From 8ff812fbb5fed2e79bf903feb4c5e4af9b4a7a7a Mon Sep 17 00:00:00 2001 From: Donato Capitella Date: Mon, 9 Feb 2026 13:30:26 +0000 Subject: [PATCH] updated benchmarks --- ...5__rocm6_4_4__hblt0__longctx32768__rpc.log | 8 + ...-00001-of-00005__rocm6_4_4__hblt0__rpc.log | 8 + ...of-00005__rocm6_4_4__longctx32768__rpc.log | 8 + ...Q4_K_XL-00001-of-00005__rocm6_4_4__rpc.log | 8 + ...m7-nightlies__hblt0__longctx32768__rpc.log | 8 + ...-of-00005__rocm7-nightlies__hblt0__rpc.log | 8 + ...05__rocm7-nightlies__longctx32768__rpc.log | 8 + ...L-00001-of-00005__rocm7-nightlies__rpc.log | 8 + ...5__rocm7_1_1__hblt0__longctx32768__rpc.log | 0 ...-00001-of-00005__rocm7_1_1__hblt0__rpc.log | 0 ...of-00005__rocm7_1_1__longctx32768__rpc.log | 0 ...Q4_K_XL-00001-of-00005__rocm7_1_1__rpc.log | 0 ...0005__vulkan_amdvlk__longctx32768__rpc.log | 19 + ..._XL-00001-of-00005__vulkan_amdvlk__rpc.log | 19 + ...-00005__vulkan_radv__longctx32768__rpc.log | 1 + ..._K_XL-00001-of-00005__vulkan_radv__rpc.log | 1 + ...4__rocm6_4_4__hblt0__longctx32768__rpc.log | 8 + ...-00001-of-00004__rocm6_4_4__hblt0__rpc.log | 8 + ...of-00004__rocm6_4_4__longctx32768__rpc.log | 8 + ...Q6_K_XL-00001-of-00004__rocm6_4_4__rpc.log | 8 + ...m7-nightlies__hblt0__longctx32768__rpc.log | 8 + ...-of-00004__rocm7-nightlies__hblt0__rpc.log | 8 + ...04__rocm7-nightlies__longctx32768__rpc.log | 8 + ...L-00001-of-00004__rocm7-nightlies__rpc.log | 8 + ...4__rocm7_1_1__hblt0__longctx32768__rpc.log | 0 ...-00001-of-00004__rocm7_1_1__hblt0__rpc.log | 0 ...of-00004__rocm7_1_1__longctx32768__rpc.log | 0 ...Q6_K_XL-00001-of-00004__rocm7_1_1__rpc.log | 0 ...0004__vulkan_amdvlk__longctx32768__rpc.log | 3 + ..._XL-00001-of-00004__vulkan_amdvlk__rpc.log | 19 + ...-00004__vulkan_radv__longctx32768__rpc.log | 19 + ..._K_XL-00001-of-00004__vulkan_radv__rpc.log | 19 + ...05__rocm-7_2__hblt0__longctx32768__rpc.log | 8 + ...L-00001-of-00005__rocm-7_2__hblt0__rpc.log | 8 + ...-of-00005__rocm-7_2__longctx32768__rpc.log | 8 + ...-Q4_K_XL-00001-of-00005__rocm-7_2__rpc.log | 8 + ...5__rocm6_4_4__hblt0__longctx32768__rpc.log | 10 +- ...-00001-of-00005__rocm6_4_4__hblt0__rpc.log | 10 +- ...of-00005__rocm6_4_4__longctx32768__rpc.log | 10 +- ...Q4_K_XL-00001-of-00005__rocm6_4_4__rpc.log | 10 +- ...m7-nightlies__hblt0__longctx32768__rpc.log | 10 +- ...-of-00005__rocm7-nightlies__hblt0__rpc.log | 10 +- ...05__rocm7-nightlies__longctx32768__rpc.log | 10 +- ...L-00001-of-00005__rocm7-nightlies__rpc.log | 10 +- ...0005__vulkan_amdvlk__longctx32768__rpc.log | 18 +- ..._XL-00001-of-00005__vulkan_amdvlk__rpc.log | 32 +- ...-00005__vulkan_radv__longctx32768__rpc.log | 2 +- ..._K_XL-00001-of-00005__vulkan_radv__rpc.log | 2 +- ...04__rocm-7_2__hblt0__longctx32768__rpc.log | 8 + ...L-00001-of-00004__rocm-7_2__hblt0__rpc.log | 8 + ...-of-00004__rocm-7_2__longctx32768__rpc.log | 8 + ...-Q6_K_XL-00001-of-00004__rocm-7_2__rpc.log | 8 + ...4__rocm6_4_4__hblt0__longctx32768__rpc.log | 10 +- ...-00001-of-00004__rocm6_4_4__hblt0__rpc.log | 10 +- ...of-00004__rocm6_4_4__longctx32768__rpc.log | 10 +- ...Q6_K_XL-00001-of-00004__rocm6_4_4__rpc.log | 10 +- ...m7-nightlies__hblt0__longctx32768__rpc.log | 10 +- ...-of-00004__rocm7-nightlies__hblt0__rpc.log | 10 +- ...04__rocm7-nightlies__longctx32768__rpc.log | 10 +- ...L-00001-of-00004__rocm7-nightlies__rpc.log | 10 +- ...0004__vulkan_amdvlk__longctx32768__rpc.log | 7 +- ..._XL-00001-of-00004__vulkan_amdvlk__rpc.log | 32 +- ...-00004__vulkan_radv__longctx32768__rpc.log | 6 +- ..._K_XL-00001-of-00004__vulkan_radv__rpc.log | 32 +- ...Q4_K_XL-00001-of-00002__rocm-7_2__fa1.log} | 0 ...of-00002__rocm-7_2__fa1__longctx32768.log} | 0 ...-00001-of-00002__rocm-7_2__hblt0__fa1.log} | 0 ...2__rocm-7_2__hblt0__fa1__longctx32768.log} | 0 ...Q4_K_XL-00001-of-00002__rocm-7_2__fa1.log} | 0 ...of-00002__rocm-7_2__fa1__longctx32768.log} | 0 ...-00001-of-00002__rocm-7_2__hblt0__fa1.log} | 0 ...2__rocm-7_2__hblt0__fa1__longctx32768.log} | 0 ...Q6_K_XL-00001-of-00003__rocm-7_2__fa1.log} | 0 ...of-00003__rocm-7_2__fa1__longctx32768.log} | 0 ...-00001-of-00003__rocm-7_2__hblt0__fa1.log} | 0 ...3__rocm-7_2__hblt0__fa1__longctx32768.log} | 0 ...Q8_K_XL-00001-of-00002__rocm-7_2__fa1.log} | 0 ...of-00002__rocm-7_2__fa1__longctx32768.log} | 0 ...-00001-of-00002__rocm-7_2__hblt0__fa1.log} | 0 ...2__rocm-7_2__hblt0__fa1__longctx32768.log} | 0 ...14B-Instruct-2512-BF16__rocm-7_2__fa1.log} | 0 ...512-BF16__rocm-7_2__fa1__longctx32768.log} | 0 ...truct-2512-BF16__rocm-7_2__hblt0__fa1.log} | 0 ...6__rocm-7_2__hblt0__fa1__longctx32768.log} | 0 ...ano-30B-A3B-UD-Q8_K_XL__rocm-7_2__fa1.log} | 0 ...-Q8_K_XL__rocm-7_2__fa1__longctx32768.log} | 0 ...-A3B-UD-Q8_K_XL__rocm-7_2__hblt0__fa1.log} | 0 ...L__rocm-7_2__hblt0__fa1__longctx32768.log} | 0 ...Q3_K_XL-00001-of-00003__rocm-7_2__fa1.log} | 0 ...of-00003__rocm-7_2__fa1__longctx32768.log} | 0 ...-00001-of-00003__rocm-7_2__hblt0__fa1.log} | 0 ...3__rocm-7_2__hblt0__fa1__longctx32768.log} | 0 ...3B-BF16-00001-of-00002__rocm-7_2__fa1.log} | 0 ...of-00002__rocm-7_2__fa1__longctx32768.log} | 0 ...-00001-of-00002__rocm-7_2__hblt0__fa1.log} | 0 ...2__rocm-7_2__hblt0__fa1__longctx32768.log} | 0 ...struct-2507-UD-Q6_K_XL__rocm-7_2__fa1.log} | 0 ...-Q6_K_XL__rocm-7_2__fa1__longctx32768.log} | 0 ...2507-UD-Q6_K_XL__rocm-7_2__hblt0__fa1.log} | 0 ...L__rocm-7_2__hblt0__fa1__longctx32768.log} | 0 ...0B-A3B-Instruct-Q4_K_M__rocm-7_2__fa1.log} | 0 ...t-Q4_K_M__rocm-7_2__fa1__longctx32768.log} | 0 ...Instruct-Q4_K_M__rocm-7_2__hblt0__fa1.log} | 0 ...M__rocm-7_2__hblt0__fa1__longctx32768.log} | 0 ...Q8_K_XL-00001-of-00002__rocm-7_2__fa1.log} | 0 ...of-00002__rocm-7_2__fa1__longctx32768.log} | 0 ...-00001-of-00002__rocm-7_2__hblt0__fa1.log} | 0 ...2__rocm-7_2__hblt0__fa1__longctx32768.log} | 0 ...ma-3-12b-it-UD-Q8_K_XL__rocm-7_2__fa1.log} | 0 ...-Q8_K_XL__rocm-7_2__fa1__longctx32768.log} | 0 ...b-it-UD-Q8_K_XL__rocm-7_2__hblt0__fa1.log} | 0 ...L__rocm-7_2__hblt0__fa1__longctx32768.log} | 0 ...it-BF16-00001-of-00002__rocm-7_2__fa1.log} | 0 ...of-00002__rocm-7_2__fa1__longctx32768.log} | 0 ...-00001-of-00002__rocm-7_2__hblt0__fa1.log} | 0 ...2__rocm-7_2__hblt0__fa1__longctx32768.log} | 0 ...> gemma-3-4b-it-Q3_K_S__rocm-7_2__fa1.log} | 0 ...t-Q3_K_S__rocm-7_2__fa1__longctx32768.log} | 0 ...-3-4b-it-Q3_K_S__rocm-7_2__hblt0__fa1.log} | 0 ...S__rocm-7_2__hblt0__fa1__longctx32768.log} | 0 ...b-mxfp4-00001-of-00003__rocm-7_2__fa1.log} | 0 ...of-00003__rocm-7_2__fa1__longctx32768.log} | 0 ...-00001-of-00003__rocm-7_2__hblt0__fa1.log} | 0 ...3__rocm-7_2__hblt0__fa1__longctx32768.log} | 0 ...g => gpt-oss-20b-mxfp4__rocm-7_2__fa1.log} | 0 ...0b-mxfp4__rocm-7_2__fa1__longctx32768.log} | 0 ...t-oss-20b-mxfp4__rocm-7_2__hblt0__fa1.log} | 0 ...4__rocm-7_2__hblt0__fa1__longctx32768.log} | 0 ...log => llama-2-7b.Q4_0__rocm-7_2__fa1.log} | 0 ...-7b.Q4_0__rocm-7_2__fa1__longctx32768.log} | 0 ...llama-2-7b.Q4_0__rocm-7_2__hblt0__fa1.log} | 0 ...0__rocm-7_2__hblt0__fa1__longctx32768.log} | 0 ...Q4_K_XL-00001-of-00002__rocm-7_2__fa1.log} | 0 ...of-00002__rocm-7_2__fa1__longctx32768.log} | 0 ...-00001-of-00002__rocm-7_2__hblt0__fa1.log} | 0 ...2__rocm-7_2__hblt0__fa1__longctx32768.log} | 0 ...sh-BF16-00001-of-00002__rocm-7_2__fa1.log} | 0 ...of-00002__rocm-7_2__fa1__longctx32768.log} | 0 ...-00001-of-00002__rocm-7_2__hblt0__fa1.log} | 0 ...2__rocm-7_2__hblt0__fa1__longctx32768.log} | 0 ...M-4.7-Flash-UD-Q8_K_XL__rocm-7_2__fa1.log} | 0 ...-Q8_K_XL__rocm-7_2__fa1__longctx32768.log} | 0 ...lash-UD-Q8_K_XL__rocm-7_2__hblt0__fa1.log} | 0 ...L__rocm-7_2__hblt0__fa1__longctx32768.log} | 0 ...Q8_K_XL-00001-of-00002__rocm-7_2__fa1.log} | 0 ...of-00002__rocm-7_2__fa1__longctx32768.log} | 0 ...-00001-of-00002__rocm-7_2__hblt0__fa1.log} | 0 ...2__rocm-7_2__hblt0__fa1__longctx32768.log} | 0 ...14B-Instruct-2512-BF16__rocm-7_2__fa1.log} | 0 ...512-BF16__rocm-7_2__fa1__longctx32768.log} | 0 ...truct-2512-BF16__rocm-7_2__hblt0__fa1.log} | 0 ...6__rocm-7_2__hblt0__fa1__longctx32768.log} | 0 ...ano-30B-A3B-UD-Q8_K_XL__rocm-7_2__fa1.log} | 0 ...-Q8_K_XL__rocm-7_2__fa1__longctx32768.log} | 0 ...-A3B-UD-Q8_K_XL__rocm-7_2__hblt0__fa1.log} | 0 ...L__rocm-7_2__hblt0__fa1__longctx32768.log} | 0 ...Q3_K_XL-00001-of-00003__rocm-7_2__fa1.log} | 0 ...of-00003__rocm-7_2__fa1__longctx32768.log} | 0 ...-00001-of-00003__rocm-7_2__hblt0__fa1.log} | 0 ...3__rocm-7_2__hblt0__fa1__longctx32768.log} | 0 ...3B-BF16-00001-of-00002__rocm-7_2__fa1.log} | 0 ...of-00002__rocm-7_2__fa1__longctx32768.log} | 0 ...-00001-of-00002__rocm-7_2__hblt0__fa1.log} | 0 ...2__rocm-7_2__hblt0__fa1__longctx32768.log} | 0 ...struct-2507-UD-Q6_K_XL__rocm-7_2__fa1.log} | 0 ...-Q6_K_XL__rocm-7_2__fa1__longctx32768.log} | 0 ...2507-UD-Q6_K_XL__rocm-7_2__hblt0__fa1.log} | 0 ...L__rocm-7_2__hblt0__fa1__longctx32768.log} | 0 ...0B-A3B-Instruct-Q4_K_M__rocm-7_2__fa1.log} | 0 ...t-Q4_K_M__rocm-7_2__fa1__longctx32768.log} | 0 ...Instruct-Q4_K_M__rocm-7_2__hblt0__fa1.log} | 0 ...M__rocm-7_2__hblt0__fa1__longctx32768.log} | 0 ...Q8_K_XL-00001-of-00002__rocm-7_2__fa1.log} | 0 ...of-00002__rocm-7_2__fa1__longctx32768.log} | 0 ...-00001-of-00002__rocm-7_2__hblt0__fa1.log} | 0 ...2__rocm-7_2__hblt0__fa1__longctx32768.log} | 0 ...ma-3-12b-it-UD-Q8_K_XL__rocm-7_2__fa1.log} | 0 ...-Q8_K_XL__rocm-7_2__fa1__longctx32768.log} | 0 ...b-it-UD-Q8_K_XL__rocm-7_2__hblt0__fa1.log} | 0 ...L__rocm-7_2__hblt0__fa1__longctx32768.log} | 0 ...it-BF16-00001-of-00002__rocm-7_2__fa1.log} | 0 ...of-00002__rocm-7_2__fa1__longctx32768.log} | 0 ...-00001-of-00002__rocm-7_2__hblt0__fa1.log} | 0 ...2__rocm-7_2__hblt0__fa1__longctx32768.log} | 0 ...> gemma-3-4b-it-Q3_K_S__rocm-7_2__fa1.log} | 0 ...t-Q3_K_S__rocm-7_2__fa1__longctx32768.log} | 0 ...