixed value | -| `description` | string | optional | Human-readable purpose | - -## `inputs` and `outputs` fields - -Both are lists of tensor specs: - -```json -"inputs": [ - { - "name": "q", - "dtype": "bfloat16", - "shape": ["batch_size", "num_qo_heads", "head_dim"] - }, - { - "name": "k_cache", - "dtype": "bfloat16", - "shape": ["num_pages", "page_size", "num_kv_heads", "head_dim"] - }, - { - "name": "kv_indptr", - "dtype": "int32", - "shape": ["len_indptr"] - }, - { - "name": "sm_scale", - "dtype": "float32", - "shape": [] // scalar - } -] -``` - -Per-tensor sub-schema: - -| Sub-field | Notes | -|---|---| -| `name` | Variable name in the reference code AND wrapper signature | -| `dtype` | Strings: `bfloat16`, `float16`, `float32`, `int32`, `int64`, `fp8_e4m3fn`, `fp8_e5m2`, … | -| `shape` | List of strings; each entry is either an axis name (resolved at runtime) or a literal integer; empty `[]` for scalars | - -Special note: the input order is the function-signature order. Some tensor inputs may be passed via the workload `safetensors` `tensor_key` field; consult `reference.code` for the actual call. - -## `constraints` field - -List of natural-language strings checked at runtime by the runner before launching: - -```json -"constraints": [ - "len_indptr == batch_size + 1", - "num_kv_indices == kv_indptr[-1].item()" -] -``` - -These are not parsed; they are advisory comments. The runner doesn't enforce them — it's the workload generator's responsibility to satisfy them. - -## `tags` field — taxonomy - -Tags are strings of the form `:`: - -| Prefix | Allowed values | Purpose | -|---|---|---| -| `stage:` | `decode`, `prefill`, `mtp`, `sparse_attention`, `topk_indexer` | Inference stage | -| `status:` | `verified`, `unverified`, `reference` | Curation status (verified = was actually produced by an inference run with confirmed correctness) | -| `model:` | `deepseek-v3`, `deepseek-r1`, `qwen3-235b`, `llama-3.1-70b`, `llama-3.2-3b`, `gemma-3-27b`, `kimi-k2`, `qwen3-next`, `nemotron-h-8b`, ... | Source model(s); a Definition may serve multiple models | -| `fi_api:` | e.g. `flashinfer.mla.BatchMLAPagedAttentionWrapper`, `flashinfer.decode.BatchDecodeWithPagedKVCacheWrapper` | The FlashInfer Python API the def is modeled after | -| `tp:` | `1`, `2`, `4`, `8`, `16` | Assumed tensor-parallel slicing | - -Additional tags occasionally seen but not formally part of the taxonomy: `quant:`, `fi_module:`, etc. These are advisory. - -## `reference` field - -Embeds the PyTorch reference implementation as a string of source code: - -```json -"reference": { - "code": "import torch\n\ndef run(q, k_cache, v_cache, kv_indptr, kv_indices, sm_scale):\n ...\n return out, lse\n", - "language": "python" -} -``` - -| Sub-field | Notes | -|---|---| -| `code` | Full Python source defining `run(...)` (or whatever the entry point is). Must be self-contained — runner will exec it in an isolated context | -| `language` | Currently only `python` is used | - -Convention: the reference function is named `run`. Its signature is the canonical contract for any wrapper Solution. - -## Worked example: `mla_paged_decode_h16_ckv512_kpe64_ps1` - -```json -{ - "name": "mla_paged_decode_h16_ckv512_kpe64_ps1", - "description": "Batched Multi-head Latent Attention decode with a paged KV cache. Captured from DeepSeek V3 at TP=8.", - "axes": { - "batch_size": {"type": "var"}, - "num_qo_heads": {"type": "const", "value": 16}, - "ckv_dim": {"type": "const", "value": 512}, - "kpe_dim": {"type": "const", "value": 64}, - "page_size": {"type": "const", "value": 1}, - "num_pages": {"type": "var"}, - "kv_seqlen": {"type": "var"} - }, - "inputs": [ - {"name": "q_nope", "dtype": "bfloat16", "shape": ["batch_size", "num_qo_heads", "ckv_dim"]}, - {"name": "q_pe", "dtype": "bfloat16", "shape": ["batch_size", "num_qo_heads", "kpe_dim"]}, - {"name": "ckv_cache", "dtype": "bfloat16", "shape": ["num_pages", "page_size", "ckv_dim"]}, - {"name": "kpe_cache", "dtype": "bfloat16", "shape": ["num_pages", "page_size", "kpe_dim"]}, - {"name": "kv_indptr", "dtype": "int32", "shape": ["len_indptr"]}, - {"name": "kv_indices", "dtype": "int32", "shape": ["num_kv_indices"]}, - {"name": "sm_scale", "dtype": "float32", "shape": []} - ], - "outputs": [ - {"name": "output", "dtype": "bfloat16", "shape": ["batch_size", "num_qo_heads", "ckv_dim"]}, - {"name": "lse", "dtype": "float32", "shape": ["batch_size", "num_qo_heads"]} - ], - "constraints": [ - "len_indptr == batch_size + 1", - "num_kv_indices == kv_indptr[-1].item()" - ], - "tags": [ - "stage:decode", - "status:verified", - "model:deepseek-v3", - "model:deepseek-r1", - "fi_api:flashinfer.mla.BatchMLAPagedAttentionWrapper", - "tp:8" - ], - "reference": { - "language": "python", - "code": "import torch\n\ndef run(q_nope, q_pe, ckv_cache, kpe_cache, kv_indptr, kv_indices, sm_scale):\n ..." - } -} -``` - ---- - -## Quick lookup: where to find what when writing a Solution - -| Question while writing wrapper | Look in Definition JSON at | -|---|---| -| What are the function args? | `inputs[].name` (in order) | -| What dtype must I produce as output? | `outputs[].dtype` | -| What's the LSE shape? | `outputs[]` where `name == "lse"` | -| Is sm_scale a scalar I receive, or do I compute it? | `inputs[]` — if `sm_scale` is in inputs, USE it; do NOT recompute | -| What's the KV layout? | `inputs[].shape` for the cache tensors (paged: `[P, ps, H_kv, D]`; ragged: `[total, H_kv, D]`; MLA paged: `[P, ps, ckv_dim]` + `[P, ps, kpe_dim]`) | -| Does causal mask apply? | Definition `name` contains `causal` or `description` mentions it | -| Should I expand GQA `H_kv → H_q`? | Yes if your underlying lib only supports MHA; check `num_qo_heads` vs `num_kv_heads` axes | -| What does the reference impl look like? | `reference.code` — read it to understand exact semantics | diff --git a/skills/add-flashinfer-solution/reference/solution_schema.md b/skills/add-flashinfer-solution/reference/solution_schema.md deleted file mode 100644 index a8924ad30181dda399f5798dc966575a546a1117..0000000000000000000000000000000000000000 --- a/skills/add-flashinfer-solution/reference/solution_schema.md +++ /dev/null @@ -1,195 +0,0 @@ -# Solution JSON Schema (full reference) - -Each Solution JSON file lives at `solutions////.json`. It declares one implementation of a Definition, including its source code (embedded in the JSON) and runtime spec. - -## Top-level schema - -```json -{ - "name": "", - "definition": "", - "author": "", - "spec": { ... }, - "sources": [ ... ] -} -``` - -| Field | Type | Required | Notes | -|---|---|---|---| -| `name` | string | yes | Solution identifier (unique within `///`) | -| `definition` | string | yes | MUST equal a Definition `name` exactly | -| `author` | string | yes | Subdir name under `solutions/`; pick one stable identifier per team / lab | -| `spec` | object | yes | Runtime metadata | -| `sources` | list of objects | yes | Embedded source code files | - -## `name` conventions - -Common patterns observed in the public dataset: - -| Pattern | Used when | -|---|---| -| `flashinfer_wrapper_<6hex>` | Reserved for the official `baseline` author — calls FlashInfer Python API directly | -| `__v` | Hand-written wrapper around a third-party lib, e.g. `fa3_gqa_paged_decode_v1`, `sglang_mla_decode_v1`, `fla_gdn_decode_v1` | -| `__<6hex>` | LLM-generated implementations, e.g. `gpt-5_cuda_5eb89c`, `claude-opus-4-1_triton_a98005` | - -Pick a name pattern that conveys lib + op + variant; bump version (`v2`) when changing the underlying impl meaningfully. - -## `spec` field - -```json -"spec": { - "language": "python", - "target_hardware": ["NVIDIA H100", "NVIDIA B200"], - "entry_point": "main.py::run", - "dependencies": ["flash-attn>=3.0.0"], - "destination_passing_style": false -} -``` - -| Sub-field | Type | Notes | -|---|---|---| -| `language` | string | `python` (most common, even when wrapping CUDA/Triton) / `triton` / `cuda` / `tilelang` / `tvm_ffi` | -| `target_hardware` | list of strings | Allowed GPU arches; runner refuses to run on unlisted hardware. Common values: `"NVIDIA H100"`, `"NVIDIA H200"`, `"NVIDIA B200"`, `"NVIDIA L40"`, `"NVIDIA A100"` | -| `entry_point` | string | `::`; default convention is `main.py::run` | -| `dependencies` | list of strings | Pip-installable packages required at runtime. Empty list `[]` if all needed code is vendored in `sources` | -| `destination_passing_style` | bool | `false` (default): entry function returns a tuple of output tensors. `true`: outputs are pre-allocated and passed in as additional args; entry function modifies in place and returns nothing | - -### Notes on each sub-field - -#### `language` -The runner uses this only for documentation; the actual entry point loading uses `entry_point` and Python's import mechanism. Even Triton / CUDA solutions usually have `language: python` because their `main.py` is a Python wrapper that imports / launches the underlying kernel. - -#### `target_hardware` -The runner introspects current GPU via `torch.cuda.get_device_name(0)`. Examples of returned strings: -- `"NVIDIA H100 PCIe"` — note: PCIe vs SXM both report `H100`-prefixed strings; matching is by prefix -- `"NVIDIA H100 80GB HBM3"` -- `"NVIDIA B200"` - -Use `["NVIDIA H100", "NVIDIA B200"]` to allow both Hopper and Blackwell. - -#### `entry_point` -Format: `::`. The function is loaded via `importlib`. Default: `main.py::run`. - -#### `dependencies` -Listed packages are NOT auto-installed by the runner; they must be present in the Python environment beforehand. Use `pip install --target /tmp/pip-pkgs ` and `export PYTHONPATH=/tmp/pip-pkgs:$PYTHONPATH` for ephemeral envs. - -For a fully self-contained Solution, leave `dependencies: []` and embed all required source files into `sources` (vendor the kernel). - -#### `destination_passing_style` -- `false` (default): - ```python - def run(q, k_cache, v_cache, ..., sm_scale): - out = ... - lse = ... - return out, lse - ``` -- `true`: - ```python - def run(q, k_cache, v_cache, ..., sm_scale, output, lse): - output.copy_(...) - lse.copy_(...) - ``` - -Use `true` when working with kernels that strictly require pre-allocated output (e.g. some Triton kernels). Most wrappers prefer `false` for simplicity. - -## `sources` field - -Embeds the entire source tree of the Solution as a JSON array: - -```json -"sources": [ - {"path": "main.py", "content": ""}, - {"path": "kernel.triton", "content": "..."}, - {"path": "vendored.py", "content": "..."} -] -``` - -| Sub-field | Notes | -|---|---| -| `path` | Relative to the Solution's directory; use forward slashes for nested files | -| `content` | Raw file content as a UTF-8 string (escape backslashes and newlines per JSON spec) | - -The runner extracts `sources` to a temp dir at run time; the entry function is loaded from there. Relative imports between files in `sources` work as expected if `main.py` includes: - -```python -import os, sys -_HERE = os.path.dirname(os.path.abspath(__file__)) -if _HERE not in sys.path: - sys.path.insert(0, _HERE) -``` - -## Worked example: minimal Solution - -For a Solution at `solutions/acme-research/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps1/my_fa3_v1.json`: - -```json -{ - "name": "my_fa3_v1", - "definition": "gqa_paged_decode_h32_kv8_d128_ps1", - "author": "acme-research", - "spec": { - "language": "python", - "target_hardware": ["NVIDIA H100", "NVIDIA B200"], - "entry_point": "main.py::run", - "dependencies": ["flash-attn>=3.0.0"], - "destination_passing_style": false - }, - "sources": [ - { - "path": "main.py", - "content": "import math\nimport torch\nfrom flash_attn_3 import flash_attn_with_kvcache\n\n\ndef run(q, k_cache, v_cache, kv_indptr, kv_indices, sm_scale):\n # ... wrapper body ...\n return out, lse\n" - } - ] -} -``` - -## Building a Solution JSON from on-disk files - -There's no standalone CLI to do this in the public 0.6.x flashinfer-bench, so a tiny Python script is fine: - -```python -import json, os -from pathlib import Path - -solution_dir = Path("solutions/acme-research/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps1") -out_json = solution_dir / "my_fa3_v1.json" - -sources = [] -for p in sorted(solution_dir.glob("**/*")): - if p.is_file() and p.name != out_json.name and p.suffix in (".py", ".cu", ".triton", ".cpp", ".h"): - sources.append({ - "path": str(p.relative_to(solution_dir)), - "content": p.read_text(), - }) - -doc = { - "name": "my_fa3_v1", - "definition": "gqa_paged_decode_h32_kv8_d128_ps1", - "author": "acme-research", - "spec": { - "language": "python", - "target_hardware": ["NVIDIA H100", "NVIDIA B200"], - "entry_point": "main.py::run", - "dependencies": ["flash-attn>=3.0.0"], - "destination_passing_style": False, - }, - "sources": sources, -} - -out_json.write_text(json.dumps(doc, indent=2)) -``` - -Run from the directory containing the Solution sources; produces a JSON ready for the runner. - -## Field-level checklist before committing a Solution JSON - -- [ ] `name` is unique within the target def directory -- [ ] `definition` matches the actual Definition name exactly (no typos, no prefix/suffix drift) -- [ ] `author` matches an existing author dir (else create one) -- [ ] `spec.target_hardware` includes the hardware you actually tested on -- [ ] `spec.entry_point` resolves correctly (function exists in the named file) -- [ ] `spec.dependencies` list is honest — every `import` in `main.py` is either in `dependencies` or in `sources` -- [ ] `spec.destination_passing_style` matches the actual entry signature -- [ ] `sources` contains every file the entry function imports (use AST or `grep -E "^(import|from)"` to audit) -- [ ] `sources[].content` is valid UTF-8 and parses as the right language (`python -c "exec(open(p).read())"` for Python; CUDA / Triton parsing requires the relevant compiler) -- [ ] No absolute paths in source content — use `os.path.dirname(__file__)` for relative loads diff --git a/skills/add-flashinfer-solution/reference/visualization.md b/skills/add-flashinfer-solution/reference/visualization.md deleted file mode 100644 index d05cf14572ba63452a94ada8b1d9fbef63e5e42a..0000000000000000000000000000000000000000 --- a/skills/add-flashinfer-solution/reference/visualization.md +++ /dev/null @@ -1,207 +0,0 @@ -# Visualization Reference - -Two paths: public site (read-only, official traces only) and local web UI (sees your local traces, full functionality). - ---- - -## 1. Public site - -URL: **https://bench.flashinfer.ai** - -Reads the published `flashinfer-trace` HF dataset directly. Shows traces from authors that have been merged upstream: -- `baseline` (FlashInfer official) -- `claude-opus-4-1-20250805`, `gemini-2.5-pro`, `gpt-5-2025-08-07`, `gpt-o3` (LLM-generated) -- ... plus any author whose contributions land on the dataset's `main` branch - -### Pages -- `/` — homepage with three sections: Leaderboard / Models / Kernels -- `/kernels/` — per-Definition page: Solutions table, fast_p curve over all workloads, top-5 by speedup -- `/models/` — per-model coverage: which Definitions belong, status -- `/viewer` — paste a single Trace JSON to inspect the raw structure - -### `/viewer` page details -Accepts a **single** Trace JSON object. Most jsonl files have many lines (one Trace each); take just one line: - -```bash -sed -n '1p' traces///.jsonl > /tmp/single.json -# Then copy /tmp/single.json contents into the textarea on /viewer -``` - -Pasting the whole jsonl raises `Unexpected non-whitespace character after JSON at position …`. - ---- - -## 2. Local web UI (recommended for local development) - -The web UI is the production source of bench.flashinfer.ai, vendored under `flashinfer-bench/web/`. Running it locally lets you visualize traces from a local `flashinfer-trace` clone, including authors that haven't been merged upstream. - -### 2.1 Setup - -```bash -# 1. Get pnpm (the project uses pnpm workspaces) -mkdir -p ~/bin -curl -fsSL -o ~/bin/pnpm \ - https://github.com/pnpm/pnpm/releases/download/v9.15.0/pnpm-linuxstatic-x64 -chmod +x ~/bin/pnpm -export PATH=~/bin:$PATH - -# 2. Configure pnpm to keep its store on scratch (~ 1 GB cache, plus 1.9 GB node_modules) -SCRATCH=/home/scratch. -mkdir -p $SCRATCH/.pnpm-store -pnpm config set store-dir $SCRATCH/.pnpm-store - -# 3. Clone flashinfer-bench codebase (if not already) -cd $SCRATCH/kernel_arena -git clone https://github.com/flashinfer-ai/flashinfer-bench.git - -# 4. Install web app dependencies (~2-3 min) -cd $SCRATCH/kernel_arena/flashinfer-bench/web -pnpm install -``` - -### 2.2 Point data loader at your local dataset - -The web app's data loader resolves the trace dataset path in this priority order (`apps/web/lib/data-loader.ts`): -1. `FLASHINFER_TRACE_PATH` env var (explicit override) -2. `FIB_DATASET_PATH` env var (set by the prebuild script) -3. Local repo's `flashinfer_trace/` dir (`web/apps/web` → `../../../flashinfer_trace`) -4. `/tmp/flashinfer-trace` (fallback) - -Belt-and-suspenders: set the env var AND symlink at the fallback paths: - -```bash -export FIB_DATASET_PATH=$SCRATCH/kernel_arena/flashinfer-trace -ln -sfn $FIB_DATASET_PATH /tmp/flashinfer-trace -ln -sfn $FIB_DATASET_PATH $SCRATCH/kernel_arena/flashinfer-bench/flashinfer_trace -``` - -### 2.3 Start the dev server - -```bash -cd $SCRATCH/kernel_arena/flashinfer-bench/web/apps/web -nohup pnpm dev > /tmp/dev.log 2>&1 & - -# Wait until ready (~60-90 s on a frontend-only host) -until grep -q "Ready in" /tmp/dev.log; do sleep 2; done -echo "Open http://localhost:3000" -``` - -The dev server runs only `apps/web` (skip `apps/docs` to save resources). - -### 2.4 Caveats of dev mode - -- **First page load is slow** — Next.js compiles each route on demand. The first `GET /` typically takes 200–400 s; the home page SSR fetches all 137+ Definitions × N solutions per Definition for the leaderboard. Subsequent loads are 5–20 s -- **Build process competing for CPU**: avoid running `pnpm build` in parallel with `pnpm dev`; it will stretch compile time to many minutes -- **Browser shows endless spinner**: usually first-load compilation. Check `tail -f /tmp/dev.log`; once you see `GET / 200 in ` the page is ready - -### 2.5 Production build (recommended for sustained use) - -For more than ~30 minutes of use, build once and run a production server: - -```bash -cd $SCRATCH/kernel_arena/flashinfer-bench/web/apps/web - -# Build (takes 8–15 min on a typical frontend host; produces .next/ ~1 GB) -pnpm build - -# Start production server (instant; pages are SSG / ISR cached) -nohup pnpm start > /tmp/prod.log 2>&1 & -``` - -Production mode: -- Pages render in <500 ms (vs 200+ s in dev) -- Hot-reload on file changes is OFF; rebuild required after data changes (re-run `pnpm build`) - -### 2.6 Page reference - -| Path | Use | -|---|---| -| `http://localhost:3000` | Leaderboard table + fast_p curve + Models grid + Kernels list | -| `http://localhost:3000/kernels/` | Per-Definition Solutions table + fast_p curve over workloads + top-5 by speedup | -| `http://localhost:3000/models/` | Per-model coverage page | -| `http://localhost:3000/viewer` | Paste single Trace JSON for raw inspection | -| `http://localhost:3000/docs/api/python/` | Embedded Sphinx docs of `flashinfer_bench` Python API | - ---- - -## 3. SSH port forwarding - -If the dev server runs on a remote machine and you browse from your laptop: - -### 3.1 VS Code Remote-SSH (easiest) -- VS Code auto-detects port 3000 and forwards it to localhost on your laptop -- Look at the **PORTS** tab at the bottom; if not auto-forwarded, click `Forward a Port` and enter `3000` - -### 3.2 Manual SSH tunnel -```bash -# From your laptop: -ssh -L 3000:localhost:3000 @ -# Then open http://localhost:3000 in your browser -``` - -For HTTP keep-alive across the tunnel: -```bash -ssh -L 3000:localhost:3000 -o ServerAliveInterval=60 -o ServerAliveCountMax=3 @ -``` - ---- - -## 4. Sharing a trace with collaborators (no full UI required) - -Three options: - -### 4.1 Send the jsonl file -```bash -# One file per (author × op_type × def_name); typically 50-200 KB -scp $TRACE_ROOT/traces///.jsonl reviewer@host:/tmp/ -# Reviewer runs the local web UI or pastes individual rows into /viewer -``` - -### 4.2 Use the public Viewer -Take a single Trace JSON line and paste into https://bench.flashinfer.ai/viewer. Reviewer gets the same structured inspection without setup. - -```bash -sed -n 'p' $TRACE_ROOT/traces///.jsonl > /tmp/single_trace.json -# Open the file, copy contents, paste at https://bench.flashinfer.ai/viewer -``` - -### 4.3 Open a PR to flashinfer-trace -Once the Solution is merged on `main`, the public site picks it up automatically (next dataset rebuild). Workflow: -- Fork `huggingface.co/datasets/flashinfer-ai/flashinfer-trace` -- Add the Solution JSON, run benchmark, commit produced traces -- Open a PR; maintainers review - ---- - -## 5. Programmatic access to traces - -If you want to build custom dashboards, just read the jsonl directly: - -```python -import json -from pathlib import Path - -trace_root = Path("/home/scratch./kernel_arena/flashinfer-trace") -trace_file = trace_root / "traces///.jsonl" - -traces = [json.loads(l) for l in trace_file.read_text().splitlines()] -for t in traces: - print( - t["definition"], t["solution"], - t["evaluation"]["status"], - t["evaluation"].get("performance", {}).get("speedup_factor"), - ) -``` - -Or use the framework's helper: - -```python -from flashinfer_bench.data import TraceSet -ts = TraceSet.from_path(trace_root) -for def_name in ts.definitions: - best = ts.get_best_trace(def_name) - if best: - print(def_name, "→", best.solution, best.evaluation.performance.speedup_factor) -``` - -`TraceSet` provides filter / best / score queries; see `flashinfer_bench/data/trace_set.py` for the full API. diff --git a/skills/add-flashinfer-solution/reference/wrapper_gotchas.md b/skills/add-flashinfer-solution/reference/wrapper_gotchas.md deleted file mode 100644 index 5fd4cb578f90555a7d981c0156e5bab1a5336ce8..0000000000000000000000000000000000000000 --- a/skills/add-flashinfer-solution/reference/wrapper_gotchas.md +++ /dev/null @@ -1,243 +0,0 @@ -# Wrapper Gotchas (extended troubleshooting) - -Catalogued issues that have surfaced when wrapping kernels for `flashinfer-bench`. Each entry has: symptom → root cause → fix → diagnostic command. - ---- - -## 1. Numerical / contract mismatches - -### 1.1 LSE base mismatch (natural log vs base-2) - -- **Symptom**: `INCORRECT_NUMERICAL`, `max_relative_error` ≈ `0.301` (i.e. `log10(2)`) on the LSE output specifically; output tensor itself is correct -- **Root cause**: FlashInfer uses **base-2** LSE by convention (`lse = log2(sum_exp(logits * log2(e)))`). Many third-party libs (Flash Attention, naive Triton) emit **natural-log** LSE -- **Fix**: in the wrapper, convert before returning: - ```python - import math - lse_b2 = lse_natural * (1.0 / math.log(2.0)) - ``` -- **Diagnostic**: print both LSEs side-by-side; ratio should be exactly `log_2(e)` ≈ `1.4427` - -### 1.2 Lost / overridden `sm_scale` - -- **Symptom**: `INCORRECT_NUMERICAL` proportional to `1/sqrt(d)`; output is "almost right" but uniformly off -- **Root cause**: third-party API was called without explicitly passing the supplied `sm_scale`, so the lib defaulted to `1/sqrt(head_dim)`. FlashInfer Definitions sometimes use a different scale (e.g. for absorbed MLA prefill the scale includes the rope-dim mixing factor) -- **Fix**: always pass `sm_scale` (or whatever the lib calls it: `softmax_scale`, `scale`, `sm_scale_for_q`, …) explicitly: - ```python - out = F.scaled_dot_product_attention(q, k, v, scale=sm_scale) - out = flash_attn_with_kvcache(..., softmax_scale=sm_scale, ...) - ``` -- **Diagnostic**: temporarily set `sm_scale = 1.0` in the workload; if the wrapper still produces non-trivially scaled output, scale is being silently overridden somewhere - -### 1.3 GQA expansion along wrong axis - -- **Symptom**: `INCORRECT_NUMERICAL` only on long-KV workloads; short-KV passes -- **Root cause**: `repeat_interleave` to expand `H_kv → H_q` was applied along the wrong axis. Common confusion: `[B, S, H_kv, D]` vs `[B, H_kv, S, D]` -- **Fix**: always check the target lib's expected layout, then expand exactly along the head axis: - ```python - # If layout is [B, S, H, D]: - k = k.repeat_interleave(H_q // H_kv, dim=2) # not dim=1! - # If layout is [B, H, S, D]: - k = k.repeat_interleave(H_q // H_kv, dim=1) - ``` -- **Diagnostic**: print `k.shape` vs `q.shape` after expansion — head dims must match - -### 1.4 KV layout direction (k-first vs k-last) for recurrent state kernels - -- **Symptom**: Output garbage — recurrent state values look uniformly wrong, not just numerically off -- **Root cause**: FLA-family kernels accept state as `[B, HV, K, V]` (k-first, default) or `[B, HV, V, K]` (k-last). Many Definition specs use k-last layout (Qwen3-Next GDN), but FLA's default is k-first -- **Fix**: pass `transpose_state_layout=True` to the FLA op explicitly when the Definition uses k-last: - ```python - o, new_state = fused_recurrent_gated_delta_rule( - ..., transpose_state_layout=True - ) - ``` -- **Diagnostic**: check the Definition `inputs[]` shape for the state tensor; the last two dim names will tell you (e.g. `[batch, head, k_dim, v_dim]` = k-first) - -### 1.5 Custom gating not pre-applied - -- **Symptom**: Output bias tracks closer to zero than expected -- **Root cause**: Some lib ops require a pre-computed gating value (e.g. `beta = sigmoid(b)` for FLA gated_delta_rule), but Definitions hand you the raw input -- **Fix**: do the activation in the wrapper: - ```python - beta = torch.sigmoid(b) - out, _ = fused_recurrent_gated_delta_rule(..., beta=beta, ...) - ``` -- **Diagnostic**: read both the Definition `reference.code` AND the lib's docstring; differences in gating semantics are a common source of confusion - -### 1.6 Double L2-norm - -- **Symptom**: Output magnitude consistently ~half of reference -- **Root cause**: Definition's `reference.code` already includes L2 norm, AND the lib's `use_qk_l2norm_in_kernel=True` flag was passed → applied twice -- **Fix**: confirm with one source of truth; either: - ```python - # Lib does the norm — don't pre-norm in wrapper - out = lib_op(q_raw, k_raw, ..., use_qk_l2norm_in_kernel=True) - # OR pre-norm in wrapper, lib doesn't - q_normed = F.normalize(q_raw, dim=-1) - k_normed = F.normalize(k_raw, dim=-1) - out = lib_op(q_normed, k_normed, ..., use_qk_l2norm_in_kernel=False) - ``` -- **Diagnostic**: compare `out.abs().mean()` with reference; ratio of ~0.5 is the smoking gun - -### 1.7 Q tensor extra dim (4D vs 3D) - -- **Symptom**: shape error `Expected 3-d but got 4-d` or vice versa -- **Root cause**: FlashInfer's decode wrappers use 3-D Q (`[B, H, D]`); FA3 expects 4-D Q (`[B, S=1, H, D]`). The wrapper needs to add or remove that S dim -- **Fix**: - ```python - q_4d = q.unsqueeze(1) # [B, H, D] -> [B, 1, H, D] - out_4d = flash_attn_with_kvcache(q_4d, ...) - out = out_4d.squeeze(1) # [B, 1, H, D] -> [B, H, D] - ``` - ---- - -## 2. Runtime / framework - -### 2.1 Worker subprocess swallowed traceback - -- **Symptom**: Trace status `RUNTIME_ERROR` with no useful error message in logs (`--log-level DEBUG` shows the same) -- **Root cause**: `flashinfer-bench` runs each Solution in an isolated subprocess; uncaught exceptions become a status enum, the actual traceback is lost -- **Fix (diagnostic only)**: reproduce outside the runner. Write a minimal script that imports the wrapper directly and calls it with manually-loaded workload tensors: - ```python - # 17_direct_repro.py - import torch - from safetensors.torch import load_file - import sys; sys.path.insert(0, "solutions/acme/op/def_name") - from main import run - - inputs = load_file("blob/workloads/op/def_name/def_name_.safetensors") - q = inputs["q"]; k = inputs["k_cache"]; ... - out = run(q, k, ..., sm_scale=0.08838) - print(out) - ``` - Run with `python -u 17_direct_repro.py` to see the real traceback. - -### 2.2 Three-strikes Solution skip - -- **Symptom**: Mid-way through a long run, you see `Skipping solution X due to 3 consecutive failures` -- **Root cause**: `flashinfer-bench` short-circuits a Solution after 3 consecutive errors, treating it as broken — this affects all subsequent workloads for that Solution -- **Fix**: address the root error first; then either re-run with a smaller `--num-trials` to confirm, or delete the Solution's traces and re-run -- **Diagnostic**: count `RUNTIME_ERROR` rows in the trace jsonl per Solution; if the first 3 fail, expect skip thereafter - -### 2.3 PersistentRunner stale baseline cache - -- **Symptom**: Re-running with a corrected Solution still shows the same speedup numbers -- **Root cause**: `PersistentRunner` (default for `flashinfer-bench run`) caches the reference baseline output keyed on `(definition, workload.uuid)`. If the worker subprocess crashed mid-run, that cache might be stale -- **Fix**: explicitly use the isolated runner: - ```bash - flashinfer-bench run --runner=isolated ... - ``` - Or simply delete the trace file for that solution before re-running (so cached comparisons are regenerated) - -### 2.4 Vendored module import error - -- **Symptom**: `ModuleNotFoundError: No module named 'sglang_decode'` despite the file being in `sources` -- **Root cause**: When the runner extracts `sources` to a temp dir, that dir is not on `sys.path` -- **Fix**: at the top of `main.py`, add: - ```python - import os, sys - _HERE = os.path.dirname(os.path.abspath(__file__)) - if _HERE not in sys.path: - sys.path.insert(0, _HERE) - from vendored_module import ... - ``` - -### 2.5 `flashinfer-bench report visualize` AttributeError - -- **Symptom**: Calling `flashinfer-bench report visualize --local .` raises `AttributeError: 'Evaluation' object has no attribute 'get'` -- **Root cause**: A bug in the public 0.6.x release: `cli/main.py:202` calls `trace.evaluation.get("status", ...)` but `trace.evaluation` is a Pydantic object, not a dict -- **Workaround**: skip the CLI `visualize` subcommand; use `report best` / `report summary` for CLI-level inspection, and use the local web UI (Step 7 in the main SKILL doc) for the full visualization - -### 2.6 No `--max-workloads` flag - -- **Symptom**: You want to run only the first N workloads of a large Definition for quick iteration; flag does not exist -- **Workaround**: temp-trim the workloads jsonl: - ```bash - WL=workloads/op/def.jsonl - cp "$WL" "$WL.bak" - head -10 "$WL.bak" > "$WL" - flashinfer-bench run ... - mv "$WL.bak" "$WL" - ``` - Use a `bash trap "mv \"$WL.bak\" \"$WL\"" EXIT` to restore on script exit (including failure). - ---- - -## 3. Environment / disk - -### 3.1 git-lfs returns pointer files instead of tensors - -- **Symptom**: `safetensors.SafetensorError: invalid header` when loading a workload tensor -- **Root cause**: `git-lfs install --local` was not run; the file is the LFS pointer, not the actual tensor -- **Fix**: - ```bash - cd flashinfer-trace - git-lfs install --local - git-lfs pull --include="blob/workloads///*" - ``` - Verify: the file size should be > 1KB (pointer files are ~130 bytes): - ```bash - find blob/workloads -name "*.safetensors" -size +1k | wc -l - ``` - -### 3.2 home quota full mid-run - -- **Symptom**: `OSError: [Errno 122] Disk quota exceeded` or `pip install` fails with `No space left on device` -- **Root cause**: Many corporate environments have a 5-GB home quota -- **Fix**: relocate everything to scratch: - ```bash - SCRATCH=/home/scratch./kernel_arena - mkdir -p $SCRATCH - mv ~/kernel_arena/flashinfer-trace $SCRATCH/ - ln -sfn $SCRATCH/flashinfer-trace ~/kernel_arena/flashinfer-trace - # Also pip target: - export PIP_TARGET=/tmp/pip-pkgs # /tmp is usually large - ``` - -### 3.3 Stale flashinfer-bench install - -- **Symptom**: `ImportError: cannot import name 'apply' from 'flashinfer_bench'` -- **Root cause**: pip cache picked up an older version -- **Fix**: - ```bash - rm -rf $PIP_TARGET/flashinfer_bench* - pip install --target $PIP_TARGET --no-cache-dir flashinfer-bench - ``` - ---- - -## 4. Hardware-specific - -### 4.1 FlashInfer GDN CuTe DSL ABI mismatch (NGC PyTorch 24.10) - -- **Symptom**: `flashinfer.gdn` import succeeds but calling `gdn_decode(...)` raises `TypeError: 'NoneType' object is not callable`; introspection shows `flashinfer.gdn.run_pretranspose_decode is None` -- **Root cause**: FlashInfer 0.6.x GDN uses CuTe DSL kernels that require NVRTC + a specific PyTorch ABI. NGC PyTorch 24.10 ships torch 2.11 with `_GLIBCXX_USE_CXX11_ABI=1`, mismatching the FlashInfer-shipped ABI; the kernel registry silently leaves the entry as `None` -- **Workaround**: use FLA's `fused_recurrent_gated_delta_rule` instead (see `templates/linear_attention_main.py`); FLA is pure Triton, no ABI dependency - -### 4.2 FA3 wheel `undefined symbol` - -- **Symptom**: `import flash_attn_3` raises `undefined symbol: _ZNxxxxx` or similar -- **Root cause**: The prebuilt FA3 wheel was compiled against a different libstdc++ ABI than the PyTorch in your container -- **Fix**: source-build FA3: - ```bash - git clone https://github.com/Dao-AILab/flash-attention.git - cd flash-attention/hopper - python setup.py install - ``` - This compiles against the active PyTorch's ABI. Takes ~15–25 min on H100 with 64 cores; one-time cost. - -### 4.3 FA3 page_size=1 slow path - -- **Symptom**: FA3 GQA decode passes correctness but is slower than FlashInfer baseline at page_size=1 -- **Root cause**: FA3 supports arbitrary page_size, but the TMA fast path requires `page_size % kBlockN == 0` (kBlockN typically 64 or 128). page_size=1 falls back to a `cp.async`-based slow path -- **Fix / workaround**: this is a known limitation; benchmark both ps=1 and ps=64 variants of the Definition and report both. For production deployment with FA3, prefer ps=64 if the workload allows. - -### 4.4 SM-arch wrapper hardcoded but runtime is different SM - -- **Symptom**: Solution `RUNTIME_ERROR` only on B200 / SM100, works on H100 -- **Root cause**: Wrapper hardcodes `arch_capability(9, 0)` or builds a Triton kernel without `tl.constexpr` arch dispatch -- **Fix**: - - In `spec.target_hardware`, include only the GPUs you actually verified - - If wrapping a Triton kernel, ensure all `triton.autotune` configs cover the target archs; or use `triton.heuristics` - - For CUDA: build with `-gencode arch=compute_90,code=sm_90 -gencode arch=compute_100,code=sm_100` diff --git a/skills/add-flashinfer-solution/templates/dense_baseline_main.py b/skills/add-flashinfer-solution/templates/dense_baseline_main.py deleted file mode 100644 index 182752c1e26f54a3b8ce4bf40902c3e19fd0381d..0000000000000000000000000000000000000000 --- a/skills/add-flashinfer-solution/templates/dense_baseline_main.py +++ /dev/null @@ -1,94 +0,0 @@ -""" -Template A: Dense Baseline Wrapper -================================== - -Use this template when wrapping a generic dense attention impl (PyTorch SDPA, -cuDNN frontend, etc.) as a reference baseline Solution. Demonstrates: - - * paged KV cache → contiguous batch (per-sequence index lookup) - * GQA expansion (replicate kv heads to query heads via repeat_interleave) - * sm_scale forwarding - * 3-D Q tensor (FlashInfer convention) — no S=1 unsqueeze needed - * No LSE output (this template is decode-only, no LSE) - -Target Definition example: gqa_paged_decode_h32_kv8_d128_ps1 - inputs: q [B, H_q, D], k_cache [P, ps, H_kv, D], v_cache [P, ps, H_kv, D], - kv_indptr [B+1], kv_indices [num_kv_indices], sm_scale (scalar) - outputs: output [B, H_q, D] -""" -from __future__ import annotations - -import torch -import torch.nn.functional as F - - -def run( - q, # [B, H_q, D] bfloat16 - k_cache, # [num_pages, page_size, H_kv, D] bfloat16 - v_cache, # [num_pages, page_size, H_kv, D] bfloat16 - kv_indptr, # [B+1] int32 — flat index into kv_indices - kv_indices, # [num_kv_indices] int32 — physical page numbers - sm_scale, # scalar float32 -): - B, H_q, D = q.shape - _, page_size, H_kv, _ = k_cache.shape - assert H_q % H_kv == 0, f"GQA: H_q ({H_q}) must be multiple of H_kv ({H_kv})" - G = H_q // H_kv # group size - - # --- 1. Paged → contiguous: gather pages per batch and concat --- - # SDPA wants dense [B, H, S, D]; we build it from the paged store. - # No batching of variable-length sequences is supported by SDPA, so we - # process each batch row separately and pad to max length, OR we loop. - # Looping is simplest for a baseline: - outs = [] - for b in range(B): - s_start = kv_indptr[b].item() - s_end = kv_indptr[b + 1].item() - seq_pages = kv_indices[s_start:s_end] # [n_pages_b] - - # Gather pages → [n_pages_b * page_size, H_kv, D] - k_b = k_cache[seq_pages].reshape(-1, H_kv, D) # [S_b, H_kv, D] - v_b = v_cache[seq_pages].reshape(-1, H_kv, D) - - # GQA expand: [S_b, H_kv, D] → [S_b, H_q, D] - k_b = k_b.repeat_interleave(G, dim=1) - v_b = v_b.repeat_interleave(G, dim=1) - - # Reshape for SDPA: [B=1, H, S, D] - q_b = q[b : b + 1].unsqueeze(2) # [1, H_q, 1, D] - k_b = k_b.permute(1, 0, 2).unsqueeze(0) # [1, H_q, S_b, D] - v_b = v_b.permute(1, 0, 2).unsqueeze(0) # [1, H_q, S_b, D] - - # --- 2. SDPA call (decode = causal mask is irrelevant since we have S_kv >> S_q=1) --- - out_b = F.scaled_dot_product_attention( - q_b, k_b, v_b, - attn_mask=None, - dropout_p=0.0, - is_causal=False, - scale=float(sm_scale), # ← MUST pass; SDPA defaults to 1/sqrt(d) - ) # [1, H_q, 1, D] - - outs.append(out_b.squeeze(2).squeeze(0)) # [H_q, D] - - output = torch.stack(outs, dim=0) # [B, H_q, D] - return output - - -# Notes on adapting this template: -# -# * For GQA paged PREFILL (instead of decode): -# - Q is [total_q, H_q, D] (ragged), or [B, S_q, H_q, D] (paged); check -# the Definition `inputs[0].shape`. -# - Apply causal mask: `is_causal=True` in F.scaled_dot_product_attention, -# or build a custom mask if the def specifies one. -# -# * For MLA paged decode (no head dim in cache): -# - cache layout is [P, ps, ckv_dim] + [P, ps, kpe_dim], not [P, ps, H, D] -# - Q has 2 components: q_nope [B, H, ckv_dim] and q_pe [B, H, kpe_dim] -# - Reference pseudo-code: out = softmax(q_nope @ k_nope^T / sqrt(d)) @ v -# + softmax(q_pe @ k_pe^T / sqrt(d)) @ v -# (Read Definition.reference.code for exact formula.) -# -# * If the def expects an LSE output: -# - SDPA does not return LSE; switch to FA / cuDNN / manual softmax + log -# - Remember LSE base-2: lse_b2 = lse_natural / math.log(2) diff --git a/skills/add-flashinfer-solution/templates/dense_baseline_solution.json b/skills/add-flashinfer-solution/templates/dense_baseline_solution.json deleted file mode 100644 index cf77708866a1a1f22d01f7c23c3213bcc166620c..0000000000000000000000000000000000000000 --- a/skills/add-flashinfer-solution/templates/dense_baseline_solution.json +++ /dev/null @@ -1,18 +0,0 @@ -{ - "name": "sdpa_paged_decode_v1", - "definition": "gqa_paged_decode_h32_kv8_d128_ps1", - "author": "", - "spec": { - "language": "python", - "target_hardware": ["NVIDIA H100", "NVIDIA H200", "NVIDIA B200", "NVIDIA L40", "NVIDIA A100"], - "entry_point": "main.py::run", - "dependencies": [], - "destination_passing_style": false - }, - "sources": [ - { - "path": "main.py", - "content": "" - } - ] -} diff --git a/skills/add-flashinfer-solution/templates/linear_attention_main.py b/skills/add-flashinfer-solution/templates/linear_attention_main.py deleted file mode 100644 index cbcfdc2dbfa0da141eb48ba23433e357b91ab55a..0000000000000000000000000000000000000000 --- a/skills/add-flashinfer-solution/templates/linear_attention_main.py +++ /dev/null @@ -1,86 +0,0 @@ -""" -Template B: External Python Lib Wrapper (FLA-style linear attention) -==================================================================== - -Use this template when wrapping a pip-installable third-party Python lib -(`flash-linear-attention` aka `fla-core`, `flash-attn`, `xformers`, etc.) as -a Solution. Demonstrates: - - * Calling a third-party op via Python API - * Pre-activations the lib expects (e.g. sigmoid for beta gating) - * Fused-mode flags that must be set explicitly (e.g. use_gate_in_kernel) - * Layout flags (k-first vs k-last) - * Recurrent state input/output handling - -Target Definition example: gdn_decode_qk4_v8_d128_k_last - inputs: q [B, T=1, HQ=4, K=128], k [B, T=1, HQ=4, K=128], v [B, T=1, HV=8, V=128], - state [B, HV, V, K] (k-last layout!), A_log [HV], a [B, T=1, HV], - dt_bias [HV], b [B, T=1, HV], scale (scalar) - outputs: output [B, T=1, HV, V], new_state [B, HV, V, K] -""" -from __future__ import annotations - -import torch - -# fla-core comes from: pip install fla-core -from fla.ops.gated_delta_rule import fused_recurrent_gated_delta_rule - - -def run( - q, # [B, T, H_q, K] - k, # [B, T, H_q, K] - v, # [B, T, H_v, V] - state, # [B, H_v, V, K] ← k-last layout - A_log, # [H_v] - a, # [B, T, H_v] - dt_bias, # [H_v] - b, # [B, T, H_v] - scale, # scalar float32 -): - # --- 1. Pre-compute beta = sigmoid(b) (FLA's op does NOT do this internally) --- - beta = torch.sigmoid(b) - - # --- 2. Call the FLA op with explicit fused-mode flags --- - output, new_state = fused_recurrent_gated_delta_rule( - q, k, v, - state, # initial state, in-place updated by op - g=a, A_log=A_log, # gating raw inputs (NOT pre-applied) - dt_bias=dt_bias, - beta=beta, - scale=float(scale), - - # ↓ All four flags below MUST be set explicitly for the standard def: - use_gate_in_kernel=True, # let FLA fuse g_eff = exp(-exp(A_log)*softplus(g+dt_bias)) - use_qk_l2norm_in_kernel=True, # FLA does the qk l2norm; do NOT pre-norm in this wrapper - transpose_state_layout=True, # def uses k-last layout; FLA's default is k-first - ) - - # output: [B, T, H_v, V] - # new_state: [B, H_v, V, K] (k-last, in-place modified) - return output, new_state - - -# Notes on adapting this template to other linear-attention variants: -# -# * For Mamba2 SSU (selective state update): use `fla.layers.mamba2`, but FLA -# delegates Mamba2 to upstream `mamba_ssm`. Install both: `pip install mamba-ssm fla-core` -# -# * For RetNet, GLA, RWKV4/6/7: use the corresponding op under `fla.ops.