-3-4b-it-Q3_K_S__rocm-7_2__hblt0__fa1.log} | 0 ...S__rocm-7_2__hblt0__fa1__longctx32768.log} | 0 ...b-mxfp4-00001-of-00003__rocm-7_2__fa1.log} | 0 ...of-00003__rocm-7_2__fa1__longctx32768.log} | 0 ...-00001-of-00003__rocm-7_2__hblt0__fa1.log} | 0 ...3__rocm-7_2__hblt0__fa1__longctx32768.log} | 0 ...g => gpt-oss-20b-mxfp4__rocm-7_2__fa1.log} | 0 ...0b-mxfp4__rocm-7_2__fa1__longctx32768.log} | 0 ...t-oss-20b-mxfp4__rocm-7_2__hblt0__fa1.log} | 0 ...4__rocm-7_2__hblt0__fa1__longctx32768.log} | 0 ...log => llama-2-7b.Q4_0__rocm-7_2__fa1.log} | 0 ...-7b.Q4_0__rocm-7_2__fa1__longctx32768.log} | 0 ...llama-2-7b.Q4_0__rocm-7_2__hblt0__fa1.log} | 0 ...0__rocm-7_2__hblt0__fa1__longctx32768.log} | 0 benchmark/run_benchmarks.sh | 3 +- benchmark/run_rpc_benchmarks.sh | 8 +- docs/index.html | 5 +- docs/results.json | 2422 ++++++++--------- 204 files changed, 1645 insertions(+), 1376 deletions(-) create mode 100644 benchmark/results-rpc/09-01-2026/GLM-4.6-UD-Q4_K_XL-00001-of-00005__rocm6_4_4__hblt0__longctx32768__rpc.log create mode 100644 benchmark/results-rpc/09-01-2026/GLM-4.6-UD-Q4_K_XL-00001-of-00005__rocm6_4_4__hblt0__rpc.log create mode 100644 benchmark/results-rpc/09-01-2026/GLM-4.6-UD-Q4_K_XL-00001-of-00005__rocm6_4_4__longctx32768__rpc.log create mode 100644 benchmark/results-rpc/09-01-2026/GLM-4.6-UD-Q4_K_XL-00001-of-00005__rocm6_4_4__rpc.log create mode 100644 benchmark/results-rpc/09-01-2026/GLM-4.6-UD-Q4_K_XL-00001-of-00005__rocm7-nightlies__hblt0__longctx32768__rpc.log create mode 100644 benchmark/results-rpc/09-01-2026/GLM-4.6-UD-Q4_K_XL-00001-of-00005__rocm7-nightlies__hblt0__rpc.log create mode 100644 benchmark/results-rpc/09-01-2026/GLM-4.6-UD-Q4_K_XL-00001-of-00005__rocm7-nightlies__longctx32768__rpc.log create mode 100644 benchmark/results-rpc/09-01-2026/GLM-4.6-UD-Q4_K_XL-00001-of-00005__rocm7-nightlies__rpc.log rename benchmark/results-rpc/{ => 09-01-2026}/GLM-4.6-UD-Q4_K_XL-00001-of-00005__rocm7_1_1__hblt0__longctx32768__rpc.log (100%) rename benchmark/results-rpc/{ => 09-01-2026}/GLM-4.6-UD-Q4_K_XL-00001-of-00005__rocm7_1_1__hblt0__rpc.log (100%) rename benchmark/results-rpc/{ => 09-01-2026}/GLM-4.6-UD-Q4_K_XL-00001-of-00005__rocm7_1_1__longctx32768__rpc.log (100%) rename benchmark/results-rpc/{ => 09-01-2026}/GLM-4.6-UD-Q4_K_XL-00001-of-00005__rocm7_1_1__rpc.log (100%) create mode 100644 benchmark/results-rpc/09-01-2026/GLM-4.6-UD-Q4_K_XL-00001-of-00005__vulkan_amdvlk__longctx32768__rpc.log create mode 100644 benchmark/results-rpc/09-01-2026/GLM-4.6-UD-Q4_K_XL-00001-of-00005__vulkan_amdvlk__rpc.log create mode 100644 benchmark/results-rpc/09-01-2026/GLM-4.6-UD-Q4_K_XL-00001-of-00005__vulkan_radv__longctx32768__rpc.log create mode 100644 benchmark/results-rpc/09-01-2026/GLM-4.6-UD-Q4_K_XL-00001-of-00005__vulkan_radv__rpc.log create mode 100644 benchmark/results-rpc/09-01-2026/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__rocm6_4_4__hblt0__longctx32768__rpc.log create mode 100644 benchmark/results-rpc/09-01-2026/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__rocm6_4_4__hblt0__rpc.log create mode 100644 benchmark/results-rpc/09-01-2026/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__rocm6_4_4__longctx32768__rpc.log create mode 100644 benchmark/results-rpc/09-01-2026/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__rocm6_4_4__rpc.log create mode 100644 benchmark/results-rpc/09-01-2026/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__rocm7-nightlies__hblt0__longctx32768__rpc.log create mode 100644 benchmark/results-rpc/09-01-2026/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__rocm7-nightlies__hblt0__rpc.log create mode 100644 benchmark/results-rpc/09-01-2026/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__rocm7-nightlies__longctx32768__rpc.log create mode 100644 benchmark/results-rpc/09-01-2026/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__rocm7-nightlies__rpc.log rename benchmark/results-rpc/{ => 09-01-2026}/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__rocm7_1_1__hblt0__longctx32768__rpc.log (100%) rename benchmark/results-rpc/{ => 09-01-2026}/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__rocm7_1_1__hblt0__rpc.log (100%) rename benchmark/results-rpc/{ => 09-01-2026}/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__rocm7_1_1__longctx32768__rpc.log (100%) rename benchmark/results-rpc/{ => 09-01-2026}/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__rocm7_1_1__rpc.log (100%) create mode 100644 benchmark/results-rpc/09-01-2026/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__vulkan_amdvlk__longctx32768__rpc.log create mode 100644 benchmark/results-rpc/09-01-2026/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__vulkan_amdvlk__rpc.log create mode 100644 benchmark/results-rpc/09-01-2026/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__vulkan_radv__longctx32768__rpc.log create mode 100644 benchmark/results-rpc/09-01-2026/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__vulkan_radv__rpc.log create mode 100644 benchmark/results-rpc/GLM-4.6-UD-Q4_K_XL-00001-of-00005__rocm-7_2__hblt0__longctx32768__rpc.log create mode 100644 benchmark/results-rpc/GLM-4.6-UD-Q4_K_XL-00001-of-00005__rocm-7_2__hblt0__rpc.log create mode 100644 benchmark/results-rpc/GLM-4.6-UD-Q4_K_XL-00001-of-00005__rocm-7_2__longctx32768__rpc.log create mode 100644 benchmark/results-rpc/GLM-4.6-UD-Q4_K_XL-00001-of-00005__rocm-7_2__rpc.log create mode 100644 benchmark/results-rpc/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__rocm-7_2__hblt0__longctx32768__rpc.log create mode 100644 benchmark/results-rpc/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__rocm-7_2__hblt0__rpc.log create mode 100644 benchmark/results-rpc/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__rocm-7_2__longctx32768__rpc.log create mode 100644 benchmark/results-rpc/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__rocm-7_2__rpc.log rename benchmark/results/09-01-2026/{Devstral-2-123B-Instruct-2512-UD-Q4_K_XL-00001-of-00002__rocm-7.2__fa1.log => Devstral-2-123B-Instruct-2512-UD-Q4_K_XL-00001-of-00002__rocm-7_2__fa1.log} (100%) rename benchmark/results/09-01-2026/{Devstral-2-123B-Instruct-2512-UD-Q4_K_XL-00001-of-00002__rocm-7.2__fa1__longctx32768.log => Devstral-2-123B-Instruct-2512-UD-Q4_K_XL-00001-of-00002__rocm-7_2__fa1__longctx32768.log} (100%) rename benchmark/results/09-01-2026/{Devstral-2-123B-Instruct-2512-UD-Q4_K_XL-00001-of-00002__rocm-7.2__hblt0__fa1.log => Devstral-2-123B-Instruct-2512-UD-Q4_K_XL-00001-of-00002__rocm-7_2__hblt0__fa1.log} (100%) rename benchmark/results/09-01-2026/{Devstral-2-123B-Instruct-2512-UD-Q4_K_XL-00001-of-00002__rocm-7.2__hblt0__fa1__longctx32768.log => Devstral-2-123B-Instruct-2512-UD-Q4_K_XL-00001-of-00002__rocm-7_2__hblt0__fa1__longctx32768.log} (100%) rename benchmark/results/09-01-2026/{GLM-4.5-Air-UD-Q4_K_XL-00001-of-00002__rocm-7.2__fa1.log => GLM-4.5-Air-UD-Q4_K_XL-00001-of-00002__rocm-7_2__fa1.log} (100%) rename benchmark/results/09-01-2026/{GLM-4.5-Air-UD-Q4_K_XL-00001-of-00002__rocm-7.2__fa1__longctx32768.log => GLM-4.5-Air-UD-Q4_K_XL-00001-of-00002__rocm-7_2__fa1__longctx32768.log} (100%) rename benchmark/results/09-01-2026/{GLM-4.5-Air-UD-Q4_K_XL-00001-of-00002__rocm-7.2__hblt0__fa1.log => GLM-4.5-Air-UD-Q4_K_XL-00001-of-00002__rocm-7_2__hblt0__fa1.log} (100%) rename benchmark/results/09-01-2026/{GLM-4.5-Air-UD-Q4_K_XL-00001-of-00002__rocm-7.2__hblt0__fa1__longctx32768.log => GLM-4.5-Air-UD-Q4_K_XL-00001-of-00002__rocm-7_2__hblt0__fa1__longctx32768.log} (100%) rename benchmark/results/09-01-2026/{GLM-4.5-Air-UD-Q6_K_XL-00001-of-00003__rocm-7.2__fa1.log => GLM-4.5-Air-UD-Q6_K_XL-00001-of-00003__rocm-7_2__fa1.log} (100%) rename benchmark/results/09-01-2026/{GLM-4.5-Air-UD-Q6_K_XL-00001-of-00003__rocm-7.2__fa1__longctx32768.log => GLM-4.5-Air-UD-Q6_K_XL-00001-of-00003__rocm-7_2__fa1__longctx32768.log} (100%) rename benchmark/results/09-01-2026/{GLM-4.5-Air-UD-Q6_K_XL-00001-of-00003__rocm-7.2__hblt0__fa1.log => GLM-4.5-Air-UD-Q6_K_XL-00001-of-00003__rocm-7_2__hblt0__fa1.log} (100%) rename benchmark/results/09-01-2026/{GLM-4.5-Air-UD-Q6_K_XL-00001-of-00003__rocm-7.2__hblt0__fa1__longctx32768.log => GLM-4.5-Air-UD-Q6_K_XL-00001-of-00003__rocm-7_2__hblt0__fa1__longctx32768.log} (100%) rename benchmark/results/09-01-2026/{Llama-3.3-70B-Instruct-UD-Q8_K_XL-00001-of-00002__rocm-7.2__fa1.log => Llama-3.3-70B-Instruct-UD-Q8_K_XL-00001-of-00002__rocm-7_2__fa1.log} (100%) rename benchmark/results/09-01-2026/{Llama-3.3-70B-Instruct-UD-Q8_K_XL-00001-of-00002__rocm-7.2__fa1__longctx32768.log => Llama-3.3-70B-Instruct-UD-Q8_K_XL-00001-of-00002__rocm-7_2__fa1__longctx32768.log} (100%) rename benchmark/results/09-01-2026/{Llama-3.3-70B-Instruct-UD-Q8_K_XL-00001-of-00002__rocm-7.2__hblt0__fa1.log => Llama-3.3-70B-Instruct-UD-Q8_K_XL-00001-of-00002__rocm-7_2__hblt0__fa1.log} (100%) rename benchmark/results/09-01-2026/{Llama-3.3-70B-Instruct-UD-Q8_K_XL-00001-of-00002__rocm-7.2__hblt0__fa1__longctx32768.log => Llama-3.3-70B-Instruct-UD-Q8_K_XL-00001-of-00002__rocm-7_2__hblt0__fa1__longctx32768.log} (100%) rename benchmark/results/09-01-2026/{Ministral-3-14B-Instruct-2512-BF16__rocm-7.2__fa1.log => Ministral-3-14B-Instruct-2512-BF16__rocm-7_2__fa1.log} (100%) rename benchmark/results/09-01-2026/{Ministral-3-14B-Instruct-2512-BF16__rocm-7.2__fa1__longctx32768.log => Ministral-3-14B-Instruct-2512-BF16__rocm-7_2__fa1__longctx32768.log} (100%) rename benchmark/results/09-01-2026/{Ministral-3-14B-Instruct-2512-BF16__rocm-7.2__hblt0__fa1.log => Ministral-3-14B-Instruct-2512-BF16__rocm-7_2__hblt0__fa1.log} (100%) rename benchmark/results/09-01-2026/{Ministral-3-14B-Instruct-2512-BF16__rocm-7.2__hblt0__fa1__longctx32768.log => Ministral-3-14B-Instruct-2512-BF16__rocm-7_2__hblt0__fa1__longctx32768.log} (100%) rename benchmark/results/09-01-2026/{Nemotron-3-Nano-30B-A3B-UD-Q8_K_XL__rocm-7.2__fa1.log => Nemotron-3-Nano-30B-A3B-UD-Q8_K_XL__rocm-7_2__fa1.log} (100%) rename benchmark/results/09-01-2026/{Nemotron-3-Nano-30B-A3B-UD-Q8_K_XL__rocm-7.2__fa1__longctx32768.log => Nemotron-3-Nano-30B-A3B-UD-Q8_K_XL__rocm-7_2__fa1__longctx32768.log} (100%) rename benchmark/results/09-01-2026/{Nemotron-3-Nano-30B-A3B-UD-Q8_K_XL__rocm-7.2__hblt0__fa1.log => Nemotron-3-Nano-30B-A3B-UD-Q8_K_XL__rocm-7_2__hblt0__fa1.log} (100%) rename benchmark/results/09-01-2026/{Nemotron-3-Nano-30B-A3B-UD-Q8_K_XL__rocm-7.2__hblt0__fa1__longctx32768.log => Nemotron-3-Nano-30B-A3B-UD-Q8_K_XL__rocm-7_2__hblt0__fa1__longctx32768.log} (100%) rename benchmark/results/09-01-2026/{Qwen3-235B-A22B-Instruct-2507-UD-Q3_K_XL-00001-of-00003__rocm-7.2__fa1.log => Qwen3-235B-A22B-Instruct-2507-UD-Q3_K_XL-00001-of-00003__rocm-7_2__fa1.log} (100%) rename benchmark/results/09-01-2026/{Qwen3-235B-A22B-Instruct-2507-UD-Q3_K_XL-00001-of-00003__rocm-7.2__fa1__longctx32768.log => Qwen3-235B-A22B-Instruct-2507-UD-Q3_K_XL-00001-of-00003__rocm-7_2__fa1__longctx32768.log} (100%) rename benchmark/results/09-01-2026/{Qwen3-235B-A22B-Instruct-2507-UD-Q3_K_XL-00001-of-00003__rocm-7.2__hblt0__fa1.log => Qwen3-235B-A22B-Instruct-2507-UD-Q3_K_XL-00001-of-00003__rocm-7_2__hblt0__fa1.log} (100%) rename benchmark/results/09-01-2026/{Qwen3-235B-A22B-Instruct-2507-UD-Q3_K_XL-00001-of-00003__rocm-7.2__hblt0__fa1__longctx32768.log => Qwen3-235B-A22B-Instruct-2507-UD-Q3_K_XL-00001-of-00003__rocm-7_2__hblt0__fa1__longctx32768.log} (100%) rename benchmark/results/09-01-2026/{Qwen3-30B-A3B-BF16-00001-of-00002__rocm-7.2__fa1.log => Qwen3-30B-A3B-BF16-00001-of-00002__rocm-7_2__fa1.log} (100%) rename benchmark/results/09-01-2026/{Qwen3-30B-A3B-BF16-00001-of-00002__rocm-7.2__fa1__longctx32768.log => Qwen3-30B-A3B-BF16-00001-of-00002__rocm-7_2__fa1__longctx32768.log} (100%) rename benchmark/results/09-01-2026/{Qwen3-30B-A3B-BF16-00001-of-00002__rocm-7.2__hblt0__fa1.log => Qwen3-30B-A3B-BF16-00001-of-00002__rocm-7_2__hblt0__fa1.log} (100%) rename benchmark/results/09-01-2026/{Qwen3-30B-A3B-BF16-00001-of-00002__rocm-7.2__hblt0__fa1__longctx32768.log => Qwen3-30B-A3B-BF16-00001-of-00002__rocm-7_2__hblt0__fa1__longctx32768.log} (100%) rename benchmark/results/09-01-2026/{Qwen3-30B-A3B-Instruct-2507-UD-Q6_K_XL__rocm-7.2__fa1.log => Qwen3-30B-A3B-Instruct-2507-UD-Q6_K_XL__rocm-7_2__fa1.log} (100%) rename benchmark/results/09-01-2026/{Qwen3-30B-A3B-Instruct-2507-UD-Q6_K_XL__rocm-7.2__fa1__longctx32768.log => Qwen3-30B-A3B-Instruct-2507-UD-Q6_K_XL__rocm-7_2__fa1__longctx32768.log} (100%) rename benchmark/results/09-01-2026/{Qwen3-30B-A3B-Instruct-2507-UD-Q6_K_XL__rocm-7.2__hblt0__fa1.log => Qwen3-30B-A3B-Instruct-2507-UD-Q6_K_XL__rocm-7_2__hblt0__fa1.log} (100%) rename benchmark/results/09-01-2026/{Qwen3-30B-A3B-Instruct-2507-UD-Q6_K_XL__rocm-7.2__hblt0__fa1__longctx32768.log => Qwen3-30B-A3B-Instruct-2507-UD-Q6_K_XL__rocm-7_2__hblt0__fa1__longctx32768.log} (100%) rename benchmark/results/09-01-2026/{Qwen3-Coder-30B-A3B-Instruct-Q4_K_M__rocm-7.2__fa1.log => Qwen3-Coder-30B-A3B-Instruct-Q4_K_M__rocm-7_2__fa1.log} (100%) rename benchmark/results/09-01-2026/{Qwen3-Coder-30B-A3B-Instruct-Q4_K_M__rocm-7.2__fa1__longctx32768.log => Qwen3-Coder-30B-A3B-Instruct-Q4_K_M__rocm-7_2__fa1__longctx32768.log} (100%) rename benchmark/results/09-01-2026/{Qwen3-Coder-30B-A3B-Instruct-Q4_K_M__rocm-7.2__hblt0__fa1.log => Qwen3-Coder-30B-A3B-Instruct-Q4_K_M__rocm-7_2__hblt0__fa1.log} (100%) rename benchmark/results/09-01-2026/{Qwen3-Coder-30B-A3B-Instruct-Q4_K_M__rocm-7.2__hblt0__fa1__longctx32768.log => Qwen3-Coder-30B-A3B-Instruct-Q4_K_M__rocm-7_2__hblt0__fa1__longctx32768.log} (100%) rename benchmark/results/09-01-2026/{Qwen3-Next-80B-A3B-Instruct-UD-Q8_K_XL-00001-of-00002__rocm-7.2__fa1.log => Qwen3-Next-80B-A3B-Instruct-UD-Q8_K_XL-00001-of-00002__rocm-7_2__fa1.log} (100%) rename benchmark/results/09-01-2026/{Qwen3-Next-80B-A3B-Instruct-UD-Q8_K_XL-00001-of-00002__rocm-7.2__fa1__longctx32768.log => Qwen3-Next-80B-A3B-Instruct-UD-Q8_K_XL-00001-of-00002__rocm-7_2__fa1__longctx32768.log} (100%) rename benchmark/results/09-01-2026/{Qwen3-Next-80B-A3B-Instruct-UD-Q8_K_XL-00001-of-00002__rocm-7.2__hblt0__fa1.log => Qwen3-Next-80B-A3B-Instruct-UD-Q8_K_XL-00001-of-00002__rocm-7_2__hblt0__fa1.log} (100%) rename benchmark/results/09-01-2026/{Qwen3-Next-80B-A3B-Instruct-UD-Q8_K_XL-00001-of-00002__rocm-7.2__hblt0__fa1__longctx32768.log => Qwen3-Next-80B-A3B-Instruct-UD-Q8_K_XL-00001-of-00002__rocm-7_2__hblt0__fa1__longctx32768.log} (100%) rename benchmark/results/09-01-2026/{gemma-3-12b-it-UD-Q8_K_XL__rocm-7.2__fa1.log => gemma-3-12b-it-UD-Q8_K_XL__rocm-7_2__fa1.log} (100%) rename benchmark/results/09-01-2026/{gemma-3-12b-it-UD-Q8_K_XL__rocm-7.2__fa1__longctx32768.log => gemma-3-12b-it-UD-Q8_K_XL__rocm-7_2__fa1__longctx32768.log} (100%) rename benchmark/results/09-01-2026/{gemma-3-12b-it-UD-Q8_K_XL__rocm-7.2__hblt0__fa1.log => gemma-3-12b-it-UD-Q8_K_XL__rocm-7_2__hblt0__fa1.log} (100%) rename benchmark/results/09-01-2026/{gemma-3-12b-it-UD-Q8_K_XL__rocm-7.2__hblt0__fa1__longctx32768.log => gemma-3-12b-it-UD-Q8_K_XL__rocm-7_2__hblt0__fa1__longctx32768.log} (100%) rename benchmark/results/09-01-2026/{gemma-3-27b-it-BF16-00001-of-00002__rocm-7.2__fa1.log => gemma-3-27b-it-BF16-00001-of-00002__rocm-7_2__fa1.log} (100%) rename benchmark/results/09-01-2026/{gemma-3-27b-it-BF16-00001-of-00002__rocm-7.2__fa1__longctx32768.log => gemma-3-27b-it-BF16-00001-of-00002__rocm-7_2__fa1__longctx32768.log} (100%) rename benchmark/results/09-01-2026/{gemma-3-27b-it-BF16-00001-of-00002__rocm-7.2__hblt0__fa1.log => gemma-3-27b-it-BF16-00001-of-00002__rocm-7_2__hblt0__fa1.log} (100%) rename benchmark/results/09-01-2026/{gemma-3-27b-it-BF16-00001-of-00002__rocm-7.2__hblt0__fa1__longctx32768.log => gemma-3-27b-it-BF16-00001-of-00002__rocm-7_2__hblt0__fa1__longctx32768.log} (100%) rename benchmark/results/09-01-2026/{gemma-3-4b-it-Q3_K_S__rocm-7.2__fa1.log => gemma-3-4b-it-Q3_K_S__rocm-7_2__fa1.log} (100%) rename benchmark/results/09-01-2026/{gemma-3-4b-it-Q3_K_S__rocm-7.2__fa1__longctx32768.log => gemma-3-4b-it-Q3_K_S__rocm-7_2__fa1__longctx32768.log} (100%) rename benchmark/results/09-01-2026/{gemma-3-4b-it-Q3_K_S__rocm-7.2__hblt0__fa1.log => gemma-3-4b-it-Q3_K_S__rocm-7_2__hblt0__fa1.log} (100%) rename benchmark/results/09-01-2026/{gemma-3-4b-it-Q3_K_S__rocm-7.2__hblt0__fa1__longctx32768.log => gemma-3-4b-it-Q3_K_S__rocm-7_2__hblt0__fa1__longctx32768.log} (100%) rename benchmark/results/09-01-2026/{gpt-oss-120b-mxfp4-00001-of-00003__rocm-7.2__fa1.log => gpt-oss-120b-mxfp4-00001-of-00003__rocm-7_2__fa1.log} (100%) rename benchmark/results/09-01-2026/{gpt-oss-120b-mxfp4-00001-of-00003__rocm-7.2__fa1__longctx32768.log => gpt-oss-120b-mxfp4-00001-of-00003__rocm-7_2__fa1__longctx32768.log} (100%) rename benchmark/results/09-01-2026/{gpt-oss-120b-mxfp4-00001-of-00003__rocm-7.2__hblt0__fa1.log => gpt-oss-120b-mxfp4-00001-of-00003__rocm-7_2__hblt0__fa1.log} (100%) rename benchmark/results/09-01-2026/{gpt-oss-120b-mxfp4-00001-of-00003__rocm-7.2__hblt0__fa1__longctx32768.log => gpt-oss-120b-mxfp4-00001-of-00003__rocm-7_2__hblt0__fa1__longctx32768.log} (100%) rename benchmark/results/09-01-2026/{gpt-oss-20b-mxfp4__rocm-7.2__fa1.log => gpt-oss-20b-mxfp4__rocm-7_2__fa1.log} (100%) rename benchmark/results/09-01-2026/{gpt-oss-20b-mxfp4__rocm-7.2__fa1__longctx32768.log => gpt-oss-20b-mxfp4__rocm-7_2__fa1__longctx32768.log} (100%) rename benchmark/results/09-01-2026/{gpt-oss-20b-mxfp4__rocm-7.2__hblt0__fa1.log => gpt-oss-20b-mxfp4__rocm-7_2__hblt0__fa1.log} (100%) rename benchmark/results/09-01-2026/{gpt-oss-20b-mxfp4__rocm-7.2__hblt0__fa1__longctx32768.log => gpt-oss-20b-mxfp4__rocm-7_2__hblt0__fa1__longctx32768.log} (100%) rename benchmark/results/09-01-2026/{llama-2-7b.Q4_0__rocm-7.2__fa1.log => llama-2-7b.Q4_0__rocm-7_2__fa1.log} (100%) rename benchmark/results/09-01-2026/{llama-2-7b.Q4_0__rocm-7.2__fa1__longctx32768.log => llama-2-7b.Q4_0__rocm-7_2__fa1__longctx32768.log} (100%) rename benchmark/results/09-01-2026/{llama-2-7b.Q4_0__rocm-7.2__hblt0__fa1.log => llama-2-7b.Q4_0__rocm-7_2__hblt0__fa1.log} (100%) rename benchmark/results/09-01-2026/{llama-2-7b.Q4_0__rocm-7.2__hblt0__fa1__longctx32768.log => llama-2-7b.Q4_0__rocm-7_2__hblt0__fa1__longctx32768.log} (100%) rename benchmark/results/{Devstral-2-123B-Instruct-2512-UD-Q4_K_XL-00001-of-00002__rocm-7.2__fa1.log => Devstral-2-123B-Instruct-2512-UD-Q4_K_XL-00001-of-00002__rocm-7_2__fa1.log} (100%) rename benchmark/results/{Devstral-2-123B-Instruct-2512-UD-Q4_K_XL-00001-of-00002__rocm-7.2__fa1__longctx32768.log => Devstral-2-123B-Instruct-2512-UD-Q4_K_XL-00001-of-00002__rocm-7_2__fa1__longctx32768.log} (100%) rename benchmark/results/{Devstral-2-123B-Instruct-2512-UD-Q4_K_XL-00001-of-00002__rocm-7.2__hblt0__fa1.log => Devstral-2-123B-Instruct-2512-UD-Q4_K_XL-00001-of-00002__rocm-7_2__hblt0__fa1.log} (100%) rename benchmark/results/{Devstral-2-123B-Instruct-2512-UD-Q4_K_XL-00001-of-00002__rocm-7.2__hblt0__fa1__longctx32768.log => Devstral-2-123B-Instruct-2512-UD-Q4_K_XL-00001-of-00002__rocm-7_2__hblt0__fa1__longctx32768.log} (100%) rename benchmark/results/{GLM-4.7-Flash-BF16-00001-of-00002__rocm-7.2__fa1.log => GLM-4.7-Flash-BF16-00001-of-00002__rocm-7_2__fa1.log} (100%) rename benchmark/results/{GLM-4.7-Flash-BF16-00001-of-00002__rocm-7.2__fa1__longctx32768.log => GLM-4.7-Flash-BF16-00001-of-00002__rocm-7_2__fa1__longctx32768.log} (100%) rename benchmark/results/{GLM-4.7-Flash-BF16-00001-of-00002__rocm-7.2__hblt0__fa1.log => GLM-4.7-Flash-BF16-00001-of-00002__rocm-7_2__hblt0__fa1.log} (100%) rename benchmark/results/{GLM-4.7-Flash-BF16-00001-of-00002__rocm-7.2__hblt0__fa1__longctx32768.log => GLM-4.7-Flash-BF16-00001-of-00002__rocm-7_2__hblt0__fa1__longctx32768.log} (100%) rename benchmark/results/{GLM-4.7-Flash-UD-Q8_K_XL__rocm-7.2__fa1.log => GLM-4.7-Flash-UD-Q8_K_XL__rocm-7_2__fa1.log} (100%) rename benchmark/results/{GLM-4.7-Flash-UD-Q8_K_XL__rocm-7.2__fa1__longctx32768.log => GLM-4.7-Flash-UD-Q8_K_XL__rocm-7_2__fa1__longctx32768.log} (100%) rename benchmark/results/{GLM-4.7-Flash-UD-Q8_K_XL__rocm-7.2__hblt0__fa1.log => GLM-4.7-Flash-UD-Q8_K_XL__rocm-7_2__hblt0__fa1.log} (100%) rename benchmark/results/{GLM-4.7-Flash-UD-Q8_K_XL__rocm-7.2__hblt0__fa1__longctx32768.log => GLM-4.7-Flash-UD-Q8_K_XL__rocm-7_2__hblt0__fa1__longctx32768.log} (100%) rename benchmark/results/{Llama-3.3-70B-Instruct-UD-Q8_K_XL-00001-of-00002__rocm-7.2__fa1.log => Llama-3.3-70B-Instruct-UD-Q8_K_XL-00001-of-00002__rocm-7_2__fa1.log} (100%) rename benchmark/results/{Llama-3.3-70B-Instruct-UD-Q8_K_XL-00001-of-00002__rocm-7.2__fa1__longctx32768.log => Llama-3.3-70B-Instruct-UD-Q8_K_XL-00001-of-00002__rocm-7_2__fa1__longctx32768.log} (100%) rename benchmark/results/{Llama-3.3-70B-Instruct-UD-Q8_K_XL-00001-of-00002__rocm-7.2__hblt0__fa1.log => Llama-3.3-70B-Instruct-UD-Q8_K_XL-00001-of-00002__rocm-7_2__hblt0__fa1.log} (100%) rename benchmark/results/{Llama-3.3-70B-Instruct-UD-Q8_K_XL-00001-of-00002__rocm-7.2__hblt0__fa1__longctx32768.log => Llama-3.3-70B-Instruct-UD-Q8_K_XL-00001-of-00002__rocm-7_2__hblt0__fa1__longctx32768.log} (100%) rename benchmark/results/{Ministral-3-14B-Instruct-2512-BF16__rocm-7.2__fa1.log => Ministral-3-14B-Instruct-2512-BF16__rocm-7_2__fa1.log} (100%) rename benchmark/results/{Ministral-3-14B-Instruct-2512-BF16__rocm-7.2__fa1__longctx32768.log => Ministral-3-14B-Instruct-2512-BF16__rocm-7_2__fa1__longctx32768.log} (100%) rename benchmark/results/{Ministral-3-14B-Instruct-2512-BF16__rocm-7.2__hblt0__fa1.log => Ministral-3-14B-Instruct-2512-BF16__rocm-7_2__hblt0__fa1.log} (100%) rename benchmark/results/{Ministral-3-14B-Instruct-2512-BF16__rocm-7.2__hblt0__fa1__longctx32768.log => Ministral-3-14B-Instruct-2512-BF16__rocm-7_2__hblt0__fa1__longctx32768.log} (100%) rename benchmark/results/{Nemotron-3-Nano-30B-A3B-UD-Q8_K_XL__rocm-7.2__fa1.log => Nemotron-3-Nano-30B-A3B-UD-Q8_K_XL__rocm-7_2__fa1.log} (100%) rename benchmark/results/{Nemotron-3-Nano-30B-A3B-UD-Q8_K_XL__rocm-7.2__fa1__longctx32768.log => Nemotron-3-Nano-30B-A3B-UD-Q8_K_XL__rocm-7_2__fa1__longctx32768.log} (100%) rename benchmark/results/{Nemotron-3-Nano-30B-A3B-UD-Q8_K_XL__rocm-7.2__hblt0__fa1.log => Nemotron-3-Nano-30B-A3B-UD-Q8_K_XL__rocm-7_2__hblt0__fa1.log} (100%) rename benchmark/results/{Nemotron-3-Nano-30B-A3B-UD-Q8_K_XL__rocm-7.2__hblt0__fa1__longctx32768.log => Nemotron-3-Nano-30B-A3B-UD-Q8_K_XL__rocm-7_2__hblt0__fa1__longctx32768.log} (100%) rename benchmark/results/{Qwen3-235B-A22B-Instruct-2507-UD-Q3_K_XL-00001-of-00003__rocm-7.2__fa1.log => Qwen3-235B-A22B-Instruct-2507-UD-Q3_K_XL-00001-of-00003__rocm-7_2__fa1.log} (100%) rename benchmark/results/{Qwen3-235B-A22B-Instruct-2507-UD-Q3_K_XL-00001-of-00003__rocm-7.2__fa1__longctx32768.log => Qwen3-235B-A22B-Instruct-2507-UD-Q3_K_XL-00001-of-00003__rocm-7_2__fa1__longctx32768.log} (100%) rename benchmark/results/{Qwen3-235B-A22B-Instruct-2507-UD-Q3_K_XL-00001-of-00003__rocm-7.2__hblt0__fa1.log => Qwen3-235B-A22B-Instruct-2507-UD-Q3_K_XL-00001-of-00003__rocm-7_2__hblt0__fa1.log} (100%) rename benchmark/results/{Qwen3-235B-A22B-Instruct-2507-UD-Q3_K_XL-00001-of-00003__rocm-7.2__hblt0__fa1__longctx32768.log => Qwen3-235B-A22B-Instruct-2507-UD-Q3_K_XL-00001-of-00003__rocm-7_2__hblt0__fa1__longctx32768.log} (100%) rename benchmark/results/{Qwen3-30B-A3B-BF16-00001-of-00002__rocm-7.2__fa1.log => Qwen3-30B-A3B-BF16-00001-of-00002__rocm-7_2__fa1.log} (100%) rename benchmark/results/{Qwen3-30B-A3B-BF16-00001-of-00002__rocm-7.2__fa1__longctx32768.log => Qwen3-30B-A3B-BF16-00001-of-00002__rocm-7_2__fa1__longctx32768.log} (100%) rename benchmark/results/{Qwen3-30B-A3B-BF16-00001-of-00002__rocm-7.2__hblt0__fa1.log => Qwen3-30B-A3B-BF16-00001-of-00002__rocm-7_2__hblt0__fa1.log} (100%) rename benchmark/results/{Qwen3-30B-A3B-BF16-00001-of-00002__rocm-7.2__hblt0__fa1__longctx32768.log => Qwen3-30B-A3B-BF16-00001-of-00002__rocm-7_2__hblt0__fa1__longctx32768.log} (100%) rename benchmark/results/{Qwen3-30B-A3B-Instruct-2507-UD-Q6_K_XL__rocm-7.2__fa1.log => Qwen3-30B-A3B-Instruct-2507-UD-Q6_K_XL__rocm-7_2__fa1.log} (100%) rename benchmark/results/{Qwen3-30B-A3B-Instruct-2507-UD-Q6_K_XL__rocm-7.2__fa1__longctx32768.log => Qwen3-30B-A3B-Instruct-2507-UD-Q6_K_XL__rocm-7_2__fa1__longctx32768.log} (100%) rename benchmark/results/{Qwen3-30B-A3B-Instruct-2507-UD-Q6_K_XL__rocm-7.2__hblt0__fa1.log => Qwen3-30B-A3B-Instruct-2507-UD-Q6_K_XL__rocm-7_2__hblt0__fa1.log} (100%) rename benchmark/results/{Qwen3-30B-A3B-Instruct-2507-UD-Q6_K_XL__rocm-7.2__hblt0__fa1__longctx32768.log => Qwen3-30B-A3B-Instruct-2507-UD-Q6_K_XL__rocm-7_2__hblt0__fa1__longctx32768.log} (100%) rename benchmark/results/{Qwen3-Coder-30B-A3B-Instruct-Q4_K_M__rocm-7.2__fa1.log => Qwen3-Coder-30B-A3B-Instruct-Q4_K_M__rocm-7_2__fa1.log} (100%) rename benchmark/results/{Qwen3-Coder-30B-A3B-Instruct-Q4_K_M__rocm-7.2__fa1__longctx32768.log => Qwen3-Coder-30B-A3B-Instruct-Q4_K_M__rocm-7_2__fa1__longctx32768.log} (100%) rename benchmark/results/{Qwen3-Coder-30B-A3B-Instruct-Q4_K_M__rocm-7.2__hblt0__fa1.log => Qwen3-Coder-30B-A3B-Instruct-Q4_K_M__rocm-7_2__hblt0__fa1.log} (100%) rename benchmark/results/{Qwen3-Coder-30B-A3B-Instruct-Q4_K_M__rocm-7.2__hblt0__fa1__longctx32768.log => Qwen3-Coder-30B-A3B-Instruct-Q4_K_M__rocm-7_2__hblt0__fa1__longctx32768.log} (100%) rename benchmark/results/{Qwen3-Next-80B-A3B-Instruct-UD-Q8_K_XL-00001-of-00002__rocm-7.2__fa1.log => Qwen3-Next-80B-A3B-Instruct-UD-Q8_K_XL-00001-of-00002__rocm-7_2__fa1.log} (100%) rename benchmark/results/{Qwen3-Next-80B-A3B-Instruct-UD-Q8_K_XL-00001-of-00002__rocm-7.2__fa1__longctx32768.log => Qwen3-Next-80B-A3B-Instruct-UD-Q8_K_XL-00001-of-00002__rocm-7_2__fa1__longctx32768.log} (100%) rename benchmark/results/{Qwen3-Next-80B-A3B-Instruct-UD-Q8_K_XL-00001-of-00002__rocm-7.2__hblt0__fa1.log => Qwen3-Next-80B-A3B-Instruct-UD-Q8_K_XL-00001-of-00002__rocm-7_2__hblt0__fa1.log} (100%) rename benchmark/results/{Qwen3-Next-80B-A3B-Instruct-UD-Q8_K_XL-00001-of-00002__rocm-7.2__hblt0__fa1__longctx32768.log => Qwen3-Next-80B-A3B-Instruct-UD-Q8_K_XL-00001-of-00002__rocm-7_2__hblt0__fa1__longctx32768.log} (100%) rename benchmark/results/{gemma-3-12b-it-UD-Q8_K_XL__rocm-7.2__fa1.log => gemma-3-12b-it-UD-Q8_K_XL__rocm-7_2__fa1.log} (100%) rename benchmark/results/{gemma-3-12b-it-UD-Q8_K_XL__rocm-7.2__fa1__longctx32768.log => gemma-3-12b-it-UD-Q8_K_XL__rocm-7_2__fa1__longctx32768.log} (100%) rename benchmark/results/{gemma-3-12b-it-UD-Q8_K_XL__rocm-7.2__hblt0__fa1.log => gemma-3-12b-it-UD-Q8_K_XL__rocm-7_2__hblt0__fa1.log} (100%) rename benchmark/results/{gemma-3-12b-it-UD-Q8_K_XL__rocm-7.2__hblt0__fa1__longctx32768.log => gemma-3-12b-it-UD-Q8_K_XL__rocm-7_2__hblt0__fa1__longctx32768.log} (100%) rename benchmark/results/{gemma-3-27b-it-BF16-00001-of-00002__rocm-7.2__fa1.log => gemma-3-27b-it-BF16-00001-of-00002__rocm-7_2__fa1.log} (100%) rename benchmark/results/{gemma-3-27b-it-BF16-00001-of-00002__rocm-7.2__fa1__longctx32768.log => gemma-3-27b-it-BF16-00001-of-00002__rocm-7_2__fa1__longctx32768.log} (100%) rename benchmark/results/{gemma-3-27b-it-BF16-00001-of-00002__rocm-7.2__hblt0__fa1.log => gemma-3-27b-it-BF16-00001-of-00002__rocm-7_2__hblt0__fa1.log} (100%) rename benchmark/results/{gemma-3-27b-it-BF16-00001-of-00002__rocm-7.2__hblt0__fa1__longctx32768.log => gemma-3-27b-it-BF16-00001-of-00002__rocm-7_2__hblt0__fa1__longctx32768.log} (100%) rename benchmark/results/{gemma-3-4b-it-Q3_K_S__rocm-7.2__fa1.log => gemma-3-4b-it-Q3_K_S__rocm-7_2__fa1.log} (100%) rename benchmark/results/{gemma-3-4b-it-Q3_K_S__rocm-7.2__fa1__longctx32768.log => gemma-3-4b-it-Q3_K_S__rocm-7_2__fa1__longctx32768.log} (100%) rename benchmark/results/{gemma-3-4b-it-Q3_K_S__rocm-7.2__hblt0__fa1.log => gemma-3-4b-it-Q3_K_S__rocm-7_2__hblt0__fa1.log} (100%) rename benchmark/results/{gemma-3-4b-it-Q3_K_S__rocm-7.2__hblt0__fa1__longctx32768.log => gemma-3-4b-it-Q3_K_S__rocm-7_2__hblt0__fa1__longctx32768.log} (100%) rename benchmark/results/{gpt-oss-120b-mxfp4-00001-of-00003__rocm-7.2__fa1.log => gpt-oss-120b-mxfp4-00001-of-00003__rocm-7_2__fa1.log} (100%) rename benchmark/results/{gpt-oss-120b-mxfp4-00001-of-00003__rocm-7.2__fa1__longctx32768.log => gpt-oss-120b-mxfp4-00001-of-00003__rocm-7_2__fa1__longctx32768.log} (100%) rename benchmark/results/{gpt-oss-120b-mxfp4-00001-of-00003__rocm-7.2__hblt0__fa1.log => gpt-oss-120b-mxfp4-00001-of-00003__rocm-7_2__hblt0__fa1.log} (100%) rename benchmark/results/{gpt-oss-120b-mxfp4-00001-of-00003__rocm-7.2__hblt0__fa1__longctx32768.log => gpt-oss-120b-mxfp4-00001-of-00003__rocm-7_2__hblt0__fa1__longctx32768.log} (100%) rename benchmark/results/{gpt-oss-20b-mxfp4__rocm-7.2__fa1.log => gpt-oss-20b-mxfp4__rocm-7_2__fa1.log} (100%) rename benchmark/results/{gpt-oss-20b-mxfp4__rocm-7.2__fa1__longctx32768.log => gpt-oss-20b-mxfp4__rocm-7_2__fa1__longctx32768.log} (100%) rename benchmark/results/{gpt-oss-20b-mxfp4__rocm-7.2__hblt0__fa1.log => gpt-oss-20b-mxfp4__rocm-7_2__hblt0__fa1.log} (100%) rename benchmark/results/{gpt-oss-20b-mxfp4__rocm-7.2__hblt0__fa1__longctx32768.log => gpt-oss-20b-mxfp4__rocm-7_2__hblt0__fa1__longctx32768.log} (100%) rename benchmark/results/{llama-2-7b.Q4_0__rocm-7.2__fa1.log => llama-2-7b.Q4_0__rocm-7_2__fa1.log} (100%) rename benchmark/results/{llama-2-7b.Q4_0__rocm-7.2__fa1__longctx32768.log => llama-2-7b.Q4_0__rocm-7_2__fa1__longctx32768.log} (100%) rename benchmark/results/{llama-2-7b.Q4_0__rocm-7.2__hblt0__fa1.log => llama-2-7b.Q4_0__rocm-7_2__hblt0__fa1.log} (100%) rename benchmark/results/{llama-2-7b.Q4_0__rocm-7.2__hblt0__fa1__longctx32768.log => llama-2-7b.Q4_0__rocm-7_2__hblt0__fa1__longctx32768.log} (100%) diff --git a/benchmark/results-rpc/09-01-2026/GLM-4.6-UD-Q4_K_XL-00001-of-00005__rocm6_4_4__hblt0__longctx32768__rpc.log b/benchmark/results-rpc/09-01-2026/GLM-4.6-UD-Q4_K_XL-00001-of-00005__rocm6_4_4__hblt0__longctx32768__rpc.log new file mode 100644 index 0000000..019b7bd --- /dev/null +++ b/benchmark/results-rpc/09-01-2026/GLM-4.6-UD-Q4_K_XL-00001-of-00005__rocm6_4_4__hblt0__longctx32768__rpc.log @@ -0,0 +1,8 @@ +ggml_cuda_init: found 1 ROCm devices: + Device 0: Radeon 8060S Graphics, gfx1151 (0x1151), VMM: no, Wave Size: 32 +| model | size | params | backend | ngl | n_ubatch | fa | mmap | test | t/s | +| ------------------------------ | ---------: | ---------: | ---------- | --: | -------: | -: | ---: | --------------: | -------------------: | +| glm4moe 355B.A32B Q4_K - Medium | 189.69 GiB | 356.79 B | ROCm,RPC | 99 | 2048 | 1 | 0 | pp2048 @ d32768 | 18.41 ± 0.00 | +| glm4moe 355B.A32B Q4_K - Medium | 189.69 GiB | 356.79 B | ROCm,RPC | 99 | 2048 | 1 | 0 | tg32 @ d32768 | 4.12 ± 0.00 | + +build: 2656c0d26 (7693) diff --git a/benchmark/results-rpc/09-01-2026/GLM-4.6-UD-Q4_K_XL-00001-of-00005__rocm6_4_4__hblt0__rpc.log b/benchmark/results-rpc/09-01-2026/GLM-4.6-UD-Q4_K_XL-00001-of-00005__rocm6_4_4__hblt0__rpc.log new file mode 100644 index 0000000..2339073 --- /dev/null +++ b/benchmark/results-rpc/09-01-2026/GLM-4.6-UD-Q4_K_XL-00001-of-00005__rocm6_4_4__hblt0__rpc.log @@ -0,0 +1,8 @@ +ggml_cuda_init: found 1 ROCm devices: + Device 0: Radeon 8060S Graphics, gfx1151 (0x1151), VMM: no, Wave Size: 32 +| model | size | params | backend | ngl | fa | mmap | test | t/s | +| ------------------------------ | ---------: | ---------: | ---------- | --: | -: | ---: | --------------: | -------------------: | +| glm4moe 355B.A32B Q4_K - Medium | 189.69 GiB | 356.79 B | ROCm,RPC | 99 | 1 | 0 | pp512 | 101.82 ± 0.34 | +| glm4moe 355B.A32B Q4_K - Medium | 189.69 GiB | 356.79 B | ROCm,RPC | 99 | 1 | 0 | tg128 | 8.71 ± 0.00 | + +build: 2656c0d26 (7693) diff --git a/benchmark/results-rpc/09-01-2026/GLM-4.6-UD-Q4_K_XL-00001-of-00005__rocm6_4_4__longctx32768__rpc.log b/benchmark/results-rpc/09-01-2026/GLM-4.6-UD-Q4_K_XL-00001-of-00005__rocm6_4_4__longctx32768__rpc.log new file mode 100644 index 0000000..7d68b86 --- /dev/null +++ b/benchmark/results-rpc/09-01-2026/GLM-4.6-UD-Q4_K_XL-00001-of-00005__rocm6_4_4__longctx32768__rpc.log @@ -0,0 +1,8 @@ +ggml_cuda_init: found 1 ROCm devices: + Device 0: Radeon 8060S Graphics, gfx1151 (0x1151), VMM: no, Wave Size: 32 +| model | size | params | backend | ngl | n_ubatch | fa | mmap | test | t/s | +| ------------------------------ | ---------: | ---------: | ---------- | --: | -------: | -: | ---: | --------------: | -------------------: | +| glm4moe 355B.A32B Q4_K - Medium | 189.69 GiB | 356.79 B | ROCm,RPC | 99 | 2048 | 1 | 0 | pp2048 @ d32768 | 17.93 ± 0.00 | +| glm4moe 355B.A32B Q4_K - Medium | 189.69 GiB | 356.79 B | ROCm,RPC | 99 | 2048 | 1 | 0 | tg32 @ d32768 | 4.13 ± 0.00 | + +build: 2656c0d26 (7693) diff --git a/benchmark/results-rpc/09-01-2026/GLM-4.6-UD-Q4_K_XL-00001-of-00005__rocm6_4_4__rpc.log b/benchmark/results-rpc/09-01-2026/GLM-4.6-UD-Q4_K_XL-00001-of-00005__rocm6_4_4__rpc.log new file mode 100644 index 0000000..09254e0 --- /dev/null +++ b/benchmark/results-rpc/09-01-2026/GLM-4.6-UD-Q4_K_XL-00001-of-00005__rocm6_4_4__rpc.log @@ -0,0 +1,8 @@ +ggml_cuda_init: found 1 ROCm devices: + Device 0: Radeon 8060S Graphics, gfx1151 (0x1151), VMM: no, Wave Size: 32 +| model | size | params | backend | ngl | fa | mmap | test | t/s | +| ------------------------------ | ---------: | ---------: | ---------- | --: | -: | ---: | --------------: | -------------------: | +| glm4moe 355B.A32B Q4_K - Medium | 189.69 GiB | 356.79 B | ROCm,RPC | 99 | 1 | 0 | pp512 | 95.55 ± 0.26 | +| glm4moe 355B.A32B Q4_K - Medium | 189.69 GiB | 356.79 B | ROCm,RPC | 99 | 1 | 0 | tg128 | 8.78 ± 0.00 | + +build: 2656c0d26 (7693) diff --git a/benchmark/results-rpc/09-01-2026/GLM-4.6-UD-Q4_K_XL-00001-of-00005__rocm7-nightlies__hblt0__longctx32768__rpc.log b/benchmark/results-rpc/09-01-2026/GLM-4.6-UD-Q4_K_XL-00001-of-00005__rocm7-nightlies__hblt0__longctx32768__rpc.log new file mode 100644 index 0000000..d7afa2a --- /dev/null +++ b/benchmark/results-rpc/09-01-2026/GLM-4.6-UD-Q4_K_XL-00001-of-00005__rocm7-nightlies__hblt0__longctx32768__rpc.log @@ -0,0 +1,8 @@ +ggml_cuda_init: found 1 ROCm devices: + Device 0: Radeon 8060S Graphics, gfx1151 (0x1151), VMM: no, Wave Size: 32 +| model | size | params | backend | ngl | n_ubatch | fa | mmap | test | t/s | +| ------------------------------ | ---------: | ---------: | ---------- | --: | -------: | -: | ---: | --------------: | -------------------: | +| glm4moe 355B.A32B Q4_K - Medium | 189.69 GiB | 356.79 B | ROCm,RPC | 99 | 2048 | 1 | 0 | pp2048 @ d32768 | 18.59 ± 0.00 | +| glm4moe 355B.A32B Q4_K - Medium | 189.69 GiB | 356.79 B | ROCm,RPC | 99 | 2048 | 1 | 0 | tg32 @ d32768 | 3.63 ± 0.00 | + +build: 2656c0d26 (7693) diff --git a/benchmark/results-rpc/09-01-2026/GLM-4.6-UD-Q4_K_XL-00001-of-00005__rocm7-nightlies__hblt0__rpc.log b/benchmark/results-rpc/09-01-2026/GLM-4.6-UD-Q4_K_XL-00001-of-00005__rocm7-nightlies__hblt0__rpc.log new file mode 100644 index 0000000..9e62d86 --- /dev/null +++ b/benchmark/results-rpc/09-01-2026/GLM-4.6-UD-Q4_K_XL-00001-of-00005__rocm7-nightlies__hblt0__rpc.log @@ -0,0 +1,8 @@ +ggml_cuda_init: found 1 ROCm devices: + Device 0: Radeon 8060S Graphics, gfx1151 (0x1151), VMM: no, Wave Size: 32 +| model | size | params | backend | ngl | fa | mmap | test | t/s | +| ------------------------------ | ---------: | ---------: | ---------- | --: | -: | ---: | --------------: | -------------------: | +| glm4moe 355B.A32B Q4_K - Medium | 189.69 GiB | 356.79 B | ROCm,RPC | 99 | 1 | 0 | pp512 | 103.11 ± 0.08 | +| glm4moe 355B.A32B Q4_K - Medium | 189.69 GiB | 356.79 B | ROCm,RPC | 99 | 1 | 0 | tg128 | 9.11 ± 0.03 | + +build: 2656c0d26 (7693) diff --git a/benchmark/results-rpc/09-01-2026/GLM-4.6-UD-Q4_K_XL-00001-of-00005__rocm7-nightlies__longctx32768__rpc.log b/benchmark/results-rpc/09-01-2026/GLM-4.6-UD-Q4_K_XL-00001-of-00005__rocm7-nightlies__longctx32768__rpc.log new file mode 100644 index 0000000..e5e59f5 --- /dev/null +++ b/benchmark/results-rpc/09-01-2026/GLM-4.6-UD-Q4_K_XL-00001-of-00005__rocm7-nightlies__longctx32768__rpc.log @@ -0,0 +1,8 @@ +ggml_cuda_init: found 1 ROCm devices: + Device 0: Radeon 8060S Graphics, gfx1151 (0x1151), VMM: no, Wave Size: 32 +| model | size | params | backend | ngl | n_ubatch | fa | mmap | test | t/s | +| ------------------------------ | ---------: | ---------: | ---------- | --: | -------: | -: | ---: | --------------: | -------------------: | +| glm4moe 355B.A32B Q4_K - Medium | 189.69 GiB | 356.79 B | ROCm,RPC | 99 | 2048 | 1 | 0 | pp2048 @ d32768 | 18.03 ± 0.00 | +| glm4moe 355B.A32B Q4_K - Medium | 189.69 GiB | 356.79 B | ROCm,RPC | 99 | 2048 | 1 | 0 | tg32 @ d32768 | 3.63 ± 0.00 | + +build: 2656c0d26 (7693) diff --git a/benchmark/results-rpc/09-01-2026/GLM-4.6-UD-Q4_K_XL-00001-of-00005__rocm7-nightlies__rpc.log b/benchmark/results-rpc/09-01-2026/GLM-4.6-UD-Q4_K_XL-00001-of-00005__rocm7-nightlies__rpc.log new file mode 100644 index 0000000..0abf42d --- /dev/null +++ b/benchmark/results-rpc/09-01-2026/GLM-4.6-UD-Q4_K_XL-00001-of-00005__rocm7-nightlies__rpc.log @@ -0,0 +1,8 @@ +ggml_cuda_init: found 1 ROCm devices: + Device 0: Radeon 8060S Graphics, gfx1151 (0x1151), VMM: no, Wave Size: 32 +| model | size | params | backend | ngl | fa | mmap | test | t/s | +| ------------------------------ | ---------: | ---------: | ---------- | --: | -: | ---: | --------------: | -------------------: | +| glm4moe 355B.A32B Q4_K - Medium | 189.69 GiB | 356.79 B | ROCm,RPC | 99 | 1 | 0 | pp512 | 87.98 ± 0.29 | +| glm4moe 355B.A32B Q4_K - Medium | 189.69 GiB | 356.79 B | ROCm,RPC | 99 | 1 | 0 | tg128 | 9.10 ± 0.02 | + +build: 2656c0d26 (7693) diff --git a/benchmark/results-rpc/GLM-4.6-UD-Q4_K_XL-00001-of-00005__rocm7_1_1__hblt0__longctx32768__rpc.log b/benchmark/results-rpc/09-01-2026/GLM-4.6-UD-Q4_K_XL-00001-of-00005__rocm7_1_1__hblt0__longctx32768__rpc.log similarity index 100% rename from benchmark/results-rpc/GLM-4.6-UD-Q4_K_XL-00001-of-00005__rocm7_1_1__hblt0__longctx32768__rpc.log rename to benchmark/results-rpc/09-01-2026/GLM-4.6-UD-Q4_K_XL-00001-of-00005__rocm7_1_1__hblt0__longctx32768__rpc.log diff --git a/benchmark/results-rpc/GLM-4.6-UD-Q4_K_XL-00001-of-00005__rocm7_1_1__hblt0__rpc.log b/benchmark/results-rpc/09-01-2026/GLM-4.6-UD-Q4_K_XL-00001-of-00005__rocm7_1_1__hblt0__rpc.log similarity index 100% rename from benchmark/results-rpc/GLM-4.6-UD-Q4_K_XL-00001-of-00005__rocm7_1_1__hblt0__rpc.log rename to benchmark/results-rpc/09-01-2026/GLM-4.6-UD-Q4_K_XL-00001-of-00005__rocm7_1_1__hblt0__rpc.log diff --git a/benchmark/results-rpc/GLM-4.6-UD-Q4_K_XL-00001-of-00005__rocm7_1_1__longctx32768__rpc.log b/benchmark/results-rpc/09-01-2026/GLM-4.6-UD-Q4_K_XL-00001-of-00005__rocm7_1_1__longctx32768__rpc.log similarity index 100% rename from benchmark/results-rpc/GLM-4.6-UD-Q4_K_XL-00001-of-00005__rocm7_1_1__longctx32768__rpc.log rename to benchmark/results-rpc/09-01-2026/GLM-4.6-UD-Q4_K_XL-00001-of-00005__rocm7_1_1__longctx32768__rpc.log