`, -# e.g. `fla.ops.rwkv7.fused_recurrent_rwkv7`. Each variant has its own gating -# semantics; consult the FLA docs. -# -# * For DeltaNet (non-gated): use `fla.ops.delta_rule.fused_recurrent_delta_rule`, -# drop the `beta` and `g` / `A_log` / `dt_bias` args. -# -# * For PREFILL (T > 1): switch to `fla.ops.gated_delta_rule.chunk_gated_delta_rule` -# instead of `fused_recurrent_*`. The chunk variant is faster for long sequences. -# -# * For MTP (multi-token prediction): pass `cu_seqlens` to handle variable-length; -# note FLA does not have a dedicated MTP entry, but vLLM's fork adds -# `num_accepted_tokens` and `ssm_state_indices` parameters for spec decode. -# -# Common mistakes: -# * Forgetting `transpose_state_layout=True` → state values silently corrupted -# * Forgetting `use_gate_in_kernel=True` → gating not applied, output drifts -# * Pre-applying L2 norm in wrapper while also setting `use_qk_l2norm_in_kernel=True` -# → output magnitude is half of reference diff --git a/skills/add-flashinfer-solution/templates/linear_attention_solution.json b/skills/add-flashinfer-solution/templates/linear_attention_solution.json deleted file mode 100644 index 8145efc6e661145231801fd649a9ef57eb095b78..0000000000000000000000000000000000000000 --- a/skills/add-flashinfer-solution/templates/linear_attention_solution.json +++ /dev/null @@ -1,18 +0,0 @@ -{ - "name": "fla_gdn_decode_v1", - "definition": "gdn_decode_qk4_v8_d128_k_last", - "author": "", - "spec": { - "language": "python", - "target_hardware": ["NVIDIA H100", "NVIDIA H200", "NVIDIA B200"], - "entry_point": "main.py::run", - "dependencies": ["fla-core>=0.5.0"], - "destination_passing_style": false - }, - "sources": [ - { - "path": "main.py", - "content": "" - } - ] -} diff --git a/skills/add-flashinfer-solution/templates/vendored_kernel_main.py b/skills/add-flashinfer-solution/templates/vendored_kernel_main.py deleted file mode 100644 index 9720c8364f30bbafc8b64a540bff0c706f66feb0..0000000000000000000000000000000000000000 --- a/skills/add-flashinfer-solution/templates/vendored_kernel_main.py +++ /dev/null @@ -1,139 +0,0 @@ -""" -Template C: Vendored Kernel Wrapper (SGLang-style MLA decode) -============================================================= - -Use this template when bringing in a Triton (or CUDA) kernel from another -project (SGLang / vLLM / private) WITHOUT depending on the upstream package. -The kernel source is co-located in the Solution sources. - -Demonstrates: - - * Vendored module import (sys.path manipulation for runner-extracted dirs) - * Workspace caching (avoid re-allocating per call) - * Split-K + LSE reduction (typical MLA decode pattern) - * MLA-specific tensor fusion (ckv + kpe → fused k buffer) - * LSE base conversion (natural-log → base-2) - -Target Definition example: mla_paged_decode_h16_ckv512_kpe64_ps1 - inputs: q_nope [B, H=16, ckv_dim=512], q_pe [B, H=16, kpe_dim=64], - ckv_cache [P, ps=1, 512], kpe_cache [P, ps=1, 64], - kv_indptr [B+1], kv_indices [num_kv_indices], sm_scale (scalar) - outputs: output [B, H=16, ckv_dim=512], lse [B, H=16] (base-2) -""" -from __future__ import annotations - -import math -import os -import sys -from typing import Tuple - -import torch - -# --- Vendored module loading --- -# When the runner extracts this Solution's `sources` to a temp dir, that dir -# is NOT on sys.path automatically. Insert it so we can import sibling files. -_HERE = os.path.dirname(os.path.abspath(__file__)) -if _HERE not in sys.path: - sys.path.insert(0, _HERE) - -from sglang_decode import decode_attention_fwd_grouped_split_k_lse # noqa: E402 - -# --- Per-process state (workspace cache + constants) --- -_MAX_KV_SPLITS = 8 -_MIN_BLOCK_KV = 64 -_LOG2 = math.log(2.0) -_ws_cache: dict = {} - - -def _get_workspace(B: int, H: int, V: int, device, dtype) -> dict: - """Cache pre-allocated work tensors keyed on (B, H, V, device, dtype).""" - key = (B, H, V, str(device), dtype) - ws = _ws_cache.get(key) - if ws is None: - ws = { - "q_fused": torch.empty((B, H, V + 64), dtype=dtype, device=device), - "attn_logits": torch.empty((B, H, _MAX_KV_SPLITS, V), dtype=torch.float32, device=device), - "attn_lse": torch.empty((B, H, _MAX_KV_SPLITS), dtype=torch.float32, device=device), - "num_kv_splits": torch.empty((B,), dtype=torch.int32, device=device), - } - _ws_cache[key] = ws - return ws - - -def run( - q_nope, # [B, H, ckv_dim] - q_pe, # [B, H, kpe_dim] - ckv_cache, # [num_pages, page_size, ckv_dim] - kpe_cache, # [num_pages, page_size, kpe_dim] - kv_indptr, # [B+1] - kv_indices, # [num_kv_indices] - sm_scale, # scalar -) -> Tuple[torch.Tensor, torch.Tensor]: - B, H, ckv_dim = q_nope.shape - kpe_dim = q_pe.shape[-1] - assert ckv_dim == 512 and kpe_dim == 64, "MLA decode template assumes ckv=512, kpe=64" - - device = q_nope.device - dtype = q_nope.dtype - ws = _get_workspace(B, H, ckv_dim, device, dtype) - - # --- Fuse Q --- - # MLA needs Q to be a single [B, H, ckv+kpe] tensor; we concat once. - ws["q_fused"][:, :, :ckv_dim].copy_(q_nope) - ws["q_fused"][:, :, ckv_dim:].copy_(q_pe) - - # --- Determine kv_splits per batch --- - # SGLang uses a heuristic: split count grows with KV length; capped at MAX. - seq_lens = kv_indptr[1:] - kv_indptr[:-1] # [B] - ws["num_kv_splits"].copy_((seq_lens.float() / _MIN_BLOCK_KV).clamp(min=1, max=_MAX_KV_SPLITS).int()) - - # --- Output buffer --- - output = torch.empty((B, H, ckv_dim), dtype=dtype, device=device) - - # --- Launch the vendored kernel --- - # Note: kernel uses ckv_cache as the V buffer (zero-copy, MLA's compressed-V property). - decode_attention_fwd_grouped_split_k_lse( - ws["q_fused"], # [B, H, ckv+kpe] - ckv_cache, # K (ckv) cache - ckv_cache, # V buffer = ckv (MLA compressed) - kpe_cache, # KPE component - output, - kv_indices, kv_indptr, - ws["num_kv_splits"], _MAX_KV_SPLITS, - ws["attn_logits"], ws["attn_lse"], - page_size=1, - sm_scale=float(sm_scale), - ) - - # --- Reduce per-split LSE to final LSE, in base-2 (FlashInfer convention) --- - # `attn_lse` stores natural-log LSE per split; reduce + convert. - lse_natural_per_split = ws["attn_lse"] # [B, H, splits], natural log - # logsumexp across split dim: - lse_natural = torch.logsumexp(lse_natural_per_split, dim=2) # [B, H], natural log - # Convert to base-2: - lse_b2 = lse_natural / _LOG2 # [B, H], base-2 - - return output, lse_b2 - - -# Notes on adapting this template: -# -# * For NON-MLA vendored kernels: drop the q_nope / q_pe fusion, keep the -# workspace caching pattern and split-K LSE reduction (most modern decode -# kernels use split-K). -# -# * If the vendored kernel returns base-2 LSE already, skip the `/ _LOG2` step. -# -# * If the kernel does not produce LSE, change the function signature to -# return only output. (Check Definition.outputs to confirm.) -# -# * For multi-file vendored sources: list every file in solution.json -# "sources"; the runner unpacks all of them to the same temp dir. -# -# * For CUDA kernels (vs Triton): vendor a pre-built .so file or compile at -# import time via cppimport / torch.utils.cpp_extension. Avoid blocking -# in run(); compile in module load time. -# -# * The `_ws_cache` global persists across run() calls within one process, -# speeding up repeated benchmarks. Be careful: in multi-process testing -# each worker has its own cache (which is what you want). diff --git a/skills/add-flashinfer-solution/templates/vendored_kernel_solution.json b/skills/add-flashinfer-solution/templates/vendored_kernel_solution.json deleted file mode 100644 index 5d21c29246f42ac9c2a9946216dbb2550801ddc3..0000000000000000000000000000000000000000 --- a/skills/add-flashinfer-solution/templates/vendored_kernel_solution.json +++ /dev/null @@ -1,22 +0,0 @@ -{ - "name": "sglang_mla_decode_v1", - "definition": "mla_paged_decode_h16_ckv512_kpe64_ps1", - "author": "", - "spec": { - "language": "python", - "target_hardware": ["NVIDIA H100", "NVIDIA H200", "NVIDIA B200"], - "entry_point": "main.py::run", - "dependencies": [], - "destination_passing_style": false - }, - "sources": [ - { - "path": "main.py", - "content": "" - }, - { - "path": "sglang_decode.py", - "content": "" - } - ] -} diff --git a/solutions/baseline/gemm/gemm_n16384_k2048/torch_matmul_0a4e73.json b/solutions/baseline/gemm/gemm_n16384_k2048/torch_matmul_0a4e73.json deleted file mode 100644 index 24a5895bb134084de71e1f6b885dfd024e621080..0000000000000000000000000000000000000000 --- a/solutions/baseline/gemm/gemm_n16384_k2048/torch_matmul_0a4e73.json +++ /dev/null @@ -1,21 +0,0 @@ -{ - "name": "torch_matmul_0a4e73", - "definition": "gemm_n16384_k2048", - "author": "baseline", - "spec": { - "language": "python", - "target_hardware": [ - "NVIDIA B200" - ], - "entry_point": "main.py::run", - "dependencies": [], - "destination_passing_style": false - }, - "sources": [ - { - "path": "main.py", - "content": "import torch\nimport torch.nn.functional as F\n\ndef run(A: torch.Tensor, B: torch.Tensor):\n return torch.nn.functional.linear(A, B)\n" - } - ], - "description": "Baseline GEMM implemented with torch.nn.functional.linear." -} diff --git a/solutions/baseline/gemm/gemm_n2048_k2048/torch_matmul_756dea.json b/solutions/baseline/gemm/gemm_n2048_k2048/torch_matmul_756dea.json deleted file mode 100644 index e7402654c581a4810eaed5edabd61ce779eb4b15..0000000000000000000000000000000000000000 --- a/solutions/baseline/gemm/gemm_n2048_k2048/torch_matmul_756dea.json +++ /dev/null @@ -1,21 +0,0 @@ -{ - "name": "torch_matmul_756dea", - "definition": "gemm_n2048_k2048", - "author": "baseline", - "spec": { - "language": "python", - "target_hardware": [ - "NVIDIA B200" - ], - "entry_point": "main.py::run", - "dependencies": [], - "destination_passing_style": false - }, - "sources": [ - { - "path": "main.py", - "content": "import torch\nimport torch.nn.functional as F\n\ndef run(A: torch.Tensor, B: torch.Tensor):\n return F.linear(A, B)\n" - } - ], - "description": "Baseline GEMM implemented with torch.nn.functional.linear." -} diff --git a/solutions/baseline/gemm/gemm_n2048_k8192/torch_matmul_ff95bc.json b/solutions/baseline/gemm/gemm_n2048_k8192/torch_matmul_ff95bc.json deleted file mode 100644 index e3e704306c2f8ec2a8c5489b410daf1600d21e62..0000000000000000000000000000000000000000 --- a/solutions/baseline/gemm/gemm_n2048_k8192/torch_matmul_ff95bc.json +++ /dev/null @@ -1,21 +0,0 @@ -{ - "name": "torch_matmul_ff95bc", - "definition": "gemm_n2048_k8192", - "author": "baseline", - "spec": { - "language": "python", - "target_hardware": [ - "NVIDIA B200" - ], - "entry_point": "main.py::run", - "dependencies": [], - "destination_passing_style": false - }, - "sources": [ - { - "path": "main.py", - "content": "import torch\nfrom torch.nn.functional import linear\n\ndef run(A: torch.Tensor, B: torch.Tensor):\n return linear(A, B)\n" - } - ], - "description": "Baseline GEMM implemented with torch.nn.functional.linear." -} diff --git a/solutions/baseline/gemm/gemm_n3072_k2048/torch_matmul_73342c.json b/solutions/baseline/gemm/gemm_n3072_k2048/torch_matmul_73342c.json deleted file mode 100644 index 4641093b24dc0ab2a24b4f9de4ae55a5fef9f5a5..0000000000000000000000000000000000000000 --- a/solutions/baseline/gemm/gemm_n3072_k2048/torch_matmul_73342c.json +++ /dev/null @@ -1,21 +0,0 @@ -{ - "name": "torch_matmul_73342c", - "definition": "gemm_n3072_k2048", - "author": "baseline", - "spec": { - "language": "python", - "target_hardware": [ - "NVIDIA B200" - ], - "entry_point": "main.py::run", - "dependencies": [], - "destination_passing_style": false - }, - "sources": [ - { - "path": "main.py", - "content": "import torch\nimport torch.nn.functional as F\n\ndef run(A: torch.Tensor, B: torch.Tensor):\n C = F.linear(A, B)\n return C\n" - } - ], - "description": "Baseline GEMM implemented with torch.nn.functional.linear." -} diff --git a/solutions/baseline/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/flashinfer_wrapper_ad4135.json b/solutions/baseline/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/flashinfer_wrapper_ad4135.json deleted file mode 100644 index 58eafe30d1b9d01ad7069018eea1388a2548e75c..0000000000000000000000000000000000000000 --- a/solutions/baseline/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/flashinfer_wrapper_ad4135.json +++ /dev/null @@ -1,27 +0,0 @@ -{ - "name": "flashinfer_wrapper_ad4135", - "definition": "gqa_paged_decode_h32_kv8_d128_ps64", - "author": "baseline", - "spec": { - "language": "python", - "target_hardware": [ - "NVIDIA A100", - "NVIDIA H20", - "NVIDIA H100", - "NVIDIA H200", - "NVIDIA B200" - ], - "entry_point": "main.py::run", - "dependencies": [ - "flashinfer" - ], - "destination_passing_style": false - }, - "sources": [ - { - "path": "main.py", - "content": "import torch\nimport flashinfer\n\n_WORKSPACE_SIZE_BYTES = 128 * 1024 * 1024\n_workspace_cache = {}\n_wrapper_cache = {}\n_plan_state = {}\n\n\ndef _get_workspace(device):\n key = str(device)\n buffer = _workspace_cache.get(key)\n if buffer is None or buffer.device != device or buffer.numel() < _WORKSPACE_SIZE_BYTES:\n buffer = torch.empty(_WORKSPACE_SIZE_BYTES, dtype=torch.uint8, device=device)\n _workspace_cache[key] = buffer\n return buffer\n\n\ndef _get_wrapper(key, device):\n wrapper = _wrapper_cache.get(key)\n if wrapper is None:\n workspace = _get_workspace(device)\n wrapper = flashinfer.BatchDecodeWithPagedKVCacheWrapper(\n workspace,\n kv_layout=\"NHD\",\n )\n _wrapper_cache[key] = wrapper\n return wrapper\n\n\ndef run(q, k_cache, v_cache, kv_indptr, kv_indices, kv_last_page_len, sm_scale):\n batch_size, num_qo_heads, head_dim = q.shape\n _, page_size, num_kv_heads, _ = k_cache.shape\n len_indptr = kv_indptr.shape[0]\n num_kv_indices = kv_indices.shape[0]\n\n device = q.device\n wrapper_key = (\n str(device),\n num_qo_heads,\n num_kv_heads,\n head_dim,\n page_size,\n q.dtype,\n k_cache.dtype,\n )\n\n wrapper = _get_wrapper(wrapper_key, device)\n state = _plan_state.get(wrapper_key)\n\n needs_plan = True\n if state is not None:\n needs_plan = (\n state.get(\"batch_size\") != batch_size\n or state.get(\"len_indptr\") != len_indptr\n or state.get(\"num_kv_indices\") != num_kv_indices\n or state.get(\"sm_scale\") != sm_scale\n or state.get(\"kv_indptr_ptr\") != kv_indptr.data_ptr()\n or state.get(\"kv_indices_ptr\") != kv_indices.data_ptr()\n or state.get(\"last_page_ptr\") != kv_last_page_len.data_ptr()\n )\n\n if needs_plan:\n wrapper.plan(\n indptr=kv_indptr,\n indices=kv_indices,\n last_page_len=kv_last_page_len,\n num_qo_heads=num_qo_heads,\n num_kv_heads=num_kv_heads,\n head_dim=head_dim,\n page_size=page_size,\n pos_encoding_mode=\"NONE\",\n q_data_type=q.dtype,\n kv_data_type=k_cache.dtype,\n sm_scale=sm_scale,\n )\n _plan_state[wrapper_key] = {\n \"batch_size\": batch_size,\n \"len_indptr\": len_indptr,\n \"num_kv_indices\": num_kv_indices,\n \"sm_scale\": sm_scale,\n \"kv_indptr_ptr\": kv_indptr.data_ptr(),\n \"kv_indices_ptr\": kv_indices.data_ptr(),\n \"last_page_ptr\": kv_last_page_len.data_ptr(),\n }\n\n output, lse = wrapper.run(\n q,\n (k_cache, v_cache),\n return_lse=True,\n )\n\n return output, lse\n" - } - ], - "description": "FlashInfer BatchDecodeWithPagedKVCacheWrapper baseline for gqa_paged_decode_h32_kv8_d128_ps64 (group_size=4, power-of-2)." -} \ No newline at end of file diff --git a/solutions/baseline/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps1/flashinfer_wrapper_3f9411.json b/solutions/baseline/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps1/flashinfer_wrapper_3f9411.json deleted file mode 100644 index f36c050f88f46b65e58358c4607eabf4983790fd..0000000000000000000000000000000000000000 --- a/solutions/baseline/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps1/flashinfer_wrapper_3f9411.json +++ /dev/null @@ -1,23 +0,0 @@ -{ - "name": "flashinfer_wrapper_3f9411", - "definition": "gqa_paged_decode_h32_kv8_d64_ps1", - "author": "baseline", - "spec": { - "language": "python", - "target_hardware": [ - "NVIDIA B200" - ], - "entry_point": "main.py::run", - "dependencies": [ - "flashinfer" - ], - "destination_passing_style": false - }, - "sources": [ - { - "path": "main.py", - "content": "import torch\nimport flashinfer\n\n_WORKSPACE_SIZE_BYTES = 128 * 1024 * 1024\n_workspace_cache = {}\n_wrapper_cache = {}\n_plan_state = {}\n\n\ndef _get_workspace(device):\n key = str(device)\n buffer = _workspace_cache.get(key)\n if buffer is None or buffer.device != device or buffer.numel() < _WORKSPACE_SIZE_BYTES:\n buffer = torch.empty(_WORKSPACE_SIZE_BYTES, dtype=torch.uint8, device=device)\n _workspace_cache[key] = buffer\n return buffer\n\n\ndef _get_wrapper(key, device):\n wrapper = _wrapper_cache.get(key)\n if wrapper is None:\n workspace = _get_workspace(device)\n wrapper = flashinfer.BatchDecodeWithPagedKVCacheWrapper(\n workspace,\n kv_layout=\"NHD\",\n )\n _wrapper_cache[key] = wrapper\n return wrapper\n\n\ndef run(q, k_cache, v_cache, kv_indptr, kv_indices, sm_scale):\n batch_size, num_qo_heads, head_dim = q.shape\n _, page_size, num_kv_heads, _ = k_cache.shape\n len_indptr = kv_indptr.shape[0]\n num_kv_indices = kv_indices.shape[0]\n\n device = q.device\n wrapper_key = (\n str(device),\n num_qo_heads,\n num_kv_heads,\n head_dim,\n page_size,\n q.dtype,\n k_cache.dtype,\n )\n\n wrapper = _get_wrapper(wrapper_key, device)\n state = _plan_state.get(wrapper_key)\n\n needs_plan = True\n if state is not None:\n needs_plan = (\n state.get(\"batch_size\") != batch_size\n or state.get(\"len_indptr\") != len_indptr\n or state.get(\"num_kv_indices\") != num_kv_indices\n or state.get(\"sm_scale\") != sm_scale\n or state.get(\"kv_indptr_ptr\") != kv_indptr.data_ptr()\n or state.get(\"kv_indices_ptr\") != kv_indices.data_ptr()\n )\n\n if needs_plan:\n kv_last_page_len = torch.ones(batch_size, dtype=torch.int32, device=device)\n wrapper.plan(\n indptr=kv_indptr,\n indices=kv_indices,\n last_page_len=kv_last_page_len,\n num_qo_heads=num_qo_heads,\n num_kv_heads=num_kv_heads,\n head_dim=head_dim,\n page_size=page_size,\n pos_encoding_mode=\"NONE\",\n q_data_type=q.dtype,\n kv_data_type=k_cache.dtype,\n sm_scale=sm_scale,\n )\n _plan_state[wrapper_key] = {\n \"batch_size\": batch_size,\n \"len_indptr\": len_indptr,\n \"num_kv_indices\": num_kv_indices,\n \"sm_scale\": sm_scale,\n \"kv_indptr_ptr\": kv_indptr.data_ptr(),\n \"kv_indices_ptr\": kv_indices.data_ptr(),\n }\n\n output, lse = wrapper.run(\n q,\n (k_cache, v_cache),\n return_lse=True,\n )\n\n return output, lse\n" - } - ], - "description": "FlashInfer BatchDecodeWithPagedKVCacheWrapper baseline for gqa_paged_decode_h32_kv8_d64_ps1 (decode, page_size=1, Llama 3.2 1B, 32q/8kv heads, head_dim=64, group_size=4 power-of-2)." -} diff --git a/solutions/baseline/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps64/flashinfer_wrapper_68ff49.json b/solutions/baseline/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps64/flashinfer_wrapper_68ff49.json deleted file mode 100644 index 0cde95b3685185c86bbc8bfd89381f52b12d65ba..0000000000000000000000000000000000000000 --- a/solutions/baseline/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps64/flashinfer_wrapper_68ff49.json +++ /dev/null @@ -1,23 +0,0 @@ -{ - "name": "flashinfer_wrapper_68ff49", - "definition": "gqa_paged_decode_h32_kv8_d64_ps64", - "author": "baseline", - "spec": { - "language": "python", - "target_hardware": [ - "NVIDIA B200" - ], - "entry_point": "main.py::run", - "dependencies": [ - "flashinfer" - ], - "destination_passing_style": false - }, - "sources": [ - { - "path": "main.py", - "content": "import torch\nimport flashinfer\n\n_WORKSPACE_SIZE_BYTES = 128 * 1024 * 1024\n_workspace_cache = {}\n_wrapper_cache = {}\n_plan_state = {}\n\n\ndef _get_workspace(device):\n key = str(device)\n buffer = _workspace_cache.get(key)\n if buffer is None or buffer.device != device or buffer.numel() < _WORKSPACE_SIZE_BYTES:\n buffer = torch.empty(_WORKSPACE_SIZE_BYTES, dtype=torch.uint8, device=device)\n _workspace_cache[key] = buffer\n return buffer\n\n\ndef _get_wrapper(key, device):\n wrapper = _wrapper_cache.get(key)\n if wrapper is None:\n workspace = _get_workspace(device)\n wrapper = flashinfer.BatchDecodeWithPagedKVCacheWrapper(\n workspace,\n kv_layout=\"NHD\",\n )\n _wrapper_cache[key] = wrapper\n return wrapper\n\n\ndef run(q, k_cache, v_cache, kv_indptr, kv_indices, kv_last_page_len, sm_scale):\n batch_size, num_qo_heads, head_dim = q.shape\n _, page_size, num_kv_heads, _ = k_cache.shape\n len_indptr = kv_indptr.shape[0]\n num_kv_indices = kv_indices.shape[0]\n\n device = q.device\n wrapper_key = (\n str(device),\n num_qo_heads,\n num_kv_heads,\n head_dim,\n page_size,\n q.dtype,\n k_cache.dtype,\n )\n\n wrapper = _get_wrapper(wrapper_key, device)\n state = _plan_state.get(wrapper_key)\n\n needs_plan = True\n if state is not None:\n needs_plan = (\n state.get(\"batch_size\") != batch_size\n or state.get(\"len_indptr\") != len_indptr\n or state.get(\"num_kv_indices\") != num_kv_indices\n or state.get(\"sm_scale\") != sm_scale\n or state.get(\"kv_indptr_ptr\") != kv_indptr.data_ptr()\n or state.get(\"kv_indices_ptr\") != kv_indices.data_ptr()\n or state.get(\"last_page_ptr\") != kv_last_page_len.data_ptr()\n )\n\n if needs_plan:\n wrapper.plan(\n indptr=kv_indptr,\n indices=kv_indices,\n last_page_len=kv_last_page_len,\n num_qo_heads=num_qo_heads,\n num_kv_heads=num_kv_heads,\n head_dim=head_dim,\n page_size=page_size,\n pos_encoding_mode=\"NONE\",\n q_data_type=q.dtype,\n kv_data_type=k_cache.dtype,\n sm_scale=sm_scale,\n )\n _plan_state[wrapper_key] = {\n \"batch_size\": batch_size,\n \"len_indptr\": len_indptr,\n \"num_kv_indices\": num_kv_indices,\n \"sm_scale\": sm_scale,\n \"kv_indptr_ptr\": kv_indptr.data_ptr(),\n \"kv_indices_ptr\": kv_indices.data_ptr(),\n \"last_page_ptr\": kv_last_page_len.data_ptr(),\n }\n\n output, lse = wrapper.run(\n q,\n (k_cache, v_cache),\n return_lse=True,\n )\n\n return output, lse\n" - } - ], - "description": "FlashInfer BatchDecodeWithPagedKVCacheWrapper baseline for gqa_paged_decode_h32_kv8_d64_ps64 (decode, page_size=64, Llama 3.2 1B, 32q/8kv heads, head_dim=64, group_size=4 power-of-2)." -} diff --git a/solutions/baseline/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps1/flashinfer_wrapper_ece89a.json b/solutions/baseline/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps1/flashinfer_wrapper_ece89a.json deleted file mode 100644 index b612e2fb3a0f3b0d786f41933b2d639c29a88521..0000000000000000000000000000000000000000 --- a/solutions/baseline/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps1/flashinfer_wrapper_ece89a.json +++ /dev/null @@ -1,23 +0,0 @@ -{ - "name": "flashinfer_wrapper_ece89a", - "definition": "gqa_paged_prefill_causal_h32_kv8_d64_ps1", - "author": "baseline", - "spec": { - "language": "python", - "target_hardware": [ - "NVIDIA B200" - ], - "entry_point": "main.py::run", - "dependencies": [ - "flashinfer" - ], - "destination_passing_style": false - }, - "sources": [ - { - "path": "main.py", - "content": "import torch\nimport flashinfer\n\n_WORKSPACE_SIZE_BYTES = 128 * 1024 * 1024\n_workspace_cache = {}\n_wrapper_cache = {}\n_plan_state = {}\n\n\ndef _get_workspace(device):\n key = str(device)\n buffer = _workspace_cache.get(key)\n if buffer is None or buffer.device != device or buffer.numel() < _WORKSPACE_SIZE_BYTES:\n buffer = torch.empty(_WORKSPACE_SIZE_BYTES, dtype=torch.uint8, device=device)\n _workspace_cache[key] = buffer\n return buffer\n\n\ndef _get_wrapper(key, device):\n wrapper = _wrapper_cache.get(key)\n if wrapper is None:\n workspace = _get_workspace(device)\n wrapper = flashinfer.BatchPrefillWithPagedKVCacheWrapper(\n workspace,\n kv_layout=\"NHD\",\n )\n _wrapper_cache[key] = wrapper\n return wrapper\n\n\ndef run(q, k_cache, v_cache, qo_indptr, kv_indptr, kv_indices, sm_scale):\n total_q, num_qo_heads, head_dim = q.shape\n _, page_size, num_kv_heads, _ = k_cache.shape\n batch_size = qo_indptr.shape[0] - 1\n num_kv_indices = kv_indices.shape[0]\n\n device = q.device\n wrapper_key = (\n str(device),\n num_qo_heads,\n num_kv_heads,\n head_dim,\n page_size,\n q.dtype,\n k_cache.dtype,\n )\n\n wrapper = _get_wrapper(wrapper_key, device)\n state = _plan_state.get(wrapper_key)\n\n if isinstance(sm_scale, torch.Tensor):\n sm_scale_value = float(sm_scale.item())\n else:\n sm_scale_value = float(sm_scale)\n\n needs_plan = True\n if state is not None:\n needs_plan = (\n state.get(\"total_q\") != total_q\n or state.get(\"batch_size\") != batch_size\n or state.get(\"num_kv_indices\") != num_kv_indices\n or state.get(\"sm_scale\") != sm_scale_value\n or state.get(\"qo_indptr_ptr\") != qo_indptr.data_ptr()\n or state.get(\"kv_indptr_ptr\") != kv_indptr.data_ptr()\n or state.get(\"kv_indices_ptr\") != kv_indices.data_ptr()\n )\n\n if needs_plan:\n last_page_len = torch.ones(batch_size, dtype=torch.int32, device=device)\n wrapper.plan(\n qo_indptr=qo_indptr,\n paged_kv_indptr=kv_indptr,\n paged_kv_indices=kv_indices,\n paged_kv_last_page_len=last_page_len,\n num_qo_heads=num_qo_heads,\n num_kv_heads=num_kv_heads,\n head_dim_qk=head_dim,\n page_size=page_size,\n causal=True,\n sm_scale=sm_scale,\n q_data_type=q.dtype,\n kv_data_type=k_cache.dtype,\n )\n _plan_state[wrapper_key] = {\n \"total_q\": total_q,\n \"batch_size\": batch_size,\n \"num_kv_indices\": num_kv_indices,\n \"sm_scale\": sm_scale_value,\n \"qo_indptr_ptr\": qo_indptr.data_ptr(),\n \"kv_indptr_ptr\": kv_indptr.data_ptr(),\n \"kv_indices_ptr\": kv_indices.data_ptr(),\n }\n\n output, lse = wrapper.run(\n q,\n (k_cache, v_cache),\n return_lse=True,\n )\n\n return output, lse\n" - } - ], - "description": "FlashInfer BatchPrefillWithPagedKVCacheWrapper baseline for gqa_paged_prefill_causal_h32_kv8_d64_ps1 (causal prefill, page_size=1, Llama 3.2 1B, 32q/8kv heads, head_dim=64)." -} diff --git a/solutions/baseline/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps64/flashinfer_wrapper_06ced8.json b/solutions/baseline/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps64/flashinfer_wrapper_06ced8.json deleted file mode 100644 index 028208d91318d7aeb238ac603132f87045196741..0000000000000000000000000000000000000000 --- a/solutions/baseline/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps64/flashinfer_wrapper_06ced8.json +++ /dev/null @@ -1,23 +0,0 @@ -{ - "name": "flashinfer_wrapper_06ced8", - "definition": "gqa_paged_prefill_causal_h32_kv8_d64_ps64", - "author": "baseline", - "spec": { - "language": "python", - "target_hardware": [ - "NVIDIA B200" - ], - "entry_point": "main.py::run", - "dependencies": [ - "flashinfer" - ], - "destination_passing_style": false - }, - "sources": [ - { - "path": "main.py", - "content": "import torch\nimport flashinfer\n\n_WORKSPACE_SIZE_BYTES = 256 * 1024 * 1024\n_workspace_cache = {}\n_wrapper_cache = {}\n_plan_state = {}\n\n\ndef _get_workspace(device):\n key = str(device)\n buf = _workspace_cache.get(key)\n if buf is None:\n buf = torch.empty(_WORKSPACE_SIZE_BYTES, dtype=torch.uint8, device=device)\n _workspace_cache[key] = buf\n return buf\n\n\ndef _get_wrapper(key, device):\n w = _wrapper_cache.get(key)\n if w is None:\n w = flashinfer.BatchPrefillWithPagedKVCacheWrapper(_get_workspace(device), kv_layout=\"NHD\")\n _wrapper_cache[key] = w\n return w\n\n\ndef run(q, k_cache, v_cache, qo_indptr, kv_indptr, kv_indices, kv_last_page_len, sm_scale):\n total_q, num_qo_heads, head_dim = q.shape\n _, page_size, num_kv_heads, _ = k_cache.shape\n batch_size = qo_indptr.shape[0] - 1\n device = q.device\n\n paged_kv = torch.stack([k_cache, v_cache], dim=1) # [num_pages, 2, page_size, kv_h, d]\n wkey = (str(device), num_qo_heads, num_kv_heads, head_dim, page_size, q.dtype, k_cache.dtype)\n wrapper = _get_wrapper(wkey, device)\n state = _plan_state.get(wkey)\n needs_plan = (\n state is None\n or state[\"batch_size\"] != batch_size\n or state[\"qo_ptr\"] != qo_indptr.data_ptr()\n or state[\"kv_ptr\"] != kv_indptr.data_ptr()\n or state[\"last_page_ptr\"] != kv_last_page_len.data_ptr()\n )\n if needs_plan:\n wrapper.plan(\n qo_indptr=qo_indptr,\n paged_kv_indptr=kv_indptr,\n paged_kv_indices=kv_indices,\n paged_kv_last_page_len=kv_last_page_len[:batch_size],\n num_qo_heads=num_qo_heads,\n num_kv_heads=num_kv_heads,\n head_dim_qk=head_dim,\n page_size=page_size,\n causal=True,\n sm_scale=float(sm_scale),\n q_data_type=q.dtype,\n kv_data_type=k_cache.dtype,\n )\n _plan_state[wkey] = {\n \"batch_size\": batch_size,\n \"qo_ptr\": qo_indptr.data_ptr(),\n \"kv_ptr\": kv_indptr.data_ptr(),\n \"last_page_ptr\": kv_last_page_len.data_ptr(),\n }\n output, lse = wrapper.run(q, paged_kv, return_lse=True)\n return output, lse\n" - } - ], - "description": "FlashInfer BatchPrefillWithPagedKVCacheWrapper baseline for gqa_paged_prefill_causal_h32_kv8_d64_ps64 (causal prefill, page_size=64, Llama 3.2 1B, 32q/8kv heads, head_dim=64)." -} diff --git a/solutions/baseline/gqa_ragged/gqa_ragged_prefill_causal_h32_kv8_d64/flashinfer_wrapper_44b034.json b/solutions/baseline/gqa_ragged/gqa_ragged_prefill_causal_h32_kv8_d64/flashinfer_wrapper_44b034.json deleted file mode 100644 index f5056b0e543d90653170d9837e11bbc6e806df02..0000000000000000000000000000000000000000 --- a/solutions/baseline/gqa_ragged/gqa_ragged_prefill_causal_h32_kv8_d64/flashinfer_wrapper_44b034.json +++ /dev/null @@ -1,23 +0,0 @@ -{ - "name": "flashinfer_wrapper_44b034", - "definition": "gqa_ragged_prefill_causal_h32_kv8_d64", - "author": "baseline", - "spec": { - "language": "python", - "target_hardware": [ - "NVIDIA B200" - ], - "entry_point": "main.py::run", - "dependencies": [ - "flashinfer" - ], - "destination_passing_style": false - }, - "sources": [ - { - "path": "main.py", - "content": "import torch\nimport flashinfer\n\n_WORKSPACE_SIZE_BYTES = 128 * 1024 * 1024\n_workspace_cache = {}\n_wrapper_cache = {}\n_plan_state = {}\n\n\ndef _get_workspace(device):\n key = str(device)\n buffer = _workspace_cache.get(key)\n if buffer is None or buffer.device != device or buffer.numel() < _WORKSPACE_SIZE_BYTES:\n buffer = torch.empty(_WORKSPACE_SIZE_BYTES, dtype=torch.uint8, device=device)\n _workspace_cache[key] = buffer\n return buffer\n\n\ndef _get_wrapper(key, device):\n wrapper = _wrapper_cache.get(key)\n if wrapper is None:\n workspace = _get_workspace(device)\n wrapper = flashinfer.BatchPrefillWithRaggedKVCacheWrapper(\n workspace,\n kv_layout=\"NHD\",\n )\n _wrapper_cache[key] = wrapper\n return wrapper\n\n\ndef run(q, k, v, qo_indptr, kv_indptr, sm_scale):\n total_q, num_qo_heads, head_dim = q.shape\n total_kv, num_kv_heads, _ = k.shape\n batch_size = qo_indptr.shape[0] - 1\n\n device = q.device\n wrapper_key = (\n str(device),\n num_qo_heads,\n num_kv_heads,\n head_dim,\n q.dtype,\n k.dtype,\n v.dtype,\n )\n\n wrapper = _get_wrapper(wrapper_key, device)\n state = _plan_state.get(wrapper_key)\n\n needs_plan = True\n if state is not None:\n needs_plan = (\n state.get(\"total_q\") != total_q\n or state.get(\"total_kv\") != total_kv\n or state.get(\"batch_size\") != batch_size\n or state.get(\"sm_scale\") != sm_scale\n or state.get(\"qo_indptr_ptr\") != qo_indptr.data_ptr()\n or state.get(\"kv_indptr_ptr\") != kv_indptr.data_ptr()\n )\n\n if needs_plan:\n wrapper.plan(\n qo_indptr=qo_indptr,\n kv_indptr=kv_indptr,\n num_qo_heads=num_qo_heads,\n num_kv_heads=num_kv_heads,\n head_dim_qk=head_dim,\n causal=True,\n sm_scale=sm_scale,\n q_data_type=q.dtype,\n kv_data_type=k.dtype,\n )\n _plan_state[wrapper_key] = {\n \"total_q\": total_q,\n \"total_kv\": total_kv,\n \"batch_size\": batch_size,\n \"sm_scale\": sm_scale,\n \"qo_indptr_ptr\": qo_indptr.data_ptr(),\n \"kv_indptr_ptr\": kv_indptr.data_ptr(),\n }\n\n output, lse = wrapper.run(\n q,\n k,\n v,\n return_lse=True,\n )\n\n return output, lse\n" - } - ], - "description": "FlashInfer BatchPrefillWithRaggedKVCacheWrapper baseline for gqa_ragged_prefill_causal_h32_kv8_d64 (ragged causal prefill, Llama 3.2 1B, 32q/8kv heads, head_dim=64)." -} diff --git a/tests/references/test_gqa_paged_decode_h32_kv8_d64_ps1.py b/tests/references/test_gqa_paged_decode_h32_kv8_d64_ps1.py deleted file mode 100644 index 53f65dcf00c3df6d80ce3c8748bdc2f390a0e0a4..0000000000000000000000000000000000000000 --- a/tests/references/test_gqa_paged_decode_h32_kv8_d64_ps1.py +++ /dev/null @@ -1,111 +0,0 @@ -"""Reference test for gqa_paged_decode_h32_kv8_d64_ps1 (Llama 3.2 1B).""" - -import math -from pathlib import Path - -import flashinfer -import torch - -from flashinfer_bench.data import Definition, load_json_file - -DEFINITIONS_DIR = Path(__file__).parent.parent.parent / "definitions" - -NUM_QO_HEADS = 32 -NUM_KV_HEADS = 8 -HEAD_DIM = 64 -PAGE_SIZE = 1 - - -def load_definition(name: str) -> Definition: - for op_dir in DEFINITIONS_DIR.iterdir(): - if op_dir.is_dir(): - def_file = op_dir / f"{name}.json" - if def_file.exists(): - return load_json_file(Definition, def_file) - raise FileNotFoundError(f"Definition {name} not found") - - -def compile_reference(reference_code: str): - namespace = {"torch": torch, "math": math} - exec(reference_code, namespace) - return namespace["run"] - - -def generate_random_inputs(batch_size, max_seq_len, device="cuda"): - seq_lens = torch.randint(1, max_seq_len + 1, (batch_size,), dtype=torch.int32, device=device) - total_pages = seq_lens.sum().item() - - kv_indptr = torch.zeros(batch_size + 1, dtype=torch.int32, device=device) - kv_indptr[1:] = torch.cumsum(seq_lens, dim=0) - kv_indices = torch.arange(total_pages, dtype=torch.int32, device=device) - kv_last_page_len = torch.ones(batch_size, dtype=torch.int32, device=device) - - q = torch.randn(batch_size, NUM_QO_HEADS, HEAD_DIM, dtype=torch.bfloat16, device=device) - num_cache_pages = total_pages + 100 - k_cache = torch.randn( - num_cache_pages, PAGE_SIZE, NUM_KV_HEADS, HEAD_DIM, dtype=torch.bfloat16, device=device - ) - v_cache = torch.randn( - num_cache_pages, PAGE_SIZE, NUM_KV_HEADS, HEAD_DIM, dtype=torch.bfloat16, device=device - ) - sm_scale = torch.tensor(1.0 / math.sqrt(HEAD_DIM), dtype=torch.float32, device=device) - - return { - "q": q, - "k_cache": k_cache, - "v_cache": v_cache, - "kv_indptr": kv_indptr, - "kv_indices": kv_indices, - "kv_last_page_len": kv_last_page_len, - "sm_scale": sm_scale, - } - - -def test_correctness(batch_size=4, max_seq_len=64, atol=1e-2, rtol=5e-2): - device = "cuda" if torch.cuda.is_available() else "cpu" - if device == "cpu": - return False - - definition = load_definition("gqa_paged_decode_h32_kv8_d64_ps1") - run = compile_reference(definition.reference) - inputs = generate_random_inputs(batch_size, max_seq_len, device) - - ref_o, ref_lse = run( - inputs["q"], - inputs["k_cache"], - inputs["v_cache"], - inputs["kv_indptr"], - inputs["kv_indices"], - inputs["sm_scale"], - ) - - workspace = torch.empty(128 * 1024 * 1024, dtype=torch.uint8, device=device) - wrapper = flashinfer.BatchDecodeWithPagedKVCacheWrapper(workspace, kv_layout="NHD") - wrapper.plan( - indptr=inputs["kv_indptr"], - indices=inputs["kv_indices"], - last_page_len=inputs["kv_last_page_len"], - num_qo_heads=NUM_QO_HEADS, - num_kv_heads=NUM_KV_HEADS, - head_dim=HEAD_DIM, - page_size=PAGE_SIZE, - pos_encoding_mode="NONE", - q_data_type=torch.bfloat16, - kv_data_type=torch.bfloat16, - sm_scale=inputs["sm_scale"].item(), - ) - fi_o, fi_lse = wrapper.run(inputs["q"], (inputs["k_cache"], inputs["v_cache"]), return_lse=True) - - out_ok = torch.allclose(ref_o.float(), fi_o.float(), atol=atol, rtol=rtol) - lse_ok = torch.allclose(ref_lse, fi_lse, atol=atol, rtol=rtol) - return out_ok and lse_ok - - -def main(): - configs = [(1, 16), (4, 64), (8, 128)] - passed = sum(1 for b, s in configs if test_correctness(b, s)) - print(f"{passed}/{len(configs)} passed") - - -if __name__ == "__main__": - main() diff --git a/tests/references/test_gqa_paged_decode_h32_kv8_d64_ps64.py b/tests/references/test_gqa_paged_decode_h32_kv8_d64_ps64.py deleted file mode 100644 index e9d9f778a388d97ce782f6d4b3272accb991aa1e..0000000000000000000000000000000000000000 --- a/tests/references/test_gqa_paged_decode_h32_kv8_d64_ps64.py +++ /dev/null @@ -1,113 +0,0 @@ -"""Reference test for gqa_paged_decode_h32_kv8_d64_ps64 (Llama 3.2 1B).""" - -import math -from pathlib import Path - -import flashinfer -import torch - -from flashinfer_bench.data import Definition, load_json_file - -DEFINITIONS_DIR = Path(__file__).parent.parent.parent / "definitions" - -NUM_QO_HEADS = 32 -NUM_KV_HEADS = 8 -HEAD_DIM = 64 -PAGE_SIZE = 64 - - -def load_definition(name: str) -> Definition: - for op_dir in DEFINITIONS_DIR.iterdir(): - if op_dir.is_dir(): - def_file = op_dir / f"{name}.json" - if def_file.exists(): - return load_json_file(Definition, def_file) - raise FileNotFoundError(f"Definition {name} not found") - - -def compile_reference(reference_code: str): - namespace = {"torch": torch, "math": math} - exec(reference_code, namespace) - return namespace["run"] - - -def generate_random_inputs(batch_size, max_seq_len, device="cuda"): - seq_lens = torch.randint(1, max_seq_len + 1, (batch_size,), dtype=torch.int32, device=device) - num_pages_per_seq = (seq_lens + PAGE_SIZE - 1) // PAGE_SIZE - total_pages = num_pages_per_seq.sum().item() - - kv_indptr = torch.zeros(batch_size + 1, dtype=torch.int32, device=device) - kv_indptr[1:] = torch.cumsum(num_pages_per_seq, dim=0) - kv_indices = torch.arange(total_pages, dtype=torch.int32, device=device) - kv_last_page_len = (seq_lens - 1) % PAGE_SIZE + 1 - - q = torch.randn(batch_size, NUM_QO_HEADS, HEAD_DIM, dtype=torch.bfloat16, device=device) - num_cache_pages = total_pages + 100 - k_cache = torch.randn( - num_cache_pages, PAGE_SIZE, NUM_KV_HEADS, HEAD_DIM, dtype=torch.bfloat16, device=device - ) - v_cache = torch.randn( - num_cache_pages, PAGE_SIZE, NUM_KV_HEADS, HEAD_DIM, dtype=torch.bfloat16, device=device - ) - sm_scale = torch.tensor(1.0 / math.sqrt(HEAD_DIM), dtype=torch.float32, device=device) - - return { - "q": q, - "k_cache": k_cache, - "v_cache": v_cache, - "kv_indptr": kv_indptr, - "kv_indices": kv_indices, - "kv_last_page_len": kv_last_page_len, - "sm_scale": sm_scale, - } - - -def test_correctness(batch_size=4, max_seq_len=256, atol=1e-2, rtol=5e-2): - device = "cuda" if torch.cuda.is_available() else "cpu" - if device == "cpu": - return False - - definition = load_definition("gqa_paged_decode_h32_kv8_d64_ps64") - run = compile_reference(definition.reference) - inputs = generate_random_inputs(batch_size, max_seq_len, device) - - ref_o, ref_lse = run( - inputs["q"], - inputs["k_cache"], - inputs["v_cache"], - inputs["kv_indptr"], - inputs["kv_indices"], - inputs["kv_last_page_len"], - inputs["sm_scale"], - ) - - workspace = torch.empty(128 * 1024 * 1024, dtype=torch.uint8, device=device) - wrapper = flashinfer.BatchDecodeWithPagedKVCacheWrapper(workspace, kv_layout="NHD") - wrapper.plan( - indptr=inputs["kv_indptr"], - indices=inputs["kv_indices"], - last_page_len=inputs["kv_last_page_len"], - num_qo_heads=NUM_QO_HEADS, - num_kv_heads=NUM_KV_HEADS, - head_dim=HEAD_DIM, - page_size=PAGE_SIZE, - pos_encoding_mode="NONE", - q_data_type=torch.bfloat16, - kv_data_type=torch.bfloat16, - sm_scale=inputs["sm_scale"].item(), - ) - fi_o, fi_lse = wrapper.run(inputs["q"], (inputs["k_cache"], inputs["v_cache"]), return_lse=True) - - out_ok = torch.allclose(ref_o.float(), fi_o.float(), atol=atol, rtol=rtol) - lse_ok = torch.allclose(ref_lse, fi_lse, atol=atol, rtol=rtol) - return out_ok and lse_ok - - -def main(): - configs = [(1, 16), (4, 256), (8, 512)] - passed = sum(1 for b, s in configs if test_correctness(b, s)) - print(f"{passed}/{len(configs)} passed") - - -if __name__ == "__main__": - main() diff --git a/tests/references/test_gqa_paged_prefill_causal_h32_kv8_d64_ps1.py b/tests/references/test_gqa_paged_prefill_causal_h32_kv8_d64_ps1.py deleted file mode 100644 index cb70bd6e33dec6ddd48d28e0179bf166dfac1e2d..0000000000000000000000000000000000000000 --- a/tests/references/test_gqa_paged_prefill_causal_h32_kv8_d64_ps1.py +++ /dev/null @@ -1,119 +0,0 @@ -"""Reference test for gqa_paged_prefill_causal_h32_kv8_d64_ps1 (Llama 3.2 1B).""" - -import math -from pathlib import Path - -import flashinfer -import torch - -from flashinfer_bench.data import Definition, load_json_file - -DEFINITIONS_DIR = Path(__file__).parent.parent.parent / "definitions" - -NUM_QO_HEADS = 32 -NUM_KV_HEADS = 8 -HEAD_DIM = 64 -PAGE_SIZE = 1 - - -def load_definition(name: str) -> Definition: - for op_dir in DEFINITIONS_DIR.iterdir(): - if op_dir.is_dir(): - def_file = op_dir / f"{name}.json" - if def_file.exists(): - return load_json_file(Definition, def_file) - raise FileNotFoundError(f"Definition {name} not found") - - -def compile_reference(reference_code: str): - namespace = {"torch": torch, "math": math} - exec(reference_code, namespace) - return namespace["run"] - - -def generate_random_inputs(batch_size, max_seq_len, device="cuda"): - total_q_per_seq = torch.randint( - 1, max_seq_len + 1, (batch_size,), dtype=torch.int32, device=device - ) - total_q = total_q_per_seq.sum().item() - total_pages = total_q_per_seq.sum().item() - kv_indptr = torch.zeros(batch_size + 1, dtype=torch.int32, device=device) - kv_indptr[1:] = torch.cumsum(total_q_per_seq, dim=0) - kv_indices = torch.arange(total_pages, dtype=torch.int32, device=device) - - qo_indptr = torch.zeros(batch_size + 1, dtype=torch.int32, device=device) - qo_indptr[1:] = torch.cumsum(total_q_per_seq, dim=0) - - q = torch.randn(total_q, NUM_QO_HEADS, HEAD_DIM, dtype=torch.bfloat16, device=device) - num_cache_pages = total_pages + 100 - k_cache = torch.randn( - num_cache_pages, PAGE_SIZE, NUM_KV_HEADS, HEAD_DIM, dtype=torch.bfloat16, device=device - ) - v_cache = torch.randn( - num_cache_pages, PAGE_SIZE, NUM_KV_HEADS, HEAD_DIM, dtype=torch.bfloat16, device=device - ) - sm_scale = torch.tensor(1.0 / math.sqrt(HEAD_DIM), dtype=torch.float32, device=device) - - return { - "q": q, - "k_cache": k_cache, - "v_cache": v_cache, - "qo_indptr": qo_indptr, - "kv_indptr": kv_indptr, - "kv_indices": kv_indices, - "sm_scale": sm_scale, - } - - -def test_correctness(batch_size=2, max_seq_len=64, atol=1e-2, rtol=5e-2): - device = "cuda" if torch.cuda.is_available() else "cpu" - if device == "cpu": - return False - - definition = load_definition("gqa_paged_prefill_causal_h32_kv8_d64_ps1") - run = compile_reference(definition.reference) - inputs = generate_random_inputs(batch_size, max_seq_len, device) - - ref_o, ref_lse = run( - inputs["q"], - inputs["k_cache"], - inputs["v_cache"], - inputs["qo_indptr"], - inputs["kv_indptr"], - inputs["kv_indices"], - inputs["sm_scale"], - ) - - workspace = torch.empty(512 * 1024 * 1024, dtype=torch.uint8, device=device) - wrapper = flashinfer.BatchPrefillWithPagedKVCacheWrapper(workspace, kv_layout="NHD") - wrapper.plan( - qo_indptr=inputs["qo_indptr"], - paged_kv_indptr=inputs["kv_indptr"], - paged_kv_indices=inputs["kv_indices"], - paged_kv_last_page_len=torch.ones( - inputs["kv_indptr"].shape[0] - 1, dtype=torch.int32, device=device - ), - num_qo_heads=NUM_QO_HEADS, - num_kv_heads=NUM_KV_HEADS, - head_dim_qk=HEAD_DIM, - page_size=PAGE_SIZE, - causal=True, - q_data_type=torch.bfloat16, - kv_data_type=torch.bfloat16, - sm_scale=inputs["sm_scale"].item(), - ) - fi_o, fi_lse = wrapper.run(inputs["q"], (inputs["k_cache"], inputs["v_cache"]), return_lse=True) - - out_ok = torch.allclose(ref_o.float(), fi_o.float(), atol=atol, rtol=rtol) - lse_ok = torch.allclose(ref_lse, fi_lse, atol=atol, rtol=rtol) - return out_ok and lse_ok - - -def main(): - configs = [(1, 16), (2, 64)] - passed = sum(1 for b, s in configs if test_correctness(b, s)) - print(f"{passed}/{len(configs)} passed") - - -if __name__ == "__main__": - main() diff --git a/tests/references/test_gqa_paged_prefill_causal_h32_kv8_d64_ps64.py b/tests/references/test_gqa_paged_prefill_causal_h32_kv8_d64_ps64.py deleted file mode 100644 index c7140296d51b372e9514023142a37f81f94be018..0000000000000000000000000000000000000000 --- a/tests/references/test_gqa_paged_prefill_causal_h32_kv8_d64_ps64.py +++ /dev/null @@ -1,132 +0,0 @@ -"""Reference test for gqa_paged_prefill_causal_h32_kv8_d64_ps64 (Llama 3.2 1B).""" - -import math -from pathlib import Path - -import flashinfer -import torch - -from flashinfer_bench.data import Definition, load_json_file - -DEFINITIONS_DIR = Path(__file__).parent.parent.parent / "definitions" - -NUM_QO_HEADS = 32 -NUM_KV_HEADS = 8 -HEAD_DIM = 64 -PAGE_SIZE = 64 - - -def load_definition(name: str) -> Definition: - for op_dir in DEFINITIONS_DIR.iterdir(): - if op_dir.is_dir(): - def_file = op_dir / f"{name}.json" - if def_file.exists(): - return load_json_file(Definition, def_file) - raise FileNotFoundError(f"Definition {name} not found") - - -def compile_reference(reference_code: str): - namespace = {"torch": torch, "math": math} - exec(reference_code, namespace) - return namespace["run"] - - -def generate_random_inputs(batch_size, max_q_len, max_kv_len, max_pages, device="cuda"): - q_lens = torch.randint(1, max_q_len + 1, (batch_size,), dtype=torch.int32) - kv_lens = torch.zeros(batch_size, dtype=torch.int32) - for i in range(batch_size): - kv_lens[i] = torch.randint(q_lens[i].item(), max_kv_len + 1, (1,)).item() - - qo_indptr = torch.zeros(batch_size + 1, dtype=torch.int32, device=device) - qo_indptr[1:] = torch.cumsum(q_lens.to(device), dim=0) - - kv_pages_per_seq = (kv_lens + PAGE_SIZE - 1) // PAGE_SIZE - kv_indptr = torch.zeros(batch_size + 1, dtype=torch.int32, device=device) - kv_indptr[1:] = torch.cumsum(kv_pages_per_seq.to(device), dim=0) - - total_q = int(qo_indptr[-1].item()) - num_kv_pages = int(kv_indptr[-1].item()) - - kv_indices = torch.arange(num_kv_pages, dtype=torch.int32, device=device) - kv_last_page_len = ((kv_lens - 1) % PAGE_SIZE + 1).to(torch.int32).to(device) - - k_cache = torch.randn( - max_pages, PAGE_SIZE, NUM_KV_HEADS, HEAD_DIM, dtype=torch.bfloat16, device=device - ) - v_cache = torch.randn( - max_pages, PAGE_SIZE, NUM_KV_HEADS, HEAD_DIM, dtype=torch.bfloat16, device=device - ) - q = torch.randn(total_q, NUM_QO_HEADS, HEAD_DIM, dtype=torch.bfloat16, device=device) - - sm_scale = torch.tensor(1.0 / math.sqrt(HEAD_DIM), dtype=torch.float32, device=device) - - return { - "q": q, - "k_cache": k_cache, - "v_cache": v_cache, - "qo_indptr": qo_indptr, - "kv_indptr": kv_indptr, - "kv_indices": kv_indices, - "kv_last_page_len": kv_last_page_len, - "q_lens": q_lens, - "kv_lens": kv_lens, - "sm_scale": sm_scale, - } - - -def test_correctness(batch_size=4, max_q_len=32, max_kv_len=128, atol=1e-2, rtol=5e-2): - device = "cuda" if torch.cuda.is_available() else "cpu" - if device == "cpu": - return False - - definition = load_definition("gqa_paged_prefill_causal_h32_kv8_d64_ps64") - run = compile_reference(definition.reference) - - max_pages = (max_kv_len * batch_size * 2 + PAGE_SIZE - 1) // PAGE_SIZE + 10 - inputs = generate_random_inputs(batch_size, max_q_len, max_kv_len, max_pages, device) - - ref_o, ref_lse = run( - inputs["q"], - inputs["k_cache"], - inputs["v_cache"], - inputs["qo_indptr"], - inputs["kv_indptr"], - inputs["kv_indices"], - inputs["kv_last_page_len"], - inputs["sm_scale"], - ) - - workspace = torch.empty(512 * 1024 * 1024, dtype=torch.uint8, device=device) - wrapper = flashinfer.BatchPrefillWithPagedKVCacheWrapper(workspace, kv_layout="NHD") - paged_kv_cache = torch.stack([inputs["k_cache"], inputs["v_cache"]], dim=1) - - wrapper.plan( - qo_indptr=inputs["qo_indptr"], - paged_kv_indptr=inputs["kv_indptr"], - paged_kv_indices=inputs["kv_indices"], - paged_kv_last_page_len=inputs["kv_last_page_len"], - num_qo_heads=NUM_QO_HEADS, - num_kv_heads=NUM_KV_HEADS, - head_dim_qk=HEAD_DIM, - page_size=PAGE_SIZE, - causal=True, - q_data_type=torch.bfloat16, - kv_data_type=torch.bfloat16, - sm_scale=inputs["sm_scale"].item(), - ) - fi_o, fi_lse = wrapper.run(inputs["q"], paged_kv_cache, return_lse=True) - - out_ok = torch.allclose(ref_o.float(), fi_o.float(), atol=atol, rtol=rtol) - lse_ok = torch.allclose(ref_lse, fi_lse, atol=atol, rtol=rtol) - assert out_ok and lse_ok, f"output_close={out_ok}, lse_close={lse_ok}" - return out_ok and lse_ok - - -def main(): - configs = [(1, 16, 64), (4, 32, 128), (8, 64, 256)] - passed = sum(1 for b, q, k in configs if test_correctness(b, q, k)) - print(f"{passed}/{len(configs)} passed") - - -if __name__ == "__main__": - main() diff --git a/tests/references/test_gqa_ragged_prefill_causal_h32_kv8_d64.py b/tests/references/test_gqa_ragged_prefill_causal_h32_kv8_d64.py deleted file mode 100644 index 74cc6b1a073226bbd008063c72ab2b1d8c660e76..0000000000000000000000000000000000000000 --- a/tests/references/test_gqa_ragged_prefill_causal_h32_kv8_d64.py +++ /dev/null @@ -1,111 +0,0 @@ -"""Reference test for gqa_ragged_prefill_causal_h32_kv8_d64 (Llama 3.2 1B).""" - -import math -from pathlib import Path - -import flashinfer -import torch - -from flashinfer_bench.data import Definition, load_json_file - -DEFINITIONS_DIR = Path(__file__).parent.parent.parent / "definitions" - -NUM_QO_HEADS = 32 -NUM_KV_HEADS = 8 -HEAD_DIM = 64 - - -def load_definition(name: str) -> Definition: - for op_dir in DEFINITIONS_DIR.iterdir(): - if op_dir.is_dir(): - def_file = op_dir / f"{name}.json" - if def_file.exists(): - return load_json_file(Definition, def_file) - raise FileNotFoundError(f"Definition {name} not found in {DEFINITIONS_DIR}") - - -def compile_reference(reference_code: str): - namespace = {"torch": torch, "math": math} - exec(reference_code, namespace) - return namespace["run"] - - -def generate_random_inputs(batch_size, max_q_len, max_kv_len, device="cuda"): - q_lens = torch.randint(1, max_q_len + 1, (batch_size,), dtype=torch.int32) - kv_lens = torch.zeros(batch_size, dtype=torch.int32) - for i in range(batch_size): - kv_lens[i] = torch.randint(q_lens[i].item(), max_kv_len + 1, (1,)).item() - - qo_indptr = torch.zeros(batch_size + 1, dtype=torch.int32, device=device) - qo_indptr[1:] = torch.cumsum(q_lens.to(device), dim=0) - - kv_indptr = torch.zeros(batch_size + 1, dtype=torch.int32, device=device) - kv_indptr[1:] = torch.cumsum(kv_lens.to(device), dim=0) - - total_q = int(qo_indptr[-1].item()) - total_kv = int(kv_indptr[-1].item()) - - q = torch.randn(total_q, NUM_QO_HEADS, HEAD_DIM, dtype=torch.bfloat16, device=device) - k = torch.randn(total_kv, NUM_KV_HEADS, HEAD_DIM, dtype=torch.bfloat16, device=device) - v = torch.randn(total_kv, NUM_KV_HEADS, HEAD_DIM, dtype=torch.bfloat16, device=device) - sm_scale = torch.tensor(1.0 / math.sqrt(HEAD_DIM), dtype=torch.float32, device=device) - - return { - "q": q, - "k": k, - "v": v, - "qo_indptr": qo_indptr, - "kv_indptr": kv_indptr, - "sm_scale": sm_scale, - } - - -def test_correctness(batch_size=4, max_q_len=32, max_kv_len=64, atol=1e-2, rtol=5e-2): - device = "cuda" if torch.cuda.is_available() else "cpu" - if device == "cpu": - return False - - definition = load_definition("gqa_ragged_prefill_causal_h32_kv8_d64") - run = compile_reference(definition.reference) - inputs = generate_random_inputs(batch_size, max_q_len, max_kv_len, device) - - ref_o, ref_lse = run( - inputs["q"], - inputs["k"], - inputs["v"], - inputs["qo_indptr"], - inputs["kv_indptr"], - inputs["sm_scale"], - ) - - workspace_buffer = torch.empty(128 * 1024 * 1024, dtype=torch.uint8, device=device) - prefill_wrapper = flashinfer.prefill.BatchPrefillWithRaggedKVCacheWrapper( - workspace_buffer, kv_layout="NHD" - ) - prefill_wrapper.plan( - qo_indptr=inputs["qo_indptr"], - kv_indptr=inputs["kv_indptr"], - num_qo_heads=NUM_QO_HEADS, - num_kv_heads=NUM_KV_HEADS, - head_dim_qk=HEAD_DIM, - head_dim_vo=HEAD_DIM, - causal=True, - sm_scale=inputs["sm_scale"].item(), - q_data_type=torch.bfloat16, - kv_data_type=torch.bfloat16, - ) - fi_output, fi_lse = prefill_wrapper.run(inputs["q"], inputs["k"], inputs["v"], return_lse=True) - - out_ok = torch.allclose(ref_o.float(), fi_output.float(), atol=atol, rtol=rtol) - lse_ok = torch.allclose(ref_lse, fi_lse, atol=atol, rtol=rtol) - return out_ok and lse_ok - - -def main(): - configs = [(1, 16, 32), (4, 32, 64), (8, 64, 128)] - passed = sum(1 for b, q, k in configs if test_correctness(b, q, k)) - print(f"{passed}/{len(configs)} passed") - - -if __name__ == "__main__": - main() diff --git a/tests/references/test_top_k_sampling_from_probs_v163840.py b/tests/references/test_top_k_sampling_from_probs_v163840.py deleted file mode 100644 index 6240c4a970af38bf29da923a229cac51a57115f4..0000000000000000000000000000000000000000 --- a/tests/references/test_top_k_sampling_from_probs_v163840.py +++ /dev/null @@ -1,91 +0,0 @@ -"""Reference test for top_k_sampling_from_probs_v163840 (Kimi K2.5).""" - -import math -from pathlib import Path - -import flashinfer -import torch - -from flashinfer_bench.data import Definition, load_json_file - -DEFINITIONS_DIR = Path(__file__).parent.parent.parent / "definitions" - -VOCAB_SIZE = 163840 - - -def load_definition(name: str) -> Definition: - for op_dir in DEFINITIONS_DIR.iterdir(): - if op_dir.is_dir(): - def_file = op_dir / f"{name}.json" - if def_file.exists(): - return load_json_file(Definition, def_file) - raise FileNotFoundError(f"Definition {name} not found in {DEFINITIONS_DIR}") - - -def compile_reference(reference_code: str): - namespace = {"torch": torch, "math": math} - exec(reference_code, namespace) - return namespace["run"] - - -def generate_random_inputs(batch_size, distribution="peaked", device="cuda"): - if distribution == "peaked": - logits = torch.randn(batch_size, VOCAB_SIZE, device=device) * 0.1 - peak_indices = torch.randint(0, VOCAB_SIZE, (batch_size,), device=device) - for i in range(batch_size): - logits[i, peak_indices[i]] += 5.0 - else: - logits = torch.randn(batch_size, VOCAB_SIZE, device=device) - - probs = torch.softmax(logits, dim=-1).to(torch.float32) - top_k = torch.randint( - 10, min(500, VOCAB_SIZE // 2), (batch_size,), dtype=torch.int32, device=device - ) - return probs, top_k - - -def test_correctness(batch_size=4, num_trials=5000): - device = "cuda" if torch.cuda.is_available() else "cpu" - if device == "cpu": - print("WARNING: CUDA not available, skipping test") - return False - - definition = load_definition("top_k_sampling_from_probs_v163840") - run = compile_reference(definition.reference) - - torch.manual_seed(42) - probs, top_k = generate_random_inputs(batch_size, "peaked", device) - - ref_counter = torch.zeros(batch_size, VOCAB_SIZE, dtype=torch.int32, device=device) - fi_counter = torch.zeros(batch_size, VOCAB_SIZE, dtype=torch.int32, device=device) - - for _ in range(num_trials): - ref_samples = run(probs.clone(), top_k) - fi_samples = flashinfer.sampling.top_k_sampling_from_probs(probs, top_k) - - for i in range(batch_size): - ref_counter[i, ref_samples[i]] += 1 - fi_counter[i, fi_samples[i]] += 1 - - ref_freq = ref_counter.float() / num_trials - fi_freq = fi_counter.float() / num_trials - - nonzero_mask = probs > 1e-6 - freq_diff = torch.abs(ref_freq[nonzero_mask] - fi_freq[nonzero_mask]).max().item() - - passed = freq_diff < 0.05 - print( - f"batch_size={batch_size}: max_freq_diff={freq_diff:.4f} " - f"{'PASSED' if passed else 'FAILED'}" - ) - return passed - - -def main(): - test_configs = [(1, 5000), (4, 5000), (8, 3000)] - passed = sum(1 for b, t in test_configs if test_correctness(b, t)) - print(f"\nSummary: {passed}/{len(test_configs)} tests passed") - - -if __name__ == "__main__": - main() diff --git a/tests/references/test_top_k_top_p_sampling_from_probs_v163840.py b/tests/references/test_top_k_top_p_sampling_from_probs_v163840.py deleted file mode 100644 index 79bd78280276add04f9d903446ce50ff6d5c5249..0000000000000000000000000000000000000000 --- a/tests/references/test_top_k_top_p_sampling_from_probs_v163840.py +++ /dev/null @@ -1,92 +0,0 @@ -"""Reference test for top_k_top_p_sampling_from_probs_v163840 (Kimi K2.5).""" - -import math -from pathlib import Path - -import flashinfer -import torch - -from flashinfer_bench.data import Definition, load_json_file - -DEFINITIONS_DIR = Path(__file__).parent.parent.parent / "definitions" - -VOCAB_SIZE = 163840 - - -def load_definition(name: str) -> Definition: - for op_dir in DEFINITIONS_DIR.iterdir(): - if op_dir.is_dir(): - def_file = op_dir / f"{name}.json" - if def_file.exists(): - return load_json_file(Definition, def_file) - raise FileNotFoundError(f"Definition {name} not found in {DEFINITIONS_DIR}") - - -def compile_reference(reference_code: str): - namespace = {"torch": torch, "math": math} - exec(reference_code, namespace) - return namespace["run"] - - -def generate_random_inputs(batch_size, distribution="peaked", device="cuda"): - if distribution == "peaked": - logits = torch.randn(batch_size, VOCAB_SIZE, device=device) * 0.1 - peak_indices = torch.randint(0, VOCAB_SIZE, (batch_size,), device=device) - for i in range(batch_size): - logits[i, peak_indices[i]] += 5.0 - else: - logits = torch.randn(batch_size, VOCAB_SIZE, device=device) - - probs = torch.softmax(logits, dim=-1).to(torch.float32) - top_k = torch.randint( - 10, min(500, VOCAB_SIZE // 2), (batch_size,), dtype=torch.int32, device=device - ) - top_p = torch.rand(batch_size, device=device) * 0.8 + 0.1 # Range [0.1, 0.9] - return probs, top_k, top_p - - -def test_correctness(batch_size=4, num_trials=5000): - device = "cuda" if torch.cuda.is_available() else "cpu" - if device == "cpu": - print("WARNING: CUDA not available, skipping test") - return False - - definition = load_definition("top_k_top_p_sampling_from_probs_v163840") - run = compile_reference(definition.reference) - - torch.manual_seed(42) - probs, top_k, top_p = generate_random_inputs(batch_size, "peaked", device) - - ref_counter = torch.zeros(batch_size, VOCAB_SIZE, dtype=torch.int32, device=device) - fi_counter = torch.zeros(batch_size, VOCAB_SIZE, dtype=torch.int32, device=device) - - for _ in range(num_trials): - ref_samples = run(probs.clone(), top_k, top_p) - fi_samples = flashinfer.sampling.top_k_top_p_sampling_from_probs(probs, top_k, top_p) - - for i in range(batch_size): - ref_counter[i, ref_samples[i]] += 1 - fi_counter[i, fi_samples[i]] += 1 - - ref_freq = ref_counter.float() / num_trials - fi_freq = fi_counter.float() / num_trials - - nonzero_mask = probs > 1e-6 - freq_diff = torch.abs(ref_freq[nonzero_mask] - fi_freq[nonzero_mask]).max().item() - - passed = freq_diff < 0.05 - print( - f"batch_size={batch_size}: max_freq_diff={freq_diff:.4f} " - f"{'PASSED' if passed else 'FAILED'}" - ) - return passed - - -def main(): - test_configs = [(1, 5000), (4, 5000), (8, 3000)] - passed = sum(1 for b, t in test_configs if test_correctness(b, t)) - print(f"\nSummary: {passed}/{len(test_configs)} tests passed") - - -if __name__ == "__main__": - main() diff --git a/tests/references/test_top_p_sampling_from_probs_v163840.py b/tests/references/test_top_p_sampling_from_probs_v163840.py deleted file mode 100644 index 40c23efec5b6b9b98cb035cbe280531d9a77d1e3..0000000000000000000000000000000000000000 --- a/tests/references/test_top_p_sampling_from_probs_v163840.py +++ /dev/null @@ -1,89 +0,0 @@ -"""Reference test for top_p_sampling_from_probs_v163840 (Kimi K2.5).""" - -import math -from pathlib import Path - -import flashinfer -import torch - -from flashinfer_bench.data import Definition, load_json_file - -DEFINITIONS_DIR = Path(__file__).parent.parent.parent / "definitions" - -VOCAB_SIZE = 163840 - - -def load_definition(name: str) -> Definition: - for op_dir in DEFINITIONS_DIR.iterdir(): - if op_dir.is_dir(): - def_file = op_dir / f"{name}.json" - if def_file.exists(): - return load_json_file(Definition, def_file) - raise FileNotFoundError(f"Definition {name} not found in {DEFINITIONS_DIR}") - - -def compile_reference(reference_code: str): - namespace = {"torch": torch, "math": math} - exec(reference_code, namespace) - return namespace["run"] - - -def generate_random_inputs(batch_size, distribution="peaked", device="cuda"): - if distribution == "peaked": - logits = torch.randn(batch_size, VOCAB_SIZE, device=device) * 0.1 - peak_indices = torch.randint(0, VOCAB_SIZE, (batch_size,), device=device) - for i in range(batch_size): - logits[i, peak_indices[i]] += 5.0 - else: - logits = torch.randn(batch_size, VOCAB_SIZE, device=device) - - probs = torch.softmax(logits, dim=-1).to(torch.float32) - top_p = torch.rand(batch_size, device=device) * 0.8 + 0.1 # Range [0.1, 0.9] - return probs, top_p - - -def test_correctness(batch_size=4, num_trials=5000): - device = "cuda" if torch.cuda.is_available() else "cpu" - if device == "cpu": - print("WARNING: CUDA not available, skipping test") - return False - - definition = load_definition("top_p_sampling_from_probs_v163840") - run = compile_reference(definition.reference) - - torch.manual_seed(42) - probs, top_p = generate_random_inputs(batch_size, "peaked", device) - - ref_counter = torch.zeros(batch_size, VOCAB_SIZE, dtype=torch.int32, device=device) - fi_counter = torch.zeros(batch_size, VOCAB_SIZE, dtype=torch.int32, device=device) - - for _ in range(num_trials): - ref_samples = run(probs.clone(), top_p) - fi_samples = flashinfer.sampling.top_p_sampling_from_probs(probs, top_p) - - for i in range(batch_size): - ref_counter[i, ref_samples[i]] += 1 - fi_counter[i, fi_samples[i]] += 1 - - ref_freq = ref_counter.float() / num_trials - fi_freq = fi_counter.float() / num_trials - - nonzero_mask = probs > 1e-6 - freq_diff = torch.abs(ref_freq[nonzero_mask] - fi_freq[nonzero_mask]).max().item() - - passed = freq_diff < 0.05 - print( - f"batch_size={batch_size}: max_freq_diff={freq_diff:.4f} " - f"{'PASSED' if passed else 'FAILED'}" - ) - return passed - - -def main(): - test_configs = [(1, 5000), (4, 5000), (8, 3000)] - passed = sum(1 for b, t in test_configs if test_correctness(b, t)) - print(f"\nSummary: {passed}/{len(test_configs)} tests passed") - - -if __name__ == "__main__": - main() diff --git a/traces/baseline/gemm/gemm_n16384_k2048.jsonl b/traces/baseline/gemm/gemm_n16384_k2048.jsonl deleted file mode 100644 index 2a7fd5f3681274f24655221f606d43400749951b..0000000000000000000000000000000000000000 --- a/traces/baseline/gemm/gemm_n16384_k2048.jsonl +++ /dev/null @@ -1,43 +0,0 @@ -{"definition": "gemm_n16384_k2048", "workload": {"axes": {"M": 1}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}, "uuid": "63ed933a-ed07-4d64-bb08-d22a18f781bc"}, "solution": "torch_matmul_0a4e73", "evaluation": {"status": "PASSED", "environment": {"hardware": "NVIDIA B200", "libs": {"torch": "2.11.0+cu130", "triton": "3.6.0", "cuda": "13.0"}}, "timestamp": "2026-04-20T07:33:05.365834", "log": "", "correctness": {"max_relative_error": 0.0, "max_absolute_error": 0.0, "extra": null}, "performance": {"latency_ms": 0.020986666282018025, "reference_latency_ms": 0.021130666447182495, "speedup_factor": 1.006861507360407}}} -{"definition": "gemm_n16384_k2048", "workload": {"axes": {"M": 2}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}, "uuid": "9bcca24f-2a17-41c5-acbe-2aa18a0d69a3"}, "solution": "torch_matmul_0a4e73", "evaluation": {"status": "PASSED", "environment": {"hardware": "NVIDIA B200", "libs": {"torch": "2.11.0+cu130", "triton": "3.6.0", "cuda": "13.0"}}, "timestamp": "2026-04-20T07:33:05.495486", "log": "", "correctness": {"max_relative_error": 0.0, "max_absolute_error": 0.0, "extra": null}, "performance": {"latency_ms": 0.02110399988790353, "reference_latency_ms": 0.02102399927874406, "speedup_factor": 0.9962092205466072}}} -{"definition": "gemm_n16384_k2048", "workload": {"axes": {"M": 4}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}, "uuid": "56fe737e-6e40-493e-b0ca-b34a6fb73425"}, "solution": "torch_matmul_0a4e73", "evaluation": {"status": "PASSED", "environment": {"hardware": "NVIDIA B200", "libs": {"torch": "2.11.0+cu130", "triton": "3.6.0", "cuda": "13.0"}}, "timestamp": "2026-04-20T07:33:05.540435", "log": "", "correctness": {"max_relative_error": 0.0, "max_absolute_error": 0.0, "extra": null}, "performance": {"latency_ms": 0.02099199965596199, "reference_latency_ms": 0.020938665916522343, "speedup_factor": 0.9974593302060911}}} -{"definition": "gemm_n16384_k2048", "workload": {"axes": {"M": 7}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}, "uuid": "787ef2c4-9368-4b71-96e2-d31ac9ee0ec9"}, "solution": "torch_matmul_0a4e73", "evaluation": {"status": "PASSED", "environment": {"hardware": "NVIDIA B200", "libs": {"torch": "2.11.0+cu130", "triton": "3.6.0", "cuda": "13.0"}}, "timestamp": "2026-04-20T07:33:05.584596", "log": "", "correctness": {"max_relative_error": 0.0, "max_absolute_error": 0.0, "extra": null}, "performance": {"latency_ms": 0.02099199965596199, "reference_latency_ms": 0.02096533278624217, "speedup_factor": 0.9987296651030456}}} -{"definition": "gemm_n16384_k2048", "workload": {"axes": {"M": 8}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}, "uuid": "46ce6192-af6c-4c02-b232-b4832028f974"}, "solution": "torch_matmul_0a4e73", "evaluation": {"status": "PASSED", "environment": {"hardware": "NVIDIA B200", "libs": {"torch": "2.11.0+cu130", "triton": "3.6.0", "cuda": "13.0"}}, "timestamp": "2026-04-20T07:33:05.628180", "log": "", "correctness": {"max_relative_error": 0.0, "max_absolute_error": 0.0, "extra": null}, "performance": {"latency_ms": 0.020981332908074062, "reference_latency_ms": 0.02093866653740406, "speedup_factor": 0.9979664604314256}}} -{"definition": "gemm_n16384_k2048", "workload": {"axes": {"M": 15}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}, "uuid": "4614c89c-dd77-4fc7-88a6-70c2b83c39fb"}, "solution": "torch_matmul_0a4e73", "evaluation": {"status": "PASSED", "environment": {"hardware": "NVIDIA B200", "libs": {"torch": "2.11.0+cu130", "triton": "3.6.0", "cuda": "13.0"}}, "timestamp": "2026-04-20T07:33:05.678377", "log": "", "correctness": {"max_relative_error": 0.0, "max_absolute_error": 0.0, "extra": null}, "performance": {"latency_ms": 0.020949332664410274, "reference_latency_ms": 0.02097066647062699, "speedup_factor": 1.0010183525441343}}} -{"definition": "gemm_n16384_k2048", "workload": {"axes": {"M": 16}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}, "uuid": "af46e3cb-fec0-42d7-979d-aa487341c1ee"}, "solution": "torch_matmul_0a4e73", "evaluation": {"status": "PASSED", "environment": {"hardware": "NVIDIA B200", "libs": {"torch": "2.11.0+cu130", "triton": "3.6.0", "cuda": "13.0"}}, "timestamp": "2026-04-20T07:33:05.724127", "log": "", "correctness": {"max_relative_error": 0.0, "max_absolute_error": 0.0, "extra": null}, "performance": {"latency_ms": 0.020986666282018025, "reference_latency_ms": 0.02094399929046631, "speedup_factor": 0.9979669476333993}}} -{"definition": "gemm_n16384_k2048", "workload": {"axes": {"M": 24}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}, "uuid": "ab02c26e-50e2-42fa-b666-b3a4c4bfbf6f"}, "solution": "torch_matmul_0a4e73", "evaluation": {"status": "PASSED", "environment": {"hardware": "NVIDIA B200", "libs": {"torch": "2.11.0+cu130", "triton": "3.6.0", "cuda": "13.0"}}, "timestamp": "2026-04-20T07:33:05.769622", "log": "", "correctness": {"max_relative_error": 0.0, "max_absolute_error": 0.0, "extra": null}, "performance": {"latency_ms": 0.021029333273569744, "reference_latency_ms": 0.021071999644239742, "speedup_factor": 1.0020288979262897}}} -{"definition": "gemm_n16384_k2048", "workload": {"axes": {"M": 32}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}, "uuid": "079da838-4ea5-42ff-8d57-808ae35c54d4"}, "solution": "torch_matmul_0a4e73", "evaluation": {"status": "PASSED", "environment": {"hardware": "NVIDIA B200", "libs": {"torch": "2.11.0+cu130", "triton": "3.6.0", "cuda": "13.0"}}, "timestamp": "2026-04-20T07:33:05.815192", "log": "", "correctness": {"max_relative_error": 0.0, "max_absolute_error": 0.0, "extra": null}, "performance": {"latency_ms": 0.020975999534130096, "reference_latency_ms": 0.02092266610513131, "speedup_factor": 0.9974574070278744}}} -{"definition": "gemm_n16384_k2048", "workload": {"axes": {"M": 35}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}, "uuid": "d2b99ac4-8563-437f-898f-a1f7a6f54456"}, "solution": "torch_matmul_0a4e73", "evaluation": {"status": "PASSED", "environment": {"hardware": "NVIDIA B200", "libs": {"torch": "2.11.0+cu130", "triton": "3.6.0", "cuda": "13.0"}}, "timestamp": "2026-04-20T07:33:05.860938", "log": "", "correctness": {"max_relative_error": 0.0, "max_absolute_error": 0.0, "extra": null}, "performance": {"latency_ms": 0.020975999534130096, "reference_latency_ms": 0.020949332664410274, "speedup_factor": 0.998728696114031}}} -{"definition": "gemm_n16384_k2048", "workload": {"axes": {"M": 40}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}, "uuid": "170d7c38-7775-4034-bee3-1c823b1c75cc"}, "solution": "torch_matmul_0a4e73", "evaluation": {"status": "PASSED", "environment": {"hardware": "NVIDIA B200", "libs": {"torch": "2.11.0+cu130", "triton": "3.6.0", "cuda": "13.0"}}, "timestamp": "2026-04-20T07:33:05.904442", "log": "", "correctness": {"max_relative_error": 0.0, "max_absolute_error": 0.0, "extra": null}, "performance": {"latency_ms": 0.020943999600907166, "reference_latency_ms": 0.020949332664410274, "speedup_factor": 1.0002546344349088}}} -{"definition": "gemm_n16384_k2048", "workload": {"axes": {"M": 48}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}, "uuid": "16349ad2-0ee1-4a5b-8d6d-5f0b7f92bc5b"}, "solution": "torch_matmul_0a4e73", "evaluation": {"status": "PASSED", "environment": {"hardware": "NVIDIA B200", "libs": {"torch": "2.11.0+cu130", "triton": "3.6.0", "cuda": "13.0"}}, "timestamp": "2026-04-20T07:33:05.968044", "log": "", "correctness": {"max_relative_error": 0.0, "max_absolute_error": 0.0, "extra": null}, "performance": {"latency_ms": 0.020965333407123882, "reference_latency_ms": 0.02104533277451992, "speedup_factor": 1.0038157927585762}}} -{"definition": "gemm_n16384_k2048", "workload": {"axes": {"M": 56}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}, "uuid": "47c493ac-b794-4825-8f10-f82a8c53f89e"}, "solution": "torch_matmul_0a4e73", "evaluation": {"status": "PASSED", "environment": {"hardware": "NVIDIA B200", "libs": {"torch": "2.11.0+cu130", "triton": "3.6.0", "cuda": "13.0"}}, "timestamp": "2026-04-20T07:33:06.013183", "log": "", "correctness": {"max_relative_error": 0.0, "max_absolute_error": 0.0, "extra": null}, "performance": {"latency_ms": 0.021007999777793884, "reference_latency_ms": 0.021061333517233532, "speedup_factor": 1.002538734767887}}} -{"definition": "gemm_n16384_k2048", "workload": {"axes": {"M": 64}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}, "uuid": "a5f2bcd9-1c45-418a-9482-68ec5be2cf7f"}, "solution": "torch_matmul_0a4e73", "evaluation": {"status": "PASSED", "environment": {"hardware": "NVIDIA B200", "libs": {"torch": "2.11.0+cu130", "triton": "3.6.0", "cuda": "13.0"}}, "timestamp": "2026-04-20T07:33:06.058560", "log": "", "correctness": {"max_relative_error": 0.0, "max_absolute_error": 0.0, "extra": null}, "performance": {"latency_ms": 0.020959999412298203, "reference_latency_ms": 0.021151999632517498, "speedup_factor": 1.0091603161070053}}} -{"definition": "gemm_n16384_k2048", "workload": {"axes": {"M": 70}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}, "uuid": "0638f5da-9fd9-4557-acd1-0dcd00deda64"}, "solution": "torch_matmul_0a4e73", "evaluation": {"status": "PASSED", "environment": {"hardware": "NVIDIA B200", "libs": {"torch": "2.11.0+cu130", "triton": "3.6.0", "cuda": "13.0"}}, "timestamp": "2026-04-20T07:33:06.103898", "log": "", "correctness": {"max_relative_error": 0.0, "max_absolute_error": 0.0, "extra": null}, "performance": {"latency_ms": 0.020949332664410274, "reference_latency_ms": 0.020943999600907166, "speedup_factor": 0.9997454303872806}}} -{"definition": "gemm_n16384_k2048", "workload": {"axes": {"M": 72}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}, "uuid": "be8d6610-82ac-4470-ad6a-8ae9f9c0efc1"}, "solution": "torch_matmul_0a4e73", "evaluation": {"status": "PASSED", "environment": {"hardware": "NVIDIA B200", "libs": {"torch": "2.11.0+cu130", "triton": "3.6.0", "cuda": "13.0"}}, "timestamp": "2026-04-20T07:33:06.147826", "log": "", "correctness": {"max_relative_error": 0.0, "max_absolute_error": 0.0, "extra": null}, "performance": {"latency_ms": 0.02100266640384992, "reference_latency_ms": 0.021141332884629566, "speedup_factor": 1.006602327443254}}} -{"definition": "gemm_n16384_k2048", "workload": {"axes": {"M": 80}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}, "uuid": "b2131315-fbe6-4c76-89f4-10a8ced7cbd3"}, "solution": "torch_matmul_0a4e73", "evaluation": {"status": "PASSED", "environment": {"hardware": "NVIDIA B200", "libs": {"torch": "2.11.0+cu130", "triton": "3.6.0", "cuda": "13.0"}}, "timestamp": "2026-04-20T07:33:06.195181", "log": "", "correctness": {"max_relative_error": 0.0, "max_absolute_error": 0.0, "extra": null}, "performance": {"latency_ms": 0.02094399929046631, "reference_latency_ms": 0.020927999168634415, "speedup_factor": 0.9992360522166759}}} -{"definition": "gemm_n16384_k2048", "workload": {"axes": {"M": 88}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}, "uuid": "a5a4c94b-587c-4d20-a45e-51eab114c101"}, "solution": "torch_matmul_0a4e73", "evaluation": {"status": "PASSED", "environment": {"hardware": "NVIDIA B200", "libs": {"torch": "2.11.0+cu130", "triton": "3.6.0", "cuda": "13.0"}}, "timestamp": "2026-04-20T07:33:06.241473", "log": "", "correctness": {"max_relative_error": 0.0, "max_absolute_error": 0.0, "extra": null}, "performance": {"latency_ms": 0.02165866643190384, "reference_latency_ms": 0.021882666895786922, "speedup_factor": 1.0103423017565443}}} -{"definition": "gemm_n16384_k2048", "workload": {"axes": {"M": 96}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}, "uuid": "d075a83c-1ef6-456b-809a-02c8c789cb38"}, "solution": "torch_matmul_0a4e73", "evaluation": {"status": "PASSED", "environment": {"hardware": "NVIDIA B200", "libs": {"torch": "2.11.0+cu130", "triton": "3.6.0", "cuda": "13.0"}}, "timestamp": "2026-04-20T07:33:06.287908", "log": "", "correctness": {"max_relative_error": 0.0, "max_absolute_error": 0.0, "extra": null}, "performance": {"latency_ms": 0.02258133329451084, "reference_latency_ms": 0.02247999981045723, "speedup_factor": 0.9955125110314791}}} -{"definition": "gemm_n16384_k2048", "workload": {"axes": {"M": 104}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}, "uuid": "b4fd51aa-621f-48d0-9a73-a9b7f0f63cca"}, "solution": "torch_matmul_0a4e73", "evaluation": {"status": "PASSED", "environment": {"hardware": "NVIDIA B200", "libs": {"torch": "2.11.0+cu130", "triton": "3.6.0", "cuda": "13.0"}}, "timestamp": "2026-04-20T07:33:06.389110", "log": "", "correctness": {"max_relative_error": 0.0, "max_absolute_error": 0.0, "extra": null}, "performance": {"latency_ms": 0.02168533330162366, "reference_latency_ms": 0.021882666895786922, "speedup_factor": 1.0090998644760736}}} -{"definition": "gemm_n16384_k2048", "workload": {"axes": {"M": 112}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}, "uuid": "6ca48022-63e7-41d6-aa97-c9ca71d0d4ce"}, "solution": "torch_matmul_0a4e73", "evaluation": {"status": "PASSED", "environment": {"hardware": "NVIDIA B200", "libs": {"torch": "2.11.0+cu130", "triton": "3.6.0", "cuda": "13.0"}}, "timestamp": "2026-04-20T07:33:06.435201", "log": "", "correctness": {"max_relative_error": 0.0, "max_absolute_error": 0.0, "extra": null}, "performance": {"latency_ms": 0.022240000156064827, "reference_latency_ms": 0.022522666802008946, "speedup_factor": 1.0127098311133347}}} -{"definition": "gemm_n16384_k2048", "workload": {"axes": {"M": 120}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}, "uuid": "ca3f0979-aea6-4964-a6c3-d140d8dc2185"}, "solution": "torch_matmul_0a4e73", "evaluation": {"status": "PASSED", "environment": {"hardware": "NVIDIA B200", "libs": {"torch": "2.11.0+cu130", "triton": "3.6.0", "cuda": "13.0"}}, "timestamp": "2026-04-20T07:33:06.481551", "log": "", "correctness": {"max_relative_error": 0.0, "max_absolute_error": 0.0, "extra": null}, "performance": {"latency_ms": 0.022474666436513264, "reference_latency_ms": 0.022848000129063923, "speedup_factor": 1.016611311834383}}} -{"definition": "gemm_n16384_k2048", "workload": {"axes": {"M": 128}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}, "uuid": "2ec930dd-2fed-4a4d-ac20-26c5e49c9b99"}, "solution": "torch_matmul_0a4e73", "evaluation": {"status": "PASSED", "environment": {"hardware": "NVIDIA B200", "libs": {"torch": "2.11.0+cu130", "triton": "3.6.0", "cuda": "13.0"}}, "timestamp": "2026-04-20T07:33:06.525355", "log": "", "correctness": {"max_relative_error": 0.0, "max_absolute_error": 0.0, "extra": null}, "performance": {"latency_ms": 0.022485333184401195, "reference_latency_ms": 0.022650666534900665, "speedup_factor": 1.00735294198861}}} -{"definition": "gemm_n16384_k2048", "workload": {"axes": {"M": 136}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}, "uuid": "c6e4803f-b5cb-4e9f-8f8b-a32abb32b085"}, "solution": "torch_matmul_0a4e73", "evaluation": {"status": "PASSED", "environment": {"hardware": "NVIDIA B200", "libs": {"torch": "2.11.0+cu130", "triton": "3.6.0", "cuda": "13.0"}}, "timestamp": "2026-04-20T07:33:06.573929", "log": "", "correctness": {"max_relative_error": 0.0, "max_absolute_error": 0.0, "extra": null}, "performance": {"latency_ms": 0.022474666436513264, "reference_latency_ms": 0.022848000129063923, "speedup_factor": 1.016611311834383}}} -{"definition": "gemm_n16384_k2048", "workload": {"axes": {"M": 144}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}, "uuid": "15185b21-70c1-4400-b24a-dc5eb3528df9"}, "solution": "torch_matmul_0a4e73", "evaluation": {"status": "PASSED", "environment": {"hardware": "NVIDIA B200", "libs": {"torch": "2.11.0+cu130", "triton": "3.6.0", "cuda": "13.0"}}, "timestamp": "2026-04-20T07:33:06.620601", "log": "", "correctness": {"max_relative_error": 0.0, "max_absolute_error": 0.0, "extra": null}, "performance": {"latency_ms": 0.02250133330623309, "reference_latency_ms": 0.022431999755402405, "speedup_factor": 0.9969186914443209}}} -{"definition": "gemm_n16384_k2048", "workload": {"axes": {"M": 152}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}, "uuid": "158af161-680f-41ad-a351-4b142799fffd"}, "solution": "torch_matmul_0a4e73", "evaluation": {"status": "PASSED", "environment": {"hardware": "NVIDIA B200", "libs": {"torch": "2.11.0+cu130", "triton": "3.6.0", "cuda": "13.0"}}, "timestamp": "2026-04-20T07:33:06.687522", "log": "", "correctness": {"max_relative_error": 0.0, "max_absolute_error": 0.0, "extra": null}, "performance": {"latency_ms": 0.023274666319290798, "reference_latency_ms": 0.023834666858116787, "speedup_factor": 1.0240605184686082}}} -{"definition": "gemm_n16384_k2048", "workload": {"axes": {"M": 160}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}, "uuid": "01b190bf-ca0f-4f73-bd8f-a1482dc880c0"}, "solution": "torch_matmul_0a4e73", "evaluation": {"status": "PASSED", "environment": {"hardware": "NVIDIA B200", "libs": {"torch": "2.11.0+cu130", "triton": "3.6.0", "cuda": "13.0"}}, "timestamp": "2026-04-20T07:33:06.746275", "log": "", "correctness": {"max_relative_error": 0.0, "max_absolute_error": 0.0, "extra": null}, "performance": {"latency_ms": 0.023061333845059078, "reference_latency_ms": 0.02314133383333683, "speedup_factor": 1.0034690095904792}}} -{"definition": "gemm_n16384_k2048", "workload": {"axes": {"M": 168}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}, "uuid": "1f837f5a-536a-4292-a95c-0b5855c92aac"}, "solution": "torch_matmul_0a4e73", "evaluation": {"status": "PASSED", "environment": {"hardware": "NVIDIA B200", "libs": {"torch": "2.11.0+cu130", "triton": "3.6.0", "cuda": "13.0"}}, "timestamp": "2026-04-20T07:33:06.799739", "log": "", "correctness": {"max_relative_error": 0.0, "max_absolute_error": 0.0, "extra": null}, "performance": {"latency_ms": 0.023274666319290798, "reference_latency_ms": 0.023680000255505245, "speedup_factor": 1.0174152415615296}}} -{"definition": "gemm_n16384_k2048", "workload": {"axes": {"M": 176}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}, "uuid": "644334e2-c032-4753-b822-84061fca1434"}, "solution": "torch_matmul_0a4e73", "evaluation": {"status": "PASSED", "environment": {"hardware": "NVIDIA B200", "libs": {"torch": "2.11.0+cu130", "triton": "3.6.0", "cuda": "13.0"}}, "timestamp": "2026-04-20T07:33:06.846986", "log": "", "correctness": {"max_relative_error": 0.0, "max_absolute_error": 0.0, "extra": null}, "performance": {"latency_ms": 0.023013333479563396, "reference_latency_ms": 0.023055999850233395, "speedup_factor": 1.0018539848087582}}} -{"definition": "gemm_n16384_k2048", "workload": {"axes": {"M": 184}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}, "uuid": "6cf68f96-3a12-4340-b4f0-abbcab986a96"}, "solution": "torch_matmul_0a4e73", "evaluation": {"status": "PASSED", "environment": {"hardware": "NVIDIA B200", "libs": {"torch": "2.11.0+cu130", "triton": "3.6.0", "cuda": "13.0"}}, "timestamp": "2026-04-20T07:33:06.971884", "log": "", "correctness": {"max_relative_error": 0.0, "max_absolute_error": 0.0, "extra": null}, "performance": {"latency_ms": 0.023685333629449207, "reference_latency_ms": 0.023957333837946255, "speedup_factor": 1.011483908681736}}} -{"definition": "gemm_n16384_k2048", "workload": {"axes": {"M": 192}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}, "uuid": "9478353c-290b-4e44-9e3f-721d69185dd6"}, "solution": "torch_matmul_0a4e73", "evaluation": {"status": "PASSED", "environment": {"hardware": "NVIDIA B200", "libs": {"torch": "2.11.0+cu130", "triton": "3.6.0", "cuda": "13.0"}}, "timestamp": "2026-04-20T07:33:07.021174", "log": "", "correctness": {"max_relative_error": 0.0, "max_absolute_error": 0.0, "extra": null}, "performance": {"latency_ms": 0.023205333078900974, "reference_latency_ms": 0.023306666562954586, "speedup_factor": 1.0043668187700243}}} -{"definition": "gemm_n16384_k2048", "workload": {"axes": {"M": 200}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}, "uuid": "8df72ae2-ba5d-457e-a834-faeec76470c2"}, "solution": "torch_matmul_0a4e73", "evaluation": {"status": "PASSED", "environment": {"hardware": "NVIDIA B200", "libs": {"torch": "2.11.0+cu130", "triton": "3.6.0", "cuda": "13.0"}}, "timestamp": "2026-04-20T07:33:07.068577", "log": "", "correctness": {"max_relative_error": 0.0, "max_absolute_error": 0.0, "extra": null}, "performance": {"latency_ms": 0.02350933353106181, "reference_latency_ms": 0.023898667345444363, "speedup_factor": 1.0165608188708601}}} -{"definition": "gemm_n16384_k2048", "workload": {"axes": {"M": 208}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}, "uuid": "7bc262bc-b2c8-462d-881d-655a358fd8bd"}, "solution": "torch_matmul_0a4e73", "evaluation": {"status": "PASSED", "environment": {"hardware": "NVIDIA B200", "libs": {"torch": "2.11.0+cu130", "triton": "3.6.0", "cuda": "13.0"}}, "timestamp": "2026-04-20T07:33:07.121770", "log": "", "correctness": {"max_relative_error": 0.0, "max_absolute_error": 0.0, "extra": null}, "performance": {"latency_ms": 0.02311466634273529, "reference_latency_ms": 0.023306666562954586, "speedup_factor": 1.0083064240414459}}} -{"definition": "gemm_n16384_k2048", "workload": {"axes": {"M": 216}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}, "uuid": "6c3dfea2-4783-4eea-a9a5-4feee25a3c26"}, "solution": "torch_matmul_0a4e73", "evaluation": {"status": "PASSED", "environment": {"hardware": "NVIDIA B200", "libs": {"torch": "2.11.0+cu130", "triton": "3.6.0", "cuda": "13.0"}}, "timestamp": "2026-04-20T07:33:07.173681", "log": "", "correctness": {"max_relative_error": 0.0, "max_absolute_error": 0.0, "extra": null}, "performance": {"latency_ms": 0.024394667086501915, "reference_latency_ms": 0.024853333209951717, "speedup_factor": 1.0188019013263596}}} -{"definition": "gemm_n16384_k2048", "workload": {"axes": {"M": 224}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}, "uuid": "2606ccca-7b5f-4e63-8bad-9470cfa78f61"}, "solution": "torch_matmul_0a4e73", "evaluation": {"status": "PASSED", "environment": {"hardware": "NVIDIA B200", "libs": {"torch": "2.11.0+cu130", "triton": "3.6.0", "cuda": "13.0"}}, "timestamp": "2026-04-20T07:33:07.218502", "log": "", "correctness": {"max_relative_error": 0.0, "max_absolute_error": 0.0, "extra": null}, "performance": {"latency_ms": 0.023978667023281258, "reference_latency_ms": 0.024346666410565376, "speedup_factor": 1.015346949308184}}} -{"definition": "gemm_n16384_k2048", "workload": {"axes": {"M": 232}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}, "uuid": "16046c0c-623a-4f31-b6ac-bba47e516380"}, "solution": "torch_matmul_0a4e73", "evaluation": {"status": "PASSED", "environment": {"hardware": "NVIDIA B200", "libs": {"torch": "2.11.0+cu130", "triton": "3.6.0", "cuda": "13.0"}}, "timestamp": "2026-04-20T07:33:07.313934", "log": "", "correctness": {"max_relative_error": 0.0, "max_absolute_error": 0.0, "extra": null}, "performance": {"latency_ms": 0.025503999864061672, "reference_latency_ms": 0.026047999660174053, "speedup_factor": 1.0213299795723}}} -{"definition": "gemm_n16384_k2048", "workload": {"axes": {"M": 240}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}, "uuid": "417e59f3-7664-4eed-9ae2-1cc129015d30"}, "solution": "torch_matmul_0a4e73", "evaluation": {"status": "PASSED", "environment": {"hardware": "NVIDIA B200", "libs": {"torch": "2.11.0+cu130", "triton": "3.6.0", "cuda": "13.0"}}, "timestamp": "2026-04-20T07:33:07.362133", "log": "", "correctness": {"max_relative_error": 0.0, "max_absolute_error": 0.0, "extra": null}, "performance": {"latency_ms": 0.02491733307639758, "reference_latency_ms": 0.02500266705950101, "speedup_factor": 1.0034246836465923}}} -{"definition": "gemm_n16384_k2048", "workload": {"axes": {"M": 248}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}, "uuid": "72f3eccc-3708-4760-8b1f-a1149caa5824"}, "solution": "torch_matmul_0a4e73", "evaluation": {"status": "PASSED", "environment": {"hardware": "NVIDIA B200", "libs": {"torch": "2.11.0+cu130", "triton": "3.6.0", "cuda": "13.0"}}, "timestamp": "2026-04-20T07:33:07.408022", "log": "", "correctness": {"max_relative_error": 0.0, "max_absolute_error": 0.0, "extra": null}, "performance": {"latency_ms": 0.024938667193055153, "reference_latency_ms": 0.025136000166336697, "speedup_factor": 1.0079127313321898}}} -{"definition": "gemm_n16384_k2048", "workload": {"axes": {"M": 256}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}, "uuid": "7537c621-a3a8-453c-b010-ec5b5f94af08"}, "solution": "torch_matmul_0a4e73", "evaluation": {"status": "PASSED", "environment": {"hardware": "NVIDIA B200", "libs": {"torch": "2.11.0+cu130", "triton": "3.6.0", "cuda": "13.0"}}, "timestamp": "2026-04-20T07:33:07.453610", "log": "", "correctness": {"max_relative_error": 0.0, "max_absolute_error": 0.0, "extra": null}, "performance": {"latency_ms": 0.02499733368555705, "reference_latency_ms": 0.024970666815837223, "speedup_factor": 0.9989332114354567}}} -{"definition": "gemm_n16384_k2048", "workload": {"axes": {"M": 972}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}, "uuid": "46cdb071-9248-43ff-9d73-bfb826f99342"}, "solution": "torch_matmul_0a4e73", "evaluation": {"status": "PASSED", "environment": {"hardware": "NVIDIA B200", "libs": {"torch": "2.11.0+cu130", "triton": "3.6.0", "cuda": "13.0"}}, "timestamp": "2026-04-20T07:33:07.515322", "log": "", "correctness": {"max_relative_error": 0.0, "max_absolute_error": 0.0, "extra": null}, "performance": {"latency_ms": 0.05378666644295057, "reference_latency_ms": 0.05394666704038779, "speedup_factor": 1.00297472604306}}} -{"definition": "gemm_n16384_k2048", "workload": {"axes": {"M": 2053}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}, "uuid": "1198c852-140b-4c3d-8893-afa944f29929"}, "solution": "torch_matmul_0a4e73", "evaluation": {"status": "PASSED", "environment": {"hardware": "NVIDIA B200", "libs": {"torch": "2.11.0+cu130", "triton": "3.6.0", "cuda": "13.0"}}, "timestamp": "2026-04-20T07:33:07.595778", "log": "", "correctness": {"max_relative_error": 0.0, "max_absolute_error": 0.0, "extra": null}, "performance": {"latency_ms": 0.09344533458352089, "reference_latency_ms": 0.09061333040396373, "speedup_factor": 0.9696934663225382}}} -{"definition": "gemm_n16384_k2048", "workload": {"axes": {"M": 2379}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}, "uuid": "b6eaeb01-6c46-4598-9cbc-8fdb92f3f4f0"}, "solution": "torch_matmul_0a4e73", "evaluation": {"status": "PASSED", "environment": {"hardware": "NVIDIA B200", "libs": {"torch": "2.11.0+cu130", "triton": "3.6.0", "cuda": "13.0"}}, "timestamp": "2026-04-20T07:33:07.677332", "log": "", "correctness": {"max_relative_error": 0.0, "max_absolute_error": 0.0, "extra": null}, "performance": {"latency_ms": 0.10474666704734166, "reference_latency_ms": 0.10451733196775119, "speedup_factor": 0.9978105739680785}}} -{"definition": "gemm_n16384_k2048", "workload": {"axes": {"M": 8192}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}, "uuid": "11cb2b3b-7bc9-4f48-9a55-4827179a25d9"}, "solution": "torch_matmul_0a4e73", "evaluation": {"status": "PASSED", "environment": {"hardware": "NVIDIA B200", "libs": {"torch": "2.11.0+cu130", "triton": "3.6.0", "cuda": "13.0"}}, "timestamp": "2026-04-20T07:33:07.874210", "log": "", "correctness": {"max_relative_error": 0.0, "max_absolute_error": 0.0, "extra": null}, "performance": {"latency_ms": 0.38468799988428753, "reference_latency_ms": 0.3319626599550247, "speedup_factor": 0.8629399930719899}}} diff --git a/traces/baseline/gemm/gemm_n2048_k2048.jsonl b/traces/baseline/gemm/gemm_n2048_k2048.jsonl deleted file mode 100644 index 87e19d6950e8ab0b3a9cdb2c36e3a4daee416736..0000000000000000000000000000000000000000 --- a/traces/baseline/gemm/gemm_n2048_k2048.jsonl +++ /dev/null @@ -1,43 +0,0 @@ -{"definition":"gemm_n2048_k2048","workload":{"axes":{"M":1},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"a98d8497-44c8-4f1e-a238-2a9b25f0d5c7"},"solution":"torch_matmul_756dea","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:07.932347","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.015349333174526691,"reference_latency_ms":0.015349333174526691,"speedup_factor":1.0}}} -{"definition":"gemm_n2048_k2048","workload":{"axes":{"M":2},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"1b4fe44c-9cc9-4622-8904-828b8bc4067d"},"solution":"torch_matmul_756dea","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:07.976499","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.012336000179251036,"reference_latency_ms":0.013141333280752102,"speedup_factor":1.0652831622729404}}} -{"definition":"gemm_n2048_k2048","workload":{"axes":{"M":4},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"b921b0d9-9c4f-4b11-b0f8-3fde74156fa4"},"solution":"torch_matmul_756dea","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:08.016185","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.011615999974310398,"reference_latency_ms":0.01229866687208414,"speedup_factor":1.0587695333405225}}} -{"definition":"gemm_n2048_k2048","workload":{"axes":{"M":7},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"bfea2711-b0c0-49c9-98b4-12a92d3d8b1e"},"solution":"torch_matmul_756dea","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:08.055195","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.010981333442032337,"reference_latency_ms":0.011434666812419891,"speedup_factor":1.0412821787791606}}} -{"definition":"gemm_n2048_k2048","workload":{"axes":{"M":8},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"d97806ad-497a-477d-b4e0-d18361810fad"},"solution":"torch_matmul_756dea","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:08.093975","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.010922666639089584,"reference_latency_ms":0.010901333453754583,"speedup_factor":0.9980468885447209}}} -{"definition":"gemm_n2048_k2048","workload":{"axes":{"M":15},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"614bb62c-9e3b-4340-8de5-e25b809a37b9"},"solution":"torch_matmul_756dea","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:08.134640","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.010751999914646149,"reference_latency_ms":0.01071999967098236,"speedup_factor":0.9970237868380004}}} -{"definition":"gemm_n2048_k2048","workload":{"axes":{"M":16},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"bab3d06b-c5a7-4dce-83d4-91e2b7d4056f"},"solution":"torch_matmul_756dea","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:08.173476","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.01071999967098236,"reference_latency_ms":0.010741333166758219,"speedup_factor":1.0019900649655433}}} -{"definition":"gemm_n2048_k2048","workload":{"axes":{"M":24},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"d68fb9b3-55bd-4b54-995a-4d952c4eb157"},"solution":"torch_matmul_756dea","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:08.214048","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.010762666662534079,"reference_latency_ms":0.010730666729311148,"speedup_factor":0.9970267653708605}}} -{"definition":"gemm_n2048_k2048","workload":{"axes":{"M":32},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"95640f6d-b61a-4621-aa0b-697cffa434b8"},"solution":"torch_matmul_756dea","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:08.252758","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.010778666784365972,"reference_latency_ms":0.01073066641887029,"speedup_factor":0.9955467251696375}}} -{"definition":"gemm_n2048_k2048","workload":{"axes":{"M":35},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"43afb86e-39cb-4b8e-a805-2e1e17aedb9d"},"solution":"torch_matmul_756dea","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:08.291752","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.010778666784365972,"reference_latency_ms":0.010735999792814255,"speedup_factor":0.9960415334841222}}} -{"definition":"gemm_n2048_k2048","workload":{"axes":{"M":40},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"bfec76cf-b9f8-4265-8f4e-bdbb8e9541d3"},"solution":"torch_matmul_756dea","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:08.330742","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.010741333166758219,"reference_latency_ms":0.010757333288590113,"speedup_factor":1.0014895843545204}}} -{"definition":"gemm_n2048_k2048","workload":{"axes":{"M":48},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"8005b86e-50e8-48ea-bc09-a68458331326"},"solution":"torch_matmul_756dea","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:08.369543","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.01071999967098236,"reference_latency_ms":0.010741333166758219,"speedup_factor":1.0019900649655433}}} -{"definition":"gemm_n2048_k2048","workload":{"axes":{"M":56},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"bf82ffcb-f158-480a-9d5c-89d3cdefd810"},"solution":"torch_matmul_756dea","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:08.409925","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.010672000236809254,"reference_latency_ms":0.01065600011497736,"speedup_factor":0.9985007382424237}}} -{"definition":"gemm_n2048_k2048","workload":{"axes":{"M":64},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"44c3c77f-f938-43a7-b5e2-975851500b47"},"solution":"torch_matmul_756dea","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:08.448824","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.010736000103255114,"reference_latency_ms":0.010714666607479254,"speedup_factor":0.9980129009341765}}} -{"definition":"gemm_n2048_k2048","workload":{"axes":{"M":70},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"5e3dc6d5-bdc5-4b65-b9ac-32b13b0e6804"},"solution":"torch_matmul_756dea","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:08.491518","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.01146666674564282,"reference_latency_ms":0.011605333226422468,"speedup_factor":1.0120930069614467}}} -{"definition":"gemm_n2048_k2048","workload":{"axes":{"M":72},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"f415e56b-b5cf-4ffe-9bd9-e3c9805e5c14"},"solution":"torch_matmul_756dea","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:08.530618","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.01145600030819575,"reference_latency_ms":0.01084266696125269,"speedup_factor":0.9464618252057593}}} -{"definition":"gemm_n2048_k2048","workload":{"axes":{"M":80},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"abc630bf-d88d-4988-9684-162130a2e9d0"},"solution":"torch_matmul_756dea","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:08.572039","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.01071999967098236,"reference_latency_ms":0.010741333166758219,"speedup_factor":1.0019900649655433}}} -{"definition":"gemm_n2048_k2048","workload":{"axes":{"M":88},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"013d1993-1545-4950-a3d2-44022105e87f"},"solution":"torch_matmul_756dea","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:08.611195","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.010746666540702185,"reference_latency_ms":0.010735999792814255,"speedup_factor":0.9990074365992906}}} -{"definition":"gemm_n2048_k2048","workload":{"axes":{"M":96},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"5dd9fe00-896a-4215-b59a-602e4456d759"},"solution":"torch_matmul_756dea","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:08.650963","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.010880000268419584,"reference_latency_ms":0.01084266680603226,"speedup_factor":0.9965686156740559}}} -{"definition":"gemm_n2048_k2048","workload":{"axes":{"M":104},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"15473932-acfb-4329-8c34-deacbe7e8953"},"solution":"torch_matmul_756dea","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:08.691601","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.01069866664086779,"reference_latency_ms":0.01071999998142322,"speedup_factor":1.0019940186260163}}} -{"definition":"gemm_n2048_k2048","workload":{"axes":{"M":112},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"f9087d88-02f1-433a-87af-718e446b761f"},"solution":"torch_matmul_756dea","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:08.737000","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.01073066641887029,"reference_latency_ms":0.010688000358641148,"speedup_factor":0.996023913281461}}} -{"definition":"gemm_n2048_k2048","workload":{"axes":{"M":120},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"b6cc480a-4fe5-4435-82a8-0d42c92e1a2c"},"solution":"torch_matmul_756dea","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:08.777162","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.01071999967098236,"reference_latency_ms":0.01069866664086779,"speedup_factor":0.9980099784730109}}} -{"definition":"gemm_n2048_k2048","workload":{"axes":{"M":128},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"cc7a63bd-0f43-4ad6-b775-0ecb51eab349"},"solution":"torch_matmul_756dea","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:08.815999","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.010682666829476753,"reference_latency_ms":0.010682666674256325,"speedup_factor":0.9999999854698802}}} -{"definition":"gemm_n2048_k2048","workload":{"axes":{"M":136},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"b6f1dd0b-fd1f-4115-8041-2184d5332cd6"},"solution":"torch_matmul_756dea","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:08.857947","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.010709333543976149,"reference_latency_ms":0.010725333355367184,"speedup_factor":1.0014940062633528}}} -{"definition":"gemm_n2048_k2048","workload":{"axes":{"M":144},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"28c94467-7fae-44fa-926d-bdda7ae79a5d"},"solution":"torch_matmul_756dea","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:08.897067","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.010730666729311148,"reference_latency_ms":0.010746666540702185,"speedup_factor":1.0014910360925975}}} -{"definition":"gemm_n2048_k2048","workload":{"axes":{"M":152},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"8ffbbaba-bc09-4372-bb18-bbe100a02607"},"solution":"torch_matmul_756dea","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:08.947553","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.010757333288590113,"reference_latency_ms":0.01071999998142322,"speedup_factor":0.9965295016743145}}} -{"definition":"gemm_n2048_k2048","workload":{"axes":{"M":160},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"a4348a72-4519-4f2e-a8a8-67a56c11b569"},"solution":"torch_matmul_756dea","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:08.989405","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.0107893335322539,"reference_latency_ms":0.010768000036478043,"speedup_factor":0.9980227234877778}}} -{"definition":"gemm_n2048_k2048","workload":{"axes":{"M":168},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"c0206d05-c49f-45b1-af48-756d6fa73a59"},"solution":"torch_matmul_756dea","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:09.030938","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.01071999998142322,"reference_latency_ms":0.010794666595757008,"speedup_factor":1.0069651692596249}}} -{"definition":"gemm_n2048_k2048","workload":{"axes":{"M":176},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"2a7605d5-09b3-4cb0-b8e7-f9159c67b148"},"solution":"torch_matmul_756dea","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:09.070212","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.010816000401973724,"reference_latency_ms":0.010757333288590113,"speedup_factor":0.9945758957837219}}} -{"definition":"gemm_n2048_k2048","workload":{"axes":{"M":184},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"51fb2600-dea5-4405-896e-8ca595b5e2ba"},"solution":"torch_matmul_756dea","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:09.110145","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.010741333477199078,"reference_latency_ms":0.01076266635209322,"speedup_factor":1.0019860546122532}}} -{"definition":"gemm_n2048_k2048","workload":{"axes":{"M":192},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"2003b8f5-86c2-42c9-95a1-2ee24a23fac0"},"solution":"torch_matmul_756dea","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:09.149666","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.010810666717588902,"reference_latency_ms":0.010757333288590113,"speedup_factor":0.9950665920620774}}} -{"definition":"gemm_n2048_k2048","workload":{"axes":{"M":200},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"57164c57-2976-4a2f-9610-56eb68e238e1"},"solution":"torch_matmul_756dea","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:09.191455","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.010762666507313648,"reference_latency_ms":0.010837333276867867,"speedup_factor":1.0069375716049065}}} -{"definition":"gemm_n2048_k2048","workload":{"axes":{"M":208},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"f5727a6a-f497-4b33-904e-234bbee68117"},"solution":"torch_matmul_756dea","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:09.231042","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.010773333410422007,"reference_latency_ms":0.01080000028014183,"speedup_factor":1.0024752663547967}}} -{"definition":"gemm_n2048_k2048","workload":{"axes":{"M":216},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"6c675b98-21f7-475e-98c4-869362f1abc7"},"solution":"torch_matmul_756dea","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:09.283090","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.010741333477199078,"reference_latency_ms":0.010805333343644937,"speedup_factor":1.0059582794428377}}} -{"definition":"gemm_n2048_k2048","workload":{"axes":{"M":224},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"35f373ea-2d0c-4298-9759-fa64f0141881"},"solution":"torch_matmul_756dea","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:09.323087","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.010794666595757008,"reference_latency_ms":0.010768000036478043,"speedup_factor":0.9975296542007656}}} -{"definition":"gemm_n2048_k2048","workload":{"axes":{"M":232},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"f95d25ff-c150-454b-a52e-4e300ee5ff61"},"solution":"torch_matmul_756dea","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:09.363484","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.010767999881257614,"reference_latency_ms":0.01084266665081183,"speedup_factor":1.0069341354362549}}} -{"definition":"gemm_n2048_k2048","workload":{"axes":{"M":240},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"62124acc-2375-4565-b08e-3e3d5e09a42a"},"solution":"torch_matmul_756dea","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:09.402944","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.010848000024755796,"reference_latency_ms":0.0107893335322539,"speedup_factor":0.9945919531371668}}} -{"definition":"gemm_n2048_k2048","workload":{"axes":{"M":248},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"fac43762-cfe9-48ae-9aa6-9931b1d13cc9"},"solution":"torch_matmul_756dea","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:09.445047","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.010725333510587612,"reference_latency_ms":0.01084266665081183,"speedup_factor":1.0109398127441345}}} -{"definition":"gemm_n2048_k2048","workload":{"axes":{"M":256},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"72e3421f-b9df-4581-ab6f-9bb09a19dc37"},"solution":"torch_matmul_756dea","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:09.484664","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.010848000024755796,"reference_latency_ms":0.0107893335322539,"speedup_factor":0.9945919531371668}}} -{"definition":"gemm_n2048_k2048","workload":{"axes":{"M":972},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"d00f0dc1-1e57-4f09-88c5-6ea4be9447fc"},"solution":"torch_matmul_756dea","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:09.530715","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.014816000126302242,"reference_latency_ms":0.014816000126302242,"speedup_factor":1.0}}} -{"definition":"gemm_n2048_k2048","workload":{"axes":{"M":2053},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"3c061ce6-901e-42ab-9a5a-13f5d9259266"},"solution":"torch_matmul_756dea","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:09.576748","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.019589333174129326,"reference_latency_ms":0.019546666493018467,"speedup_factor":0.9978219431600047}}} -{"definition":"gemm_n2048_k2048","workload":{"axes":{"M":2379},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"78ead1f3-8c0a-45dd-b832-42861e7bf658"},"solution":"torch_matmul_756dea","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:09.634594","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.02092266610513131,"reference_latency_ms":0.020927999789516132,"speedup_factor":1.000254923744327}}} -{"definition":"gemm_n2048_k2048","workload":{"axes":{"M":8192},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"23bb0f3f-abe2-40ac-80af-dff54fed8646"},"solution":"torch_matmul_756dea","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:09.694237","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.05116266570985317,"reference_latency_ms":0.05146133402983347,"speedup_factor":1.0058376223333254}}} diff --git a/traces/baseline/gemm/gemm_n2048_k8192.jsonl b/traces/baseline/gemm/gemm_n2048_k8192.jsonl deleted file mode 100644 index c7135f4133483ded7997b2050db404419a1ba350..0000000000000000000000000000000000000000 --- a/traces/baseline/gemm/gemm_n2048_k8192.jsonl +++ /dev/null @@ -1,43 +0,0 @@ -{"definition":"gemm_n2048_k8192","workload":{"axes":{"M":1},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"19d02aae-74b7-4b82-aa34-f7d4c0bceebd"},"solution":"torch_matmul_ff95bc","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:09.787481","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.016309333654741447,"reference_latency_ms":0.016197333112359047,"speedup_factor":0.9931327333934431}}} -{"definition":"gemm_n2048_k8192","workload":{"axes":{"M":2},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"927a928e-e89d-45bb-9b3a-8b07063cf6b0"},"solution":"torch_matmul_ff95bc","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:09.831401","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.015898666034142177,"reference_latency_ms":0.016842666392525036,"speedup_factor":1.0593760732098927}}} -{"definition":"gemm_n2048_k8192","workload":{"axes":{"M":4},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"2a6b32c1-3c52-47b7-8d44-5a1e4260b2e8"},"solution":"torch_matmul_ff95bc","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:09.874373","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.016703999911745388,"reference_latency_ms":0.016917333627740543,"speedup_factor":1.0127714150576084}}} -{"definition":"gemm_n2048_k8192","workload":{"axes":{"M":7},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"66014bc9-8e6d-4e68-9657-d46ec4e913d5"},"solution":"torch_matmul_ff95bc","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:09.921596","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.016896000131964684,"reference_latency_ms":0.016650666793187458,"speedup_factor":0.9854797977710067}}} -{"definition":"gemm_n2048_k8192","workload":{"axes":{"M":8},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"91615ba2-0906-4a91-b970-23adc00b277d"},"solution":"torch_matmul_ff95bc","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:09.963415","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.01686933357268572,"reference_latency_ms":0.016954667245348293,"speedup_factor":1.0050585088198591}}} -{"definition":"gemm_n2048_k8192","workload":{"axes":{"M":15},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"3ab6c6a6-22cc-4f1b-9840-9e875b591b77"},"solution":"torch_matmul_ff95bc","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:10.009163","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.017605333278576534,"reference_latency_ms":0.017946666727463405,"speedup_factor":1.0193880708468173}}} -{"definition":"gemm_n2048_k8192","workload":{"axes":{"M":16},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"22113298-fdca-480d-a948-841ad8369fe7"},"solution":"torch_matmul_ff95bc","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:10.052736","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.016490666816631954,"reference_latency_ms":0.016815999522805214,"speedup_factor":1.0197282929666094}}} -{"definition":"gemm_n2048_k8192","workload":{"axes":{"M":24},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"3d1ee002-d32e-4da2-93e4-102a9410aef7"},"solution":"torch_matmul_ff95bc","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:10.097178","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.017893333608905476,"reference_latency_ms":0.018090666582187016,"speedup_factor":1.0110282956543843}}} -{"definition":"gemm_n2048_k8192","workload":{"axes":{"M":32},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"5d1101a2-2dc1-437d-bdb9-b63bcb52b26f"},"solution":"torch_matmul_ff95bc","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:10.142158","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.01844266677896182,"reference_latency_ms":0.018192000376681488,"speedup_factor":0.9864083429319302}}} -{"definition":"gemm_n2048_k8192","workload":{"axes":{"M":35},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"59f6b86a-fa4b-4f2a-a600-2982b68e0bdb"},"solution":"torch_matmul_ff95bc","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:10.188689","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.01891733395556609,"reference_latency_ms":0.01826133330663045,"speedup_factor":0.9653227748436178}}} -{"definition":"gemm_n2048_k8192","workload":{"axes":{"M":40},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"e4c9c6ee-f59e-4187-8ca8-fdfc5f870e35"},"solution":"torch_matmul_ff95bc","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:10.231637","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.018949332957466442,"reference_latency_ms":0.018623999630411465,"speedup_factor":0.9828314100667702}}} -{"definition":"gemm_n2048_k8192","workload":{"axes":{"M":48},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"4571a224-a05b-4a03-8f39-378316788096"},"solution":"torch_matmul_ff95bc","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:10.281075","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.018853333468238514,"reference_latency_ms":0.018581333259741466,"speedup_factor":0.9855728320429236}}} -{"definition":"gemm_n2048_k8192","workload":{"axes":{"M":56},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"948d16e9-2aac-4233-a67a-c3a0c2d10ad7"},"solution":"torch_matmul_ff95bc","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:10.323455","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.01898133320113023,"reference_latency_ms":0.019007999449968338,"speedup_factor":1.0014048670109494}}} -{"definition":"gemm_n2048_k8192","workload":{"axes":{"M":64},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"d2f2fdd7-c19b-4131-a342-4a1b9e95747e"},"solution":"torch_matmul_ff95bc","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:10.367039","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.018511999708910782,"reference_latency_ms":0.018842666720350582,"speedup_factor":1.017862306430387}}} -{"definition":"gemm_n2048_k8192","workload":{"axes":{"M":70},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"5221701f-4873-4fe3-a73a-a01001c56f9c"},"solution":"torch_matmul_ff95bc","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:10.414650","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.01698133411506812,"reference_latency_ms":0.016976000120242436,"speedup_factor":0.9996858907086134}}} -{"definition":"gemm_n2048_k8192","workload":{"axes":{"M":72},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"f7b9af39-1fbd-43c9-8da6-29797fb3c671"},"solution":"torch_matmul_ff95bc","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:10.455950","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.01699200024207433,"reference_latency_ms":0.01701333311696847,"speedup_factor":1.001255465783323}}} -{"definition":"gemm_n2048_k8192","workload":{"axes":{"M":80},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"96b8a5ba-0ef5-41d3-9962-b7ce024bb338"},"solution":"torch_matmul_ff95bc","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:10.500296","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.019029332945744198,"reference_latency_ms":0.019007999449968338,"speedup_factor":0.9988789152075543}}} -{"definition":"gemm_n2048_k8192","workload":{"axes":{"M":88},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"1227c633-8ed4-4a4c-bc0a-c0bf6a1a29bc"},"solution":"torch_matmul_ff95bc","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:10.542307","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.01903466631968816,"reference_latency_ms":0.019018666197856266,"speedup_factor":0.9991594219954703}}} -{"definition":"gemm_n2048_k8192","workload":{"axes":{"M":96},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"63901d6e-2f7a-4fa9-86b7-b4dce10e7bcf"},"solution":"torch_matmul_ff95bc","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:10.586766","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.018975999827186268,"reference_latency_ms":0.01897599920630455,"speedup_factor":0.9999999672806849}}} -{"definition":"gemm_n2048_k8192","workload":{"axes":{"M":104},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"790a0497-0605-4b00-ac02-fb27a1516f7a"},"solution":"torch_matmul_ff95bc","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:10.657700","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.019018666197856266,"reference_latency_ms":0.019018666197856266,"speedup_factor":1.0}}} -{"definition":"gemm_n2048_k8192","workload":{"axes":{"M":112},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"013339f2-443f-4fe3-b511-b77427f4705a"},"solution":"torch_matmul_ff95bc","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:10.699975","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.019039999693632126,"reference_latency_ms":0.019029332945744198,"speedup_factor":0.9994397716355271}}} -{"definition":"gemm_n2048_k8192","workload":{"axes":{"M":120},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"f6f1b907-5f13-4d58-a00b-ad0d97b8d6cc"},"solution":"torch_matmul_ff95bc","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:10.754419","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.01899733270208041,"reference_latency_ms":0.019034666940569878,"speedup_factor":1.0019652358083606}}} -{"definition":"gemm_n2048_k8192","workload":{"axes":{"M":128},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"d21e0f96-99e4-426c-ba72-8ef19ae3123f"},"solution":"torch_matmul_ff95bc","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:10.798616","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.020693333819508553,"reference_latency_ms":0.020821332931518555,"speedup_factor":1.006185523953097}}} -{"definition":"gemm_n2048_k8192","workload":{"axes":{"M":136},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"39d5ff01-b6bf-477c-a514-e96c4a244d7a"},"solution":"torch_matmul_ff95bc","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:10.845291","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.020693333198626835,"reference_latency_ms":0.020639999769628048,"speedup_factor":0.9974226757725853}}} -{"definition":"gemm_n2048_k8192","workload":{"axes":{"M":144},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"5ec080a0-0622-421a-a0e6-8d4dce250c58"},"solution":"torch_matmul_ff95bc","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:10.890091","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.019061333189407986,"reference_latency_ms":0.019120000302791595,"speedup_factor":1.0030778074545283}}} -{"definition":"gemm_n2048_k8192","workload":{"axes":{"M":152},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"90a6b71b-de24-4d97-a307-13f82d5f75b4"},"solution":"torch_matmul_ff95bc","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:10.962620","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.020143999718129635,"reference_latency_ms":0.020005333858231705,"speedup_factor":0.9931162697657739}}} -{"definition":"gemm_n2048_k8192","workload":{"axes":{"M":160},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"cfcf6b3e-bb34-4cf7-9c1a-cb42efa90a59"},"solution":"torch_matmul_ff95bc","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:11.007290","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.019904000063737232,"reference_latency_ms":0.01993066631257534,"speedup_factor":1.00133974320502}}} -{"definition":"gemm_n2048_k8192","workload":{"axes":{"M":168},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"dba698ce-382a-40c1-ba21-7a0313a1ee1a"},"solution":"torch_matmul_ff95bc","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:11.052593","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.023082667340834934,"reference_latency_ms":0.022890666810174782,"speedup_factor":0.9916820474936842}}} -{"definition":"gemm_n2048_k8192","workload":{"axes":{"M":176},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"10404bdb-ca42-4284-8ada-8c034d3c8fcd"},"solution":"torch_matmul_ff95bc","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:11.095275","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.02094933328529199,"reference_latency_ms":0.020959999412298203,"speedup_factor":1.0005091392103491}}} -{"definition":"gemm_n2048_k8192","workload":{"axes":{"M":184},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"5d307ee8-c3d7-4313-814d-c54041b52693"},"solution":"torch_matmul_ff95bc","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:11.138208","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.02096533278624217,"reference_latency_ms":0.020938665916522343,"speedup_factor":0.9987280492996837}}} -{"definition":"gemm_n2048_k8192","workload":{"axes":{"M":192},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"b54d5b5e-f7a9-4990-a304-921b8bac4ec9"},"solution":"torch_matmul_ff95bc","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:11.183148","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.020922666415572166,"reference_latency_ms":0.02097066616018613,"speedup_factor":1.0022941504519824}}} -{"definition":"gemm_n2048_k8192","workload":{"axes":{"M":200},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"1be9cecb-d10f-4384-abd7-933f3cd304dc"},"solution":"torch_matmul_ff95bc","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:11.226314","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.02099199965596199,"reference_latency_ms":0.020954666038354237,"speedup_factor":0.9982215311442638}}} -{"definition":"gemm_n2048_k8192","workload":{"axes":{"M":208},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"f0f9da24-b32d-474e-a17f-a0d9f7b9e526"},"solution":"torch_matmul_ff95bc","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:11.273505","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.02099199965596199,"reference_latency_ms":0.02097066616018613,"speedup_factor":0.9989837320824364}}} -{"definition":"gemm_n2048_k8192","workload":{"axes":{"M":216},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"af55a0d2-7155-4e3b-897b-ff2e42d7f040"},"solution":"torch_matmul_ff95bc","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:11.317409","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.020943999911348026,"reference_latency_ms":0.020986666282018025,"speedup_factor":1.0020371643836228}}} -{"definition":"gemm_n2048_k8192","workload":{"axes":{"M":224},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"c18392ce-a799-4299-9c54-99e6fa1af0fc"},"solution":"torch_matmul_ff95bc","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:11.360598","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.020981332908074062,"reference_latency_ms":0.021018666525681812,"speedup_factor":1.0017793730155906}}} -{"definition":"gemm_n2048_k8192","workload":{"axes":{"M":232},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"478c93fa-d2bc-4927-8d9a-e6da75976747"},"solution":"torch_matmul_ff95bc","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:11.404414","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.021023999899625778,"reference_latency_ms":0.02099199965596199,"speedup_factor":0.9984779183877205}}} -{"definition":"gemm_n2048_k8192","workload":{"axes":{"M":240},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"88000cdb-65f3-4fc8-a645-fe50d1f73b8f"},"solution":"torch_matmul_ff95bc","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:11.448010","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.02100266640384992,"reference_latency_ms":0.021007999467353027,"speedup_factor":1.0002539231638765}}} -{"definition":"gemm_n2048_k8192","workload":{"axes":{"M":248},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"91b3bfa3-ed8b-4295-af25-4ec4d00d85ae"},"solution":"torch_matmul_ff95bc","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:11.491925","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.021029333273569744,"reference_latency_ms":0.02101333315173785,"speedup_factor":0.9992391522059331}}} -{"definition":"gemm_n2048_k8192","workload":{"axes":{"M":256},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"1822180b-656c-4695-afb4-f0e002526d46"},"solution":"torch_matmul_ff95bc","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:11.535169","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.021056000143289566,"reference_latency_ms":0.021077333018183708,"speedup_factor":1.0010131494466645}}} -{"definition":"gemm_n2048_k8192","workload":{"axes":{"M":972},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"3ff729a7-9620-4914-872b-bcd59b73b850"},"solution":"torch_matmul_ff95bc","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:11.583647","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.03421333308021227,"reference_latency_ms":0.033877333626151085,"speedup_factor":0.9901792832264121}}} -{"definition":"gemm_n2048_k8192","workload":{"axes":{"M":2053},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"d49a821d-f75d-47cb-9572-9b3d37569e01"},"solution":"torch_matmul_ff95bc","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:11.640886","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.05612266684571902,"reference_latency_ms":0.05661333352327347,"speedup_factor":1.008742754133607}}} -{"definition":"gemm_n2048_k8192","workload":{"axes":{"M":2379},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"da657596-8823-4be8-aee7-81aa5328bcce"},"solution":"torch_matmul_ff95bc","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:11.700183","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.059722667559981346,"reference_latency_ms":0.05985066791375478,"speedup_factor":1.0021432457558075}}} -{"definition":"gemm_n2048_k8192","workload":{"axes":{"M":8192},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"bdbd4c26-24c8-4dee-9457-72bfaad13d8a"},"solution":"torch_matmul_ff95bc","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:11.815353","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.19470933079719543,"reference_latency_ms":0.18278933068116507,"speedup_factor":0.938780539857918}}} diff --git a/traces/baseline/gemm/gemm_n3072_k2048.jsonl b/traces/baseline/gemm/gemm_n3072_k2048.jsonl deleted file mode 100644 index b4f7dea59d5d0758dfa85f00babe5d606794c00b..0000000000000000000000000000000000000000 --- a/traces/baseline/gemm/gemm_n3072_k2048.jsonl +++ /dev/null @@ -1,43 +0,0 @@ -{"definition":"gemm_n3072_k2048","workload":{"axes":{"M":1},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"87baad53-95d2-420c-adcd-81d651da4141"},"solution":"torch_matmul_73342c","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:11.865870","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.012671999943753084,"reference_latency_ms":0.013279999606311321,"speedup_factor":1.0479797715638377}}} -{"definition":"gemm_n3072_k2048","workload":{"axes":{"M":2},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"94896a55-6e36-43ee-a833-f493eed6c8ef"},"solution":"torch_matmul_73342c","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:11.905765","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.011408000097920498,"reference_latency_ms":0.011733333424975475,"speedup_factor":1.028517998269853}}} -{"definition":"gemm_n3072_k2048","workload":{"axes":{"M":4},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"f84f00bb-a362-41ca-8fcf-c62c3ca07a36"},"solution":"torch_matmul_73342c","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:11.945033","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.011194666692366203,"reference_latency_ms":0.011978666763752699,"speedup_factor":1.070033355429967}}} -{"definition":"gemm_n3072_k2048","workload":{"axes":{"M":7},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"5dc703eb-0f70-47f5-b89f-69b504b38a0a"},"solution":"torch_matmul_73342c","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:11.983944","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.01086400014658769,"reference_latency_ms":0.011360000198086103,"speedup_factor":1.0456553796765369}}} -{"definition":"gemm_n3072_k2048","workload":{"axes":{"M":8},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"26a1b2cf-5645-403d-b359-2025bec2cc5d"},"solution":"torch_matmul_73342c","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:12.022229","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.011023999967922768,"reference_latency_ms":0.011066666649033626,"speedup_factor":1.0038703448144966}}} -{"definition":"gemm_n3072_k2048","workload":{"axes":{"M":15},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"27384021-9f80-4763-8ff8-ed4ee32fa5cb"},"solution":"torch_matmul_73342c","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:12.062477","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.010762666662534079,"reference_latency_ms":0.010794666595757008,"speedup_factor":1.0029732346291393}}} -{"definition":"gemm_n3072_k2048","workload":{"axes":{"M":16},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"344fefad-0983-4f85-a744-fd194c60e36c"},"solution":"torch_matmul_73342c","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:12.101624","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.010805333654085795,"reference_latency_ms":0.01081066687280933,"speedup_factor":1.0004935727941653}}} -{"definition":"gemm_n3072_k2048","workload":{"axes":{"M":24},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"a48073de-4b7b-4bcd-9d6a-ce5c89c2ae74"},"solution":"torch_matmul_73342c","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:12.140697","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.010858666772643724,"reference_latency_ms":0.010879999957978725,"speedup_factor":1.0019646228935533}}} -{"definition":"gemm_n3072_k2048","workload":{"axes":{"M":32},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"c5391284-499b-48c5-b6e3-263246d33367"},"solution":"torch_matmul_73342c","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:12.179592","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.010853333088258902,"reference_latency_ms":0.010821333465476831,"speedup_factor":0.997051631740973}}} -{"definition":"gemm_n3072_k2048","workload":{"axes":{"M":35},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"e6d9bc7c-297c-49d0-8b6f-fefad9788867"},"solution":"torch_matmul_73342c","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:12.221181","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.010863999836146832,"reference_latency_ms":0.010874666894475618,"speedup_factor":1.0009818720995647}}} -{"definition":"gemm_n3072_k2048","workload":{"axes":{"M":40},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"442f657e-0ea5-4995-ba54-1943b0edc22f"},"solution":"torch_matmul_73342c","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:12.260185","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.01085333339869976,"reference_latency_ms":0.010960000101476908,"speedup_factor":1.0098280130959514}}} -{"definition":"gemm_n3072_k2048","workload":{"axes":{"M":48},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"20859e9e-e357-4210-8b3d-52e30bf45ec6"},"solution":"torch_matmul_73342c","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:12.299197","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.010917333575586477,"reference_latency_ms":0.010890667016307512,"speedup_factor":0.9975574109653847}}} -{"definition":"gemm_n3072_k2048","workload":{"axes":{"M":56},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"417ccb5d-1132-4309-87e9-d7feaee11be5"},"solution":"torch_matmul_73342c","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:12.338106","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.010912000201642513,"reference_latency_ms":0.010944000134865442,"speedup_factor":1.0029325451458582}}} -{"definition":"gemm_n3072_k2048","workload":{"axes":{"M":64},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"c4c3576e-99cf-41a7-a0d9-7a78e9f53108"},"solution":"torch_matmul_73342c","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:12.376886","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.010944000134865442,"reference_latency_ms":0.01088533333192269,"speedup_factor":0.9946393638322562}}} -{"definition":"gemm_n3072_k2048","workload":{"axes":{"M":70},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"6a696da8-b670-44df-9f4b-9d3892c89b10"},"solution":"torch_matmul_73342c","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:12.419473","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.011018666749199232,"reference_latency_ms":0.011109333330144485,"speedup_factor":1.0082284529525174}}} -{"definition":"gemm_n3072_k2048","workload":{"axes":{"M":72},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"bbc5595a-d522-4bbc-850c-21cd87cb22e5"},"solution":"torch_matmul_73342c","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:12.459453","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.010954666882753372,"reference_latency_ms":0.011061333430310091,"speedup_factor":1.009737087279637}}} -{"definition":"gemm_n3072_k2048","workload":{"axes":{"M":80},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"50a7a6ec-04cc-4eca-9c55-a0ca1fc360eb"},"solution":"torch_matmul_73342c","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:12.498906","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.011066666804254055,"reference_latency_ms":0.011061333430310091,"speedup_factor":0.999518068625513}}} -{"definition":"gemm_n3072_k2048","workload":{"axes":{"M":88},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"aa056b8d-75c3-497d-a080-aee5c936483b"},"solution":"torch_matmul_73342c","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:12.538820","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.011072000178198019,"reference_latency_ms":0.011098666427036127,"speedup_factor":1.0024084400658353}}} -{"definition":"gemm_n3072_k2048","workload":{"axes":{"M":96},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"b24b592f-881f-4fb8-8720-97782af98f5d"},"solution":"torch_matmul_73342c","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:12.577800","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.011066666493813196,"reference_latency_ms":0.011045333308478197,"speedup_factor":0.9980723024998607}}} -{"definition":"gemm_n3072_k2048","workload":{"axes":{"M":104},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"cd026c2a-883e-4c5c-97ca-bbdd293637d1"},"solution":"torch_matmul_73342c","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:12.620230","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.012213333509862423,"reference_latency_ms":0.012069333189477524,"speedup_factor":0.9882095809249279}}} -{"definition":"gemm_n3072_k2048","workload":{"axes":{"M":112},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"b1dea987-a7f1-4f1e-9c2a-22d185d12c43"},"solution":"torch_matmul_73342c","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:12.678576","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.011360000042865673,"reference_latency_ms":0.011909333212922016,"speedup_factor":1.0483567929562938}}} -{"definition":"gemm_n3072_k2048","workload":{"axes":{"M":120},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"771606fd-e670-4f52-b479-58a879b3c378"},"solution":"torch_matmul_73342c","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:12.719171","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.011455999997754892,"reference_latency_ms":0.011312000143031279,"speedup_factor":0.9874301802765508}}} -{"definition":"gemm_n3072_k2048","workload":{"axes":{"M":128},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"d5c39f45-6a6f-49a7-9cc5-5886d3e7a104"},"solution":"torch_matmul_73342c","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:12.769277","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.011370666635533174,"reference_latency_ms":0.011530666767309109,"speedup_factor":1.01407130618674}}} -{"definition":"gemm_n3072_k2048","workload":{"axes":{"M":136},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"54172e8f-382f-4360-aacf-e580a623b0d6"},"solution":"torch_matmul_73342c","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:12.811731","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.011253333184868097,"reference_latency_ms":0.011658666810641686,"speedup_factor":1.0360189838081595}}} -{"definition":"gemm_n3072_k2048","workload":{"axes":{"M":144},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"f4301d94-f87e-4ee7-bf77-cd6746a9055e"},"solution":"torch_matmul_73342c","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:12.853494","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.011215999877701202,"reference_latency_ms":0.01168533336992065,"speedup_factor":1.041844997979408}}} -{"definition":"gemm_n3072_k2048","workload":{"axes":{"M":152},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"b86fbc83-1a20-4b27-b761-e6ff69569d76"},"solution":"torch_matmul_73342c","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:12.893801","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.012170666828751564,"reference_latency_ms":0.012336000024030605,"speedup_factor":1.0135845634101546}}} -{"definition":"gemm_n3072_k2048","workload":{"axes":{"M":160},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"44977a71-5f00-4c17-9de8-1837a124109f"},"solution":"torch_matmul_73342c","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:12.974965","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.011973333545029163,"reference_latency_ms":0.012288000124196211,"speedup_factor":1.0262806158354858}}} -{"definition":"gemm_n3072_k2048","workload":{"axes":{"M":168},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"187b0190-f52b-45f3-b19b-904db0ed6071"},"solution":"torch_matmul_73342c","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:13.015766","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.011808000039309263,"reference_latency_ms":0.011386666602144638,"speedup_factor":0.9643179678385848}}} -{"definition":"gemm_n3072_k2048","workload":{"axes":{"M":176},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"d25a7831-ea39-41e5-9439-f106eb540557"},"solution":"torch_matmul_73342c","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:13.055834","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.01135466651370128,"reference_latency_ms":0.01212266661847631,"speedup_factor":1.067637398583949}}} -{"definition":"gemm_n3072_k2048","workload":{"axes":{"M":184},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"6eb7fe5f-aa3b-4e7f-ac7a-61c63df11296"},"solution":"torch_matmul_73342c","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:13.103106","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.011344000231474638,"reference_latency_ms":0.011440000031143427,"speedup_factor":1.0084626055809158}}} -{"definition":"gemm_n3072_k2048","workload":{"axes":{"M":192},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"0dc54c87-e26b-4a37-8db3-0ee2befd174a"},"solution":"torch_matmul_73342c","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:13.143353","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.01148800008619825,"reference_latency_ms":0.011920000116030375,"speedup_factor":1.0376044591391613}}} -{"definition":"gemm_n3072_k2048","workload":{"axes":{"M":200},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"13db2d38-8290-4831-bace-25997cabc258"},"solution":"torch_matmul_73342c","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:13.186231","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.012639999700089296,"reference_latency_ms":0.012656000287582478,"speedup_factor":1.0012658692937364}}} -{"definition":"gemm_n3072_k2048","workload":{"axes":{"M":208},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"9293595b-ddfa-4791-9977-0a9459205d3d"},"solution":"torch_matmul_73342c","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:13.226306","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.012661333506306013,"reference_latency_ms":0.012549333119144043,"speedup_factor":0.9911541397194704}}} -{"definition":"gemm_n3072_k2048","workload":{"axes":{"M":216},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"2f5f76ed-e14c-467e-8651-3eebd8476bb1"},"solution":"torch_matmul_73342c","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:13.304309","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.012213333509862423,"reference_latency_ms":0.012181333421419064,"speedup_factor":0.9973799054600844}}} -{"definition":"gemm_n3072_k2048","workload":{"axes":{"M":224},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"495a45c0-b0f8-49da-a3c1-140719a72a19"},"solution":"torch_matmul_73342c","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:13.344221","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.011797333446641764,"reference_latency_ms":0.01229866687208414,"speedup_factor":1.0424954866038043}}} -{"definition":"gemm_n3072_k2048","workload":{"axes":{"M":232},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"38c1053f-1a32-4790-8ae5-b581427c1826"},"solution":"torch_matmul_73342c","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:13.384045","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.011855999939143658,"reference_latency_ms":0.012229333476473888,"speedup_factor":1.0314889962252476}}} -{"definition":"gemm_n3072_k2048","workload":{"axes":{"M":240},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"b3959a7c-687e-4a5c-80bd-0186d30425bb"},"solution":"torch_matmul_73342c","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:13.423968","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.01208000024780631,"reference_latency_ms":0.011818666631976763,"speedup_factor":0.9783664229744528}}} -{"definition":"gemm_n3072_k2048","workload":{"axes":{"M":248},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"38b8d208-9268-40e9-9123-314e97170347"},"solution":"torch_matmul_73342c","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:13.466241","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.012240000224361816,"reference_latency_ms":0.012442666726807753,"speedup_factor":1.0165577204845602}}} -{"definition":"gemm_n3072_k2048","workload":{"axes":{"M":256},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"ab069f30-7a43-4b22-acc2-ab6c5033bb82"},"solution":"torch_matmul_73342c","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:13.506450","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.012357333364586035,"reference_latency_ms":0.01241600001230836,"speedup_factor":1.0047475167976332}}} -{"definition":"gemm_n3072_k2048","workload":{"axes":{"M":972},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"9a43b59c-66b6-46af-8e71-a76a3f70e39c"},"solution":"torch_matmul_73342c","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:13.553623","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.016847999766469002,"reference_latency_ms":0.016853333140412968,"speedup_factor":1.000316558286912}}} -{"definition":"gemm_n3072_k2048","workload":{"axes":{"M":2053},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"c4e5fc75-2a7d-434c-92c1-557ec2b9dec7"},"solution":"torch_matmul_73342c","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:13.609164","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.02714133324722449,"reference_latency_ms":0.02711999975144863,"speedup_factor":0.9992139849733417}}} -{"definition":"gemm_n3072_k2048","workload":{"axes":{"M":2379},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"59e7eeae-b0e1-4b69-bbf8-7d038745eb55"},"solution":"torch_matmul_73342c","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:13.661019","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.02709333288172881,"reference_latency_ms":0.02706133263806502,"speedup_factor":0.9988188886246119}}} -{"definition":"gemm_n3072_k2048","workload":{"axes":{"M":8192},"inputs":{"A":{"type":"random"},"B":{"type":"random"}},"uuid":"6ffbca81-ec08-44fe-ac35-d823455a2f32"},"solution":"torch_matmul_73342c","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:13.726068","log":"","correctness":{"max_relative_error":0.0,"max_absolute_error":0.0,"extra":null},"performance":{"latency_ms":0.07023466502626736,"reference_latency_ms":0.07029333213965099,"speedup_factor":1.0008353013908684}}} diff --git a/traces/baseline/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64.jsonl b/traces/baseline/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64.jsonl deleted file mode 100644 index cc47c69d39adc0d1cea8d24e613daf43a87afc4d..0000000000000000000000000000000000000000 --- a/traces/baseline/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64.jsonl +++ /dev/null @@ -1,18 +0,0 @@ -{"definition": "gqa_paged_decode_h32_kv8_d128_ps64", "workload": {"axes": {"batch_size": 8, "num_pages": 910, "len_indptr": 9, "num_kv_indices": 640}, "inputs": {"q": {"type": "random"}, "k_cache": {"type": "random"}, "v_cache": {"type": "random"}, "kv_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_1a1f0089-58e8-4bb5-81a6-fbe229e3b845.safetensors", "tensor_key": "kv_indptr"}, "kv_indices": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_1a1f0089-58e8-4bb5-81a6-fbe229e3b845.safetensors", "tensor_key": "kv_indices"}, "kv_last_page_len": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_1a1f0089-58e8-4bb5-81a6-fbe229e3b845.safetensors", "tensor_key": "kv_last_page_len"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831843}}, "uuid": "1a1f0089-58e8-4bb5-81a6-fbe229e3b845"}, "solution": "flashinfer_wrapper_ad4135", "evaluation": {"status": "PASSED", "environment": {"hardware": "NVIDIA B200", "libs": {"torch": "2.9.1+cu128", "triton": "3.5.1", "cuda": "12.8"}}, "timestamp": "2026-04-30T01:30:00.391266", "log": "", "correctness": {"max_relative_error": 661.892822265625, "max_absolute_error": 0.0009765625, "extra": null}, "performance": {"latency_ms": 0.0640105, "reference_latency_ms": 9.019708166666668, "speedup_factor": 140.90982208648063}}} -{"definition": "gqa_paged_decode_h32_kv8_d128_ps64", "workload": {"axes": {"batch_size": 8, "num_pages": 839, "len_indptr": 9, "num_kv_indices": 528}, "inputs": {"q": {"type": "random"}, "k_cache": {"type": "random"}, "v_cache": {"type": "random"}, "kv_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_ed7c988c-a1a9-4e62-b410-2e33b2bdc1d8.safetensors", "tensor_key": "kv_indptr"}, "kv_indices": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_ed7c988c-a1a9-4e62-b410-2e33b2bdc1d8.safetensors", "tensor_key": "kv_indices"}, "kv_last_page_len": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_ed7c988c-a1a9-4e62-b410-2e33b2bdc1d8.safetensors", "tensor_key": "kv_last_page_len"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831843}}, "uuid": "ed7c988c-a1a9-4e62-b410-2e33b2bdc1d8"}, "solution": "flashinfer_wrapper_ad4135", "evaluation": {"status": "PASSED", "environment": {"hardware": "NVIDIA B200", "libs": {"torch": "2.9.1+cu128", "triton": "3.5.1", "cuda": "12.8"}}, "timestamp": "2026-04-30T01:30:04.252270", "log": "", "correctness": {"max_relative_error": 183.28530883789062, "max_absolute_error": 0.001953125, "extra": null}, "performance": {"latency_ms": 0.057812833333333334, "reference_latency_ms": 9.4367515, "speedup_factor": 163.22935507398876}}} -{"definition": "gqa_paged_decode_h32_kv8_d128_ps64", "workload": {"axes": {"batch_size": 8, "num_pages": 902, "len_indptr": 9, "num_kv_indices": 576}, "inputs": {"q": {"type": "random"}, "k_cache": {"type": "random"}, "v_cache": {"type": "random"}, "kv_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_1735d27c-3a9a-44b4-9fa1-e9a0e3703d38.safetensors", "tensor_key": "kv_indptr"}, "kv_indices": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_1735d27c-3a9a-44b4-9fa1-e9a0e3703d38.safetensors", "tensor_key": "kv_indices"}, "kv_last_page_len": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_1735d27c-3a9a-44b4-9fa1-e9a0e3703d38.safetensors", "tensor_key": "kv_last_page_len"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831843}}, "uuid": "1735d27c-3a9a-44b4-9fa1-e9a0e3703d38"}, "solution": "flashinfer_wrapper_ad4135", "evaluation": {"status": "PASSED", "environment": {"hardware": "NVIDIA B200", "libs": {"torch": "2.9.1+cu128", "triton": "3.5.1", "cuda": "12.8"}}, "timestamp": "2026-04-30T01:30:08.099438", "log": "", "correctness": {"max_relative_error": 138.99830627441406, "max_absolute_error": 0.001953125, "extra": null}, "performance": {"latency_ms": 0.0584905, "reference_latency_ms": 9.897810166666666, "speedup_factor": 169.22081648586806}}} -{"definition": "gqa_paged_decode_h32_kv8_d128_ps64", "workload": {"axes": {"batch_size": 31, "num_pages": 15233, "len_indptr": 32, "num_kv_indices": 14145}, "inputs": {"q": {"type": "random"}, "k_cache": {"type": "random"}, "v_cache": {"type": "random"}, "kv_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_1bdf652e-366a-4024-9236-ce54682a7cf7.safetensors", "tensor_key": "kv_indptr"}, "kv_indices": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_1bdf652e-366a-4024-9236-ce54682a7cf7.safetensors", "tensor_key": "kv_indices"}, "kv_last_page_len": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_1bdf652e-366a-4024-9236-ce54682a7cf7.safetensors", "tensor_key": "kv_last_page_len"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831843}}, "uuid": "1bdf652e-366a-4024-9236-ce54682a7cf7"}, "solution": "flashinfer_wrapper_ad4135", "evaluation": {"status": "PASSED", "environment": {"hardware": "NVIDIA B200", "libs": {"torch": "2.9.1+cu128", "triton": "3.5.1", "cuda": "12.8"}}, "timestamp": "2026-04-30T01:30:28.148941", "log": "", "correctness": {"max_relative_error": 543.0467529296875, "max_absolute_error": 0.00048828125, "extra": null}, "performance": {"latency_ms": 1.262074, "reference_latency_ms": 60.46130016666666, "speedup_factor": 47.906303565929306}}} -{"definition": "gqa_paged_decode_h32_kv8_d128_ps64", "workload": {"axes": {"batch_size": 32, "num_pages": 15114, "len_indptr": 33, "num_kv_indices": 13783}, "inputs": {"q": {"type": "random"}, "k_cache": {"type": "random"}, "v_cache": {"type": "random"}, "kv_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_721ff730-4d9e-49d8-9fd4-7ba713b7cfdb.safetensors", "tensor_key": "kv_indptr"}, "kv_indices": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_721ff730-4d9e-49d8-9fd4-7ba713b7cfdb.safetensors", "tensor_key": "kv_indices"}, "kv_last_page_len": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_721ff730-4d9e-49d8-9fd4-7ba713b7cfdb.safetensors", "tensor_key": "kv_last_page_len"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831843}}, "uuid": "721ff730-4d9e-49d8-9fd4-7ba713b7cfdb"}, "solution": "flashinfer_wrapper_ad4135", "evaluation": {"status": "PASSED", "environment": {"hardware": "NVIDIA B200", "libs": {"torch": "2.9.1+cu128", "triton": "3.5.1", "cuda": "12.8"}}, "timestamp": "2026-04-30T01:30:40.918306", "log": "", "correctness": {"max_relative_error": 425.1051025390625, "max_absolute_error": 0.00048828125, "extra": null}, "performance": {"latency_ms": 1.2502615, "reference_latency_ms": 48.024005666666675, "speedup_factor": 38.41116891679595}}} -{"definition": "gqa_paged_decode_h32_kv8_d128_ps64", "workload": {"axes": {"batch_size": 32, "num_pages": 15178, "len_indptr": 33, "num_kv_indices": 14103}, "inputs": {"q": {"type": "random"}, "k_cache": {"type": "random"}, "v_cache": {"type": "random"}, "kv_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_ff1d78b9-0383-4845-94df-904c49112856.safetensors", "tensor_key": "kv_indptr"}, "kv_indices": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_ff1d78b9-0383-4845-94df-904c49112856.safetensors", "tensor_key": "kv_indices"}, "kv_last_page_len": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_ff1d78b9-0383-4845-94df-904c49112856.safetensors", "tensor_key": "kv_last_page_len"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831843}}, "uuid": "ff1d78b9-0383-4845-94df-904c49112856"}, "solution": "flashinfer_wrapper_ad4135", "evaluation": {"status": "PASSED", "environment": {"hardware": "NVIDIA B200", "libs": {"torch": "2.9.1+cu128", "triton": "3.5.1", "cuda": "12.8"}}, "timestamp": "2026-04-30T01:30:53.452444", "log": "", "correctness": {"max_relative_error": 43.10818099975586, "max_absolute_error": 0.00048828125, "extra": null}, "performance": {"latency_ms": 1.2630401666666666, "reference_latency_ms": 48.34570716666667, "speedup_factor": 38.2772523333581}}} -{"definition": "gqa_paged_decode_h32_kv8_d128_ps64", "workload": {"axes": {"batch_size": 63, "num_pages": 29569, "len_indptr": 64, "num_kv_indices": 27392}, "inputs": {"q": {"type": "random"}, "k_cache": {"type": "random"}, "v_cache": {"type": "random"}, "kv_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_16d8ee84-f85b-4192-a58a-2228a5d9641e.safetensors", "tensor_key": "kv_indptr"}, "kv_indices": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_16d8ee84-f85b-4192-a58a-2228a5d9641e.safetensors", "tensor_key": "kv_indices"}, "kv_last_page_len": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_16d8ee84-f85b-4192-a58a-2228a5d9641e.safetensors", "tensor_key": "kv_last_page_len"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831843}}, "uuid": "16d8ee84-f85b-4192-a58a-2228a5d9641e"}, "solution": "flashinfer_wrapper_ad4135", "evaluation": {"status": "PASSED", "environment": {"hardware": "NVIDIA B200", "libs": {"torch": "2.9.1+cu128", "triton": "3.5.1", "cuda": "12.8"}}, "timestamp": "2026-04-30T01:31:17.621445", "log": "", "correctness": {"max_relative_error": 155.3499298095703, "max_absolute_error": 0.00048828125, "extra": null}, "performance": {"latency_ms": 2.8545226666666665, "reference_latency_ms": 93.65224183333333, "speedup_factor": 32.808372106112785}}} -{"definition": "gqa_paged_decode_h32_kv8_d128_ps64", "workload": {"axes": {"batch_size": 64, "num_pages": 29189, "len_indptr": 65, "num_kv_indices": 26582}, "inputs": {"q": {"type": "random"}, "k_cache": {"type": "random"}, "v_cache": {"type": "random"}, "kv_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_51393402-8fec-4b52-88ca-3cdb2270d293.safetensors", "tensor_key": "kv_indptr"}, "kv_indices": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_51393402-8fec-4b52-88ca-3cdb2270d293.safetensors", "tensor_key": "kv_indices"}, "kv_last_page_len": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_51393402-8fec-4b52-88ca-3cdb2270d293.safetensors", "tensor_key": "kv_last_page_len"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831843}}, "uuid": "51393402-8fec-4b52-88ca-3cdb2270d293"}, "solution": "flashinfer_wrapper_ad4135", "evaluation": {"status": "PASSED", "environment": {"hardware": "NVIDIA B200", "libs": {"torch": "2.9.1+cu128", "triton": "3.5.1", "cuda": "12.8"}}, "timestamp": "2026-04-30T01:31:41.029542", "log": "", "correctness": {"max_relative_error": 769.1510009765625, "max_absolute_error": 0.0009765625, "extra": null}, "performance": {"latency_ms": 2.6916171666666666, "reference_latency_ms": 92.9195085, "speedup_factor": 34.52181448785777}}} -{"definition": "gqa_paged_decode_h32_kv8_d128_ps64", "workload": {"axes": {"batch_size": 64, "num_pages": 29441, "len_indptr": 65, "num_kv_indices": 27222}, "inputs": {"q": {"type": "random"}, "k_cache": {"type": "random"}, "v_cache": {"type": "random"}, "kv_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_8c88a86f-1843-42eb-9251-8a0ef4b6a1ee.safetensors", "tensor_key": "kv_indptr"}, "kv_indices": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_8c88a86f-1843-42eb-9251-8a0ef4b6a1ee.safetensors", "tensor_key": "kv_indices"}, "kv_last_page_len": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_8c88a86f-1843-42eb-9251-8a0ef4b6a1ee.safetensors", "tensor_key": "kv_last_page_len"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831843}}, "uuid": "8c88a86f-1843-42eb-9251-8a0ef4b6a1ee"}, "solution": "flashinfer_wrapper_ad4135", "evaluation": {"status": "PASSED", "environment": {"hardware": "NVIDIA B200", "libs": {"torch": "2.9.1+cu128", "triton": "3.5.1", "cuda": "12.8"}}, "timestamp": "2026-04-30T01:32:05.423734", "log": "", "correctness": {"max_relative_error": 242.13320922851562, "max_absolute_error": 0.00048828125, "extra": null}, "performance": {"latency_ms": 2.7251739999999995, "reference_latency_ms": 94.36251966666667, "speedup_factor": 34.62623658770658}}} -{"definition": "gqa_paged_decode_h32_kv8_d128_ps64", "workload": {"axes": {"batch_size": 127, "num_pages": 52481, "len_indptr": 128, "num_kv_indices": 48120}, "inputs": {"q": {"type": "random"}, "k_cache": {"type": "random"}, "v_cache": {"type": "random"}, "kv_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_7832673f-b1c1-4185-b298-d510edbceac4.safetensors", "tensor_key": "kv_indptr"}, "kv_indices": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_7832673f-b1c1-4185-b298-d510edbceac4.safetensors", "tensor_key": "kv_indices"}, "kv_last_page_len": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_7832673f-b1c1-4185-b298-d510edbceac4.safetensors", "tensor_key": "kv_last_page_len"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831843}}, "uuid": "7832673f-b1c1-4185-b298-d510edbceac4"}, "solution": "flashinfer_wrapper_ad4135", "evaluation": {"status": "PASSED", "environment": {"hardware": "NVIDIA B200", "libs": {"torch": "2.9.1+cu128", "triton": "3.5.1", "cuda": "12.8"}}, "timestamp": "2026-04-30T01:32:52.890706", "log": "", "correctness": {"max_relative_error": 76.90428161621094, "max_absolute_error": 0.00048828125, "extra": null}, "performance": {"latency_ms": 6.960167166666667, "reference_latency_ms": 170.976936, "speedup_factor": 24.565061715591455}}} -{"definition": "gqa_paged_decode_h32_kv8_d128_ps64", "workload": {"axes": {"batch_size": 16, "num_pages": 1794, "len_indptr": 17, "num_kv_indices": 1257}, "inputs": {"q": {"type": "random"}, "k_cache": {"type": "random"}, "v_cache": {"type": "random"}, "kv_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_ee6efcf9-bf19-432f-bcd3-33f1c9ec599e.safetensors", "tensor_key": "kv_indptr"}, "kv_indices": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_ee6efcf9-bf19-432f-bcd3-33f1c9ec599e.safetensors", "tensor_key": "kv_indices"}, "kv_last_page_len": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_ee6efcf9-bf19-432f-bcd3-33f1c9ec599e.safetensors", "tensor_key": "kv_last_page_len"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831843}}, "uuid": "ee6efcf9-bf19-432f-bcd3-33f1c9ec599e"}, "solution": "flashinfer_wrapper_ad4135", "evaluation": {"status": "PASSED", "environment": {"hardware": "NVIDIA B200", "libs": {"torch": "2.9.1+cu128", "triton": "3.5.1", "cuda": "12.8"}}, "timestamp": "2026-04-30T01:33:49.907646", "log": "", "correctness": {"max_relative_error": 285.8631591796875, "max_absolute_error": 0.0009765625, "extra": null}, "performance": {"latency_ms": 0.11598366666666667, "reference_latency_ms": 8.0027205, "speedup_factor": 68.99868515969203}}} -{"definition": "gqa_paged_decode_h32_kv8_d128_ps64", "workload": {"axes": {"batch_size": 16, "num_pages": 1686, "len_indptr": 17, "num_kv_indices": 1033}, "inputs": {"q": {"type": "random"}, "k_cache": {"type": "random"}, "v_cache": {"type": "random"}, "kv_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_1196e756-48f7-4f94-81a3-94a68f170b59.safetensors", "tensor_key": "kv_indptr"}, "kv_indices": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_1196e756-48f7-4f94-81a3-94a68f170b59.safetensors", "tensor_key": "kv_indices"}, "kv_last_page_len": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_1196e756-48f7-4f94-81a3-94a68f170b59.safetensors", "tensor_key": "kv_last_page_len"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831843}}, "uuid": "1196e756-48f7-4f94-81a3-94a68f170b59"}, "solution": "flashinfer_wrapper_ad4135", "evaluation": {"status": "PASSED", "environment": {"hardware": "NVIDIA B200", "libs": {"torch": "2.9.1+cu128", "triton": "3.5.1", "cuda": "12.8"}}, "timestamp": "2026-04-30T01:33:52.667322", "log": "", "correctness": {"max_relative_error": 180.0902099609375, "max_absolute_error": 0.0009765625, "extra": null}, "performance": {"latency_ms": 0.10291133333333334, "reference_latency_ms": 7.839893333333333, "speedup_factor": 76.18104905841274}}} -{"definition": "gqa_paged_decode_h32_kv8_d128_ps64", "workload": {"axes": {"batch_size": 16, "num_pages": 1735, "len_indptr": 17, "num_kv_indices": 1145}, "inputs": {"q": {"type": "random"}, "k_cache": {"type": "random"}, "v_cache": {"type": "random"}, "kv_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_09bcbbfb-0d48-42d8-a39c-e7d01300a646.safetensors", "tensor_key": "kv_indptr"}, "kv_indices": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_09bcbbfb-0d48-42d8-a39c-e7d01300a646.safetensors", "tensor_key": "kv_indices"}, "kv_last_page_len": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_09bcbbfb-0d48-42d8-a39c-e7d01300a646.safetensors", "tensor_key": "kv_last_page_len"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831843}}, "uuid": "09bcbbfb-0d48-42d8-a39c-e7d01300a646"}, "solution": "flashinfer_wrapper_ad4135", "evaluation": {"status": "PASSED", "environment": {"hardware": "NVIDIA B200", "libs": {"torch": "2.9.1+cu128", "triton": "3.5.1", "cuda": "12.8"}}, "timestamp": "2026-04-30T01:33:55.469433", "log": "", "correctness": {"max_relative_error": 260.6885070800781, "max_absolute_error": 0.001953125, "extra": null}, "performance": {"latency_ms": 0.11564783333333334, "reference_latency_ms": 7.895311333333333, "speedup_factor": 68.27029184867276}}} -{"definition": "gqa_paged_decode_h32_kv8_d128_ps64", "workload": {"axes": {"batch_size": 128, "num_pages": 51792, "len_indptr": 129, "num_kv_indices": 46414}, "inputs": {"q": {"type": "random"}, "k_cache": {"type": "random"}, "v_cache": {"type": "random"}, "kv_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_c81ef23a-da17-4cdd-88c2-327a4af240d0.safetensors", "tensor_key": "kv_indptr"}, "kv_indices": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_c81ef23a-da17-4cdd-88c2-327a4af240d0.safetensors", "tensor_key": "kv_indices"}, "kv_last_page_len": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_c81ef23a-da17-4cdd-88c2-327a4af240d0.safetensors", "tensor_key": "kv_last_page_len"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831843}}, "uuid": "c81ef23a-da17-4cdd-88c2-327a4af240d0"}, "solution": "flashinfer_wrapper_ad4135", "evaluation": {"status": "PASSED", "environment": {"hardware": "NVIDIA B200", "libs": {"torch": "2.9.1+cu128", "triton": "3.5.1", "cuda": "12.8"}}, "timestamp": "2026-04-30T01:40:34.811864", "log": "", "correctness": {"max_relative_error": 196.78802490234375, "max_absolute_error": 0.0009765625, "extra": null}, "performance": {"latency_ms": 7.1282915000000004, "reference_latency_ms": 167.85268583333334, "speedup_factor": 23.54739362627543}}} -{"definition": "gqa_paged_decode_h32_kv8_d128_ps64", "workload": {"axes": {"batch_size": 128, "num_pages": 52033, "len_indptr": 129, "num_kv_indices": 47438}, "inputs": {"q": {"type": "random"}, "k_cache": {"type": "random"}, "v_cache": {"type": "random"}, "kv_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_2b180048-9203-476c-a5b4-255118f15059.safetensors", "tensor_key": "kv_indptr"}, "kv_indices": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_2b180048-9203-476c-a5b4-255118f15059.safetensors", "tensor_key": "kv_indices"}, "kv_last_page_len": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_2b180048-9203-476c-a5b4-255118f15059.safetensors", "tensor_key": "kv_last_page_len"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831843}}, "uuid": "2b180048-9203-476c-a5b4-255118f15059"}, "solution": "flashinfer_wrapper_ad4135", "evaluation": {"status": "PASSED", "environment": {"hardware": "NVIDIA B200", "libs": {"torch": "2.9.1+cu128", "triton": "3.5.1", "cuda": "12.8"}}, "timestamp": "2026-04-30T01:43:31.325401", "log": "", "correctness": {"max_relative_error": 57.499549865722656, "max_absolute_error": 0.00048828125, "extra": null}, "performance": {"latency_ms": 7.188028333333333, "reference_latency_ms": 168.88721466666667, "speedup_factor": 23.49562450713768}}} -{"definition":"gqa_paged_decode_h32_kv8_d128_ps64","workload":{"axes":{"batch_size":4,"num_pages":398,"len_indptr":5,"num_kv_indices":268},"inputs":{"q":{"type":"random"},"k_cache":{"type":"random"},"v_cache":{"type":"random"},"kv_indptr":{"type":"safetensors","path":"./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_580fa22c-dee6-484e-8847-eb301b21057f.safetensors","tensor_key":"kv_indptr"},"kv_indices":{"type":"safetensors","path":"./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_580fa22c-dee6-484e-8847-eb301b21057f.safetensors","tensor_key":"kv_indices"},"kv_last_page_len":{"type":"safetensors","path":"./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_580fa22c-dee6-484e-8847-eb301b21057f.safetensors","tensor_key":"kv_last_page_len"},"sm_scale":{"type":"scalar","value":0.08838834764831843}},"uuid":"580fa22c-dee6-484e-8847-eb301b21057f"},"solution":"flashinfer_wrapper_ad4135","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.9.1+cu128","triton":"3.5.1","cuda":"12.8"}},"timestamp":"2026-04-30T01:56:33.332047","log":"","correctness":{"max_relative_error":114.7226333618164,"max_absolute_error":0.00048828125,"extra":null},"performance":{"latency_ms":0.03439966666666667,"reference_latency_ms":29.673828,"speedup_factor":862.6196377871878}}} -{"definition":"gqa_paged_decode_h32_kv8_d128_ps64","workload":{"axes":{"batch_size":4,"num_pages":349,"len_indptr":5,"num_kv_indices":212},"inputs":{"q":{"type":"random"},"k_cache":{"type":"random"},"v_cache":{"type":"random"},"kv_indptr":{"type":"safetensors","path":"./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_b013fc3f-eca8-4c18-a5a1-7231c26eb231.safetensors","tensor_key":"kv_indptr"},"kv_indices":{"type":"safetensors","path":"./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_b013fc3f-eca8-4c18-a5a1-7231c26eb231.safetensors","tensor_key":"kv_indices"},"kv_last_page_len":{"type":"safetensors","path":"./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_b013fc3f-eca8-4c18-a5a1-7231c26eb231.safetensors","tensor_key":"kv_last_page_len"},"sm_scale":{"type":"scalar","value":0.08838834764831843}},"uuid":"b013fc3f-eca8-4c18-a5a1-7231c26eb231"},"solution":"flashinfer_wrapper_ad4135","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.9.1+cu128","triton":"3.5.1","cuda":"12.8"}},"timestamp":"2026-04-30T01:56:40.666622","log":"","correctness":{"max_relative_error":55.278594970703125,"max_absolute_error":0.0009765625,"extra":null},"performance":{"latency_ms":0.028751666666666665,"reference_latency_ms":29.503797333333335,"speedup_factor":1026.1595501710046}}} -{"definition":"gqa_paged_decode_h32_kv8_d128_ps64","workload":{"axes":{"batch_size":4,"num_pages":389,"len_indptr":5,"num_kv_indices":232},"inputs":{"q":{"type":"random"},"k_cache":{"type":"random"},"v_cache":{"type":"random"},"kv_indptr":{"type":"safetensors","path":"./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_1e0c7af6-f077-42f6-b5a0-3cbb2e30c5c3.safetensors","tensor_key":"kv_indptr"},"kv_indices":{"type":"safetensors","path":"./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_1e0c7af6-f077-42f6-b5a0-3cbb2e30c5c3.safetensors","tensor_key":"kv_indices"},"kv_last_page_len":{"type":"safetensors","path":"./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_1e0c7af6-f077-42f6-b5a0-3cbb2e30c5c3.safetensors","tensor_key":"kv_last_page_len"},"sm_scale":{"type":"scalar","value":0.08838834764831843}},"uuid":"1e0c7af6-f077-42f6-b5a0-3cbb2e30c5c3"},"solution":"flashinfer_wrapper_ad4135","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.9.1+cu128","triton":"3.5.1","cuda":"12.8"}},"timestamp":"2026-04-30T01:56:47.993725","log":"","correctness":{"max_relative_error":110.45864868164062,"max_absolute_error":0.0009765625,"extra":null},"performance":{"latency_ms":0.031407833333333336,"reference_latency_ms":29.568117166666667,"speedup_factor":941.4249258412178}}} diff --git a/traces/baseline/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps1.jsonl b/traces/baseline/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps1.jsonl deleted file mode 100644 index a536d451fa788fd3c26ef9a01e099e4e620ee2a3..0000000000000000000000000000000000000000 --- a/traces/baseline/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps1.jsonl +++ /dev/null @@ -1,8 +0,0 @@ -{"definition": "gqa_paged_decode_h32_kv8_d64_ps1", "workload": {"axes": {"batch_size": 1, "num_pages": 9, "len_indptr": 2, "num_kv_indices": 8}, "inputs": {"q": {"type": "random"}, "k_cache": {"type": "random"}, "v_cache": {"type": "random"}, "kv_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps1/gqa_paged_decode_h32_kv8_d64_ps1_8342a570-03f6-42ab-b7fb-bc1eadc8759e.safetensors", "tensor_key": "kv_indptr"}, "kv_indices": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps1/gqa_paged_decode_h32_kv8_d64_ps1_8342a570-03f6-42ab-b7fb-bc1eadc8759e.safetensors", "tensor_key": "kv_indices"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831843}}, "uuid": "8342a570-03f6-42ab-b7fb-bc1eadc8759e"}, "solution": "flashinfer_wrapper_3f9411", "evaluation": {"status": "PASSED", "environment": {"hardware": "NVIDIA B200", "libs": {"torch": "2.11.0+cu130", "triton": "3.6.0", "cuda": "13.0"}}, "timestamp": "2026-04-20T07:33:20.340034", "log": "", "correctness": {"max_relative_error": 0.004031917080283165, "max_absolute_error": 4.76837158203125e-07, "extra": null}, "performance": {"latency_ms": 0.0081599997356534, "reference_latency_ms": 0.21507733315229416, "speedup_factor": 26.357517171545858}}} -{"definition": "gqa_paged_decode_h32_kv8_d64_ps1", "workload": {"axes": {"batch_size": 1, "num_pages": 73, "len_indptr": 2, "num_kv_indices": 56}, "inputs": {"q": {"type": "random"}, "k_cache": {"type": "random"}, "v_cache": {"type": "random"}, "kv_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps1/gqa_paged_decode_h32_kv8_d64_ps1_5d4bb19c-8f86-41c2-a3dc-ab974ba60bdf.safetensors", "tensor_key": "kv_indptr"}, "kv_indices": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps1/gqa_paged_decode_h32_kv8_d64_ps1_5d4bb19c-8f86-41c2-a3dc-ab974ba60bdf.safetensors", "tensor_key": "kv_indices"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831843}}, "uuid": "5d4bb19c-8f86-41c2-a3dc-ab974ba60bdf"}, "solution": "flashinfer_wrapper_3f9411", "evaluation": {"status": "PASSED", "environment": {"hardware": "NVIDIA B200", "libs": {"torch": "2.11.0+cu130", "triton": "3.6.0", "cuda": "13.0"}}, "timestamp": "2026-04-20T07:33:20.423031", "log": "", "correctness": {"max_relative_error": 0.04477168619632721, "max_absolute_error": 9.5367431640625e-07, "extra": null}, "performance": {"latency_ms": 0.010208000428974628, "reference_latency_ms": 0.19693333407243094, "speedup_factor": 19.2920577778828}}} -{"definition": "gqa_paged_decode_h32_kv8_d64_ps1", "workload": {"axes": {"batch_size": 1, "num_pages": 18, "len_indptr": 2, "num_kv_indices": 2}, "inputs": {"q": {"type": "random"}, "k_cache": {"type": "random"}, "v_cache": {"type": "random"}, "kv_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps1/gqa_paged_decode_h32_kv8_d64_ps1_3d044768-4601-4b70-8472-85fbc34a4504.safetensors", "tensor_key": "kv_indptr"}, "kv_indices": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps1/gqa_paged_decode_h32_kv8_d64_ps1_3d044768-4601-4b70-8472-85fbc34a4504.safetensors", "tensor_key": "kv_indices"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831843}}, "uuid": "3d044768-4601-4b70-8472-85fbc34a4504"}, "solution": "flashinfer_wrapper_3f9411", "evaluation": {"status": "PASSED", "environment": {"hardware": "NVIDIA B200", "libs": {"torch": "2.11.0+cu130", "triton": "3.6.0", "cuda": "13.0"}}, "timestamp": "2026-04-20T07:33:20.506419", "log": "", "correctness": {"max_relative_error": 0.00505049666389823, "max_absolute_error": 3.0517578125e-05, "extra": null}, "performance": {"latency_ms": 0.008175999857485294, "reference_latency_ms": 0.20251733313004175, "speedup_factor": 24.76973295744777}}} -{"definition": "gqa_paged_decode_h32_kv8_d64_ps1", "workload": {"axes": {"batch_size": 1, "num_pages": 16, "len_indptr": 2, "num_kv_indices": 15}, "inputs": {"q": {"type": "random"}, "k_cache": {"type": "random"}, "v_cache": {"type": "random"}, "kv_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps1/gqa_paged_decode_h32_kv8_d64_ps1_921b9fdc-3937-4873-b11e-7f4650daf0f8.safetensors", "tensor_key": "kv_indptr"}, "kv_indices": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps1/gqa_paged_decode_h32_kv8_d64_ps1_921b9fdc-3937-4873-b11e-7f4650daf0f8.safetensors", "tensor_key": "kv_indices"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831843}}, "uuid": "921b9fdc-3937-4873-b11e-7f4650daf0f8"}, "solution": "flashinfer_wrapper_3f9411", "evaluation": {"status": "PASSED", "environment": {"hardware": "NVIDIA B200", "libs": {"torch": "2.11.0+cu130", "triton": "3.6.0", "cuda": "13.0"}}, "timestamp": "2026-04-20T07:33:20.589818", "log": "", "correctness": {"max_relative_error": 0.005291004665195942, "max_absolute_error": 0.00048828125, "extra": null}, "performance": {"latency_ms": 0.00817066648354133, "reference_latency_ms": 0.20868266622225443, "speedup_factor": 25.54047049192579}}} -{"definition": "gqa_paged_decode_h32_kv8_d64_ps1", "workload": {"axes": {"batch_size": 1, "num_pages": 9, "len_indptr": 2, "num_kv_indices": 8}, "inputs": {"q": {"type": "random"}, "k_cache": {"type": "random"}, "v_cache": {"type": "random"}, "kv_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps1/gqa_paged_decode_h32_kv8_d64_ps1_011db24f-e331-41e4-89fd-dff3bb7345e5.safetensors", "tensor_key": "kv_indptr"}, "kv_indices": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps1/gqa_paged_decode_h32_kv8_d64_ps1_011db24f-e331-41e4-89fd-dff3bb7345e5.safetensors", "tensor_key": "kv_indices"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831843}}, "uuid": "011db24f-e331-41e4-89fd-dff3bb7345e5"}, "solution": "flashinfer_wrapper_3f9411", "evaluation": {"status": "PASSED", "environment": {"hardware": "NVIDIA B200", "libs": {"torch": "2.11.0+cu130", "triton": "3.6.0", "cuda": "13.0"}}, "timestamp": "2026-04-20T07:33:20.671320", "log": "", "correctness": {"max_relative_error": 1.3320982361619826e-07, "max_absolute_error": 4.76837158203125e-07, "extra": null}, "performance": {"latency_ms": 0.008154666516929865, "reference_latency_ms": 0.20191466559966406, "speedup_factor": 24.760628185158765}}} -{"definition": "gqa_paged_decode_h32_kv8_d64_ps1", "workload": {"axes": {"batch_size": 1, "num_pages": 73, "len_indptr": 2, "num_kv_indices": 56}, "inputs": {"q": {"type": "random"}, "k_cache": {"type": "random"}, "v_cache": {"type": "random"}, "kv_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps1/gqa_paged_decode_h32_kv8_d64_ps1_7810a1b5-7bcd-4825-b36f-80c9b4ec61c6.safetensors", "tensor_key": "kv_indptr"}, "kv_indices": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps1/gqa_paged_decode_h32_kv8_d64_ps1_7810a1b5-7bcd-4825-b36f-80c9b4ec61c6.safetensors", "tensor_key": "kv_indices"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831843}}, "uuid": "7810a1b5-7bcd-4825-b36f-80c9b4ec61c6"}, "solution": "flashinfer_wrapper_3f9411", "evaluation": {"status": "PASSED", "environment": {"hardware": "NVIDIA B200", "libs": {"torch": "2.11.0+cu130", "triton": "3.6.0", "cuda": "13.0"}}, "timestamp": "2026-04-20T07:33:20.752889", "log": "", "correctness": {"max_relative_error": 1.5637542105650937e-07, "max_absolute_error": 9.5367431640625e-07, "extra": null}, "performance": {"latency_ms": 0.010208000428974628, "reference_latency_ms": 0.19841066747903824, "speedup_factor": 19.436780871978094}}} -{"definition": "gqa_paged_decode_h32_kv8_d64_ps1", "workload": {"axes": {"batch_size": 1, "num_pages": 18, "len_indptr": 2, "num_kv_indices": 2}, "inputs": {"q": {"type": "random"}, "k_cache": {"type": "random"}, "v_cache": {"type": "random"}, "kv_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps1/gqa_paged_decode_h32_kv8_d64_ps1_e97b000c-8677-4877-86bf-2a82b9cb77e4.safetensors", "tensor_key": "kv_indptr"}, "kv_indices": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps1/gqa_paged_decode_h32_kv8_d64_ps1_e97b000c-8677-4877-86bf-2a82b9cb77e4.safetensors", "tensor_key": "kv_indices"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831843}}, "uuid": "e97b000c-8677-4877-86bf-2a82b9cb77e4"}, "solution": "flashinfer_wrapper_3f9411", "evaluation": {"status": "PASSED", "environment": {"hardware": "NVIDIA B200", "libs": {"torch": "2.11.0+cu130", "triton": "3.6.0", "cuda": "13.0"}}, "timestamp": "2026-04-20T07:33:20.833093", "log": "", "correctness": {"max_relative_error": 2.020364036070532e-06, "max_absolute_error": 2.384185791015625e-07, "extra": null}, "performance": {"latency_ms": 0.008165333109597364, "reference_latency_ms": 0.19856533408164978, "speedup_factor": 24.318093507815398}}} -{"definition": "gqa_paged_decode_h32_kv8_d64_ps1", "workload": {"axes": {"batch_size": 1, "num_pages": 16, "len_indptr": 2, "num_kv_indices": 15}, "inputs": {"q": {"type": "random"}, "k_cache": {"type": "random"}, "v_cache": {"type": "random"}, "kv_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps1/gqa_paged_decode_h32_kv8_d64_ps1_11155649-fd33-4b5c-80d8-05553c0cec0b.safetensors", "tensor_key": "kv_indptr"}, "kv_indices": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps1/gqa_paged_decode_h32_kv8_d64_ps1_11155649-fd33-4b5c-80d8-05553c0cec0b.safetensors", "tensor_key": "kv_indices"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831843}}, "uuid": "11155649-fd33-4b5c-80d8-05553c0cec0b"}, "solution": "flashinfer_wrapper_3f9411", "evaluation": {"status": "PASSED", "environment": {"hardware": "NVIDIA B200", "libs": {"torch": "2.11.0+cu130", "triton": "3.6.0", "cuda": "13.0"}}, "timestamp": "2026-04-20T07:33:20.913983", "log": "", "correctness": {"max_relative_error": 2.287309399662263e-07, "max_absolute_error": 9.5367431640625e-07, "extra": null}, "performance": {"latency_ms": 0.0081599997356534, "reference_latency_ms": 0.2066026652852694, "speedup_factor": 25.318954899295225}}} diff --git a/traces/baseline/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps64.jsonl b/traces/baseline/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps64.jsonl deleted file mode 100644 index 9dcb9fd5f9d629f55467302c689e5eb8b5a4fb38..0000000000000000000000000000000000000000 --- a/traces/baseline/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps64.jsonl +++ /dev/null @@ -1,8 +0,0 @@ -{"definition":"gqa_paged_decode_h32_kv8_d64_ps64","workload":{"axes":{"batch_size":1,"num_pages":72,"len_indptr":2,"num_kv_indices":8},"inputs":{"q":{"type":"random"},"k_cache":{"type":"random"},"v_cache":{"type":"random"},"kv_indptr":{"type":"safetensors","path":"./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps64/gqa_paged_decode_h32_kv8_d64_ps64_4b40f9c1-2054-4b1c-ab80-0873f2801f7e.safetensors","tensor_key":"kv_indptr"},"kv_indices":{"type":"safetensors","path":"./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps64/gqa_paged_decode_h32_kv8_d64_ps64_4b40f9c1-2054-4b1c-ab80-0873f2801f7e.safetensors","tensor_key":"kv_indices"},"kv_last_page_len":{"type":"safetensors","path":"./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps64/gqa_paged_decode_h32_kv8_d64_ps64_4b40f9c1-2054-4b1c-ab80-0873f2801f7e.safetensors","tensor_key":"kv_last_page_len"},"sm_scale":{"type":"scalar","value":0.08838834764831843}},"uuid":"4b40f9c1-2054-4b1c-ab80-0873f2801f7e"},"solution":"flashinfer_wrapper_68ff49","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:21.034597","log":"","correctness":{"max_relative_error":3.218233823776245,"max_absolute_error":0.0009765625,"extra":null},"performance":{"latency_ms":0.013717333320528269,"reference_latency_ms":0.258213331302007,"speedup_factor":18.823872342271184}}} -{"definition":"gqa_paged_decode_h32_kv8_d64_ps64","workload":{"axes":{"batch_size":1,"num_pages":120,"len_indptr":2,"num_kv_indices":56},"inputs":{"q":{"type":"random"},"k_cache":{"type":"random"},"v_cache":{"type":"random"},"kv_indptr":{"type":"safetensors","path":"./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps64/gqa_paged_decode_h32_kv8_d64_ps64_3a4d1bf7-c9fe-40b9-a50d-9780a14fabf7.safetensors","tensor_key":"kv_indptr"},"kv_indices":{"type":"safetensors","path":"./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps64/gqa_paged_decode_h32_kv8_d64_ps64_3a4d1bf7-c9fe-40b9-a50d-9780a14fabf7.safetensors","tensor_key":"kv_indices"},"kv_last_page_len":{"type":"safetensors","path":"./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps64/gqa_paged_decode_h32_kv8_d64_ps64_3a4d1bf7-c9fe-40b9-a50d-9780a14fabf7.safetensors","tensor_key":"kv_last_page_len"},"sm_scale":{"type":"scalar","value":0.08838834764831843}},"uuid":"3a4d1bf7-c9fe-40b9-a50d-9780a14fabf7"},"solution":"flashinfer_wrapper_68ff49","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:21.135712","log":"","correctness":{"max_relative_error":3.3653345108032227,"max_absolute_error":0.00048828125,"extra":null},"performance":{"latency_ms":0.015573333483189344,"reference_latency_ms":0.2911626646916072,"speedup_factor":18.696232569982726}}} -{"definition":"gqa_paged_decode_h32_kv8_d64_ps64","workload":{"axes":{"batch_size":1,"num_pages":66,"len_indptr":2,"num_kv_indices":2},"inputs":{"q":{"type":"random"},"k_cache":{"type":"random"},"v_cache":{"type":"random"},"kv_indptr":{"type":"safetensors","path":"./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps64/gqa_paged_decode_h32_kv8_d64_ps64_82761919-e520-4a9d-b18d-ef5b07f25245.safetensors","tensor_key":"kv_indptr"},"kv_indices":{"type":"safetensors","path":"./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps64/gqa_paged_decode_h32_kv8_d64_ps64_82761919-e520-4a9d-b18d-ef5b07f25245.safetensors","tensor_key":"kv_indices"},"kv_last_page_len":{"type":"safetensors","path":"./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps64/gqa_paged_decode_h32_kv8_d64_ps64_82761919-e520-4a9d-b18d-ef5b07f25245.safetensors","tensor_key":"kv_last_page_len"},"sm_scale":{"type":"scalar","value":0.08838834764831843}},"uuid":"82761919-e520-4a9d-b18d-ef5b07f25245"},"solution":"flashinfer_wrapper_68ff49","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:21.249392","log":"","correctness":{"max_relative_error":0.007575757801532745,"max_absolute_error":0.001953125,"extra":null},"performance":{"latency_ms":0.010197333681086699,"reference_latency_ms":0.2627573360999425,"speedup_factor":25.76725880680814}}} -{"definition":"gqa_paged_decode_h32_kv8_d64_ps64","workload":{"axes":{"batch_size":1,"num_pages":79,"len_indptr":2,"num_kv_indices":15},"inputs":{"q":{"type":"random"},"k_cache":{"type":"random"},"v_cache":{"type":"random"},"kv_indptr":{"type":"safetensors","path":"./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps64/gqa_paged_decode_h32_kv8_d64_ps64_9d298dcb-2965-4e58-8c82-ea1b1fa49ccd.safetensors","tensor_key":"kv_indptr"},"kv_indices":{"type":"safetensors","path":"./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps64/gqa_paged_decode_h32_kv8_d64_ps64_9d298dcb-2965-4e58-8c82-ea1b1fa49ccd.safetensors","tensor_key":"kv_indices"},"kv_last_page_len":{"type":"safetensors","path":"./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps64/gqa_paged_decode_h32_kv8_d64_ps64_9d298dcb-2965-4e58-8c82-ea1b1fa49ccd.safetensors","tensor_key":"kv_last_page_len"},"sm_scale":{"type":"scalar","value":0.08838834764831843}},"uuid":"9d298dcb-2965-4e58-8c82-ea1b1fa49ccd"},"solution":"flashinfer_wrapper_68ff49","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:21.860684","log":"","correctness":{"max_relative_error":21.320280075073242,"max_absolute_error":0.0009765625,"extra":null},"performance":{"latency_ms":0.014463999929527441,"reference_latency_ms":0.2807413289944331,"speedup_factor":19.409660561551544}}} -{"definition":"gqa_paged_decode_h32_kv8_d64_ps64","workload":{"axes":{"batch_size":1,"num_pages":72,"len_indptr":2,"num_kv_indices":8},"inputs":{"q":{"type":"random"},"k_cache":{"type":"random"},"v_cache":{"type":"random"},"kv_indptr":{"type":"safetensors","path":"./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps64/gqa_paged_decode_h32_kv8_d64_ps64_f5ae694d-0c9c-4534-8ab2-b4cdd38d6a35.safetensors","tensor_key":"kv_indptr"},"kv_indices":{"type":"safetensors","path":"./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps64/gqa_paged_decode_h32_kv8_d64_ps64_f5ae694d-0c9c-4534-8ab2-b4cdd38d6a35.safetensors","tensor_key":"kv_indices"},"kv_last_page_len":{"type":"safetensors","path":"./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps64/gqa_paged_decode_h32_kv8_d64_ps64_f5ae694d-0c9c-4534-8ab2-b4cdd38d6a35.safetensors","tensor_key":"kv_last_page_len"},"sm_scale":{"type":"scalar","value":0.08838834764831843}},"uuid":"f5ae694d-0c9c-4534-8ab2-b4cdd38d6a35"},"solution":"flashinfer_wrapper_68ff49","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:21.963083","log":"","correctness":{"max_relative_error":24.489131927490234,"max_absolute_error":0.0009765625,"extra":null},"performance":{"latency_ms":0.013418666707972685,"reference_latency_ms":0.25382400552431744,"speedup_factor":18.91573962214206}}} -{"definition":"gqa_paged_decode_h32_kv8_d64_ps64","workload":{"axes":{"batch_size":1,"num_pages":120,"len_indptr":2,"num_kv_indices":56},"inputs":{"q":{"type":"random"},"k_cache":{"type":"random"},"v_cache":{"type":"random"},"kv_indptr":{"type":"safetensors","path":"./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps64/gqa_paged_decode_h32_kv8_d64_ps64_ea8adb18-8557-4dd2-9191-c1a045c7416d.safetensors","tensor_key":"kv_indptr"},"kv_indices":{"type":"safetensors","path":"./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps64/gqa_paged_decode_h32_kv8_d64_ps64_ea8adb18-8557-4dd2-9191-c1a045c7416d.safetensors","tensor_key":"kv_indices"},"kv_last_page_len":{"type":"safetensors","path":"./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps64/gqa_paged_decode_h32_kv8_d64_ps64_ea8adb18-8557-4dd2-9191-c1a045c7416d.safetensors","tensor_key":"kv_last_page_len"},"sm_scale":{"type":"scalar","value":0.08838834764831843}},"uuid":"ea8adb18-8557-4dd2-9191-c1a045c7416d"},"solution":"flashinfer_wrapper_68ff49","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:22.065815","log":"","correctness":{"max_relative_error":5.504561424255371,"max_absolute_error":0.00048828125,"extra":null},"performance":{"latency_ms":0.015770666611691315,"reference_latency_ms":0.29241599639256793,"speedup_factor":18.54176513856366}}} -{"definition":"gqa_paged_decode_h32_kv8_d64_ps64","workload":{"axes":{"batch_size":1,"num_pages":66,"len_indptr":2,"num_kv_indices":2},"inputs":{"q":{"type":"random"},"k_cache":{"type":"random"},"v_cache":{"type":"random"},"kv_indptr":{"type":"safetensors","path":"./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps64/gqa_paged_decode_h32_kv8_d64_ps64_68af5fd6-8da8-4070-ad1b-503ba3a56b94.safetensors","tensor_key":"kv_indptr"},"kv_indices":{"type":"safetensors","path":"./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps64/gqa_paged_decode_h32_kv8_d64_ps64_68af5fd6-8da8-4070-ad1b-503ba3a56b94.safetensors","tensor_key":"kv_indices"},"kv_last_page_len":{"type":"safetensors","path":"./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps64/gqa_paged_decode_h32_kv8_d64_ps64_68af5fd6-8da8-4070-ad1b-503ba3a56b94.safetensors","tensor_key":"kv_last_page_len"},"sm_scale":{"type":"scalar","value":0.08838834764831843}},"uuid":"68af5fd6-8da8-4070-ad1b-503ba3a56b94"},"solution":"flashinfer_wrapper_68ff49","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:22.158867","log":"","correctness":{"max_relative_error":0.004366811830550432,"max_absolute_error":0.00048828125,"extra":null},"performance":{"latency_ms":0.0102186668664217,"reference_latency_ms":0.25706133246421814,"speedup_factor":25.156053702945897}}} -{"definition":"gqa_paged_decode_h32_kv8_d64_ps64","workload":{"axes":{"batch_size":1,"num_pages":79,"len_indptr":2,"num_kv_indices":15},"inputs":{"q":{"type":"random"},"k_cache":{"type":"random"},"v_cache":{"type":"random"},"kv_indptr":{"type":"safetensors","path":"./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps64/gqa_paged_decode_h32_kv8_d64_ps64_687cdb3b-d2e0-4215-887b-fbf56b8569e7.safetensors","tensor_key":"kv_indptr"},"kv_indices":{"type":"safetensors","path":"./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps64/gqa_paged_decode_h32_kv8_d64_ps64_687cdb3b-d2e0-4215-887b-fbf56b8569e7.safetensors","tensor_key":"kv_indices"},"kv_last_page_len":{"type":"safetensors","path":"./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps64/gqa_paged_decode_h32_kv8_d64_ps64_687cdb3b-d2e0-4215-887b-fbf56b8569e7.safetensors","tensor_key":"kv_last_page_len"},"sm_scale":{"type":"scalar","value":0.08838834764831843}},"uuid":"687cdb3b-d2e0-4215-887b-fbf56b8569e7"},"solution":"flashinfer_wrapper_68ff49","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:22.251507","log":"","correctness":{"max_relative_error":42.51358413696289,"max_absolute_error":0.0009765625,"extra":null},"performance":{"latency_ms":0.01468266686424613,"reference_latency_ms":0.2542613297700882,"speedup_factor":17.317108133076413}}} diff --git a/traces/baseline/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps1.jsonl b/traces/baseline/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps1.jsonl deleted file mode 100644 index 59b9df7ace738f782c9d6156208ff3b039fc90ae..0000000000000000000000000000000000000000 --- a/traces/baseline/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps1.jsonl +++ /dev/null @@ -1,8 +0,0 @@ -{"definition": "gqa_paged_prefill_causal_h32_kv8_d64_ps1", "workload": {"axes": {"len_indptr": 2, "total_q": 4, "num_kv_indices": 4, "num_pages": 1}, "inputs": {"q": {"type": "random"}, "k_cache": {"type": "random"}, "v_cache": {"type": "random"}, "qo_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps1/gqa_paged_prefill_causal_h32_kv8_d64_ps1_25732cc4-7e33-44bc-9a26-8f5f1ebd1da5.safetensors", "tensor_key": "qo_indptr"}, "kv_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps1/gqa_paged_prefill_causal_h32_kv8_d64_ps1_25732cc4-7e33-44bc-9a26-8f5f1ebd1da5.safetensors", "tensor_key": "kv_indptr"}, "kv_indices": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps1/gqa_paged_prefill_causal_h32_kv8_d64_ps1_25732cc4-7e33-44bc-9a26-8f5f1ebd1da5.safetensors", "tensor_key": "kv_indices"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831843}}, "uuid": "25732cc4-7e33-44bc-9a26-8f5f1ebd1da5"}, "solution": "flashinfer_wrapper_ece89a", "evaluation": {"status": "PASSED", "environment": {"hardware": "NVIDIA B200", "libs": {"torch": "2.11.0+cu130", "triton": "3.6.0", "cuda": "13.0"}}, "timestamp": "2026-04-20T07:33:35.874208", "log": "", "correctness": {"max_relative_error": 1.0990966984536499e-05, "max_absolute_error": 4.76837158203125e-07, "extra": null}, "performance": {"latency_ms": 0.012341333242754141, "reference_latency_ms": 0.26654399931430817, "speedup_factor": 21.59766648152069}}} -{"definition": "gqa_paged_prefill_causal_h32_kv8_d64_ps1", "workload": {"axes": {"len_indptr": 2, "total_q": 8192, "num_kv_indices": 8192, "num_pages": 1}, "inputs": {"q": {"type": "random"}, "k_cache": {"type": "random"}, "v_cache": {"type": "random"}, "qo_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps1/gqa_paged_prefill_causal_h32_kv8_d64_ps1_f8641e85-8041-426f-9996-e189e4aa7004.safetensors", "tensor_key": "qo_indptr"}, "kv_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps1/gqa_paged_prefill_causal_h32_kv8_d64_ps1_f8641e85-8041-426f-9996-e189e4aa7004.safetensors", "tensor_key": "kv_indptr"}, "kv_indices": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps1/gqa_paged_prefill_causal_h32_kv8_d64_ps1_f8641e85-8041-426f-9996-e189e4aa7004.safetensors", "tensor_key": "kv_indices"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831843}}, "uuid": "f8641e85-8041-426f-9996-e189e4aa7004"}, "solution": "flashinfer_wrapper_ece89a", "evaluation": {"status": "PASSED", "environment": {"hardware": "NVIDIA B200", "libs": {"torch": "2.11.0+cu130", "triton": "3.6.0", "cuda": "13.0"}}, "timestamp": "2026-04-20T07:33:43.001593", "log": "", "correctness": {"max_relative_error": 0.0011211346136406064, "max_absolute_error": 1.1444091796875e-05, "extra": null}, "performance": {"latency_ms": 1.6083786884943645, "reference_latency_ms": 33.26183001200358, "speedup_factor": 20.680347389544586}}} -{"definition": "gqa_paged_prefill_causal_h32_kv8_d64_ps1", "workload": {"axes": {"len_indptr": 2, "total_q": 576, "num_kv_indices": 576, "num_pages": 1}, "inputs": {"q": {"type": "random"}, "k_cache": {"type": "random"}, "v_cache": {"type": "random"}, "qo_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps1/gqa_paged_prefill_causal_h32_kv8_d64_ps1_8f24a1a5-f13a-4902-a0ee-f457a5074862.safetensors", "tensor_key": "qo_indptr"}, "kv_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps1/gqa_paged_prefill_causal_h32_kv8_d64_ps1_8f24a1a5-f13a-4902-a0ee-f457a5074862.safetensors", "tensor_key": "kv_indptr"}, "kv_indices": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps1/gqa_paged_prefill_causal_h32_kv8_d64_ps1_8f24a1a5-f13a-4902-a0ee-f457a5074862.safetensors", "tensor_key": "kv_indices"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831843}}, "uuid": "8f24a1a5-f13a-4902-a0ee-f457a5074862"}, "solution": "flashinfer_wrapper_ece89a", "evaluation": {"status": "PASSED", "environment": {"hardware": "NVIDIA B200", "libs": {"torch": "2.11.0+cu130", "triton": "3.6.0", "cuda": "13.0"}}, "timestamp": "2026-04-20T07:33:43.157412", "log": "", "correctness": {"max_relative_error": 0.00015423737931996584, "max_absolute_error": 1.9073486328125e-06, "extra": null}, "performance": {"latency_ms": 0.09592000146706899, "reference_latency_ms": 0.4317973256111145, "speedup_factor": 4.5016401064105285}}} -{"definition": "gqa_paged_prefill_causal_h32_kv8_d64_ps1", "workload": {"axes": {"len_indptr": 2, "total_q": 2304, "num_kv_indices": 2304, "num_pages": 1}, "inputs": {"q": {"type": "random"}, "k_cache": {"type": "random"}, "v_cache": {"type": "random"}, "qo_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps1/gqa_paged_prefill_causal_h32_kv8_d64_ps1_297c3b37-1681-41f6-a052-1f449af8c580.safetensors", "tensor_key": "qo_indptr"}, "kv_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps1/gqa_paged_prefill_causal_h32_kv8_d64_ps1_297c3b37-1681-41f6-a052-1f449af8c580.safetensors", "tensor_key": "kv_indptr"}, "kv_indices": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps1/gqa_paged_prefill_causal_h32_kv8_d64_ps1_297c3b37-1681-41f6-a052-1f449af8c580.safetensors", "tensor_key": "kv_indices"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831843}}, "uuid": "297c3b37-1681-41f6-a052-1f449af8c580"}, "solution": "flashinfer_wrapper_ece89a", "evaluation": {"status": "PASSED", "environment": {"hardware": "NVIDIA B200", "libs": {"torch": "2.11.0+cu130", "triton": "3.6.0", "cuda": "13.0"}}, "timestamp": "2026-04-20T07:33:43.857094", "log": "", "correctness": {"max_relative_error": 1.7800892237573862e-05, "max_absolute_error": 3.337860107421875e-06, "extra": null}, "performance": {"latency_ms": 0.15729066481192908, "reference_latency_ms": 3.115493337313334, "speedup_factor": 19.80723611943849}}} -{"definition": "gqa_paged_prefill_causal_h32_kv8_d64_ps1", "workload": {"axes": {"len_indptr": 2, "total_q": 4, "num_kv_indices": 4, "num_pages": 1}, "inputs": {"q": {"type": "random"}, "k_cache": {"type": "random"}, "v_cache": {"type": "random"}, "qo_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps1/gqa_paged_prefill_causal_h32_kv8_d64_ps1_efb8c6d0-1c04-4804-a29c-b5339ccdbe3a.safetensors", "tensor_key": "qo_indptr"}, "kv_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps1/gqa_paged_prefill_causal_h32_kv8_d64_ps1_efb8c6d0-1c04-4804-a29c-b5339ccdbe3a.safetensors", "tensor_key": "kv_indptr"}, "kv_indices": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps1/gqa_paged_prefill_causal_h32_kv8_d64_ps1_efb8c6d0-1c04-4804-a29c-b5339ccdbe3a.safetensors", "tensor_key": "kv_indices"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831843}}, "uuid": "efb8c6d0-1c04-4804-a29c-b5339ccdbe3a"}, "solution": "flashinfer_wrapper_ece89a", "evaluation": {"status": "PASSED", "environment": {"hardware": "NVIDIA B200", "libs": {"torch": "2.11.0+cu130", "triton": "3.6.0", "cuda": "13.0"}}, "timestamp": "2026-04-20T07:33:43.952591", "log": "", "correctness": {"max_relative_error": 1.2201169738546014e-05, "max_absolute_error": 4.76837158203125e-07, "extra": null}, "performance": {"latency_ms": 0.012485333252698183, "reference_latency_ms": 0.26441067457199097, "speedup_factor": 21.177702606764594}}} -{"definition": "gqa_paged_prefill_causal_h32_kv8_d64_ps1", "workload": {"axes": {"len_indptr": 2, "total_q": 8192, "num_kv_indices": 8192, "num_pages": 1}, "inputs": {"q": {"type": "random"}, "k_cache": {"type": "random"}, "v_cache": {"type": "random"}, "qo_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps1/gqa_paged_prefill_causal_h32_kv8_d64_ps1_940a5fb1-7b82-40ad-b1ca-153bd7c170d9.safetensors", "tensor_key": "qo_indptr"}, "kv_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps1/gqa_paged_prefill_causal_h32_kv8_d64_ps1_940a5fb1-7b82-40ad-b1ca-153bd7c170d9.safetensors", "tensor_key": "kv_indptr"}, "kv_indices": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps1/gqa_paged_prefill_causal_h32_kv8_d64_ps1_940a5fb1-7b82-40ad-b1ca-153bd7c170d9.safetensors", "tensor_key": "kv_indices"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831843}}, "uuid": "940a5fb1-7b82-40ad-b1ca-153bd7c170d9"}, "solution": "flashinfer_wrapper_ece89a", "evaluation": {"status": "PASSED", "environment": {"hardware": "NVIDIA B200", "libs": {"torch": "2.11.0+cu130", "triton": "3.6.0", "cuda": "13.0"}}, "timestamp": "2026-04-20T07:33:51.032655", "log": "", "correctness": {"max_relative_error": 4.9039576879295055e-06, "max_absolute_error": 1.049041748046875e-05, "extra": null}, "performance": {"latency_ms": 1.596554656823476, "reference_latency_ms": 33.26037343343099, "speedup_factor": 20.832593041071465}}} -{"definition": "gqa_paged_prefill_causal_h32_kv8_d64_ps1", "workload": {"axes": {"len_indptr": 2, "total_q": 576, "num_kv_indices": 576, "num_pages": 1}, "inputs": {"q": {"type": "random"}, "k_cache": {"type": "random"}, "v_cache": {"type": "random"}, "qo_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps1/gqa_paged_prefill_causal_h32_kv8_d64_ps1_46907c3e-0b63-41a2-9ebf-63382ffffb60.safetensors", "tensor_key": "qo_indptr"}, "kv_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps1/gqa_paged_prefill_causal_h32_kv8_d64_ps1_46907c3e-0b63-41a2-9ebf-63382ffffb60.safetensors", "tensor_key": "kv_indptr"}, "kv_indices": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps1/gqa_paged_prefill_causal_h32_kv8_d64_ps1_46907c3e-0b63-41a2-9ebf-63382ffffb60.safetensors", "tensor_key": "kv_indices"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831843}}, "uuid": "46907c3e-0b63-41a2-9ebf-63382ffffb60"}, "solution": "flashinfer_wrapper_ece89a", "evaluation": {"status": "PASSED", "environment": {"hardware": "NVIDIA B200", "libs": {"torch": "2.11.0+cu130", "triton": "3.6.0", "cuda": "13.0"}}, "timestamp": "2026-04-20T07:33:51.179989", "log": "", "correctness": {"max_relative_error": 9.49550940276822e-06, "max_absolute_error": 1.9073486328125e-06, "extra": null}, "performance": {"latency_ms": 0.09571733325719833, "reference_latency_ms": 0.42634666462739307, "speedup_factor": 4.454226315329675}}} -{"definition": "gqa_paged_prefill_causal_h32_kv8_d64_ps1", "workload": {"axes": {"len_indptr": 2, "total_q": 2304, "num_kv_indices": 2304, "num_pages": 1}, "inputs": {"q": {"type": "random"}, "k_cache": {"type": "random"}, "v_cache": {"type": "random"}, "qo_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps1/gqa_paged_prefill_causal_h32_kv8_d64_ps1_0b2f34f7-531b-436a-8c39-3d13f486cb78.safetensors", "tensor_key": "qo_indptr"}, "kv_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps1/gqa_paged_prefill_causal_h32_kv8_d64_ps1_0b2f34f7-531b-436a-8c39-3d13f486cb78.safetensors", "tensor_key": "kv_indptr"}, "kv_indices": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps1/gqa_paged_prefill_causal_h32_kv8_d64_ps1_0b2f34f7-531b-436a-8c39-3d13f486cb78.safetensors", "tensor_key": "kv_indices"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831843}}, "uuid": "0b2f34f7-531b-436a-8c39-3d13f486cb78"}, "solution": "flashinfer_wrapper_ece89a", "evaluation": {"status": "PASSED", "environment": {"hardware": "NVIDIA B200", "libs": {"torch": "2.11.0+cu130", "triton": "3.6.0", "cuda": "13.0"}}, "timestamp": "2026-04-20T07:33:51.887178", "log": "", "correctness": {"max_relative_error": 3.327755848658853e-06, "max_absolute_error": 2.86102294921875e-06, "extra": null}, "performance": {"latency_ms": 0.158202663064003, "reference_latency_ms": 3.125445326169332, "speedup_factor": 19.755959006232985}}} diff --git a/traces/baseline/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps64.jsonl b/traces/baseline/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps64.jsonl deleted file mode 100644 index 0e42891745f90dee2690ab9ba9a99925ee334b5a..0000000000000000000000000000000000000000 --- a/traces/baseline/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps64.jsonl +++ /dev/null @@ -1,8 +0,0 @@ -{"definition": "gqa_paged_prefill_causal_h32_kv8_d64_ps64", "workload": {"axes": {"len_indptr": 2, "total_q": 4, "num_kv_indices": 4, "num_pages": 1, "batch_size": 2}, "inputs": {"q": {"type": "random"}, "k_cache": {"type": "random"}, "v_cache": {"type": "random"}, "qo_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps64/gqa_paged_prefill_causal_h32_kv8_d64_ps64_83cca546-a68d-47ee-ac1d-1883ea9a8cd3.safetensors", "tensor_key": "qo_indptr"}, "kv_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps64/gqa_paged_prefill_causal_h32_kv8_d64_ps64_83cca546-a68d-47ee-ac1d-1883ea9a8cd3.safetensors", "tensor_key": "kv_indptr"}, "kv_indices": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps64/gqa_paged_prefill_causal_h32_kv8_d64_ps64_83cca546-a68d-47ee-ac1d-1883ea9a8cd3.safetensors", "tensor_key": "kv_indices"}, "kv_last_page_len": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps64/gqa_paged_prefill_causal_h32_kv8_d64_ps64_83cca546-a68d-47ee-ac1d-1883ea9a8cd3.safetensors", "tensor_key": "kv_last_page_len"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831843}}, "uuid": "83cca546-a68d-47ee-ac1d-1883ea9a8cd3"}, "solution": "flashinfer_wrapper_06ced8", "evaluation": {"status": "PASSED", "environment": {"hardware": "NVIDIA B200", "libs": {"torch": "2.11.0+cu130", "triton": "3.6.0", "cuda": "13.0"}}, "timestamp": "2026-04-20T06:59:05.353212", "log": "", "correctness": {"max_relative_error": 100.85750579833984, "max_absolute_error": 0.00390625, "extra": null}, "performance": {"latency_ms": 0.019466665883858997, "reference_latency_ms": 0.32366933425267536, "speedup_factor": 16.626850030905878}}} -{"definition": "gqa_paged_prefill_causal_h32_kv8_d64_ps64", "workload": {"axes": {"len_indptr": 2, "total_q": 8192, "num_kv_indices": 8192, "num_pages": 1, "batch_size": 2}, "inputs": {"q": {"type": "random"}, "k_cache": {"type": "random"}, "v_cache": {"type": "random"}, "qo_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps64/gqa_paged_prefill_causal_h32_kv8_d64_ps64_ea6b783f-fdb1-4be4-97fe-180de19df270.safetensors", "tensor_key": "qo_indptr"}, "kv_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps64/gqa_paged_prefill_causal_h32_kv8_d64_ps64_ea6b783f-fdb1-4be4-97fe-180de19df270.safetensors", "tensor_key": "kv_indptr"}, "kv_indices": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps64/gqa_paged_prefill_causal_h32_kv8_d64_ps64_ea6b783f-fdb1-4be4-97fe-180de19df270.safetensors", "tensor_key": "kv_indices"}, "kv_last_page_len": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps64/gqa_paged_prefill_causal_h32_kv8_d64_ps64_ea6b783f-fdb1-4be4-97fe-180de19df270.safetensors", "tensor_key": "kv_last_page_len"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831843}}, "uuid": "ea6b783f-fdb1-4be4-97fe-180de19df270"}, "solution": "flashinfer_wrapper_06ced8", "evaluation": {"status": "PASSED", "environment": {"hardware": "NVIDIA B200", "libs": {"torch": "2.11.0+cu130", "triton": "3.6.0", "cuda": "13.0"}}, "timestamp": "2026-04-20T07:06:41.219732", "log": "", "correctness": {"max_relative_error": 27948.244140625, "max_absolute_error": 0.0078125, "extra": null}, "performance": {"latency_ms": 112.56131362915039, "reference_latency_ms": 2155.3841145833335, "speedup_factor": 19.148533764314084}}} -{"definition": "gqa_paged_prefill_causal_h32_kv8_d64_ps64", "workload": {"axes": {"len_indptr": 2, "total_q": 576, "num_kv_indices": 576, "num_pages": 1, "batch_size": 2}, "inputs": {"q": {"type": "random"}, "k_cache": {"type": "random"}, "v_cache": {"type": "random"}, "qo_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps64/gqa_paged_prefill_causal_h32_kv8_d64_ps64_b5f2086c-8df8-4ee3-9883-b2ba121c7b69.safetensors", "tensor_key": "qo_indptr"}, "kv_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps64/gqa_paged_prefill_causal_h32_kv8_d64_ps64_b5f2086c-8df8-4ee3-9883-b2ba121c7b69.safetensors", "tensor_key": "kv_indptr"}, "kv_indices": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps64/gqa_paged_prefill_causal_h32_kv8_d64_ps64_b5f2086c-8df8-4ee3-9883-b2ba121c7b69.safetensors", "tensor_key": "kv_indices"}, "kv_last_page_len": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps64/gqa_paged_prefill_causal_h32_kv8_d64_ps64_b5f2086c-8df8-4ee3-9883-b2ba121c7b69.safetensors", "tensor_key": "kv_last_page_len"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831843}}, "uuid": "b5f2086c-8df8-4ee3-9883-b2ba121c7b69"}, "solution": "flashinfer_wrapper_06ced8", "evaluation": {"status": "PASSED", "environment": {"hardware": "NVIDIA B200", "libs": {"torch": "2.11.0+cu130", "triton": "3.6.0", "cuda": "13.0"}}, "timestamp": "2026-04-20T07:06:43.658030", "log": "", "correctness": {"max_relative_error": 17755.462890625, "max_absolute_error": 0.00390625, "extra": null}, "performance": {"latency_ms": 0.6632533371448517, "reference_latency_ms": 11.246698697408041, "speedup_factor": 16.956867108761806}}} -{"definition": "gqa_paged_prefill_causal_h32_kv8_d64_ps64", "workload": {"axes": {"len_indptr": 2, "total_q": 2304, "num_kv_indices": 2304, "num_pages": 1, "batch_size": 2}, "inputs": {"q": {"type": "random"}, "k_cache": {"type": "random"}, "v_cache": {"type": "random"}, "qo_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps64/gqa_paged_prefill_causal_h32_kv8_d64_ps64_93cd7fca-8cef-4d07-aa0b-af998636b6bf.safetensors", "tensor_key": "qo_indptr"}, "kv_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps64/gqa_paged_prefill_causal_h32_kv8_d64_ps64_93cd7fca-8cef-4d07-aa0b-af998636b6bf.safetensors", "tensor_key": "kv_indptr"}, "kv_indices": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps64/gqa_paged_prefill_causal_h32_kv8_d64_ps64_93cd7fca-8cef-4d07-aa0b-af998636b6bf.safetensors", "tensor_key": "kv_indices"}, "kv_last_page_len": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps64/gqa_paged_prefill_causal_h32_kv8_d64_ps64_93cd7fca-8cef-4d07-aa0b-af998636b6bf.safetensors", "tensor_key": "kv_last_page_len"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831843}}, "uuid": "93cd7fca-8cef-4d07-aa0b-af998636b6bf"}, "solution": "flashinfer_wrapper_06ced8", "evaluation": {"status": "PASSED", "environment": {"hardware": "NVIDIA B200", "libs": {"torch": "2.11.0+cu130", "triton": "3.6.0", "cuda": "13.0"}}, "timestamp": "2026-04-20T07:07:20.346005", "log": "", "correctness": {"max_relative_error": 3974.53955078125, "max_absolute_error": 0.00390625, "extra": null}, "performance": {"latency_ms": 9.154325326283773, "reference_latency_ms": 173.10760498046875, "speedup_factor": 18.909924960110885}}} -{"definition": "gqa_paged_prefill_causal_h32_kv8_d64_ps64", "workload": {"axes": {"len_indptr": 2, "total_q": 4, "num_kv_indices": 4, "num_pages": 1, "batch_size": 2}, "inputs": {"q": {"type": "random"}, "k_cache": {"type": "random"}, "v_cache": {"type": "random"}, "qo_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps64/gqa_paged_prefill_causal_h32_kv8_d64_ps64_a10cccb9-d773-4921-b85e-b9bc75a9ed39.safetensors", "tensor_key": "qo_indptr"}, "kv_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps64/gqa_paged_prefill_causal_h32_kv8_d64_ps64_a10cccb9-d773-4921-b85e-b9bc75a9ed39.safetensors", "tensor_key": "kv_indptr"}, "kv_indices": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps64/gqa_paged_prefill_causal_h32_kv8_d64_ps64_a10cccb9-d773-4921-b85e-b9bc75a9ed39.safetensors", "tensor_key": "kv_indices"}, "kv_last_page_len": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps64/gqa_paged_prefill_causal_h32_kv8_d64_ps64_a10cccb9-d773-4921-b85e-b9bc75a9ed39.safetensors", "tensor_key": "kv_last_page_len"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831843}}, "uuid": "a10cccb9-d773-4921-b85e-b9bc75a9ed39"}, "solution": "flashinfer_wrapper_06ced8", "evaluation": {"status": "PASSED", "environment": {"hardware": "NVIDIA B200", "libs": {"torch": "2.11.0+cu130", "triton": "3.6.0", "cuda": "13.0"}}, "timestamp": "2026-04-20T07:07:20.453079", "log": "", "correctness": {"max_relative_error": 25.3980655670166, "max_absolute_error": 0.00390625, "extra": null}, "performance": {"latency_ms": 0.01945599913597107, "reference_latency_ms": 0.3065440058708191, "speedup_factor": 15.755757580399335}}} -{"definition": "gqa_paged_prefill_causal_h32_kv8_d64_ps64", "workload": {"axes": {"len_indptr": 2, "total_q": 8192, "num_kv_indices": 8192, "num_pages": 1, "batch_size": 2}, "inputs": {"q": {"type": "random"}, "k_cache": {"type": "random"}, "v_cache": {"type": "random"}, "qo_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps64/gqa_paged_prefill_causal_h32_kv8_d64_ps64_59f2a01c-bbcd-44b9-aac1-10878d8ec2b7.safetensors", "tensor_key": "qo_indptr"}, "kv_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps64/gqa_paged_prefill_causal_h32_kv8_d64_ps64_59f2a01c-bbcd-44b9-aac1-10878d8ec2b7.safetensors", "tensor_key": "kv_indptr"}, "kv_indices": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps64/gqa_paged_prefill_causal_h32_kv8_d64_ps64_59f2a01c-bbcd-44b9-aac1-10878d8ec2b7.safetensors", "tensor_key": "kv_indices"}, "kv_last_page_len": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps64/gqa_paged_prefill_causal_h32_kv8_d64_ps64_59f2a01c-bbcd-44b9-aac1-10878d8ec2b7.safetensors", "tensor_key": "kv_last_page_len"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831843}}, "uuid": "59f2a01c-bbcd-44b9-aac1-10878d8ec2b7"}, "solution": "flashinfer_wrapper_06ced8", "evaluation": {"status": "PASSED", "environment": {"hardware": "NVIDIA B200", "libs": {"torch": "2.11.0+cu130", "triton": "3.6.0", "cuda": "13.0"}}, "timestamp": "2026-04-20T07:14:56.501707", "log": "", "correctness": {"max_relative_error": 23643.0703125, "max_absolute_error": 0.0078125, "extra": null}, "performance": {"latency_ms": 112.53980763753255, "reference_latency_ms": 2156.205769856771, "speedup_factor": 19.159494005902907}}} -{"definition": "gqa_paged_prefill_causal_h32_kv8_d64_ps64", "workload": {"axes": {"len_indptr": 2, "total_q": 576, "num_kv_indices": 576, "num_pages": 1, "batch_size": 2}, "inputs": {"q": {"type": "random"}, "k_cache": {"type": "random"}, "v_cache": {"type": "random"}, "qo_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps64/gqa_paged_prefill_causal_h32_kv8_d64_ps64_58b5a85a-8410-46bc-bf0f-fbc41ad6ad84.safetensors", "tensor_key": "qo_indptr"}, "kv_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps64/gqa_paged_prefill_causal_h32_kv8_d64_ps64_58b5a85a-8410-46bc-bf0f-fbc41ad6ad84.safetensors", "tensor_key": "kv_indptr"}, "kv_indices": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps64/gqa_paged_prefill_causal_h32_kv8_d64_ps64_58b5a85a-8410-46bc-bf0f-fbc41ad6ad84.safetensors", "tensor_key": "kv_indices"}, "kv_last_page_len": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps64/gqa_paged_prefill_causal_h32_kv8_d64_ps64_58b5a85a-8410-46bc-bf0f-fbc41ad6ad84.safetensors", "tensor_key": "kv_last_page_len"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831843}}, "uuid": "58b5a85a-8410-46bc-bf0f-fbc41ad6ad84"}, "solution": "flashinfer_wrapper_06ced8", "evaluation": {"status": "PASSED", "environment": {"hardware": "NVIDIA B200", "libs": {"torch": "2.11.0+cu130", "triton": "3.6.0", "cuda": "13.0"}}, "timestamp": "2026-04-20T07:14:58.936787", "log": "", "correctness": {"max_relative_error": 832.6439208984375, "max_absolute_error": 0.00390625, "extra": null}, "performance": {"latency_ms": 0.6638666689395905, "reference_latency_ms": 11.245392004648844, "speedup_factor": 16.939232726672916}}} -{"definition": "gqa_paged_prefill_causal_h32_kv8_d64_ps64", "workload": {"axes": {"len_indptr": 2, "total_q": 2304, "num_kv_indices": 2304, "num_pages": 1, "batch_size": 2}, "inputs": {"q": {"type": "random"}, "k_cache": {"type": "random"}, "v_cache": {"type": "random"}, "qo_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps64/gqa_paged_prefill_causal_h32_kv8_d64_ps64_90a5a745-f1f4-42d6-8d63-a3ac06d27326.safetensors", "tensor_key": "qo_indptr"}, "kv_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps64/gqa_paged_prefill_causal_h32_kv8_d64_ps64_90a5a745-f1f4-42d6-8d63-a3ac06d27326.safetensors", "tensor_key": "kv_indptr"}, "kv_indices": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps64/gqa_paged_prefill_causal_h32_kv8_d64_ps64_90a5a745-f1f4-42d6-8d63-a3ac06d27326.safetensors", "tensor_key": "kv_indices"}, "kv_last_page_len": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps64/gqa_paged_prefill_causal_h32_kv8_d64_ps64_90a5a745-f1f4-42d6-8d63-a3ac06d27326.safetensors", "tensor_key": "kv_last_page_len"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831843}}, "uuid": "90a5a745-f1f4-42d6-8d63-a3ac06d27326"}, "solution": "flashinfer_wrapper_06ced8", "evaluation": {"status": "PASSED", "environment": {"hardware": "NVIDIA B200", "libs": {"torch": "2.11.0+cu130", "triton": "3.6.0", "cuda": "13.0"}}, "timestamp": "2026-04-20T07:15:35.633645", "log": "", "correctness": {"max_relative_error": 11580.685546875, "max_absolute_error": 0.0078125, "extra": null}, "performance": {"latency_ms": 9.153866608937582, "reference_latency_ms": 173.1889623006185, "speedup_factor": 18.919760326365427}}} diff --git a/traces/baseline/gqa_ragged/gqa_ragged_prefill_causal_h32_kv8_d64.jsonl b/traces/baseline/gqa_ragged/gqa_ragged_prefill_causal_h32_kv8_d64.jsonl deleted file mode 100644 index 6b8aa902ab761cdaa845d88c3e270b0d5d54fbdc..0000000000000000000000000000000000000000 --- a/traces/baseline/gqa_ragged/gqa_ragged_prefill_causal_h32_kv8_d64.jsonl +++ /dev/null @@ -1,8 +0,0 @@ -{"definition":"gqa_ragged_prefill_causal_h32_kv8_d64","workload":{"axes":{"len_indptr":2,"total_q":1,"total_kv":1},"inputs":{"q":{"type":"random"},"k":{"type":"random"},"v":{"type":"random"},"qo_indptr":{"type":"safetensors","path":"./blob/workloads/gqa_ragged/gqa_ragged_prefill_causal_h32_kv8_d64/gqa_ragged_prefill_causal_h32_kv8_d64_1bf692f3-4891-4382-b3a0-452aff438463.safetensors","tensor_key":"qo_indptr"},"kv_indptr":{"type":"safetensors","path":"./blob/workloads/gqa_ragged/gqa_ragged_prefill_causal_h32_kv8_d64/gqa_ragged_prefill_causal_h32_kv8_d64_1bf692f3-4891-4382-b3a0-452aff438463.safetensors","tensor_key":"kv_indptr"},"sm_scale":{"type":"scalar","value":0.08838834764831843}},"uuid":"1bf692f3-4891-4382-b3a0-452aff438463"},"solution":"flashinfer_wrapper_44b034","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:52.159735","log":"","correctness":{"max_relative_error":0.00001401659392286092,"max_absolute_error":2.384185791015625e-7,"extra":null},"performance":{"latency_ms":0.00899199986209472,"reference_latency_ms":0.2665066619714101,"speedup_factor":29.638196848162135}}} -{"definition":"gqa_ragged_prefill_causal_h32_kv8_d64","workload":{"axes":{"len_indptr":32,"total_q":13716,"total_kv":13716},"inputs":{"q":{"type":"random"},"k":{"type":"random"},"v":{"type":"random"},"qo_indptr":{"type":"safetensors","path":"./blob/workloads/gqa_ragged/gqa_ragged_prefill_causal_h32_kv8_d64/gqa_ragged_prefill_causal_h32_kv8_d64_db7cafbd-5dc4-4a6b-a073-38b22ea8c0b4.safetensors","tensor_key":"qo_indptr"},"kv_indptr":{"type":"safetensors","path":"./blob/workloads/gqa_ragged/gqa_ragged_prefill_causal_h32_kv8_d64/gqa_ragged_prefill_causal_h32_kv8_d64_db7cafbd-5dc4-4a6b-a073-38b22ea8c0b4.safetensors","tensor_key":"kv_indptr"},"sm_scale":{"type":"scalar","value":0.08838834764831843}},"uuid":"db7cafbd-5dc4-4a6b-a073-38b22ea8c0b4"},"solution":"flashinfer_wrapper_44b034","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:56.256230","log":"","correctness":{"max_relative_error":10175.771484375,"max_absolute_error":0.015625,"extra":null},"performance":{"latency_ms":0.3059253344933192,"reference_latency_ms":19.796122868855793,"speedup_factor":64.70900130465691}}} -{"definition":"gqa_ragged_prefill_causal_h32_kv8_d64","workload":{"axes":{"len_indptr":2,"total_q":35,"total_kv":35},"inputs":{"q":{"type":"random"},"k":{"type":"random"},"v":{"type":"random"},"qo_indptr":{"type":"safetensors","path":"./blob/workloads/gqa_ragged/gqa_ragged_prefill_causal_h32_kv8_d64/gqa_ragged_prefill_causal_h32_kv8_d64_eaf31898-d557-474a-8967-08912443e85e.safetensors","tensor_key":"qo_indptr"},"kv_indptr":{"type":"safetensors","path":"./blob/workloads/gqa_ragged/gqa_ragged_prefill_causal_h32_kv8_d64/gqa_ragged_prefill_causal_h32_kv8_d64_eaf31898-d557-474a-8967-08912443e85e.safetensors","tensor_key":"kv_indptr"},"sm_scale":{"type":"scalar","value":0.08838834764831843}},"uuid":"eaf31898-d557-474a-8967-08912443e85e"},"solution":"flashinfer_wrapper_44b034","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:56.356343","log":"","correctness":{"max_relative_error":46.273372650146484,"max_absolute_error":0.015625,"extra":null},"performance":{"latency_ms":0.010096000197033087,"reference_latency_ms":0.292821337779363,"speedup_factor":29.003697708466213}}} -{"definition":"gqa_ragged_prefill_causal_h32_kv8_d64","workload":{"axes":{"len_indptr":13,"total_q":5932,"total_kv":5932},"inputs":{"q":{"type":"random"},"k":{"type":"random"},"v":{"type":"random"},"qo_indptr":{"type":"safetensors","path":"./blob/workloads/gqa_ragged/gqa_ragged_prefill_causal_h32_kv8_d64/gqa_ragged_prefill_causal_h32_kv8_d64_d22f94aa-fc21-4bdd-ac49-f36bae0caa5c.safetensors","tensor_key":"qo_indptr"},"kv_indptr":{"type":"safetensors","path":"./blob/workloads/gqa_ragged/gqa_ragged_prefill_causal_h32_kv8_d64/gqa_ragged_prefill_causal_h32_kv8_d64_d22f94aa-fc21-4bdd-ac49-f36bae0caa5c.safetensors","tensor_key":"kv_indptr"},"sm_scale":{"type":"scalar","value":0.08838834764831843}},"uuid":"d22f94aa-fc21-4bdd-ac49-f36bae0caa5c"},"solution":"flashinfer_wrapper_44b034","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:58.043803","log":"","correctness":{"max_relative_error":6859.591796875,"max_absolute_error":0.015625,"extra":null},"performance":{"latency_ms":0.13855466494957605,"reference_latency_ms":7.034613291422526,"speedup_factor":50.77139260509648}}} -{"definition":"gqa_ragged_prefill_causal_h32_kv8_d64","workload":{"axes":{"len_indptr":2,"total_q":1,"total_kv":1},"inputs":{"q":{"type":"random"},"k":{"type":"random"},"v":{"type":"random"},"qo_indptr":{"type":"safetensors","path":"./blob/workloads/gqa_ragged/gqa_ragged_prefill_causal_h32_kv8_d64/gqa_ragged_prefill_causal_h32_kv8_d64_17e47b78-d996-4fc1-af3e-56bd9c789a82.safetensors","tensor_key":"qo_indptr"},"kv_indptr":{"type":"safetensors","path":"./blob/workloads/gqa_ragged/gqa_ragged_prefill_causal_h32_kv8_d64/gqa_ragged_prefill_causal_h32_kv8_d64_17e47b78-d996-4fc1-af3e-56bd9c789a82.safetensors","tensor_key":"kv_indptr"},"sm_scale":{"type":"scalar","value":0.08838834764831843}},"uuid":"17e47b78-d996-4fc1-af3e-56bd9c789a82"},"solution":"flashinfer_wrapper_44b034","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:33:58.146735","log":"","correctness":{"max_relative_error":0.00001982858157134615,"max_absolute_error":3.5762786865234375e-7,"extra":null},"performance":{"latency_ms":0.008933333369592825,"reference_latency_ms":0.2792053272326787,"speedup_factor":31.25432754844172}}} -{"definition":"gqa_ragged_prefill_causal_h32_kv8_d64","workload":{"axes":{"len_indptr":55,"total_q":10930,"total_kv":10930},"inputs":{"q":{"type":"random"},"k":{"type":"random"},"v":{"type":"random"},"qo_indptr":{"type":"safetensors","path":"./blob/workloads/gqa_ragged/gqa_ragged_prefill_causal_h32_kv8_d64/gqa_ragged_prefill_causal_h32_kv8_d64_7adaac59-352e-4574-a959-d3470a966090.safetensors","tensor_key":"qo_indptr"},"kv_indptr":{"type":"safetensors","path":"./blob/workloads/gqa_ragged/gqa_ragged_prefill_causal_h32_kv8_d64/gqa_ragged_prefill_causal_h32_kv8_d64_7adaac59-352e-4574-a959-d3470a966090.safetensors","tensor_key":"kv_indptr"},"sm_scale":{"type":"scalar","value":0.08838834764831843}},"uuid":"7adaac59-352e-4574-a959-d3470a966090"},"solution":"flashinfer_wrapper_44b034","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:34:04.295946","log":"","correctness":{"max_relative_error":32424.927734375,"max_absolute_error":0.015625,"extra":null},"performance":{"latency_ms":0.4598186711470286,"reference_latency_ms":29.843242327372234,"speedup_factor":64.90219775749331}}} -{"definition":"gqa_ragged_prefill_causal_h32_kv8_d64","workload":{"axes":{"len_indptr":2,"total_q":35,"total_kv":35},"inputs":{"q":{"type":"random"},"k":{"type":"random"},"v":{"type":"random"},"qo_indptr":{"type":"safetensors","path":"./blob/workloads/gqa_ragged/gqa_ragged_prefill_causal_h32_kv8_d64/gqa_ragged_prefill_causal_h32_kv8_d64_319d7c83-25e5-47c3-8de0-bf65e9e80a35.safetensors","tensor_key":"qo_indptr"},"kv_indptr":{"type":"safetensors","path":"./blob/workloads/gqa_ragged/gqa_ragged_prefill_causal_h32_kv8_d64/gqa_ragged_prefill_causal_h32_kv8_d64_319d7c83-25e5-47c3-8de0-bf65e9e80a35.safetensors","tensor_key":"kv_indptr"},"sm_scale":{"type":"scalar","value":0.08838834764831843}},"uuid":"319d7c83-25e5-47c3-8de0-bf65e9e80a35"},"solution":"flashinfer_wrapper_44b034","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:34:04.395984","log":"","correctness":{"max_relative_error":39.81319046020508,"max_absolute_error":0.015625,"extra":null},"performance":{"latency_ms":0.010090666512648264,"reference_latency_ms":0.295935998360316,"speedup_factor":29.327695845400452}}} -{"definition":"gqa_ragged_prefill_causal_h32_kv8_d64","workload":{"axes":{"len_indptr":10,"total_q":8819,"total_kv":8819},"inputs":{"q":{"type":"random"},"k":{"type":"random"},"v":{"type":"random"},"qo_indptr":{"type":"safetensors","path":"./blob/workloads/gqa_ragged/gqa_ragged_prefill_causal_h32_kv8_d64/gqa_ragged_prefill_causal_h32_kv8_d64_f75c1eec-0a27-4e06-bfb6-d6238049271d.safetensors","tensor_key":"qo_indptr"},"kv_indptr":{"type":"safetensors","path":"./blob/workloads/gqa_ragged/gqa_ragged_prefill_causal_h32_kv8_d64/gqa_ragged_prefill_causal_h32_kv8_d64_f75c1eec-0a27-4e06-bfb6-d6238049271d.safetensors","tensor_key":"kv_indptr"},"sm_scale":{"type":"scalar","value":0.08838834764831843}},"uuid":"f75c1eec-0a27-4e06-bfb6-d6238049271d"},"solution":"flashinfer_wrapper_44b034","evaluation":{"status":"PASSED","environment":{"hardware":"NVIDIA B200","libs":{"torch":"2.11.0+cu130","triton":"3.6.0","cuda":"13.0"}},"timestamp":"2026-04-20T07:34:09.579545","log":"","correctness":{"max_relative_error":7014.21044921875,"max_absolute_error":0.015625,"extra":null},"performance":{"latency_ms":0.4923146665096283,"reference_latency_ms":25.040037790934246,"speedup_factor":50.86185623609597}}} diff --git a/workloads/gemm/gemm_n16384_k2048.jsonl b/workloads/gemm/gemm_n16384_k2048.jsonl deleted file mode 100644 index 08ed7139635d3d819a29bc53a932ded88343ff92..0000000000000000000000000000000000000000 --- a/workloads/gemm/gemm_n16384_k2048.jsonl +++ /dev/null @@ -1,43 +0,0 @@ -{"definition": "gemm_n16384_k2048", "solution": null, "workload": {"uuid": "63ed933a-ed07-4d64-bb08-d22a18f781bc", "axes": {"M": 1}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n16384_k2048", "solution": null, "workload": {"uuid": "9bcca24f-2a17-41c5-acbe-2aa18a0d69a3", "axes": {"M": 2}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n16384_k2048", "solution": null, "workload": {"uuid": "56fe737e-6e40-493e-b0ca-b34a6fb73425", "axes": {"M": 4}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n16384_k2048", "solution": null, "workload": {"uuid": "787ef2c4-9368-4b71-96e2-d31ac9ee0ec9", "axes": {"M": 7}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n16384_k2048", "solution": null, "workload": {"uuid": "46ce6192-af6c-4c02-b232-b4832028f974", "axes": {"M": 8}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n16384_k2048", "solution": null, "workload": {"uuid": "4614c89c-dd77-4fc7-88a6-70c2b83c39fb", "axes": {"M": 15}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n16384_k2048", "solution": null, "workload": {"uuid": "af46e3cb-fec0-42d7-979d-aa487341c1ee", "axes": {"M": 16}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n16384_k2048", "solution": null, "workload": {"uuid": "ab02c26e-50e2-42fa-b666-b3a4c4bfbf6f", "axes": {"M": 24}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n16384_k2048", "solution": null, "workload": {"uuid": "079da838-4ea5-42ff-8d57-808ae35c54d4", "axes": {"M": 32}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n16384_k2048", "solution": null, "workload": {"uuid": "d2b99ac4-8563-437f-898f-a1f7a6f54456", "axes": {"M": 35}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n16384_k2048", "solution": null, "workload": {"uuid": "170d7c38-7775-4034-bee3-1c823b1c75cc", "axes": {"M": 40}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n16384_k2048", "solution": null, "workload": {"uuid": "16349ad2-0ee1-4a5b-8d6d-5f0b7f92bc5b", "axes": {"M": 48}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n16384_k2048", "solution": null, "workload": {"uuid": "47c493ac-b794-4825-8f10-f82a8c53f89e", "axes": {"M": 56}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n16384_k2048", "solution": null, "workload": {"uuid": "a5f2bcd9-1c45-418a-9482-68ec5be2cf7f", "axes": {"M": 64}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n16384_k2048", "solution": null, "workload": {"uuid": "0638f5da-9fd9-4557-acd1-0dcd00deda64", "axes": {"M": 70}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n16384_k2048", "solution": null, "workload": {"uuid": "be8d6610-82ac-4470-ad6a-8ae9f9c0efc1", "axes": {"M": 72}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n16384_k2048", "solution": null, "workload": {"uuid": "b2131315-fbe6-4c76-89f4-10a8ced7cbd3", "axes": {"M": 80}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n16384_k2048", "solution": null, "workload": {"uuid": "a5a4c94b-587c-4d20-a45e-51eab114c101", "axes": {"M": 88}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n16384_k2048", "solution": null, "workload": {"uuid": "d075a83c-1ef6-456b-809a-02c8c789cb38", "axes": {"M": 96}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n16384_k2048", "solution": null, "workload": {"uuid": "b4fd51aa-621f-48d0-9a73-a9b7f0f63cca", "axes": {"M": 104}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n16384_k2048", "solution": null, "workload": {"uuid": "6ca48022-63e7-41d6-aa97-c9ca71d0d4ce", "axes": {"M": 112}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n16384_k2048", "solution": null, "workload": {"uuid": "ca3f0979-aea6-4964-a6c3-d140d8dc2185", "axes": {"M": 120}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n16384_k2048", "solution": null, "workload": {"uuid": "2ec930dd-2fed-4a4d-ac20-26c5e49c9b99", "axes": {"M": 128}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n16384_k2048", "solution": null, "workload": {"uuid": "c6e4803f-b5cb-4e9f-8f8b-a32abb32b085", "axes": {"M": 136}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n16384_k2048", "solution": null, "workload": {"uuid": "15185b21-70c1-4400-b24a-dc5eb3528df9", "axes": {"M": 144}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n16384_k2048", "solution": null, "workload": {"uuid": "158af161-680f-41ad-a351-4b142799fffd", "axes": {"M": 152}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n16384_k2048", "solution": null, "workload": {"uuid": "01b190bf-ca0f-4f73-bd8f-a1482dc880c0", "axes": {"M": 160}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n16384_k2048", "solution": null, "workload": {"uuid": "1f837f5a-536a-4292-a95c-0b5855c92aac", "axes": {"M": 168}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n16384_k2048", "solution": null, "workload": {"uuid": "644334e2-c032-4753-b822-84061fca1434", "axes": {"M": 176}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n16384_k2048", "solution": null, "workload": {"uuid": "6cf68f96-3a12-4340-b4f0-abbcab986a96", "axes": {"M": 184}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n16384_k2048", "solution": null, "workload": {"uuid": "9478353c-290b-4e44-9e3f-721d69185dd6", "axes": {"M": 192}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n16384_k2048", "solution": null, "workload": {"uuid": "8df72ae2-ba5d-457e-a834-faeec76470c2", "axes": {"M": 200}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n16384_k2048", "solution": null, "workload": {"uuid": "7bc262bc-b2c8-462d-881d-655a358fd8bd", "axes": {"M": 208}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n16384_k2048", "solution": null, "workload": {"uuid": "6c3dfea2-4783-4eea-a9a5-4feee25a3c26", "axes": {"M": 216}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n16384_k2048", "solution": null, "workload": {"uuid": "2606ccca-7b5f-4e63-8bad-9470cfa78f61", "axes": {"M": 224}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n16384_k2048", "solution": null, "workload": {"uuid": "16046c0c-623a-4f31-b6ac-bba47e516380", "axes": {"M": 232}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n16384_k2048", "solution": null, "workload": {"uuid": "417e59f3-7664-4eed-9ae2-1cc129015d30", "axes": {"M": 240}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n16384_k2048", "solution": null, "workload": {"uuid": "72f3eccc-3708-4760-8b1f-a1149caa5824", "axes": {"M": 248}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n16384_k2048", "solution": null, "workload": {"uuid": "7537c621-a3a8-453c-b010-ec5b5f94af08", "axes": {"M": 256}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n16384_k2048", "solution": null, "workload": {"uuid": "46cdb071-9248-43ff-9d73-bfb826f99342", "axes": {"M": 972}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n16384_k2048", "solution": null, "workload": {"uuid": "1198c852-140b-4c3d-8893-afa944f29929", "axes": {"M": 2053}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n16384_k2048", "solution": null, "workload": {"uuid": "b6eaeb01-6c46-4598-9cbc-8fdb92f3f4f0", "axes": {"M": 2379}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n16384_k2048", "solution": null, "workload": {"uuid": "11cb2b3b-7bc9-4f48-9a55-4827179a25d9", "axes": {"M": 8192}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} diff --git a/workloads/gemm/gemm_n2048_k2048.jsonl b/workloads/gemm/gemm_n2048_k2048.jsonl deleted file mode 100644 index f38569755d4ca310636c6a45c8f3f2bbf07f542b..0000000000000000000000000000000000000000 --- a/workloads/gemm/gemm_n2048_k2048.jsonl +++ /dev/null @@ -1,43 +0,0 @@ -{"definition": "gemm_n2048_k2048", "solution": null, "workload": {"uuid": "a98d8497-44c8-4f1e-a238-2a9b25f0d5c7", "axes": {"M": 1}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n2048_k2048", "solution": null, "workload": {"uuid": "1b4fe44c-9cc9-4622-8904-828b8bc4067d", "axes": {"M": 2}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n2048_k2048", "solution": null, "workload": {"uuid": "b921b0d9-9c4f-4b11-b0f8-3fde74156fa4", "axes": {"M": 4}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n2048_k2048", "solution": null, "workload": {"uuid": "bfea2711-b0c0-49c9-98b4-12a92d3d8b1e", "axes": {"M": 7}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n2048_k2048", "solution": null, "workload": {"uuid": "d97806ad-497a-477d-b4e0-d18361810fad", "axes": {"M": 8}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n2048_k2048", "solution": null, "workload": {"uuid": "614bb62c-9e3b-4340-8de5-e25b809a37b9", "axes": {"M": 15}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n2048_k2048", "solution": null, "workload": {"uuid": "bab3d06b-c5a7-4dce-83d4-91e2b7d4056f", "axes": {"M": 16}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n2048_k2048", "solution": null, "workload": {"uuid": "d68fb9b3-55bd-4b54-995a-4d952c4eb157", "axes": {"M": 24}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n2048_k2048", "solution": null, "workload": {"uuid": "95640f6d-b61a-4621-aa0b-697cffa434b8", "axes": {"M": 32}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n2048_k2048", "solution": null, "workload": {"uuid": "43afb86e-39cb-4b8e-a805-2e1e17aedb9d", "axes": {"M": 35}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n2048_k2048", "solution": null, "workload": {"uuid": "bfec76cf-b9f8-4265-8f4e-bdbb8e9541d3", "axes": {"M": 40}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n2048_k2048", "solution": null, "workload": {"uuid": "8005b86e-50e8-48ea-bc09-a68458331326", "axes": {"M": 48}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n2048_k2048", "solution": null, "workload": {"uuid": "bf82ffcb-f158-480a-9d5c-89d3cdefd810", "axes": {"M": 56}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n2048_k2048", "solution": null, "workload": {"uuid": "44c3c77f-f938-43a7-b5e2-975851500b47", "axes": {"M": 64}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n2048_k2048", "solution": null, "workload": {"uuid": "5e3dc6d5-bdc5-4b65-b9ac-32b13b0e6804", "axes": {"M": 70}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n2048_k2048", "solution": null, "workload": {"uuid": "f415e56b-b5cf-4ffe-9bd9-e3c9805e5c14", "axes": {"M": 72}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n2048_k2048", "solution": null, "workload": {"uuid": "abc630bf-d88d-4988-9684-162130a2e9d0", "axes": {"M": 80}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n2048_k2048", "solution": null, "workload": {"uuid": "013d1993-1545-4950-a3d2-44022105e87f", "axes": {"M": 88}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n2048_k2048", "solution": null, "workload": {"uuid": "5dd9fe00-896a-4215-b59a-602e4456d759", "axes": {"M": 96}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n2048_k2048", "solution": null, "workload": {"uuid": "15473932-acfb-4329-8c34-deacbe7e8953", "axes": {"M": 104}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n2048_k2048", "solution": null, "workload": {"uuid": "f9087d88-02f1-433a-87af-718e446b761f", "axes": {"M": 112}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n2048_k2048", "solution": null, "workload": {"uuid": "b6cc480a-4fe5-4435-82a8-0d42c92e1a2c", "axes": {"M": 120}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n2048_k2048", "solution": null, "workload": {"uuid": "cc7a63bd-0f43-4ad6-b775-0ecb51eab349", "axes": {"M": 128}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n2048_k2048", "solution": null, "workload": {"uuid": "b6f1dd0b-fd1f-4115-8041-2184d5332cd6", "axes": {"M": 136}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n2048_k2048", "solution": null, "workload": {"uuid": "28c94467-7fae-44fa-926d-bdda7ae79a5d", "axes": {"M": 144}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n2048_k2048", "solution": null, "workload": {"uuid": "8ffbbaba-bc09-4372-bb18-bbe100a02607", "axes": {"M": 152}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n2048_k2048", "solution": null, "workload": {"uuid": "a4348a72-4519-4f2e-a8a8-67a56c11b569", "axes": {"M": 160}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n2048_k2048", "solution": null, "workload": {"uuid": "c0206d05-c49f-45b1-af48-756d6fa73a59", "axes": {"M": 168}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n2048_k2048", "solution": null, "workload": {"uuid": "2a7605d5-09b3-4cb0-b8e7-f9159c67b148", "axes": {"M": 176}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n2048_k2048", "solution": null, "workload": {"uuid": "51fb2600-dea5-4405-896e-8ca595b5e2ba", "axes": {"M": 184}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n2048_k2048", "solution": null, "workload": {"uuid": "2003b8f5-86c2-42c9-95a1-2ee24a23fac0", "axes": {"M": 192}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n2048_k2048", "solution": null, "workload": {"uuid": "57164c57-2976-4a2f-9610-56eb68e238e1", "axes": {"M": 200}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n2048_k2048", "solution": null, "workload": {"uuid": "f5727a6a-f497-4b33-904e-234bbee68117", "axes": {"M": 208}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n2048_k2048", "solution": null, "workload": {"uuid": "6c675b98-21f7-475e-98c4-869362f1abc7", "axes": {"M": 216}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n2048_k2048", "solution": null, "workload": {"uuid": "35f373ea-2d0c-4298-9759-fa64f0141881", "axes": {"M": 224}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n2048_k2048", "solution": null, "workload": {"uuid": "f95d25ff-c150-454b-a52e-4e300ee5ff61", "axes": {"M": 232}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n2048_k2048", "solution": null, "workload": {"uuid": "62124acc-2375-4565-b08e-3e3d5e09a42a", "axes": {"M": 240}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n2048_k2048", "solution": null, "workload": {"uuid": "fac43762-cfe9-48ae-9aa6-9931b1d13cc9", "axes": {"M": 248}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n2048_k2048", "solution": null, "workload": {"uuid": "72e3421f-b9df-4581-ab6f-9bb09a19dc37", "axes": {"M": 256}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n2048_k2048", "solution": null, "workload": {"uuid": "d00f0dc1-1e57-4f09-88c5-6ea4be9447fc", "axes": {"M": 972}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n2048_k2048", "solution": null, "workload": {"uuid": "3c061ce6-901e-42ab-9a5a-13f5d9259266", "axes": {"M": 2053}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n2048_k2048", "solution": null, "workload": {"uuid": "78ead1f3-8c0a-45dd-b832-42861e7bf658", "axes": {"M": 2379}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n2048_k2048", "solution": null, "workload": {"uuid": "23bb0f3f-abe2-40ac-80af-dff54fed8646", "axes": {"M": 8192}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} diff --git a/workloads/gemm/gemm_n2048_k8192.jsonl b/workloads/gemm/gemm_n2048_k8192.jsonl deleted file mode 100644 index 6a51263fa21be05d1b166c49358c1920fa1ba52b..0000000000000000000000000000000000000000 --- a/workloads/gemm/gemm_n2048_k8192.jsonl +++ /dev/null @@ -1,43 +0,0 @@ -{"definition": "gemm_n2048_k8192", "solution": null, "workload": {"uuid": "19d02aae-74b7-4b82-aa34-f7d4c0bceebd", "axes": {"M": 1}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n2048_k8192", "solution": null, "workload": {"uuid": "927a928e-e89d-45bb-9b3a-8b07063cf6b0", "axes": {"M": 2}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n2048_k8192", "solution": null, "workload": {"uuid": "2a6b32c1-3c52-47b7-8d44-5a1e4260b2e8", "axes": {"M": 4}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n2048_k8192", "solution": null, "workload": {"uuid": "66014bc9-8e6d-4e68-9657-d46ec4e913d5", "axes": {"M": 7}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n2048_k8192", "solution": null, "workload": {"uuid": "91615ba2-0906-4a91-b970-23adc00b277d", "axes": {"M": 8}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n2048_k8192", "solution": null, "workload": {"uuid": "3ab6c6a6-22cc-4f1b-9840-9e875b591b77", "axes": {"M": 15}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n2048_k8192", "solution": null, "workload": {"uuid": "22113298-fdca-480d-a948-841ad8369fe7", "axes": {"M": 16}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n2048_k8192", "solution": null, "workload": {"uuid": "3d1ee002-d32e-4da2-93e4-102a9410aef7", "axes": {"M": 24}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n2048_k8192", "solution": null, "workload": {"uuid": "5d1101a2-2dc1-437d-bdb9-b63bcb52b26f", "axes": {"M": 32}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n2048_k8192", "solution": null, "workload": {"uuid": "59f6b86a-fa4b-4f2a-a600-2982b68e0bdb", "axes": {"M": 35}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n2048_k8192", "solution": null, "workload": {"uuid": "e4c9c6ee-f59e-4187-8ca8-fdfc5f870e35", "axes": {"M": 40}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n2048_k8192", "solution": null, "workload": {"uuid": "4571a224-a05b-4a03-8f39-378316788096", "axes": {"M": 48}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n2048_k8192", "solution": null, "workload": {"uuid": "948d16e9-2aac-4233-a67a-c3a0c2d10ad7", "axes": {"M": 56}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n2048_k8192", "solution": null, "workload": {"uuid": "d2f2fdd7-c19b-4131-a342-4a1b9e95747e", "axes": {"M": 64}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n2048_k8192", "solution": null, "workload": {"uuid": "5221701f-4873-4fe3-a73a-a01001c56f9c", "axes": {"M": 70}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n2048_k8192", "solution": null, "workload": {"uuid": "f7b9af39-1fbd-43c9-8da6-29797fb3c671", "axes": {"M": 72}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n2048_k8192", "solution": null, "workload": {"uuid": "96b8a5ba-0ef5-41d3-9962-b7ce024bb338", "axes": {"M": 80}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n2048_k8192", "solution": null, "workload": {"uuid": "1227c633-8ed4-4a4c-bc0a-c0bf6a1a29bc", "axes": {"M": 88}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n2048_k8192", "solution": null, "workload": {"uuid": "63901d6e-2f7a-4fa9-86b7-b4dce10e7bcf", "axes": {"M": 96}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n2048_k8192", "solution": null, "workload": {"uuid": "790a0497-0605-4b00-ac02-fb27a1516f7a", "axes": {"M": 104}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n2048_k8192", "solution": null, "workload": {"uuid": "013339f2-443f-4fe3-b511-b77427f4705a", "axes": {"M": 112}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n2048_k8192", "solution": null, "workload": {"uuid": "f6f1b907-5f13-4d58-a00b-ad0d97b8d6cc", "axes": {"M": 120}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n2048_k8192", "solution": null, "workload": {"uuid": "d21e0f96-99e4-426c-ba72-8ef19ae3123f", "axes": {"M": 128}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n2048_k8192", "solution": null, "workload": {"uuid": "39d5ff01-b6bf-477c-a514-e96c4a244d7a", "axes": {"M": 136}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n2048_k8192", "solution": null, "workload": {"uuid": "5ec080a0-0622-421a-a0e6-8d4dce250c58", "axes": {"M": 144}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n2048_k8192", "solution": null, "workload": {"uuid": "90a6b71b-de24-4d97-a307-13f82d5f75b4", "axes": {"M": 152}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n2048_k8192", "solution": null, "workload": {"uuid": "cfcf6b3e-bb34-4cf7-9c1a-cb42efa90a59", "axes": {"M": 160}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n2048_k8192", "solution": null, "workload": {"uuid": "dba698ce-382a-40c1-ba21-7a0313a1ee1a", "axes": {"M": 168}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n2048_k8192", "solution": null, "workload": {"uuid": "10404bdb-ca42-4284-8ada-8c034d3c8fcd", "axes": {"M": 176}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n2048_k8192", "solution": null, "workload": {"uuid": "5d307ee8-c3d7-4313-814d-c54041b52693", "axes": {"M": 184}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n2048_k8192", "solution": null, "workload": {"uuid": "b54d5b5e-f7a9-4990-a304-921b8bac4ec9", "axes": {"M": 192}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n2048_k8192", "solution": null, "workload": {"uuid": "1be9cecb-d10f-4384-abd7-933f3cd304dc", "axes": {"M": 200}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n2048_k8192", "solution": null, "workload": {"uuid": "f0f9da24-b32d-474e-a17f-a0d9f7b9e526", "axes": {"M": 208}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n2048_k8192", "solution": null, "workload": {"uuid": "af55a0d2-7155-4e3b-897b-ff2e42d7f040", "axes": {"M": 216}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n2048_k8192", "solution": null, "workload": {"uuid": "c18392ce-a799-4299-9c54-99e6fa1af0fc", "axes": {"M": 224}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n2048_k8192", "solution": null, "workload": {"uuid": "478c93fa-d2bc-4927-8d9a-e6da75976747", "axes": {"M": 232}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n2048_k8192", "solution": null, "workload": {"uuid": "88000cdb-65f3-4fc8-a645-fe50d1f73b8f", "axes": {"M": 240}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n2048_k8192", "solution": null, "workload": {"uuid": "91b3bfa3-ed8b-4295-af25-4ec4d00d85ae", "axes": {"M": 248}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n2048_k8192", "solution": null, "workload": {"uuid": "1822180b-656c-4695-afb4-f0e002526d46", "axes": {"M": 256}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n2048_k8192", "solution": null, "workload": {"uuid": "3ff729a7-9620-4914-872b-bcd59b73b850", "axes": {"M": 972}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n2048_k8192", "solution": null, "workload": {"uuid": "d49a821d-f75d-47cb-9572-9b3d37569e01", "axes": {"M": 2053}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n2048_k8192", "solution": null, "workload": {"uuid": "da657596-8823-4be8-aee7-81aa5328bcce", "axes": {"M": 2379}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n2048_k8192", "solution": null, "workload": {"uuid": "bdbd4c26-24c8-4dee-9457-72bfaad13d8a", "axes": {"M": 8192}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} diff --git a/workloads/gemm/gemm_n3072_k2048.jsonl b/workloads/gemm/gemm_n3072_k2048.jsonl deleted file mode 100644 index a2fc28ab8b183e79a35a88274767bab120ba267f..0000000000000000000000000000000000000000 --- a/workloads/gemm/gemm_n3072_k2048.jsonl +++ /dev/null @@ -1,43 +0,0 @@ -{"definition": "gemm_n3072_k2048", "solution": null, "workload": {"uuid": "87baad53-95d2-420c-adcd-81d651da4141", "axes": {"M": 1}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n3072_k2048", "solution": null, "workload": {"uuid": "94896a55-6e36-43ee-a833-f493eed6c8ef", "axes": {"M": 2}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n3072_k2048", "solution": null, "workload": {"uuid": "f84f00bb-a362-41ca-8fcf-c62c3ca07a36", "axes": {"M": 4}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n3072_k2048", "solution": null, "workload": {"uuid": "5dc703eb-0f70-47f5-b89f-69b504b38a0a", "axes": {"M": 7}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n3072_k2048", "solution": null, "workload": {"uuid": "26a1b2cf-5645-403d-b359-2025bec2cc5d", "axes": {"M": 8}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n3072_k2048", "solution": null, "workload": {"uuid": "27384021-9f80-4763-8ff8-ed4ee32fa5cb", "axes": {"M": 15}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n3072_k2048", "solution": null, "workload": {"uuid": "344fefad-0983-4f85-a744-fd194c60e36c", "axes": {"M": 16}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n3072_k2048", "solution": null, "workload": {"uuid": "a48073de-4b7b-4bcd-9d6a-ce5c89c2ae74", "axes": {"M": 24}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n3072_k2048", "solution": null, "workload": {"uuid": "c5391284-499b-48c5-b6e3-263246d33367", "axes": {"M": 32}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n3072_k2048", "solution": null, "workload": {"uuid": "e6d9bc7c-297c-49d0-8b6f-fefad9788867", "axes": {"M": 35}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n3072_k2048", "solution": null, "workload": {"uuid": "442f657e-0ea5-4995-ba54-1943b0edc22f", "axes": {"M": 40}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n3072_k2048", "solution": null, "workload": {"uuid": "20859e9e-e357-4210-8b3d-52e30bf45ec6", "axes": {"M": 48}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n3072_k2048", "solution": null, "workload": {"uuid": "417ccb5d-1132-4309-87e9-d7feaee11be5", "axes": {"M": 56}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n3072_k2048", "solution": null, "workload": {"uuid": "c4c3576e-99cf-41a7-a0d9-7a78e9f53108", "axes": {"M": 64}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n3072_k2048", "solution": null, "workload": {"uuid": "6a696da8-b670-44df-9f4b-9d3892c89b10", "axes": {"M": 70}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n3072_k2048", "solution": null, "workload": {"uuid": "bbc5595a-d522-4bbc-850c-21cd87cb22e5", "axes": {"M": 72}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n3072_k2048", "solution": null, "workload": {"uuid": "50a7a6ec-04cc-4eca-9c55-a0ca1fc360eb", "axes": {"M": 80}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n3072_k2048", "solution": null, "workload": {"uuid": "aa056b8d-75c3-497d-a080-aee5c936483b", "axes": {"M": 88}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n3072_k2048", "solution": null, "workload": {"uuid": "b24b592f-881f-4fb8-8720-97782af98f5d", "axes": {"M": 96}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n3072_k2048", "solution": null, "workload": {"uuid": "cd026c2a-883e-4c5c-97ca-bbdd293637d1", "axes": {"M": 104}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n3072_k2048", "solution": null, "workload": {"uuid": "b1dea987-a7f1-4f1e-9c2a-22d185d12c43", "axes": {"M": 112}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n3072_k2048", "solution": null, "workload": {"uuid": "771606fd-e670-4f52-b479-58a879b3c378", "axes": {"M": 120}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n3072_k2048", "solution": null, "workload": {"uuid": "d5c39f45-6a6f-49a7-9cc5-5886d3e7a104", "axes": {"M": 128}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n3072_k2048", "solution": null, "workload": {"uuid": "54172e8f-382f-4360-aacf-e580a623b0d6", "axes": {"M": 136}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n3072_k2048", "solution": null, "workload": {"uuid": "f4301d94-f87e-4ee7-bf77-cd6746a9055e", "axes": {"M": 144}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n3072_k2048", "solution": null, "workload": {"uuid": "b86fbc83-1a20-4b27-b761-e6ff69569d76", "axes": {"M": 152}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n3072_k2048", "solution": null, "workload": {"uuid": "44977a71-5f00-4c17-9de8-1837a124109f", "axes": {"M": 160}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n3072_k2048", "solution": null, "workload": {"uuid": "187b0190-f52b-45f3-b19b-904db0ed6071", "axes": {"M": 168}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n3072_k2048", "solution": null, "workload": {"uuid": "d25a7831-ea39-41e5-9439-f106eb540557", "axes": {"M": 176}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n3072_k2048", "solution": null, "workload": {"uuid": "6eb7fe5f-aa3b-4e7f-ac7a-61c63df11296", "axes": {"M": 184}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n3072_k2048", "solution": null, "workload": {"uuid": "0dc54c87-e26b-4a37-8db3-0ee2befd174a", "axes": {"M": 192}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n3072_k2048", "solution": null, "workload": {"uuid": "13db2d38-8290-4831-bace-25997cabc258", "axes": {"M": 200}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n3072_k2048", "solution": null, "workload": {"uuid": "9293595b-ddfa-4791-9977-0a9459205d3d", "axes": {"M": 208}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n3072_k2048", "solution": null, "workload": {"uuid": "2f5f76ed-e14c-467e-8651-3eebd8476bb1", "axes": {"M": 216}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n3072_k2048", "solution": null, "workload": {"uuid": "495a45c0-b0f8-49da-a3c1-140719a72a19", "axes": {"M": 224}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n3072_k2048", "solution": null, "workload": {"uuid": "38c1053f-1a32-4790-8ae5-b581427c1826", "axes": {"M": 232}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n3072_k2048", "solution": null, "workload": {"uuid": "b3959a7c-687e-4a5c-80bd-0186d30425bb", "axes": {"M": 240}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n3072_k2048", "solution": null, "workload": {"uuid": "38b8d208-9268-40e9-9123-314e97170347", "axes": {"M": 248}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n3072_k2048", "solution": null, "workload": {"uuid": "ab069f30-7a43-4b22-acc2-ab6c5033bb82", "axes": {"M": 256}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n3072_k2048", "solution": null, "workload": {"uuid": "9a43b59c-66b6-46af-8e71-a76a3f70e39c", "axes": {"M": 972}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n3072_k2048", "solution": null, "workload": {"uuid": "c4e5fc75-2a7d-434c-92c1-557ec2b9dec7", "axes": {"M": 2053}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n3072_k2048", "solution": null, "workload": {"uuid": "59e7eeae-b0e1-4b69-bbf8-7d038745eb55", "axes": {"M": 2379}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} -{"definition": "gemm_n3072_k2048", "solution": null, "workload": {"uuid": "6ffbca81-ec08-44fe-ac35-d823455a2f32", "axes": {"M": 8192}, "inputs": {"A": {"type": "random"}, "B": {"type": "random"}}}, "evaluation": null} diff --git a/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64.jsonl b/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64.jsonl deleted file mode 100644 index ef626d80077e00a3aed4cd705f1ffaa73a753daf..0000000000000000000000000000000000000000 --- a/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64.jsonl +++ /dev/null @@ -1,18 +0,0 @@ -{"definition": "gqa_paged_decode_h32_kv8_d128_ps64", "solution": null, "workload": {"uuid": "1a1f0089-58e8-4bb5-81a6-fbe229e3b845", "axes": {"batch_size": 8, "num_pages": 910, "len_indptr": 9, "num_kv_indices": 640}, "inputs": {"q": {"type": "random"}, "k_cache": {"type": "random"}, "v_cache": {"type": "random"}, "kv_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_1a1f0089-58e8-4bb5-81a6-fbe229e3b845.safetensors", "tensor_key": "kv_indptr"}, "kv_indices": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_1a1f0089-58e8-4bb5-81a6-fbe229e3b845.safetensors", "tensor_key": "kv_indices"}, "kv_last_page_len": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_1a1f0089-58e8-4bb5-81a6-fbe229e3b845.safetensors", "tensor_key": "kv_last_page_len"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831843}}}, "evaluation": null} -{"definition": "gqa_paged_decode_h32_kv8_d128_ps64", "solution": null, "workload": {"uuid": "ed7c988c-a1a9-4e62-b410-2e33b2bdc1d8", "axes": {"batch_size": 8, "num_pages": 839, "len_indptr": 9, "num_kv_indices": 528}, "inputs": {"q": {"type": "random"}, "k_cache": {"type": "random"}, "v_cache": {"type": "random"}, "kv_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_ed7c988c-a1a9-4e62-b410-2e33b2bdc1d8.safetensors", "tensor_key": "kv_indptr"}, "kv_indices": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_ed7c988c-a1a9-4e62-b410-2e33b2bdc1d8.safetensors", "tensor_key": "kv_indices"}, "kv_last_page_len": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_ed7c988c-a1a9-4e62-b410-2e33b2bdc1d8.safetensors", "tensor_key": "kv_last_page_len"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831843}}}, "evaluation": null} -{"definition": "gqa_paged_decode_h32_kv8_d128_ps64", "solution": null, "workload": {"uuid": "1735d27c-3a9a-44b4-9fa1-e9a0e3703d38", "axes": {"batch_size": 8, "num_pages": 902, "len_indptr": 9, "num_kv_indices": 576}, "inputs": {"q": {"type": "random"}, "k_cache": {"type": "random"}, "v_cache": {"type": "random"}, "kv_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_1735d27c-3a9a-44b4-9fa1-e9a0e3703d38.safetensors", "tensor_key": "kv_indptr"}, "kv_indices": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_1735d27c-3a9a-44b4-9fa1-e9a0e3703d38.safetensors", "tensor_key": "kv_indices"}, "kv_last_page_len": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_1735d27c-3a9a-44b4-9fa1-e9a0e3703d38.safetensors", "tensor_key": "kv_last_page_len"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831843}}}, "evaluation": null} -{"definition": "gqa_paged_decode_h32_kv8_d128_ps64", "solution": null, "workload": {"uuid": "1bdf652e-366a-4024-9236-ce54682a7cf7", "axes": {"batch_size": 31, "num_pages": 15233, "len_indptr": 32, "num_kv_indices": 14145}, "inputs": {"q": {"type": "random"}, "k_cache": {"type": "random"}, "v_cache": {"type": "random"}, "kv_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_1bdf652e-366a-4024-9236-ce54682a7cf7.safetensors", "tensor_key": "kv_indptr"}, "kv_indices": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_1bdf652e-366a-4024-9236-ce54682a7cf7.safetensors", "tensor_key": "kv_indices"}, "kv_last_page_len": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_1bdf652e-366a-4024-9236-ce54682a7cf7.safetensors", "tensor_key": "kv_last_page_len"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831843}}}, "evaluation": null} -{"definition": "gqa_paged_decode_h32_kv8_d128_ps64", "solution": null, "workload": {"uuid": "721ff730-4d9e-49d8-9fd4-7ba713b7cfdb", "axes": {"batch_size": 32, "num_pages": 15114, "len_indptr": 33, "num_kv_indices": 13783}, "inputs": {"q": {"type": "random"}, "k_cache": {"type": "random"}, "v_cache": {"type": "random"}, "kv_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_721ff730-4d9e-49d8-9fd4-7ba713b7cfdb.safetensors", "tensor_key": "kv_indptr"}, "kv_indices": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_721ff730-4d9e-49d8-9fd4-7ba713b7cfdb.safetensors", "tensor_key": "kv_indices"}, "kv_last_page_len": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_721ff730-4d9e-49d8-9fd4-7ba713b7cfdb.safetensors", "tensor_key": "kv_last_page_len"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831843}}}, "evaluation": null} -{"definition": "gqa_paged_decode_h32_kv8_d128_ps64", "solution": null, "workload": {"uuid": "ff1d78b9-0383-4845-94df-904c49112856", "axes": {"batch_size": 32, "num_pages": 15178, "len_indptr": 33, "num_kv_indices": 14103}, "inputs": {"q": {"type": "random"}, "k_cache": {"type": "random"}, "v_cache": {"type": "random"}, "kv_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_ff1d78b9-0383-4845-94df-904c49112856.safetensors", "tensor_key": "kv_indptr"}, "kv_indices": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_ff1d78b9-0383-4845-94df-904c49112856.safetensors", "tensor_key": "kv_indices"}, "kv_last_page_len": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_ff1d78b9-0383-4845-94df-904c49112856.safetensors", "tensor_key": "kv_last_page_len"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831843}}}, "evaluation": null} -{"definition": "gqa_paged_decode_h32_kv8_d128_ps64", "solution": null, "workload": {"uuid": "16d8ee84-f85b-4192-a58a-2228a5d9641e", "axes": {"batch_size": 63, "num_pages": 29569, "len_indptr": 64, "num_kv_indices": 27392}, "inputs": {"q": {"type": "random"}, "k_cache": {"type": "random"}, "v_cache": {"type": "random"}, "kv_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_16d8ee84-f85b-4192-a58a-2228a5d9641e.safetensors", "tensor_key": "kv_indptr"}, "kv_indices": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_16d8ee84-f85b-4192-a58a-2228a5d9641e.safetensors", "tensor_key": "kv_indices"}, "kv_last_page_len": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_16d8ee84-f85b-4192-a58a-2228a5d9641e.safetensors", "tensor_key": "kv_last_page_len"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831843}}}, "evaluation": null} -{"definition": "gqa_paged_decode_h32_kv8_d128_ps64", "solution": null, "workload": {"uuid": "51393402-8fec-4b52-88ca-3cdb2270d293", "axes": {"batch_size": 64, "num_pages": 29189, "len_indptr": 65, "num_kv_indices": 26582}, "inputs": {"q": {"type": "random"}, "k_cache": {"type": "random"}, "v_cache": {"type": "random"}, "kv_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_51393402-8fec-4b52-88ca-3cdb2270d293.safetensors", "tensor_key": "kv_indptr"}, "kv_indices": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_51393402-8fec-4b52-88ca-3cdb2270d293.safetensors", "tensor_key": "kv_indices"}, "kv_last_page_len": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_51393402-8fec-4b52-88ca-3cdb2270d293.safetensors", "tensor_key": "kv_last_page_len"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831843}}}, "evaluation": null} -{"definition": "gqa_paged_decode_h32_kv8_d128_ps64", "solution": null, "workload": {"uuid": "8c88a86f-1843-42eb-9251-8a0ef4b6a1ee", "axes": {"batch_size": 64, "num_pages": 29441, "len_indptr": 65, "num_kv_indices": 27222}, "inputs": {"q": {"type": "random"}, "k_cache": {"type": "random"}, "v_cache": {"type": "random"}, "kv_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_8c88a86f-1843-42eb-9251-8a0ef4b6a1ee.safetensors", "tensor_key": "kv_indptr"}, "kv_indices": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_8c88a86f-1843-42eb-9251-8a0ef4b6a1ee.safetensors", "tensor_key": "kv_indices"}, "kv_last_page_len": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_8c88a86f-1843-42eb-9251-8a0ef4b6a1ee.safetensors", "tensor_key": "kv_last_page_len"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831843}}}, "evaluation": null} -{"definition": "gqa_paged_decode_h32_kv8_d128_ps64", "solution": null, "workload": {"uuid": "7832673f-b1c1-4185-b298-d510edbceac4", "axes": {"batch_size": 127, "num_pages": 52481, "len_indptr": 128, "num_kv_indices": 48120}, "inputs": {"q": {"type": "random"}, "k_cache": {"type": "random"}, "v_cache": {"type": "random"}, "kv_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_7832673f-b1c1-4185-b298-d510edbceac4.safetensors", "tensor_key": "kv_indptr"}, "kv_indices": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_7832673f-b1c1-4185-b298-d510edbceac4.safetensors", "tensor_key": "kv_indices"}, "kv_last_page_len": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_7832673f-b1c1-4185-b298-d510edbceac4.safetensors", "tensor_key": "kv_last_page_len"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831843}}}, "evaluation": null} -{"definition": "gqa_paged_decode_h32_kv8_d128_ps64", "solution": null, "workload": {"uuid": "c81ef23a-da17-4cdd-88c2-327a4af240d0", "axes": {"batch_size": 128, "num_pages": 51792, "len_indptr": 129, "num_kv_indices": 46414}, "inputs": {"q": {"type": "random"}, "k_cache": {"type": "random"}, "v_cache": {"type": "random"}, "kv_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_c81ef23a-da17-4cdd-88c2-327a4af240d0.safetensors", "tensor_key": "kv_indptr"}, "kv_indices": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_c81ef23a-da17-4cdd-88c2-327a4af240d0.safetensors", "tensor_key": "kv_indices"}, "kv_last_page_len": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_c81ef23a-da17-4cdd-88c2-327a4af240d0.safetensors", "tensor_key": "kv_last_page_len"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831843}}}, "evaluation": null} -{"definition": "gqa_paged_decode_h32_kv8_d128_ps64", "solution": null, "workload": {"uuid": "2b180048-9203-476c-a5b4-255118f15059", "axes": {"batch_size": 128, "num_pages": 52033, "len_indptr": 129, "num_kv_indices": 47438}, "inputs": {"q": {"type": "random"}, "k_cache": {"type": "random"}, "v_cache": {"type": "random"}, "kv_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_2b180048-9203-476c-a5b4-255118f15059.safetensors", "tensor_key": "kv_indptr"}, "kv_indices": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_2b180048-9203-476c-a5b4-255118f15059.safetensors", "tensor_key": "kv_indices"}, "kv_last_page_len": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_2b180048-9203-476c-a5b4-255118f15059.safetensors", "tensor_key": "kv_last_page_len"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831843}}}, "evaluation": null} -{"definition": "gqa_paged_decode_h32_kv8_d128_ps64", "solution": null, "workload": {"uuid": "ee6efcf9-bf19-432f-bcd3-33f1c9ec599e", "axes": {"batch_size": 16, "num_pages": 1794, "len_indptr": 17, "num_kv_indices": 1257}, "inputs": {"q": {"type": "random"}, "k_cache": {"type": "random"}, "v_cache": {"type": "random"}, "kv_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_ee6efcf9-bf19-432f-bcd3-33f1c9ec599e.safetensors", "tensor_key": "kv_indptr"}, "kv_indices": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_ee6efcf9-bf19-432f-bcd3-33f1c9ec599e.safetensors", "tensor_key": "kv_indices"}, "kv_last_page_len": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_ee6efcf9-bf19-432f-bcd3-33f1c9ec599e.safetensors", "tensor_key": "kv_last_page_len"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831843}}}, "evaluation": null} -{"definition": "gqa_paged_decode_h32_kv8_d128_ps64", "solution": null, "workload": {"uuid": "1196e756-48f7-4f94-81a3-94a68f170b59", "axes": {"batch_size": 16, "num_pages": 1686, "len_indptr": 17, "num_kv_indices": 1033}, "inputs": {"q": {"type": "random"}, "k_cache": {"type": "random"}, "v_cache": {"type": "random"}, "kv_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_1196e756-48f7-4f94-81a3-94a68f170b59.safetensors", "tensor_key": "kv_indptr"}, "kv_indices": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_1196e756-48f7-4f94-81a3-94a68f170b59.safetensors", "tensor_key": "kv_indices"}, "kv_last_page_len": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_1196e756-48f7-4f94-81a3-94a68f170b59.safetensors", "tensor_key": "kv_last_page_len"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831843}}}, "evaluation": null} -{"definition": "gqa_paged_decode_h32_kv8_d128_ps64", "solution": null, "workload": {"uuid": "09bcbbfb-0d48-42d8-a39c-e7d01300a646", "axes": {"batch_size": 16, "num_pages": 1735, "len_indptr": 17, "num_kv_indices": 1145}, "inputs": {"q": {"type": "random"}, "k_cache": {"type": "random"}, "v_cache": {"type": "random"}, "kv_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_09bcbbfb-0d48-42d8-a39c-e7d01300a646.safetensors", "tensor_key": "kv_indptr"}, "kv_indices": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_09bcbbfb-0d48-42d8-a39c-e7d01300a646.safetensors", "tensor_key": "kv_indices"}, "kv_last_page_len": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_09bcbbfb-0d48-42d8-a39c-e7d01300a646.safetensors", "tensor_key": "kv_last_page_len"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831843}}}, "evaluation": null} -{"definition": "gqa_paged_decode_h32_kv8_d128_ps64", "solution": null, "workload": {"uuid": "580fa22c-dee6-484e-8847-eb301b21057f", "axes": {"batch_size": 4, "num_pages": 398, "len_indptr": 5, "num_kv_indices": 268}, "inputs": {"q": {"type": "random"}, "k_cache": {"type": "random"}, "v_cache": {"type": "random"}, "kv_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_580fa22c-dee6-484e-8847-eb301b21057f.safetensors", "tensor_key": "kv_indptr"}, "kv_indices": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_580fa22c-dee6-484e-8847-eb301b21057f.safetensors", "tensor_key": "kv_indices"}, "kv_last_page_len": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_580fa22c-dee6-484e-8847-eb301b21057f.safetensors", "tensor_key": "kv_last_page_len"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831843}}}, "evaluation": null} -{"definition": "gqa_paged_decode_h32_kv8_d128_ps64", "solution": null, "workload": {"uuid": "b013fc3f-eca8-4c18-a5a1-7231c26eb231", "axes": {"batch_size": 4, "num_pages": 349, "len_indptr": 5, "num_kv_indices": 212}, "inputs": {"q": {"type": "random"}, "k_cache": {"type": "random"}, "v_cache": {"type": "random"}, "kv_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_b013fc3f-eca8-4c18-a5a1-7231c26eb231.safetensors", "tensor_key": "kv_indptr"}, "kv_indices": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_b013fc3f-eca8-4c18-a5a1-7231c26eb231.safetensors", "tensor_key": "kv_indices"}, "kv_last_page_len": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_b013fc3f-eca8-4c18-a5a1-7231c26eb231.safetensors", "tensor_key": "kv_last_page_len"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831843}}}, "evaluation": null} -{"definition": "gqa_paged_decode_h32_kv8_d128_ps64", "solution": null, "workload": {"uuid": "1e0c7af6-f077-42f6-b5a0-3cbb2e30c5c3", "axes": {"batch_size": 4, "num_pages": 389, "len_indptr": 5, "num_kv_indices": 232}, "inputs": {"q": {"type": "random"}, "k_cache": {"type": "random"}, "v_cache": {"type": "random"}, "kv_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_1e0c7af6-f077-42f6-b5a0-3cbb2e30c5c3.safetensors", "tensor_key": "kv_indptr"}, "kv_indices": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_1e0c7af6-f077-42f6-b5a0-3cbb2e30c5c3.safetensors", "tensor_key": "kv_indices"}, "kv_last_page_len": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d128_ps64/gqa_paged_decode_h32_kv8_d128_ps64_1e0c7af6-f077-42f6-b5a0-3cbb2e30c5c3.safetensors", "tensor_key": "kv_last_page_len"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831843}}}, "evaluation": null} diff --git a/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps1.jsonl b/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps1.jsonl deleted file mode 100644 index ec0fe42ce6a5fef07d46e2b6b5cbcb9099daf377..0000000000000000000000000000000000000000 --- a/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps1.jsonl +++ /dev/null @@ -1,8 +0,0 @@ -{"definition": "gqa_paged_decode_h32_kv8_d64_ps1", "solution": null, "workload": {"uuid": "8342a570-03f6-42ab-b7fb-bc1eadc8759e", "axes": {"batch_size": 1, "num_pages": 9, "len_indptr": 2, "num_kv_indices": 8}, "inputs": {"q": {"type": "random"}, "k_cache": {"type": "random"}, "v_cache": {"type": "random"}, "kv_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps1/gqa_paged_decode_h32_kv8_d64_ps1_8342a570-03f6-42ab-b7fb-bc1eadc8759e.safetensors", "tensor_key": "kv_indptr"}, "kv_indices": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps1/gqa_paged_decode_h32_kv8_d64_ps1_8342a570-03f6-42ab-b7fb-bc1eadc8759e.safetensors", "tensor_key": "kv_indices"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831843}}}, "evaluation": null} -{"definition": "gqa_paged_decode_h32_kv8_d64_ps1", "solution": null, "workload": {"uuid": "5d4bb19c-8f86-41c2-a3dc-ab974ba60bdf", "axes": {"batch_size": 1, "num_pages": 73, "len_indptr": 2, "num_kv_indices": 56}, "inputs": {"q": {"type": "random"}, "k_cache": {"type": "random"}, "v_cache": {"type": "random"}, "kv_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps1/gqa_paged_decode_h32_kv8_d64_ps1_5d4bb19c-8f86-41c2-a3dc-ab974ba60bdf.safetensors", "tensor_key": "kv_indptr"}, "kv_indices": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps1/gqa_paged_decode_h32_kv8_d64_ps1_5d4bb19c-8f86-41c2-a3dc-ab974ba60bdf.safetensors", "tensor_key": "kv_indices"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831843}}}, "evaluation": null} -{"definition": "gqa_paged_decode_h32_kv8_d64_ps1", "solution": null, "workload": {"uuid": "3d044768-4601-4b70-8472-85fbc34a4504", "axes": {"batch_size": 1, "num_pages": 18, "len_indptr": 2, "num_kv_indices": 2}, "inputs": {"q": {"type": "random"}, "k_cache": {"type": "random"}, "v_cache": {"type": "random"}, "kv_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps1/gqa_paged_decode_h32_kv8_d64_ps1_3d044768-4601-4b70-8472-85fbc34a4504.safetensors", "tensor_key": "kv_indptr"}, "kv_indices": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps1/gqa_paged_decode_h32_kv8_d64_ps1_3d044768-4601-4b70-8472-85fbc34a4504.safetensors", "tensor_key": "kv_indices"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831843}}}, "evaluation": null} -{"definition": "gqa_paged_decode_h32_kv8_d64_ps1", "solution": null, "workload": {"uuid": "921b9fdc-3937-4873-b11e-7f4650daf0f8", "axes": {"batch_size": 1, "num_pages": 16, "len_indptr": 2, "num_kv_indices": 15}, "inputs": {"q": {"type": "random"}, "k_cache": {"type": "random"}, "v_cache": {"type": "random"}, "kv_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps1/gqa_paged_decode_h32_kv8_d64_ps1_921b9fdc-3937-4873-b11e-7f4650daf0f8.safetensors", "tensor_key": "kv_indptr"}, "kv_indices": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps1/gqa_paged_decode_h32_kv8_d64_ps1_921b9fdc-3937-4873-b11e-7f4650daf0f8.safetensors", "tensor_key": "kv_indices"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831843}}}, "evaluation": null} -{"definition": "gqa_paged_decode_h32_kv8_d64_ps1", "solution": null, "workload": {"uuid": "011db24f-e331-41e4-89fd-dff3bb7345e5", "axes": {"batch_size": 1, "num_pages": 9, "len_indptr": 2, "num_kv_indices": 8}, "inputs": {"q": {"type": "random"}, "k_cache": {"type": "random"}, "v_cache": {"type": "random"}, "kv_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps1/gqa_paged_decode_h32_kv8_d64_ps1_011db24f-e331-41e4-89fd-dff3bb7345e5.safetensors", "tensor_key": "kv_indptr"}, "kv_indices": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps1/gqa_paged_decode_h32_kv8_d64_ps1_011db24f-e331-41e4-89fd-dff3bb7345e5.safetensors", "tensor_key": "kv_indices"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831843}}}, "evaluation": null} -{"definition": "gqa_paged_decode_h32_kv8_d64_ps1", "solution": null, "workload": {"uuid": "7810a1b5-7bcd-4825-b36f-80c9b4ec61c6", "axes": {"batch_size": 