diff --git a/benchmark/results-rpc/GLM-4.6-UD-Q4_K_XL-00001-of-00005__rocm7_1_1__rpc.log b/benchmark/results-rpc/09-01-2026/GLM-4.6-UD-Q4_K_XL-00001-of-00005__rocm7_1_1__rpc.log similarity index 100% rename from benchmark/results-rpc/GLM-4.6-UD-Q4_K_XL-00001-of-00005__rocm7_1_1__rpc.log rename to benchmark/results-rpc/09-01-2026/GLM-4.6-UD-Q4_K_XL-00001-of-00005__rocm7_1_1__rpc.log diff --git a/benchmark/results-rpc/09-01-2026/GLM-4.6-UD-Q4_K_XL-00001-of-00005__vulkan_amdvlk__longctx32768__rpc.log b/benchmark/results-rpc/09-01-2026/GLM-4.6-UD-Q4_K_XL-00001-of-00005__vulkan_amdvlk__longctx32768__rpc.log new file mode 100644 index 0000000..bff85ad --- /dev/null +++ b/benchmark/results-rpc/09-01-2026/GLM-4.6-UD-Q4_K_XL-00001-of-00005__vulkan_amdvlk__longctx32768__rpc.log @@ -0,0 +1,19 @@ +ggml_vulkan: Found 1 Vulkan devices: +ggml_vulkan: 0 = Radeon 8060S Graphics (AMD open-source driver) | uma: 1 | fp16: 1 | bf16: 0 | warp size: 64 | shared memory: 32768 | int dot: 1 | matrix cores: KHR_coopmat +| model | size | params | backend | ngl | fa | mmap | test | t/s | +| ------------------------------ | ---------: | ---------: | ---------- | --: | -: | ---: | --------------: | -------------------: | +/opt/llama.cpp/ggml/src/ggml-rpc/ggml-rpc.cpp:724: Remote RPC server crashed or returned malformed response +/lib64/libggml-base.so.0(+0x35a5) [0x7f92f39eb5a5] +/lib64/libggml-base.so.0(ggml_print_backtrace+0x1eb) [0x7f92f39eb96b] +/lib64/libggml-base.so.0(ggml_abort+0x11f) [0x7f92f39ebaef] +/lib64/libggml-rpc.so.0(+0x5b4a) [0x7f92f7010b4a] +/lib64/libggml-base.so.0(+0x171b2) [0x7f92f39ff1b2] +/lib64/libggml-base.so.0(+0x1749f) [0x7f92f39ff49f] +/lib64/libggml-base.so.0(ggml_backend_alloc_ctx_tensors_from_buft+0x19) [0x7f92f3a00509] +/lib64/libllama.so.0(_ZN11llama_model12load_tensorsER18llama_model_loader+0x3c61) [0x7f92f72603c1] +/lib64/libllama.so.0(+0x25568) [0x7f92f71b6568] +/lib64/libllama.so.0(llama_model_load_from_file+0xac) [0x7f92f71b73cc] +/usr/sbin/llama-bench() [0x4077b5] +/lib64/libc.so.6(+0x35b5) [0x7f92f33815b5] +/lib64/libc.so.6(__libc_start_main+0x88) [0x7f92f3381668] +/usr/sbin/llama-bench() [0x409cf5] diff --git a/benchmark/results-rpc/09-01-2026/GLM-4.6-UD-Q4_K_XL-00001-of-00005__vulkan_amdvlk__rpc.log b/benchmark/results-rpc/09-01-2026/GLM-4.6-UD-Q4_K_XL-00001-of-00005__vulkan_amdvlk__rpc.log new file mode 100644 index 0000000..6f7a3a0 --- /dev/null +++ b/benchmark/results-rpc/09-01-2026/GLM-4.6-UD-Q4_K_XL-00001-of-00005__vulkan_amdvlk__rpc.log @@ -0,0 +1,19 @@ +ggml_vulkan: Found 1 Vulkan devices: +ggml_vulkan: 0 = Radeon 8060S Graphics (AMD open-source driver) | uma: 1 | fp16: 1 | bf16: 0 | warp size: 64 | shared memory: 32768 | int dot: 1 | matrix cores: KHR_coopmat +| model | size | params | backend | ngl | fa | mmap | test | t/s | +| ------------------------------ | ---------: | ---------: | ---------- | --: | -: | ---: | --------------: | -------------------: | +/opt/llama.cpp/ggml/src/ggml-rpc/ggml-rpc.cpp:724: Remote RPC server crashed or returned malformed response +/lib64/libggml-base.so.0(+0x35a5) [0x7f4efadba5a5] +/lib64/libggml-base.so.0(ggml_print_backtrace+0x1eb) [0x7f4efadba96b] +/lib64/libggml-base.so.0(ggml_abort+0x11f) [0x7f4efadbaaef] +/lib64/libggml-rpc.so.0(+0x5b4a) [0x7f4efe3dfb4a] +/lib64/libggml-base.so.0(+0x171b2) [0x7f4efadce1b2] +/lib64/libggml-base.so.0(+0x1749f) [0x7f4efadce49f] +/lib64/libggml-base.so.0(ggml_backend_alloc_ctx_tensors_from_buft+0x19) [0x7f4efadcf509] +/lib64/libllama.so.0(_ZN11llama_model12load_tensorsER18llama_model_loader+0x3c61) [0x7f4efe62f3c1] +/lib64/libllama.so.0(+0x25568) [0x7f4efe585568] +/lib64/libllama.so.0(llama_model_load_from_file+0xac) [0x7f4efe5863cc] +/usr/sbin/llama-bench() [0x4077b5] +/lib64/libc.so.6(+0x35b5) [0x7f4efa7505b5] +/lib64/libc.so.6(__libc_start_main+0x88) [0x7f4efa750668] +/usr/sbin/llama-bench() [0x409cf5] diff --git a/benchmark/results-rpc/09-01-2026/GLM-4.6-UD-Q4_K_XL-00001-of-00005__vulkan_radv__longctx32768__rpc.log b/benchmark/results-rpc/09-01-2026/GLM-4.6-UD-Q4_K_XL-00001-of-00005__vulkan_radv__longctx32768__rpc.log new file mode 100644 index 0000000..13c4edf --- /dev/null +++ b/benchmark/results-rpc/09-01-2026/GLM-4.6-UD-Q4_K_XL-00001-of-00005__vulkan_radv__longctx32768__rpc.log @@ -0,0 +1 @@ +Error: unable to find user kyuz0: no matching entries in passwd file diff --git a/benchmark/results-rpc/09-01-2026/GLM-4.6-UD-Q4_K_XL-00001-of-00005__vulkan_radv__rpc.log b/benchmark/results-rpc/09-01-2026/GLM-4.6-UD-Q4_K_XL-00001-of-00005__vulkan_radv__rpc.log new file mode 100644 index 0000000..13c4edf --- /dev/null +++ b/benchmark/results-rpc/09-01-2026/GLM-4.6-UD-Q4_K_XL-00001-of-00005__vulkan_radv__rpc.log @@ -0,0 +1 @@ +Error: unable to find user kyuz0: no matching entries in passwd file diff --git a/benchmark/results-rpc/09-01-2026/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__rocm6_4_4__hblt0__longctx32768__rpc.log b/benchmark/results-rpc/09-01-2026/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__rocm6_4_4__hblt0__longctx32768__rpc.log new file mode 100644 index 0000000..ed18edd --- /dev/null +++ b/benchmark/results-rpc/09-01-2026/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__rocm6_4_4__hblt0__longctx32768__rpc.log @@ -0,0 +1,8 @@ +ggml_cuda_init: found 1 ROCm devices: + Device 0: Radeon 8060S Graphics, gfx1151 (0x1151), VMM: no, Wave Size: 32 +| model | size | params | backend | ngl | n_ubatch | fa | mmap | test | t/s | +| ------------------------------ | ---------: | ---------: | ---------- | --: | -------: | -: | ---: | --------------: | -------------------: | +| minimax-m2 230B.A10B Q6_K | 180.94 GiB | 228.69 B | ROCm,RPC | 99 | 2048 | 1 | 0 | pp2048 @ d32768 | 59.80 ± 0.00 | +| minimax-m2 230B.A10B Q6_K | 180.94 GiB | 228.69 B | ROCm,RPC | 99 | 2048 | 1 | 0 | tg32 @ d32768 | 6.45 ± 0.00 | + +build: 2656c0d26 (7693) diff --git a/benchmark/results-rpc/09-01-2026/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__rocm6_4_4__hblt0__rpc.log b/benchmark/results-rpc/09-01-2026/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__rocm6_4_4__hblt0__rpc.log new file mode 100644 index 0000000..7578042 --- /dev/null +++ b/benchmark/results-rpc/09-01-2026/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__rocm6_4_4__hblt0__rpc.log @@ -0,0 +1,8 @@ +ggml_cuda_init: found 1 ROCm devices: + Device 0: Radeon 8060S Graphics, gfx1151 (0x1151), VMM: no, Wave Size: 32 +| model | size | params | backend | ngl | fa | mmap | test | t/s | +| ------------------------------ | ---------: | ---------: | ---------- | --: | -: | ---: | --------------: | -------------------: | +| minimax-m2 230B.A10B Q6_K | 180.94 GiB | 228.69 B | ROCm,RPC | 99 | 1 | 0 | pp512 | 172.78 ± 2.43 | +| minimax-m2 230B.A10B Q6_K | 180.94 GiB | 228.69 B | ROCm,RPC | 99 | 1 | 0 | tg128 | 18.17 ± 0.05 | + +build: 2656c0d26 (7693) diff --git a/benchmark/results-rpc/09-01-2026/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__rocm6_4_4__longctx32768__rpc.log b/benchmark/results-rpc/09-01-2026/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__rocm6_4_4__longctx32768__rpc.log new file mode 100644 index 0000000..3ea4755 --- /dev/null +++ b/benchmark/results-rpc/09-01-2026/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__rocm6_4_4__longctx32768__rpc.log @@ -0,0 +1,8 @@ +ggml_cuda_init: found 1 ROCm devices: + Device 0: Radeon 8060S Graphics, gfx1151 (0x1151), VMM: no, Wave Size: 32 +| model | size | params | backend | ngl | n_ubatch | fa | mmap | test | t/s | +| ------------------------------ | ---------: | ---------: | ---------- | --: | -------: | -: | ---: | --------------: | -------------------: | +| minimax-m2 230B.A10B Q6_K | 180.94 GiB | 228.69 B | ROCm,RPC | 99 | 2048 | 1 | 0 | pp2048 @ d32768 | 59.95 ± 0.00 | +| minimax-m2 230B.A10B Q6_K | 180.94 GiB | 228.69 B | ROCm,RPC | 99 | 2048 | 1 | 0 | tg32 @ d32768 | 6.45 ± 0.00 | + +build: 2656c0d26 (7693) diff --git a/benchmark/results-rpc/09-01-2026/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__rocm6_4_4__rpc.log b/benchmark/results-rpc/09-01-2026/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__rocm6_4_4__rpc.log new file mode 100644 index 0000000..1978468 --- /dev/null +++ b/benchmark/results-rpc/09-01-2026/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__rocm6_4_4__rpc.log @@ -0,0 +1,8 @@ +ggml_cuda_init: found 1 ROCm devices: + Device 0: Radeon 8060S Graphics, gfx1151 (0x1151), VMM: no, Wave Size: 32 +| model | size | params | backend | ngl | fa | mmap | test | t/s | +| ------------------------------ | ---------: | ---------: | ---------- | --: | -: | ---: | --------------: | -------------------: | +| minimax-m2 230B.A10B Q6_K | 180.94 GiB | 228.69 B | ROCm,RPC | 99 | 1 | 0 | pp512 | 173.98 ± 1.76 | +| minimax-m2 230B.A10B Q6_K | 180.94 GiB | 228.69 B | ROCm,RPC | 99 | 1 | 0 | tg128 | 18.17 ± 0.04 | + +build: 2656c0d26 (7693) diff --git a/benchmark/results-rpc/09-01-2026/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__rocm7-nightlies__hblt0__longctx32768__rpc.log b/benchmark/results-rpc/09-01-2026/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__rocm7-nightlies__hblt0__longctx32768__rpc.log new file mode 100644 index 0000000..d67ecf9 --- /dev/null +++ b/benchmark/results-rpc/09-01-2026/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__rocm7-nightlies__hblt0__longctx32768__rpc.log @@ -0,0 +1,8 @@ +ggml_cuda_init: found 1 ROCm devices: + Device 0: Radeon 8060S Graphics, gfx1151 (0x1151), VMM: no, Wave Size: 32 +| model | size | params | backend | ngl | n_ubatch | fa | mmap | test | t/s | +| ------------------------------ | ---------: | ---------: | ---------- | --: | -------: | -: | ---: | --------------: | -------------------: | +| minimax-m2 230B.A10B Q6_K | 180.94 GiB | 228.69 B | ROCm,RPC | 99 | 2048 | 1 | 0 | pp2048 @ d32768 | 60.12 ± 0.00 | +| minimax-m2 230B.A10B Q6_K | 180.94 GiB | 228.69 B | ROCm,RPC | 99 | 2048 | 1 | 0 | tg32 @ d32768 | 6.04 ± 0.00 | + +build: 2656c0d26 (7693) diff --git a/benchmark/results-rpc/09-01-2026/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__rocm7-nightlies__hblt0__rpc.log b/benchmark/results-rpc/09-01-2026/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__rocm7-nightlies__hblt0__rpc.log new file mode 100644 index 0000000..733c3da --- /dev/null +++ b/benchmark/results-rpc/09-01-2026/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__rocm7-nightlies__hblt0__rpc.log @@ -0,0 +1,8 @@ +ggml_cuda_init: found 1 ROCm devices: + Device 0: Radeon 8060S Graphics, gfx1151 (0x1151), VMM: no, Wave Size: 32 +| model | size | params | backend | ngl | fa | mmap | test | t/s | +| ------------------------------ | ---------: | ---------: | ---------- | --: | -: | ---: | --------------: | -------------------: | +| minimax-m2 230B.A10B Q6_K | 180.94 GiB | 228.69 B | ROCm,RPC | 99 | 1 | 0 | pp512 | 157.51 ± 1.13 | +| minimax-m2 230B.A10B Q6_K | 180.94 GiB | 228.69 B | ROCm,RPC | 99 | 1 | 0 | tg128 | 18.24 ± 0.10 | + +build: 2656c0d26 (7693) diff --git a/benchmark/results-rpc/09-01-2026/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__rocm7-nightlies__longctx32768__rpc.log b/benchmark/results-rpc/09-01-2026/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__rocm7-nightlies__longctx32768__rpc.log new file mode 100644 index 0000000..2b480f3 --- /dev/null +++ b/benchmark/results-rpc/09-01-2026/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__rocm7-nightlies__longctx32768__rpc.log @@ -0,0 +1,8 @@ +ggml_cuda_init: found 1 ROCm devices: + Device 0: Radeon 8060S Graphics, gfx1151 (0x1151), VMM: no, Wave Size: 32 +| model | size | params | backend | ngl | n_ubatch | fa | mmap | test | t/s | +| ------------------------------ | ---------: | ---------: | ---------- | --: | -------: | -: | ---: | --------------: | -------------------: | +| minimax-m2 230B.A10B Q6_K | 180.94 GiB | 228.69 B | ROCm,RPC | 99 | 2048 | 1 | 0 | pp2048 @ d32768 | 60.47 ± 0.00 | +| minimax-m2 230B.A10B Q6_K | 180.94 GiB | 228.69 B | ROCm,RPC | 99 | 2048 | 1 | 0 | tg32 @ d32768 | 6.04 ± 0.00 | + +build: 2656c0d26 (7693) diff --git a/benchmark/results-rpc/09-01-2026/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__rocm7-nightlies__rpc.log b/benchmark/results-rpc/09-01-2026/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__rocm7-nightlies__rpc.log new file mode 100644 index 0000000..6d1ea52 --- /dev/null +++ b/benchmark/results-rpc/09-01-2026/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__rocm7-nightlies__rpc.log @@ -0,0 +1,8 @@ +ggml_cuda_init: found 1 ROCm devices: + Device 0: Radeon 8060S Graphics, gfx1151 (0x1151), VMM: no, Wave Size: 32 +| model | size | params | backend | ngl | fa | mmap | test | t/s | +| ------------------------------ | ---------: | ---------: | ---------- | --: | -: | ---: | --------------: | -------------------: | +| minimax-m2 230B.A10B Q6_K | 180.94 GiB | 228.69 B | ROCm,RPC | 99 | 1 | 0 | pp512 | 162.36 ± 1.16 | +| minimax-m2 230B.A10B Q6_K | 180.94 GiB | 228.69 B | ROCm,RPC | 99 | 1 | 0 | tg128 | 18.23 ± 0.08 | + +build: 2656c0d26 (7693) diff --git a/benchmark/results-rpc/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__rocm7_1_1__hblt0__longctx32768__rpc.log b/benchmark/results-rpc/09-01-2026/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__rocm7_1_1__hblt0__longctx32768__rpc.log similarity index 100% rename from benchmark/results-rpc/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__rocm7_1_1__hblt0__longctx32768__rpc.log rename to benchmark/results-rpc/09-01-2026/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__rocm7_1_1__hblt0__longctx32768__rpc.log diff --git a/benchmark/results-rpc/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__rocm7_1_1__hblt0__rpc.log b/benchmark/results-rpc/09-01-2026/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__rocm7_1_1__hblt0__rpc.log similarity index 100% rename from benchmark/results-rpc/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__rocm7_1_1__hblt0__rpc.log rename to benchmark/results-rpc/09-01-2026/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__rocm7_1_1__hblt0__rpc.log diff --git a/benchmark/results-rpc/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__rocm7_1_1__longctx32768__rpc.log b/benchmark/results-rpc/09-01-2026/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__rocm7_1_1__longctx32768__rpc.log similarity index 100% rename from benchmark/results-rpc/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__rocm7_1_1__longctx32768__rpc.log rename to benchmark/results-rpc/09-01-2026/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__rocm7_1_1__longctx32768__rpc.log diff --git a/benchmark/results-rpc/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__rocm7_1_1__rpc.log b/benchmark/results-rpc/09-01-2026/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__rocm7_1_1__rpc.log similarity index 100% rename from benchmark/results-rpc/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__rocm7_1_1__rpc.log rename to benchmark/results-rpc/09-01-2026/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__rocm7_1_1__rpc.log diff --git a/benchmark/results-rpc/09-01-2026/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__vulkan_amdvlk__longctx32768__rpc.log b/benchmark/results-rpc/09-01-2026/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__vulkan_amdvlk__longctx32768__rpc.log new file mode 100644 index 0000000..ce783c3 --- /dev/null +++ b/benchmark/results-rpc/09-01-2026/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__vulkan_amdvlk__longctx32768__rpc.log @@ -0,0 +1,3 @@ +ggml_vulkan: Found 1 Vulkan devices: +ggml_vulkan: 0 = Radeon 8060S Graphics (AMD open-source driver) | uma: 1 | fp16: 1 | bf16: 0 | warp size: 64 | shared memory: 32768 | int dot: 1 | matrix cores: KHR_coopmat +Failed to connect to 10.0.0.1:50052 diff --git a/benchmark/results-rpc/09-01-2026/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__vulkan_amdvlk__rpc.log b/benchmark/results-rpc/09-01-2026/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__vulkan_amdvlk__rpc.log new file mode 100644 index 0000000..264c27f --- /dev/null +++ b/benchmark/results-rpc/09-01-2026/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__vulkan_amdvlk__rpc.log @@ -0,0 +1,19 @@ +ggml_vulkan: Found 1 Vulkan devices: +ggml_vulkan: 0 = Radeon 8060S Graphics (AMD open-source driver) | uma: 1 | fp16: 1 | bf16: 0 | warp size: 64 | shared memory: 32768 | int dot: 1 | matrix cores: KHR_coopmat +| model | size | params | backend | ngl | fa | mmap | test | t/s | +| ------------------------------ | ---------: | ---------: | ---------- | --: | -: | ---: | --------------: | -------------------: | +/opt/llama.cpp/ggml/src/ggml-rpc/ggml-rpc.cpp:724: Remote RPC server crashed or returned malformed response +/lib64/libggml-base.so.0(+0x35a5) [0x7f7c046f25a5] +/lib64/libggml-base.so.0(ggml_print_backtrace+0x1eb) [0x7f7c046f296b] +/lib64/libggml-base.so.0(ggml_abort+0x11f) [0x7f7c046f2aef] +/lib64/libggml-rpc.so.0(+0x5b4a) [0x7f7c07d17b4a] +/lib64/libggml-base.so.0(+0x171b2) [0x7f7c047061b2] +/lib64/libggml-base.so.0(+0x1749f) [0x7f7c0470649f] +/lib64/libggml-base.so.0(ggml_backend_alloc_ctx_tensors_from_buft+0x19) [0x7f7c04707509] +/lib64/libllama.so.0(_ZN11llama_model12load_tensorsER18llama_model_loader+0x3c61) [0x7f7c07f673c1] +/lib64/libllama.so.0(+0x25568) [0x7f7c07ebd568] +/lib64/libllama.so.0(llama_model_load_from_file+0xac) [0x7f7c07ebe3cc] +/usr/sbin/llama-bench() [0x4077b5] +/lib64/libc.so.6(+0x35b5) [0x7f7c040885b5] +/lib64/libc.so.6(__libc_start_main+0x88) [0x7f7c04088668] +/usr/sbin/llama-bench() [0x409cf5] diff --git a/benchmark/results-rpc/09-01-2026/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__vulkan_radv__longctx32768__rpc.log b/benchmark/results-rpc/09-01-2026/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__vulkan_radv__longctx32768__rpc.log new file mode 100644 index 0000000..381ed94 --- /dev/null +++ b/benchmark/results-rpc/09-01-2026/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__vulkan_radv__longctx32768__rpc.log @@ -0,0 +1,19 @@ +ggml_vulkan: Found 1 Vulkan devices: +ggml_vulkan: 0 = Radeon 8060S Graphics (RADV GFX1151) (radv) | uma: 1 | fp16: 1 | bf16: 0 | warp size: 64 | shared memory: 65536 | int dot: 1 | matrix cores: KHR_coopmat +Failed to connect to 10.0.0.1:50052 +radv/amdgpu: Failed to allocate a buffer: +radv/amdgpu: size : 990904320 bytes +radv/amdgpu: alignment : 262144 bytes +radv/amdgpu: domains : 4 +radv/amdgpu: Failed to allocate a buffer: +radv/amdgpu: size : 990904320 bytes +radv/amdgpu: alignment : 262144 bytes +radv/amdgpu: domains : 4 +radv/amdgpu: Failed to allocate a buffer: +radv/amdgpu: size : 990904320 bytes +radv/amdgpu: alignment : 262144 bytes +radv/amdgpu: domains : 4 +radv/amdgpu: Failed to allocate a buffer: +radv/amdgpu: size : 990904320 bytes +radv/amdgpu: alignment : 262144 bytes +radv/amdgpu: domains : 4 diff --git a/benchmark/results-rpc/09-01-2026/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__vulkan_radv__rpc.log b/benchmark/results-rpc/09-01-2026/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__vulkan_radv__rpc.log new file mode 100644 index 0000000..f1f44b3 --- /dev/null +++ b/benchmark/results-rpc/09-01-2026/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__vulkan_radv__rpc.log @@ -0,0 +1,19 @@ +ggml_vulkan: Found 1 Vulkan devices: +ggml_vulkan: 