1, "num_pages": 73, "len_indptr": 2, "num_kv_indices": 56}, "inputs": {"q": {"type": "random"}, "k_cache": {"type": "random"}, "v_cache": {"type": "random"}, "kv_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps1/gqa_paged_decode_h32_kv8_d64_ps1_7810a1b5-7bcd-4825-b36f-80c9b4ec61c6.safetensors", "tensor_key": "kv_indptr"}, "kv_indices": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps1/gqa_paged_decode_h32_kv8_d64_ps1_7810a1b5-7bcd-4825-b36f-80c9b4ec61c6.safetensors", "tensor_key": "kv_indices"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831843}}}, "evaluation": null} -{"definition": "gqa_paged_decode_h32_kv8_d64_ps1", "solution": null, "workload": {"uuid": "e97b000c-8677-4877-86bf-2a82b9cb77e4", "axes": {"batch_size": 1, "num_pages": 18, "len_indptr": 2, "num_kv_indices": 2}, "inputs": {"q": {"type": "random"}, "k_cache": {"type": "random"}, "v_cache": {"type": "random"}, "kv_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps1/gqa_paged_decode_h32_kv8_d64_ps1_e97b000c-8677-4877-86bf-2a82b9cb77e4.safetensors", "tensor_key": "kv_indptr"}, "kv_indices": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps1/gqa_paged_decode_h32_kv8_d64_ps1_e97b000c-8677-4877-86bf-2a82b9cb77e4.safetensors", "tensor_key": "kv_indices"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831843}}}, "evaluation": null} -{"definition": "gqa_paged_decode_h32_kv8_d64_ps1", "solution": null, "workload": {"uuid": "11155649-fd33-4b5c-80d8-05553c0cec0b", "axes": {"batch_size": 1, "num_pages": 16, "len_indptr": 2, "num_kv_indices": 15}, "inputs": {"q": {"type": "random"}, "k_cache": {"type": "random"}, "v_cache": {"type": "random"}, "kv_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps1/gqa_paged_decode_h32_kv8_d64_ps1_11155649-fd33-4b5c-80d8-05553c0cec0b.safetensors", "tensor_key": "kv_indptr"}, "kv_indices": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps1/gqa_paged_decode_h32_kv8_d64_ps1_11155649-fd33-4b5c-80d8-05553c0cec0b.safetensors", "tensor_key": "kv_indices"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831843}}}, "evaluation": null} diff --git a/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps64.jsonl b/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps64.jsonl deleted file mode 100644 index e43d8b6baafefe8676633ac96b48d205554af7e1..0000000000000000000000000000000000000000 --- a/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps64.jsonl +++ /dev/null @@ -1,8 +0,0 @@ -{"definition": "gqa_paged_decode_h32_kv8_d64_ps64", "solution": null, "workload": {"uuid": "4b40f9c1-2054-4b1c-ab80-0873f2801f7e", "axes": {"batch_size": 1, "num_pages": 72, "len_indptr": 2, "num_kv_indices": 8}, "inputs": {"q": {"type": "random"}, "k_cache": {"type": "random"}, "v_cache": {"type": "random"}, "kv_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps64/gqa_paged_decode_h32_kv8_d64_ps64_4b40f9c1-2054-4b1c-ab80-0873f2801f7e.safetensors", "tensor_key": "kv_indptr"}, "kv_indices": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps64/gqa_paged_decode_h32_kv8_d64_ps64_4b40f9c1-2054-4b1c-ab80-0873f2801f7e.safetensors", "tensor_key": "kv_indices"}, "kv_last_page_len": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps64/gqa_paged_decode_h32_kv8_d64_ps64_4b40f9c1-2054-4b1c-ab80-0873f2801f7e.safetensors", "tensor_key": "kv_last_page_len"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831843}}}, "evaluation": null} -{"definition": "gqa_paged_decode_h32_kv8_d64_ps64", "solution": null, "workload": {"uuid": "3a4d1bf7-c9fe-40b9-a50d-9780a14fabf7", "axes": {"batch_size": 1, "num_pages": 120, "len_indptr": 2, "num_kv_indices": 56}, "inputs": {"q": {"type": "random"}, "k_cache": {"type": "random"}, "v_cache": {"type": "random"}, "kv_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps64/gqa_paged_decode_h32_kv8_d64_ps64_3a4d1bf7-c9fe-40b9-a50d-9780a14fabf7.safetensors", "tensor_key": "kv_indptr"}, "kv_indices": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps64/gqa_paged_decode_h32_kv8_d64_ps64_3a4d1bf7-c9fe-40b9-a50d-9780a14fabf7.safetensors", "tensor_key": "kv_indices"}, "kv_last_page_len": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps64/gqa_paged_decode_h32_kv8_d64_ps64_3a4d1bf7-c9fe-40b9-a50d-9780a14fabf7.safetensors", "tensor_key": "kv_last_page_len"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831843}}}, "evaluation": null} -{"definition": "gqa_paged_decode_h32_kv8_d64_ps64", "solution": null, "workload": {"uuid": "82761919-e520-4a9d-b18d-ef5b07f25245", "axes": {"batch_size": 1, "num_pages": 66, "len_indptr": 2, "num_kv_indices": 2}, "inputs": {"q": {"type": "random"}, "k_cache": {"type": "random"}, "v_cache": {"type": "random"}, "kv_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps64/gqa_paged_decode_h32_kv8_d64_ps64_82761919-e520-4a9d-b18d-ef5b07f25245.safetensors", "tensor_key": "kv_indptr"}, "kv_indices": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps64/gqa_paged_decode_h32_kv8_d64_ps64_82761919-e520-4a9d-b18d-ef5b07f25245.safetensors", "tensor_key": "kv_indices"}, "kv_last_page_len": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps64/gqa_paged_decode_h32_kv8_d64_ps64_82761919-e520-4a9d-b18d-ef5b07f25245.safetensors", "tensor_key": "kv_last_page_len"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831843}}}, "evaluation": null} -{"definition": "gqa_paged_decode_h32_kv8_d64_ps64", "solution": null, "workload": {"uuid": "9d298dcb-2965-4e58-8c82-ea1b1fa49ccd", "axes": {"batch_size": 1, "num_pages": 79, "len_indptr": 2, "num_kv_indices": 15}, "inputs": {"q": {"type": "random"}, "k_cache": {"type": "random"}, "v_cache": {"type": "random"}, "kv_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps64/gqa_paged_decode_h32_kv8_d64_ps64_9d298dcb-2965-4e58-8c82-ea1b1fa49ccd.safetensors", "tensor_key": "kv_indptr"}, "kv_indices": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps64/gqa_paged_decode_h32_kv8_d64_ps64_9d298dcb-2965-4e58-8c82-ea1b1fa49ccd.safetensors", "tensor_key": "kv_indices"}, "kv_last_page_len": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps64/gqa_paged_decode_h32_kv8_d64_ps64_9d298dcb-2965-4e58-8c82-ea1b1fa49ccd.safetensors", "tensor_key": "kv_last_page_len"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831843}}}, "evaluation": null} -{"definition": "gqa_paged_decode_h32_kv8_d64_ps64", "solution": null, "workload": {"uuid": "f5ae694d-0c9c-4534-8ab2-b4cdd38d6a35", "axes": {"batch_size": 1, "num_pages": 72, "len_indptr": 2, "num_kv_indices": 8}, "inputs": {"q": {"type": "random"}, "k_cache": {"type": "random"}, "v_cache": {"type": "random"}, "kv_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps64/gqa_paged_decode_h32_kv8_d64_ps64_f5ae694d-0c9c-4534-8ab2-b4cdd38d6a35.safetensors", "tensor_key": "kv_indptr"}, "kv_indices": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps64/gqa_paged_decode_h32_kv8_d64_ps64_f5ae694d-0c9c-4534-8ab2-b4cdd38d6a35.safetensors", "tensor_key": "kv_indices"}, "kv_last_page_len": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps64/gqa_paged_decode_h32_kv8_d64_ps64_f5ae694d-0c9c-4534-8ab2-b4cdd38d6a35.safetensors", "tensor_key": "kv_last_page_len"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831843}}}, "evaluation": null} -{"definition": "gqa_paged_decode_h32_kv8_d64_ps64", "solution": null, "workload": {"uuid": "ea8adb18-8557-4dd2-9191-c1a045c7416d", "axes": {"batch_size": 1, "num_pages": 120, "len_indptr": 2, "num_kv_indices": 56}, "inputs": {"q": {"type": "random"}, "k_cache": {"type": "random"}, "v_cache": {"type": "random"}, "kv_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps64/gqa_paged_decode_h32_kv8_d64_ps64_ea8adb18-8557-4dd2-9191-c1a045c7416d.safetensors", "tensor_key": "kv_indptr"}, "kv_indices": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps64/gqa_paged_decode_h32_kv8_d64_ps64_ea8adb18-8557-4dd2-9191-c1a045c7416d.safetensors", "tensor_key": "kv_indices"}, "kv_last_page_len": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps64/gqa_paged_decode_h32_kv8_d64_ps64_ea8adb18-8557-4dd2-9191-c1a045c7416d.safetensors", "tensor_key": "kv_last_page_len"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831843}}}, "evaluation": null} -{"definition": "gqa_paged_decode_h32_kv8_d64_ps64", "solution": null, "workload": {"uuid": "68af5fd6-8da8-4070-ad1b-503ba3a56b94", "axes": {"batch_size": 1, "num_pages": 66, "len_indptr": 2, "num_kv_indices": 2}, "inputs": {"q": {"type": "random"}, "k_cache": {"type": "random"}, "v_cache": {"type": "random"}, "kv_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps64/gqa_paged_decode_h32_kv8_d64_ps64_68af5fd6-8da8-4070-ad1b-503ba3a56b94.safetensors", "tensor_key": "kv_indptr"}, "kv_indices": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps64/gqa_paged_decode_h32_kv8_d64_ps64_68af5fd6-8da8-4070-ad1b-503ba3a56b94.safetensors", "tensor_key": "kv_indices"}, "kv_last_page_len": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps64/gqa_paged_decode_h32_kv8_d64_ps64_68af5fd6-8da8-4070-ad1b-503ba3a56b94.safetensors", "tensor_key": "kv_last_page_len"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831843}}}, "evaluation": null} -{"definition": "gqa_paged_decode_h32_kv8_d64_ps64", "solution": null, "workload": {"uuid": "687cdb3b-d2e0-4215-887b-fbf56b8569e7", "axes": {"batch_size": 1, "num_pages": 79, "len_indptr": 2, "num_kv_indices": 15}, "inputs": {"q": {"type": "random"}, "k_cache": {"type": "random"}, "v_cache": {"type": "random"}, "kv_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps64/gqa_paged_decode_h32_kv8_d64_ps64_687cdb3b-d2e0-4215-887b-fbf56b8569e7.safetensors", "tensor_key": "kv_indptr"}, "kv_indices": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps64/gqa_paged_decode_h32_kv8_d64_ps64_687cdb3b-d2e0-4215-887b-fbf56b8569e7.safetensors", "tensor_key": "kv_indices"}, "kv_last_page_len": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_decode_h32_kv8_d64_ps64/gqa_paged_decode_h32_kv8_d64_ps64_687cdb3b-d2e0-4215-887b-fbf56b8569e7.safetensors", "tensor_key": "kv_last_page_len"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831843}}}, "evaluation": null} diff --git a/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps1.jsonl b/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps1.jsonl deleted file mode 100644 index 29823829855f5d126c1d5caf91a26542a01ffbdc..0000000000000000000000000000000000000000 --- a/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps1.jsonl +++ /dev/null @@ -1,8 +0,0 @@ -{"definition": "gqa_paged_prefill_causal_h32_kv8_d64_ps1", "solution": null, "workload": {"uuid": "25732cc4-7e33-44bc-9a26-8f5f1ebd1da5", "axes": {"len_indptr": 2, "total_q": 4, "num_kv_indices": 4, "num_pages": 1}, "inputs": {"q": {"type": "random"}, "k_cache": {"type": "random"}, "v_cache": {"type": "random"}, "qo_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps1/gqa_paged_prefill_causal_h32_kv8_d64_ps1_25732cc4-7e33-44bc-9a26-8f5f1ebd1da5.safetensors", "tensor_key": "qo_indptr"}, "kv_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps1/gqa_paged_prefill_causal_h32_kv8_d64_ps1_25732cc4-7e33-44bc-9a26-8f5f1ebd1da5.safetensors", "tensor_key": "kv_indptr"}, "kv_indices": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps1/gqa_paged_prefill_causal_h32_kv8_d64_ps1_25732cc4-7e33-44bc-9a26-8f5f1ebd1da5.safetensors", "tensor_key": "kv_indices"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831843}}}, "evaluation": null} -{"definition": "gqa_paged_prefill_causal_h32_kv8_d64_ps1", "solution": null, "workload": {"uuid": "f8641e85-8041-426f-9996-e189e4aa7004", "axes": {"len_indptr": 2, "total_q": 8192, "num_kv_indices": 8192, "num_pages": 1}, "inputs": {"q": {"type": "random"}, "k_cache": {"type": "random"}, "v_cache": {"type": "random"}, "qo_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps1/gqa_paged_prefill_causal_h32_kv8_d64_ps1_f8641e85-8041-426f-9996-e189e4aa7004.safetensors", "tensor_key": "qo_indptr"}, "kv_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps1/gqa_paged_prefill_causal_h32_kv8_d64_ps1_f8641e85-8041-426f-9996-e189e4aa7004.safetensors", "tensor_key": "kv_indptr"}, "kv_indices": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps1/gqa_paged_prefill_causal_h32_kv8_d64_ps1_f8641e85-8041-426f-9996-e189e4aa7004.safetensors", "tensor_key": "kv_indices"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831843}}}, "evaluation": null} -{"definition": "gqa_paged_prefill_causal_h32_kv8_d64_ps1", "solution": null, "workload": {"uuid": "8f24a1a5-f13a-4902-a0ee-f457a5074862", "axes": {"len_indptr": 2, "total_q": 576, "num_kv_indices": 576, "num_pages": 1}, "inputs": {"q": {"type": "random"}, "k_cache": {"type": "random"}, "v_cache": {"type": "random"}, "qo_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps1/gqa_paged_prefill_causal_h32_kv8_d64_ps1_8f24a1a5-f13a-4902-a0ee-f457a5074862.safetensors", "tensor_key": "qo_indptr"}, "kv_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps1/gqa_paged_prefill_causal_h32_kv8_d64_ps1_8f24a1a5-f13a-4902-a0ee-f457a5074862.safetensors", "tensor_key": "kv_indptr"}, "kv_indices": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps1/gqa_paged_prefill_causal_h32_kv8_d64_ps1_8f24a1a5-f13a-4902-a0ee-f457a5074862.safetensors", "tensor_key": "kv_indices"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831843}}}, "evaluation": null} -{"definition": "gqa_paged_prefill_causal_h32_kv8_d64_ps1", "solution": null, "workload": {"uuid": "297c3b37-1681-41f6-a052-1f449af8c580", "axes": {"len_indptr": 2, "total_q": 2304, "num_kv_indices": 2304, "num_pages": 1}, "inputs": {"q": {"type": "random"}, "k_cache": {"type": "random"}, "v_cache": {"type": "random"}, "qo_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps1/gqa_paged_prefill_causal_h32_kv8_d64_ps1_297c3b37-1681-41f6-a052-1f449af8c580.safetensors", "tensor_key": "qo_indptr"}, "kv_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps1/gqa_paged_prefill_causal_h32_kv8_d64_ps1_297c3b37-1681-41f6-a052-1f449af8c580.safetensors", "tensor_key": "kv_indptr"}, "kv_indices": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps1/gqa_paged_prefill_causal_h32_kv8_d64_ps1_297c3b37-1681-41f6-a052-1f449af8c580.safetensors", "tensor_key": "kv_indices"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831843}}}, "evaluation": null} -{"definition": "gqa_paged_prefill_causal_h32_kv8_d64_ps1", "solution": null, "workload": {"uuid": "efb8c6d0-1c04-4804-a29c-b5339ccdbe3a", "axes": {"len_indptr": 2, "total_q": 4, "num_kv_indices": 4, "num_pages": 1}, "inputs": {"q": {"type": "random"}, "k_cache": {"type": "random"}, "v_cache": {"type": "random"}, "qo_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps1/gqa_paged_prefill_causal_h32_kv8_d64_ps1_efb8c6d0-1c04-4804-a29c-b5339ccdbe3a.safetensors", "tensor_key": "qo_indptr"}, "kv_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps1/gqa_paged_prefill_causal_h32_kv8_d64_ps1_efb8c6d0-1c04-4804-a29c-b5339ccdbe3a.safetensors", "tensor_key": "kv_indptr"}, "kv_indices": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps1/gqa_paged_prefill_causal_h32_kv8_d64_ps1_efb8c6d0-1c04-4804-a29c-b5339ccdbe3a.safetensors", "tensor_key": "kv_indices"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831843}}}, "evaluation": null} -{"definition": "gqa_paged_prefill_causal_h32_kv8_d64_ps1", "solution": null, "workload": {"uuid": "940a5fb1-7b82-40ad-b1ca-153bd7c170d9", "axes": {"len_indptr": 2, "total_q": 8192, "num_kv_indices": 8192, "num_pages": 1}, "inputs": {"q": {"type": "random"}, "k_cache": {"type": "random"}, "v_cache": {"type": "random"}, "qo_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps1/gqa_paged_prefill_causal_h32_kv8_d64_ps1_940a5fb1-7b82-40ad-b1ca-153bd7c170d9.safetensors", "tensor_key": "qo_indptr"}, "kv_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps1/gqa_paged_prefill_causal_h32_kv8_d64_ps1_940a5fb1-7b82-40ad-b1ca-153bd7c170d9.safetensors", "tensor_key": "kv_indptr"}, "kv_indices": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps1/gqa_paged_prefill_causal_h32_kv8_d64_ps1_940a5fb1-7b82-40ad-b1ca-153bd7c170d9.safetensors", "tensor_key": "kv_indices"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831843}}}, "evaluation": null} -{"definition": "gqa_paged_prefill_causal_h32_kv8_d64_ps1", "solution": null, "workload": {"uuid": "46907c3e-0b63-41a2-9ebf-63382ffffb60", "axes": {"len_indptr": 2, "total_q": 576, "num_kv_indices": 576, "num_pages": 1}, "inputs": {"q": {"type": "random"}, "k_cache": {"type": "random"}, "v_cache": {"type": "random"}, "qo_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps1/gqa_paged_prefill_causal_h32_kv8_d64_ps1_46907c3e-0b63-41a2-9ebf-63382ffffb60.safetensors", "tensor_key": "qo_indptr"}, "kv_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps1/gqa_paged_prefill_causal_h32_kv8_d64_ps1_46907c3e-0b63-41a2-9ebf-63382ffffb60.safetensors", "tensor_key": "kv_indptr"}, "kv_indices": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps1/gqa_paged_prefill_causal_h32_kv8_d64_ps1_46907c3e-0b63-41a2-9ebf-63382ffffb60.safetensors", "tensor_key": "kv_indices"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831843}}}, "evaluation": null} -{"definition": "gqa_paged_prefill_causal_h32_kv8_d64_ps1", "solution": null, "workload": {"uuid": "0b2f34f7-531b-436a-8c39-3d13f486cb78", "axes": {"len_indptr": 2, "total_q": 2304, "num_kv_indices": 2304, "num_pages": 1}, "inputs": {"q": {"type": "random"}, "k_cache": {"type": "random"}, "v_cache": {"type": "random"}, "qo_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps1/gqa_paged_prefill_causal_h32_kv8_d64_ps1_0b2f34f7-531b-436a-8c39-3d13f486cb78.safetensors", "tensor_key": "qo_indptr"}, "kv_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps1/gqa_paged_prefill_causal_h32_kv8_d64_ps1_0b2f34f7-531b-436a-8c39-3d13f486cb78.safetensors", "tensor_key": "kv_indptr"}, "kv_indices": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps1/gqa_paged_prefill_causal_h32_kv8_d64_ps1_0b2f34f7-531b-436a-8c39-3d13f486cb78.safetensors", "tensor_key": "kv_indices"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831843}}}, "evaluation": null} diff --git a/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps64.jsonl b/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps64.jsonl deleted file mode 100644 index c16a30e1f015668438a1fc8aa16332c278991ae1..0000000000000000000000000000000000000000 --- a/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps64.jsonl +++ /dev/null @@ -1,8 +0,0 @@ -{"definition": "gqa_paged_prefill_causal_h32_kv8_d64_ps64", "solution": null, "workload": {"uuid": "83cca546-a68d-47ee-ac1d-1883ea9a8cd3", "axes": {"len_indptr": 2, "total_q": 4, "num_kv_indices": 4, "num_pages": 1, "batch_size": 2}, "inputs": {"q": {"type": "random"}, "k_cache": {"type": "random"}, "v_cache": {"type": "random"}, "qo_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps64/gqa_paged_prefill_causal_h32_kv8_d64_ps64_83cca546-a68d-47ee-ac1d-1883ea9a8cd3.safetensors", "tensor_key": "qo_indptr"}, "kv_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps64/gqa_paged_prefill_causal_h32_kv8_d64_ps64_83cca546-a68d-47ee-ac1d-1883ea9a8cd3.safetensors", "tensor_key": "kv_indptr"}, "kv_indices": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps64/gqa_paged_prefill_causal_h32_kv8_d64_ps64_83cca546-a68d-47ee-ac1d-1883ea9a8cd3.safetensors", "tensor_key": "kv_indices"}, "kv_last_page_len": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps64/gqa_paged_prefill_causal_h32_kv8_d64_ps64_83cca546-a68d-47ee-ac1d-1883ea9a8cd3.safetensors", "tensor_key": "kv_last_page_len"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831843}}}, "evaluation": null} -{"definition": "gqa_paged_prefill_causal_h32_kv8_d64_ps64", "solution": null, "workload": {"uuid": "ea6b783f-fdb1-4be4-97fe-180de19df270", "axes": {"len_indptr": 2, "total_q": 8192, "num_kv_indices": 8192, "num_pages": 1, "batch_size": 2}, "inputs": {"q": {"type": "random"}, "k_cache": {"type": "random"}, "v_cache": {"type": "random"}, "qo_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps64/gqa_paged_prefill_causal_h32_kv8_d64_ps64_ea6b783f-fdb1-4be4-97fe-180de19df270.safetensors", "tensor_key": "qo_indptr"}, "kv_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps64/gqa_paged_prefill_causal_h32_kv8_d64_ps64_ea6b783f-fdb1-4be4-97fe-180de19df270.safetensors", "tensor_key": "kv_indptr"}, "kv_indices": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps64/gqa_paged_prefill_causal_h32_kv8_d64_ps64_ea6b783f-fdb1-4be4-97fe-180de19df270.safetensors", "tensor_key": "kv_indices"}, "kv_last_page_len": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps64/gqa_paged_prefill_causal_h32_kv8_d64_ps64_ea6b783f-fdb1-4be4-97fe-180de19df270.safetensors", "tensor_key": "kv_last_page_len"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831843}}}, "evaluation": null} -{"definition": "gqa_paged_prefill_causal_h32_kv8_d64_ps64", "solution": null, "workload": {"uuid": "b5f2086c-8df8-4ee3-9883-b2ba121c7b69", "axes": {"len_indptr": 2, "total_q": 576, "num_kv_indices": 576, "num_pages": 1, "batch_size": 2}, "inputs": {"q": {"type": "random"}, "k_cache": {"type": "random"}, "v_cache": {"type": "random"}, "qo_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps64/gqa_paged_prefill_causal_h32_kv8_d64_ps64_b5f2086c-8df8-4ee3-9883-b2ba121c7b69.safetensors", "tensor_key": "qo_indptr"}, "kv_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps64/gqa_paged_prefill_causal_h32_kv8_d64_ps64_b5f2086c-8df8-4ee3-9883-b2ba121c7b69.safetensors", "tensor_key": "kv_indptr"}, "kv_indices": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps64/gqa_paged_prefill_causal_h32_kv8_d64_ps64_b5f2086c-8df8-4ee3-9883-b2ba121c7b69.safetensors", "tensor_key": "kv_indices"}, "kv_last_page_len": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps64/gqa_paged_prefill_causal_h32_kv8_d64_ps64_b5f2086c-8df8-4ee3-9883-b2ba121c7b69.safetensors", "tensor_key": "kv_last_page_len"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831843}}}, "evaluation": null} -{"definition": "gqa_paged_prefill_causal_h32_kv8_d64_ps64", "solution": null, "workload": {"uuid": "93cd7fca-8cef-4d07-aa0b-af998636b6bf", "axes": {"len_indptr": 2, "total_q": 2304, "num_kv_indices": 2304, "num_pages": 1, "batch_size": 2}, "inputs": {"q": {"type": "random"}, "k_cache": {"type": "random"}, "v_cache": {"type": "random"}, "qo_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps64/gqa_paged_prefill_causal_h32_kv8_d64_ps64_93cd7fca-8cef-4d07-aa0b-af998636b6bf.safetensors", "tensor_key": "qo_indptr"}, "kv_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps64/gqa_paged_prefill_causal_h32_kv8_d64_ps64_93cd7fca-8cef-4d07-aa0b-af998636b6bf.safetensors", "tensor_key": "kv_indptr"}, "kv_indices": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps64/gqa_paged_prefill_causal_h32_kv8_d64_ps64_93cd7fca-8cef-4d07-aa0b-af998636b6bf.safetensors", "tensor_key": "kv_indices"}, "kv_last_page_len": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps64/gqa_paged_prefill_causal_h32_kv8_d64_ps64_93cd7fca-8cef-4d07-aa0b-af998636b6bf.safetensors", "tensor_key": "kv_last_page_len"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831843}}}, "evaluation": null} -{"definition": "gqa_paged_prefill_causal_h32_kv8_d64_ps64", "solution": null, "workload": {"uuid": "a10cccb9-d773-4921-b85e-b9bc75a9ed39", "axes": {"len_indptr": 2, "total_q": 4, "num_kv_indices": 4, "num_pages": 1, "batch_size": 2}, "inputs": {"q": {"type": "random"}, "k_cache": {"type": "random"}, "v_cache": {"type": "random"}, "qo_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps64/gqa_paged_prefill_causal_h32_kv8_d64_ps64_a10cccb9-d773-4921-b85e-b9bc75a9ed39.safetensors", "tensor_key": "qo_indptr"}, "kv_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps64/gqa_paged_prefill_causal_h32_kv8_d64_ps64_a10cccb9-d773-4921-b85e-b9bc75a9ed39.safetensors", "tensor_key": "kv_indptr"}, "kv_indices": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps64/gqa_paged_prefill_causal_h32_kv8_d64_ps64_a10cccb9-d773-4921-b85e-b9bc75a9ed39.safetensors", "tensor_key": "kv_indices"}, "kv_last_page_len": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps64/gqa_paged_prefill_causal_h32_kv8_d64_ps64_a10cccb9-d773-4921-b85e-b9bc75a9ed39.safetensors", "tensor_key": "kv_last_page_len"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831843}}}, "evaluation": null} -{"definition": "gqa_paged_prefill_causal_h32_kv8_d64_ps64", "solution": null, "workload": {"uuid": "59f2a01c-bbcd-44b9-aac1-10878d8ec2b7", "axes": {"len_indptr": 2, "total_q": 8192, "num_kv_indices": 8192, "num_pages": 1, "batch_size": 2}, "inputs": {"q": {"type": "random"}, "k_cache": {"type": "random"}, "v_cache": {"type": "random"}, "qo_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps64/gqa_paged_prefill_causal_h32_kv8_d64_ps64_59f2a01c-bbcd-44b9-aac1-10878d8ec2b7.safetensors", "tensor_key": "qo_indptr"}, "kv_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps64/gqa_paged_prefill_causal_h32_kv8_d64_ps64_59f2a01c-bbcd-44b9-aac1-10878d8ec2b7.safetensors", "tensor_key": "kv_indptr"}, "kv_indices": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps64/gqa_paged_prefill_causal_h32_kv8_d64_ps64_59f2a01c-bbcd-44b9-aac1-10878d8ec2b7.safetensors", "tensor_key": "kv_indices"}, "kv_last_page_len": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps64/gqa_paged_prefill_causal_h32_kv8_d64_ps64_59f2a01c-bbcd-44b9-aac1-10878d8ec2b7.safetensors", "tensor_key": "kv_last_page_len"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831843}}}, "evaluation": null} -{"definition": "gqa_paged_prefill_causal_h32_kv8_d64_ps64", "solution": null, "workload": {"uuid": "58b5a85a-8410-46bc-bf0f-fbc41ad6ad84", "axes": {"len_indptr": 2, "total_q": 576, "num_kv_indices": 576, "num_pages": 1, "batch_size": 2}, "inputs": {"q": {"type": "random"}, "k_cache": {"type": "random"}, "v_cache": {"type": "random"}, "qo_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps64/gqa_paged_prefill_causal_h32_kv8_d64_ps64_58b5a85a-8410-46bc-bf0f-fbc41ad6ad84.safetensors", "tensor_key": "qo_indptr"}, "kv_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps64/gqa_paged_prefill_causal_h32_kv8_d64_ps64_58b5a85a-8410-46bc-bf0f-fbc41ad6ad84.safetensors", "tensor_key": "kv_indptr"}, "kv_indices": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps64/gqa_paged_prefill_causal_h32_kv8_d64_ps64_58b5a85a-8410-46bc-bf0f-fbc41ad6ad84.safetensors", "tensor_key": "kv_indices"}, "kv_last_page_len": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps64/gqa_paged_prefill_causal_h32_kv8_d64_ps64_58b5a85a-8410-46bc-bf0f-fbc41ad6ad84.safetensors", "tensor_key": "kv_last_page_len"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831843}}}, "evaluation": null} -{"definition": "gqa_paged_prefill_causal_h32_kv8_d64_ps64", "solution": null, "workload": {"uuid": "90a5a745-f1f4-42d6-8d63-a3ac06d27326", "axes": {"len_indptr": 2, "total_q": 2304, "num_kv_indices": 2304, "num_pages": 1, "batch_size": 2}, "inputs": {"q": {"type": "random"}, "k_cache": {"type": "random"}, "v_cache": {"type": "random"}, "qo_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps64/gqa_paged_prefill_causal_h32_kv8_d64_ps64_90a5a745-f1f4-42d6-8d63-a3ac06d27326.safetensors", "tensor_key": "qo_indptr"}, "kv_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps64/gqa_paged_prefill_causal_h32_kv8_d64_ps64_90a5a745-f1f4-42d6-8d63-a3ac06d27326.safetensors", "tensor_key": "kv_indptr"}, "kv_indices": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps64/gqa_paged_prefill_causal_h32_kv8_d64_ps64_90a5a745-f1f4-42d6-8d63-a3ac06d27326.safetensors", "tensor_key": "kv_indices"}, "kv_last_page_len": {"type": "safetensors", "path": "./blob/workloads/gqa_paged/gqa_paged_prefill_causal_h32_kv8_d64_ps64/gqa_paged_prefill_causal_h32_kv8_d64_ps64_90a5a745-f1f4-42d6-8d63-a3ac06d27326.safetensors", "tensor_key": "kv_last_page_len"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831843}}}, "evaluation": null} diff --git a/workloads/gqa_ragged/gqa_ragged_prefill_causal_h32_kv8_d64.jsonl b/workloads/gqa_ragged/gqa_ragged_prefill_causal_h32_kv8_d64.jsonl deleted file mode 100644 index d00f2e0581de2f9b94452487dd2776d6b2d078f9..0000000000000000000000000000000000000000 --- a/workloads/gqa_ragged/gqa_ragged_prefill_causal_h32_kv8_d64.jsonl +++ /dev/null @@ -1,8 +0,0 @@ -{"definition": "gqa_ragged_prefill_causal_h32_kv8_d64", "solution": null, "workload": {"uuid": "1bf692f3-4891-4382-b3a0-452aff438463", "axes": {"len_indptr": 2, "total_q": 1, "total_kv": 1}, "inputs": {"q": {"type": "random"}, "k": {"type": "random"}, "v": {"type": "random"}, "qo_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_ragged/gqa_ragged_prefill_causal_h32_kv8_d64/gqa_ragged_prefill_causal_h32_kv8_d64_1bf692f3-4891-4382-b3a0-452aff438463.safetensors", "tensor_key": "qo_indptr"}, "kv_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_ragged/gqa_ragged_prefill_causal_h32_kv8_d64/gqa_ragged_prefill_causal_h32_kv8_d64_1bf692f3-4891-4382-b3a0-452aff438463.safetensors", "tensor_key": "kv_indptr"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831843}}}, "evaluation": null} -{"definition": "gqa_ragged_prefill_causal_h32_kv8_d64", "solution": null, "workload": {"uuid": "db7cafbd-5dc4-4a6b-a073-38b22ea8c0b4", "axes": {"len_indptr": 32, "total_q": 13716, "total_kv": 13716}, "inputs": {"q": {"type": "random"}, "k": {"type": "random"}, "v": {"type": "random"}, "qo_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_ragged/gqa_ragged_prefill_causal_h32_kv8_d64/gqa_ragged_prefill_causal_h32_kv8_d64_db7cafbd-5dc4-4a6b-a073-38b22ea8c0b4.safetensors", "tensor_key": "qo_indptr"}, "kv_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_ragged/gqa_ragged_prefill_causal_h32_kv8_d64/gqa_ragged_prefill_causal_h32_kv8_d64_db7cafbd-5dc4-4a6b-a073-38b22ea8c0b4.safetensors", "tensor_key": "kv_indptr"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831843}}}, "evaluation": null} -{"definition": "gqa_ragged_prefill_causal_h32_kv8_d64", "solution": null, "workload": {"uuid": "eaf31898-d557-474a-8967-08912443e85e", "axes": {"len_indptr": 2, "total_q": 35, "total_kv": 35}, "inputs": {"q": {"type": "random"}, "k": {"type": "random"}, "v": {"type": "random"}, "qo_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_ragged/gqa_ragged_prefill_causal_h32_kv8_d64/gqa_ragged_prefill_causal_h32_kv8_d64_eaf31898-d557-474a-8967-08912443e85e.safetensors", "tensor_key": "qo_indptr"}, "kv_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_ragged/gqa_ragged_prefill_causal_h32_kv8_d64/gqa_ragged_prefill_causal_h32_kv8_d64_eaf31898-d557-474a-8967-08912443e85e.safetensors", "tensor_key": "kv_indptr"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831843}}}, "evaluation": null} -{"definition": "gqa_ragged_prefill_causal_h32_kv8_d64", "solution": null, "workload": {"uuid": "d22f94aa-fc21-4bdd-ac49-f36bae0caa5c", "axes": {"len_indptr": 13, "total_q": 5932, "total_kv": 5932}, "inputs": {"q": {"type": "random"}, "k": {"type": "random"}, "v": {"type": "random"}, "qo_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_ragged/gqa_ragged_prefill_causal_h32_kv8_d64/gqa_ragged_prefill_causal_h32_kv8_d64_d22f94aa-fc21-4bdd-ac49-f36bae0caa5c.safetensors", "tensor_key": "qo_indptr"}, "kv_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_ragged/gqa_ragged_prefill_causal_h32_kv8_d64/gqa_ragged_prefill_causal_h32_kv8_d64_d22f94aa-fc21-4bdd-ac49-f36bae0caa5c.safetensors", "tensor_key": "kv_indptr"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831843}}}, "evaluation": null} -{"definition": "gqa_ragged_prefill_causal_h32_kv8_d64", "solution": null, "workload": {"uuid": "17e47b78-d996-4fc1-af3e-56bd9c789a82", "axes": {"len_indptr": 2, "total_q": 1, "total_kv": 1}, "inputs": {"q": {"type": "random"}, "k": {"type": "random"}, "v": {"type": "random"}, "qo_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_ragged/gqa_ragged_prefill_causal_h32_kv8_d64/gqa_ragged_prefill_causal_h32_kv8_d64_17e47b78-d996-4fc1-af3e-56bd9c789a82.safetensors", "tensor_key": "qo_indptr"}, "kv_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_ragged/gqa_ragged_prefill_causal_h32_kv8_d64/gqa_ragged_prefill_causal_h32_kv8_d64_17e47b78-d996-4fc1-af3e-56bd9c789a82.safetensors", "tensor_key": "kv_indptr"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831843}}}, "evaluation": null} -{"definition": "gqa_ragged_prefill_causal_h32_kv8_d64", "solution": null, "workload": {"uuid": "7adaac59-352e-4574-a959-d3470a966090", "axes": {"len_indptr": 55, "total_q": 10930, "total_kv": 10930}, "inputs": {"q": {"type": "random"}, "k": {"type": "random"}, "v": {"type": "random"}, "qo_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_ragged/gqa_ragged_prefill_causal_h32_kv8_d64/gqa_ragged_prefill_causal_h32_kv8_d64_7adaac59-352e-4574-a959-d3470a966090.safetensors", "tensor_key": "qo_indptr"}, "kv_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_ragged/gqa_ragged_prefill_causal_h32_kv8_d64/gqa_ragged_prefill_causal_h32_kv8_d64_7adaac59-352e-4574-a959-d3470a966090.safetensors", "tensor_key": "kv_indptr"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831843}}}, "evaluation": null} -{"definition": "gqa_ragged_prefill_causal_h32_kv8_d64", "solution": null, "workload": {"uuid": "319d7c83-25e5-47c3-8de0-bf65e9e80a35", "axes": {"len_indptr": 2, "total_q": 35, "total_kv": 35}, "inputs": {"q": {"type": "random"}, "k": {"type": "random"}, "v": {"type": "random"}, "qo_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_ragged/gqa_ragged_prefill_causal_h32_kv8_d64/gqa_ragged_prefill_causal_h32_kv8_d64_319d7c83-25e5-47c3-8de0-bf65e9e80a35.safetensors", "tensor_key": "qo_indptr"}, "kv_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_ragged/gqa_ragged_prefill_causal_h32_kv8_d64/gqa_ragged_prefill_causal_h32_kv8_d64_319d7c83-25e5-47c3-8de0-bf65e9e80a35.safetensors", "tensor_key": "kv_indptr"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831843}}}, "evaluation": null} -{"definition": "gqa_ragged_prefill_causal_h32_kv8_d64", "solution": null, "workload": {"uuid": "f75c1eec-0a27-4e06-bfb6-d6238049271d", "axes": {"len_indptr": 10, "total_q": 8819, "total_kv": 8819}, "inputs": {"q": {"type": "random"}, "k": {"type": "random"}, "v": {"type": "random"}, "qo_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_ragged/gqa_ragged_prefill_causal_h32_kv8_d64/gqa_ragged_prefill_causal_h32_kv8_d64_f75c1eec-0a27-4e06-bfb6-d6238049271d.safetensors", "tensor_key": "qo_indptr"}, "kv_indptr": {"type": "safetensors", "path": "./blob/workloads/gqa_ragged/gqa_ragged_prefill_causal_h32_kv8_d64/gqa_ragged_prefill_causal_h32_kv8_d64_f75c1eec-0a27-4e06-bfb6-d6238049271d.safetensors", "tensor_key": "kv_indptr"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831843}}}, "evaluation": null} diff --git a/workloads/sampling/top_k_top_p_sampling_from_probs_v248320.jsonl b/workloads/sampling/top_k_top_p_sampling_from_probs_v248320.jsonl deleted file mode 100644 index 8e25b7de12d5e0566d4bfdec0d2c96dd70761db2..0000000000000000000000000000000000000000 --- a/workloads/sampling/top_k_top_p_sampling_from_probs_v248320.jsonl +++ /dev/null @@ -1,20 +0,0 @@ -{"definition": "top_k_top_p_sampling_from_probs_v248320", "workload": {"axes": {"batch_size": 31}, "inputs": {"probs": {"type": "random"}, "top_k": {"type": "safetensors", "path": "blob/workloads/sampling/top_k_top_p_sampling_from_probs_v248320/top_k_top_p_sampling_from_probs_v248320_21b36673-7c3f-4d58-9e9d-688e7498557a.safetensors", "tensor_key": "top_k"}, "top_p": {"type": "random"}}, "uuid": "21b36673-7c3f-4d58-9e9d-688e7498557a"}, "solution": null, "evaluation": null} -{"definition": "top_k_top_p_sampling_from_probs_v248320", "workload": {"axes": {"batch_size": 7}, "inputs": {"probs": {"type": "random"}, "top_k": {"type": "safetensors", "path": "blob/workloads/sampling/top_k_top_p_sampling_from_probs_v248320/top_k_top_p_sampling_from_probs_v248320_0d6d294f-ff2d-48a3-a241-1f0740709b61.safetensors", "tensor_key": "top_k"}, "top_p": {"type": "random"}}, "uuid": "0d6d294f-ff2d-48a3-a241-1f0740709b61"}, "solution": null, "evaluation": null} -{"definition": "top_k_top_p_sampling_from_probs_v248320", "workload": {"axes": {"batch_size": 24}, "inputs": {"probs": {"type": "random"}, "top_k": {"type": "safetensors", "path": "blob/workloads/sampling/top_k_top_p_sampling_from_probs_v248320/top_k_top_p_sampling_from_probs_v248320_975b78d0-9973-4ac9-9152-e100f8b6c2ba.safetensors", "tensor_key": "top_k"}, "top_p": {"type": "random"}}, "uuid": "975b78d0-9973-4ac9-9152-e100f8b6c2ba"}, "solution": null, "evaluation": null} -{"definition": "top_k_top_p_sampling_from_probs_v248320", "workload": {"axes": {"batch_size": 19}, "inputs": {"probs": {"type": "random"}, "top_k": {"type": "safetensors", "path": "blob/workloads/sampling/top_k_top_p_sampling_from_probs_v248320/top_k_top_p_sampling_from_probs_v248320_d6db6ce7-70e4-4775-859a-dfd6be2b6e53.safetensors", "tensor_key": "top_k"}, "top_p": {"type": "random"}}, "uuid": "d6db6ce7-70e4-4775-859a-dfd6be2b6e53"}, "solution": null, "evaluation": null} -{"definition": "top_k_top_p_sampling_from_probs_v248320", "workload": {"axes": {"batch_size": 23}, "inputs": {"probs": {"type": "random"}, "top_k": {"type": "safetensors", "path": "blob/workloads/sampling/top_k_top_p_sampling_from_probs_v248320/top_k_top_p_sampling_from_probs_v248320_1050f50f-88d7-4f23-ba26-dd654440e6ed.safetensors", "tensor_key": "top_k"}, "top_p": {"type": "random"}}, "uuid": "1050f50f-88d7-4f23-ba26-dd654440e6ed"}, "solution": null, "evaluation": null} -{"definition": "top_k_top_p_sampling_from_probs_v248320", "workload": {"axes": {"batch_size": 100}, "inputs": {"probs": {"type": "random"}, "top_k": {"type": "safetensors", "path": "blob/workloads/sampling/top_k_top_p_sampling_from_probs_v248320/top_k_top_p_sampling_from_probs_v248320_2413e30c-d007-4adc-b0ec-dfed05df6504.safetensors", "tensor_key": "top_k"}, "top_p": {"type": "random"}}, "uuid": "2413e30c-d007-4adc-b0ec-dfed05df6504"}, "solution": null, "evaluation": null} -{"definition": "top_k_top_p_sampling_from_probs_v248320", "workload": {"axes": {"batch_size": 99}, "inputs": {"probs": {"type": "random"}, "top_k": {"type": "safetensors", "path": "blob/workloads/sampling/top_k_top_p_sampling_from_probs_v248320/top_k_top_p_sampling_from_probs_v248320_59cb7944-0b44-429f-97ba-31d124e6c720.safetensors", "tensor_key": "top_k"}, "top_p": {"type": "random"}}, "uuid": "59cb7944-0b44-429f-97ba-31d124e6c720"}, "solution": null, "evaluation": null} -{"definition": "top_k_top_p_sampling_from_probs_v248320", "workload": {"axes": {"batch_size": 98}, "inputs": {"probs": {"type": "random"}, "top_k": {"type": "safetensors", "path": "blob/workloads/sampling/top_k_top_p_sampling_from_probs_v248320/top_k_top_p_sampling_from_probs_v248320_2ee765e1-0c33-4b92-a6bd-84c4ea1df0f2.safetensors", "tensor_key": "top_k"}, "top_p": {"type": "random"}}, "uuid": "2ee765e1-0c33-4b92-a6bd-84c4ea1df0f2"}, "solution": null, "evaluation": null} -{"definition": "top_k_top_p_sampling_from_probs_v248320", "workload": {"axes": {"batch_size": 97}, "inputs": {"probs": {"type": "random"}, "top_k": {"type": "safetensors", "path": "blob/workloads/sampling/top_k_top_p_sampling_from_probs_v248320/top_k_top_p_sampling_from_probs_v248320_98c33e79-e32a-4d9f-99e0-45c800dd883a.safetensors", "tensor_key": "top_k"}, "top_p": {"type": "random"}}, "uuid": "98c33e79-e32a-4d9f-99e0-45c800dd883a"}, "solution": null, "evaluation": null} -{"definition": "top_k_top_p_sampling_from_probs_v248320", "workload": {"axes": {"batch_size": 96}, "inputs": {"probs": {"type": "random"}, "top_k": {"type": "safetensors", "path": "blob/workloads/sampling/top_k_top_p_sampling_from_probs_v248320/top_k_top_p_sampling_from_probs_v248320_a2b286ac-256d-4c6c-a5b0-6076203fbefd.safetensors", "tensor_key": "top_k"}, "top_p": {"type": "random"}}, "uuid": "a2b286ac-256d-4c6c-a5b0-6076203fbefd"}, "solution": null, "evaluation": null} -{"definition": "top_k_top_p_sampling_from_probs_v248320", "workload": {"axes": {"batch_size": 2}, "inputs": {"probs": {"type": "random"}, "top_k": {"type": "safetensors", "path": "blob/workloads/sampling/top_k_top_p_sampling_from_probs_v248320/top_k_top_p_sampling_from_probs_v248320_1aa3d7a2-e19b-4898-b468-1e89f9438b1f.safetensors", "tensor_key": "top_k"}, "top_p": {"type": "random"}}, "uuid": "1aa3d7a2-e19b-4898-b468-1e89f9438b1f"}, "solution": null, "evaluation": null} -{"definition": "top_k_top_p_sampling_from_probs_v248320", "workload": {"axes": {"batch_size": 29}, "inputs": {"probs": {"type": "random"}, "top_k": {"type": "safetensors", "path": "blob/workloads/sampling/top_k_top_p_sampling_from_probs_v248320/top_k_top_p_sampling_from_probs_v248320_37e90fb8-4450-451b-8bff-5915d50e445c.safetensors", "tensor_key": "top_k"}, "top_p": {"type": "random"}}, "uuid": "37e90fb8-4450-451b-8bff-5915d50e445c"}, "solution": null, "evaluation": null} -{"definition": "top_k_top_p_sampling_from_probs_v248320", "workload": {"axes": {"batch_size": 41}, "inputs": {"probs": {"type": "random"}, "top_k": {"type": "safetensors", "path": "blob/workloads/sampling/top_k_top_p_sampling_from_probs_v248320/top_k_top_p_sampling_from_probs_v248320_5b83f0e1-4d9e-436a-b6e1-874ecdbcaf1c.safetensors", "tensor_key": "top_k"}, "top_p": {"type": "random"}}, "uuid": "5b83f0e1-4d9e-436a-b6e1-874ecdbcaf1c"}, "solution": null, "evaluation": null} -{"definition": "top_k_top_p_sampling_from_probs_v248320", "workload": {"axes": {"batch_size": 3}, "inputs": {"probs": {"type": "random"}, "top_k": {"type": "safetensors", "path": "blob/workloads/sampling/top_k_top_p_sampling_from_probs_v248320/top_k_top_p_sampling_from_probs_v248320_64a90d77-55d4-4b07-8493-baa18e016cd6.safetensors", "tensor_key": "top_k"}, "top_p": {"type": "random"}}, "uuid": "64a90d77-55d4-4b07-8493-baa18e016cd6"}, "solution": null, "evaluation": null} -{"definition": "top_k_top_p_sampling_from_probs_v248320", "workload": {"axes": {"batch_size": 28}, "inputs": {"probs": {"type": "random"}, "top_k": {"type": "safetensors", "path": "blob/workloads/sampling/top_k_top_p_sampling_from_probs_v248320/top_k_top_p_sampling_from_probs_v248320_6218b0d1-0ee3-44db-b8e7-604bc90c8335.safetensors", "tensor_key": "top_k"}, "top_p": {"type": "random"}}, "uuid": "6218b0d1-0ee3-44db-b8e7-604bc90c8335"}, "solution": null, "evaluation": null} -{"definition": "top_k_top_p_sampling_from_probs_v248320", "workload": {"axes": {"batch_size": 8}, "inputs": {"probs": {"type": "random"}, "top_k": {"type": "safetensors", "path": "blob/workloads/sampling/top_k_top_p_sampling_from_probs_v248320/top_k_top_p_sampling_from_probs_v248320_b11edc96-72c7-44fd-aec9-d53cf809382b.safetensors", "tensor_key": "top_k"}, "top_p": {"type": "random"}}, "uuid": "b11edc96-72c7-44fd-aec9-d53cf809382b"}, "solution": null, "evaluation": null} -{"definition": "top_k_top_p_sampling_from_probs_v248320", "workload": {"axes": {"batch_size": 35}, "inputs": {"probs": {"type": "random"}, "top_k": {"type": "safetensors", "path": "blob/workloads/sampling/top_k_top_p_sampling_from_probs_v248320/top_k_top_p_sampling_from_probs_v248320_ce8e6cbc-010e-49bc-9b67-8523231e2e65.safetensors", "tensor_key": "top_k"}, "top_p": {"type": "random"}}, "uuid": "ce8e6cbc-010e-49bc-9b67-8523231e2e65"}, "solution": null, "evaluation": null} -{"definition": "top_k_top_p_sampling_from_probs_v248320", "workload": {"axes": {"batch_size": 95}, "inputs": {"probs": {"type": "random"}, "top_k": {"type": "safetensors", "path": "blob/workloads/sampling/top_k_top_p_sampling_from_probs_v248320/top_k_top_p_sampling_from_probs_v248320_9def7b85-9198-4acd-a3d7-e79c608321e4.safetensors", "tensor_key": "top_k"}, "top_p": {"type": "random"}}, "uuid": "9def7b85-9198-4acd-a3d7-e79c608321e4"}, "solution": null, "evaluation": null} -{"definition": "top_k_top_p_sampling_from_probs_v248320", "workload": {"axes": {"batch_size": 6}, "inputs": {"probs": {"type": "random"}, "top_k": {"type": "safetensors", "path": "blob/workloads/sampling/top_k_top_p_sampling_from_probs_v248320/top_k_top_p_sampling_from_probs_v248320_415daee6-51bb-4248-bc39-ffc9d51f4a59.safetensors", "tensor_key": "top_k"}, "top_p": {"type": "random"}}, "uuid": "415daee6-51bb-4248-bc39-ffc9d51f4a59"}, "solution": null, "evaluation": null} -{"definition": "top_k_top_p_sampling_from_probs_v248320", "workload": {"axes": {"batch_size": 25}, "inputs": {"probs": {"type": "random"}, "top_k": {"type": "safetensors", "path": "blob/workloads/sampling/top_k_top_p_sampling_from_probs_v248320/top_k_top_p_sampling_from_probs_v248320_11ba194f-96e1-485e-bdeb-55aac8f944db.safetensors", "tensor_key": "top_k"}, "top_p": {"type": "random"}}, "uuid": "11ba194f-96e1-485e-bdeb-55aac8f944db"}, "solution": null, "evaluation": null}