0 = Radeon 8060S Graphics (RADV GFX1151) (radv) | uma: 1 | fp16: 1 | bf16: 0 | warp size: 64 | shared memory: 65536 | int dot: 1 | matrix cores: KHR_coopmat +| model | size | params | backend | ngl | fa | mmap | test | t/s | +| ------------------------------ | ---------: | ---------: | ---------- | --: | -: | ---: | --------------: | -------------------: | +/opt/llama.cpp/ggml/src/ggml-rpc/ggml-rpc.cpp:724: Remote RPC server crashed or returned malformed response +/lib64/libggml-base.so.0(+0x35a5) [0x7fe6965fe5a5] +/lib64/libggml-base.so.0(ggml_print_backtrace+0x1eb) [0x7fe6965fe96b] +/lib64/libggml-base.so.0(ggml_abort+0x11f) [0x7fe6965feaef] +/lib64/libggml-rpc.so.0(+0x5b4a) [0x7fe699c23b4a] +/lib64/libggml-base.so.0(+0x171b2) [0x7fe6966121b2] +/lib64/libggml-base.so.0(+0x1749f) [0x7fe69661249f] +/lib64/libggml-base.so.0(ggml_backend_alloc_ctx_tensors_from_buft+0x19) [0x7fe696613509] +/lib64/libllama.so.0(_ZN11llama_model12load_tensorsER18llama_model_loader+0x3c61) [0x7fe699e733c1] +/lib64/libllama.so.0(+0x25568) [0x7fe699dc9568] +/lib64/libllama.so.0(llama_model_load_from_file+0xac) [0x7fe699dca3cc] +/usr/sbin/llama-bench() [0x4077b5] +/lib64/libc.so.6(+0x35b5) [0x7fe695f945b5] +/lib64/libc.so.6(__libc_start_main+0x88) [0x7fe695f94668] +/usr/sbin/llama-bench() [0x409cf5] diff --git a/benchmark/results-rpc/GLM-4.6-UD-Q4_K_XL-00001-of-00005__rocm-7_2__hblt0__longctx32768__rpc.log b/benchmark/results-rpc/GLM-4.6-UD-Q4_K_XL-00001-of-00005__rocm-7_2__hblt0__longctx32768__rpc.log new file mode 100644 index 0000000..f092d93 --- /dev/null +++ b/benchmark/results-rpc/GLM-4.6-UD-Q4_K_XL-00001-of-00005__rocm-7_2__hblt0__longctx32768__rpc.log @@ -0,0 +1,8 @@ +ggml_cuda_init: found 1 ROCm devices: + Device 0: Radeon 8060S Graphics, gfx1151 (0x1151), VMM: no, Wave Size: 32 +| model | size | params | backend | ngl | n_ubatch | fa | test | t/s | +| ------------------------------ | ---------: | ---------: | ---------- | --: | -------: | -: | --------------: | -------------------: | +| glm4moe 355B.A32B Q4_K - Medium | 189.69 GiB | 356.79 B | ROCm,RPC | 99 | 2048 | 1 | pp2048 @ d32768 | 18.17 ± 0.00 | +| glm4moe 355B.A32B Q4_K - Medium | 189.69 GiB | 356.79 B | ROCm,RPC | 99 | 2048 | 1 | tg32 @ d32768 | 3.73 ± 0.00 | + +build: e0c93af2a (7938) diff --git a/benchmark/results-rpc/GLM-4.6-UD-Q4_K_XL-00001-of-00005__rocm-7_2__hblt0__rpc.log b/benchmark/results-rpc/GLM-4.6-UD-Q4_K_XL-00001-of-00005__rocm-7_2__hblt0__rpc.log new file mode 100644 index 0000000..9081cd2 --- /dev/null +++ b/benchmark/results-rpc/GLM-4.6-UD-Q4_K_XL-00001-of-00005__rocm-7_2__hblt0__rpc.log @@ -0,0 +1,8 @@ +ggml_cuda_init: found 1 ROCm devices: + Device 0: Radeon 8060S Graphics, gfx1151 (0x1151), VMM: no, Wave Size: 32 +| model | size | params | backend | ngl | fa | test | t/s | +| ------------------------------ | ---------: | ---------: | ---------- | --: | -: | --------------: | -------------------: | +| glm4moe 355B.A32B Q4_K - Medium | 189.69 GiB | 356.79 B | ROCm,RPC | 99 | 1 | pp512 | 97.96 ± 0.29 | +| glm4moe 355B.A32B Q4_K - Medium | 189.69 GiB | 356.79 B | ROCm,RPC | 99 | 1 | tg128 | 9.09 ± 0.00 | + +build: e0c93af2a (7938) diff --git a/benchmark/results-rpc/GLM-4.6-UD-Q4_K_XL-00001-of-00005__rocm-7_2__longctx32768__rpc.log b/benchmark/results-rpc/GLM-4.6-UD-Q4_K_XL-00001-of-00005__rocm-7_2__longctx32768__rpc.log new file mode 100644 index 0000000..9e17012 --- /dev/null +++ b/benchmark/results-rpc/GLM-4.6-UD-Q4_K_XL-00001-of-00005__rocm-7_2__longctx32768__rpc.log @@ -0,0 +1,8 @@ +ggml_cuda_init: found 1 ROCm devices: + Device 0: Radeon 8060S Graphics, gfx1151 (0x1151), VMM: no, Wave Size: 32 +| model | size | params | backend | ngl | n_ubatch | fa | test | t/s | +| ------------------------------ | ---------: | ---------: | ---------- | --: | -------: | -: | --------------: | -------------------: | +| glm4moe 355B.A32B Q4_K - Medium | 189.69 GiB | 356.79 B | ROCm,RPC | 99 | 2048 | 1 | pp2048 @ d32768 | 17.22 ± 0.00 | +| glm4moe 355B.A32B Q4_K - Medium | 189.69 GiB | 356.79 B | ROCm,RPC | 99 | 2048 | 1 | tg32 @ d32768 | 3.72 ± 0.00 | + +build: e0c93af2a (7938) diff --git a/benchmark/results-rpc/GLM-4.6-UD-Q4_K_XL-00001-of-00005__rocm-7_2__rpc.log b/benchmark/results-rpc/GLM-4.6-UD-Q4_K_XL-00001-of-00005__rocm-7_2__rpc.log new file mode 100644 index 0000000..9d7ef4b --- /dev/null +++ b/benchmark/results-rpc/GLM-4.6-UD-Q4_K_XL-00001-of-00005__rocm-7_2__rpc.log @@ -0,0 +1,8 @@ +ggml_cuda_init: found 1 ROCm devices: + Device 0: Radeon 8060S Graphics, gfx1151 (0x1151), VMM: no, Wave Size: 32 +| model | size | params | backend | ngl | fa | test | t/s | +| ------------------------------ | ---------: | ---------: | ---------- | --: | -: | --------------: | -------------------: | +| glm4moe 355B.A32B Q4_K - Medium | 189.69 GiB | 356.79 B | ROCm,RPC | 99 | 1 | pp512 | 73.57 ± 0.23 | +| glm4moe 355B.A32B Q4_K - Medium | 189.69 GiB | 356.79 B | ROCm,RPC | 99 | 1 | tg128 | 9.02 ± 0.01 | + +build: e0c93af2a (7938) diff --git a/benchmark/results-rpc/GLM-4.6-UD-Q4_K_XL-00001-of-00005__rocm6_4_4__hblt0__longctx32768__rpc.log b/benchmark/results-rpc/GLM-4.6-UD-Q4_K_XL-00001-of-00005__rocm6_4_4__hblt0__longctx32768__rpc.log index 019b7bd..a33a174 100644 --- a/benchmark/results-rpc/GLM-4.6-UD-Q4_K_XL-00001-of-00005__rocm6_4_4__hblt0__longctx32768__rpc.log +++ b/benchmark/results-rpc/GLM-4.6-UD-Q4_K_XL-00001-of-00005__rocm6_4_4__hblt0__longctx32768__rpc.log @@ -1,8 +1,8 @@ ggml_cuda_init: found 1 ROCm devices: Device 0: Radeon 8060S Graphics, gfx1151 (0x1151), VMM: no, Wave Size: 32 -| model | size | params | backend | ngl | n_ubatch | fa | mmap | test | t/s | -| ------------------------------ | ---------: | ---------: | ---------- | --: | -------: | -: | ---: | --------------: | -------------------: | -| glm4moe 355B.A32B Q4_K - Medium | 189.69 GiB | 356.79 B | ROCm,RPC | 99 | 2048 | 1 | 0 | pp2048 @ d32768 | 18.41 ± 0.00 | -| glm4moe 355B.A32B Q4_K - Medium | 189.69 GiB | 356.79 B | ROCm,RPC | 99 | 2048 | 1 | 0 | tg32 @ d32768 | 4.12 ± 0.00 | +| model | size | params | backend | ngl | n_ubatch | fa | test | t/s | +| ------------------------------ | ---------: | ---------: | ---------- | --: | -------: | -: | --------------: | -------------------: | +| glm4moe 355B.A32B Q4_K - Medium | 189.69 GiB | 356.79 B | ROCm,RPC | 99 | 2048 | 1 | pp2048 @ d32768 | 18.79 ± 0.00 | +| glm4moe 355B.A32B Q4_K - Medium | 189.69 GiB | 356.79 B | ROCm,RPC | 99 | 2048 | 1 | tg32 @ d32768 | 4.13 ± 0.00 | -build: 2656c0d26 (7693) +build: e0c93af2a (7938) diff --git a/benchmark/results-rpc/GLM-4.6-UD-Q4_K_XL-00001-of-00005__rocm6_4_4__hblt0__rpc.log b/benchmark/results-rpc/GLM-4.6-UD-Q4_K_XL-00001-of-00005__rocm6_4_4__hblt0__rpc.log index 2339073..f0507b1 100644 --- a/benchmark/results-rpc/GLM-4.6-UD-Q4_K_XL-00001-of-00005__rocm6_4_4__hblt0__rpc.log +++ b/benchmark/results-rpc/GLM-4.6-UD-Q4_K_XL-00001-of-00005__rocm6_4_4__hblt0__rpc.log @@ -1,8 +1,8 @@ ggml_cuda_init: found 1 ROCm devices: Device 0: Radeon 8060S Graphics, gfx1151 (0x1151), VMM: no, Wave Size: 32 -| model | size | params | backend | ngl | fa | mmap | test | t/s | -| ------------------------------ | ---------: | ---------: | ---------- | --: | -: | ---: | --------------: | -------------------: | -| glm4moe 355B.A32B Q4_K - Medium | 189.69 GiB | 356.79 B | ROCm,RPC | 99 | 1 | 0 | pp512 | 101.82 ± 0.34 | -| glm4moe 355B.A32B Q4_K - Medium | 189.69 GiB | 356.79 B | ROCm,RPC | 99 | 1 | 0 | tg128 | 8.71 ± 0.00 | +| model | size | params | backend | ngl | fa | test | t/s | +| ------------------------------ | ---------: | ---------: | ---------- | --: | -: | --------------: | -------------------: | +| glm4moe 355B.A32B Q4_K - Medium | 189.69 GiB | 356.79 B | ROCm,RPC | 99 | 1 | pp512 | 99.24 ± 0.14 | +| glm4moe 355B.A32B Q4_K - Medium | 189.69 GiB | 356.79 B | ROCm,RPC | 99 | 1 | tg128 | 8.55 ± 0.00 | -build: 2656c0d26 (7693) +build: e0c93af2a (7938) diff --git a/benchmark/results-rpc/GLM-4.6-UD-Q4_K_XL-00001-of-00005__rocm6_4_4__longctx32768__rpc.log b/benchmark/results-rpc/GLM-4.6-UD-Q4_K_XL-00001-of-00005__rocm6_4_4__longctx32768__rpc.log index 7d68b86..181fa3d 100644 --- a/benchmark/results-rpc/GLM-4.6-UD-Q4_K_XL-00001-of-00005__rocm6_4_4__longctx32768__rpc.log +++ b/benchmark/results-rpc/GLM-4.6-UD-Q4_K_XL-00001-of-00005__rocm6_4_4__longctx32768__rpc.log @@ -1,8 +1,8 @@ ggml_cuda_init: found 1 ROCm devices: Device 0: Radeon 8060S Graphics, gfx1151 (0x1151), VMM: no, Wave Size: 32 -| model | size | params | backend | ngl | n_ubatch | fa | mmap | test | t/s | -| ------------------------------ | ---------: | ---------: | ---------- | --: | -------: | -: | ---: | --------------: | -------------------: | -| glm4moe 355B.A32B Q4_K - Medium | 189.69 GiB | 356.79 B | ROCm,RPC | 99 | 2048 | 1 | 0 | pp2048 @ d32768 | 17.93 ± 0.00 | -| glm4moe 355B.A32B Q4_K - Medium | 189.69 GiB | 356.79 B | ROCm,RPC | 99 | 2048 | 1 | 0 | tg32 @ d32768 | 4.13 ± 0.00 | +| model | size | params | backend | ngl | n_ubatch | fa | test | t/s | +| ------------------------------ | ---------: | ---------: | ---------- | --: | -------: | -: | --------------: | -------------------: | +| glm4moe 355B.A32B Q4_K - Medium | 189.69 GiB | 356.79 B | ROCm,RPC | 99 | 2048 | 1 | pp2048 @ d32768 | 18.80 ± 0.00 | +| glm4moe 355B.A32B Q4_K - Medium | 189.69 GiB | 356.79 B | ROCm,RPC | 99 | 2048 | 1 | tg32 @ d32768 | 4.10 ± 0.00 | -build: 2656c0d26 (7693) +build: e0c93af2a (7938) diff --git a/benchmark/results-rpc/GLM-4.6-UD-Q4_K_XL-00001-of-00005__rocm6_4_4__rpc.log b/benchmark/results-rpc/GLM-4.6-UD-Q4_K_XL-00001-of-00005__rocm6_4_4__rpc.log index 09254e0..265c4d3 100644 --- a/benchmark/results-rpc/GLM-4.6-UD-Q4_K_XL-00001-of-00005__rocm6_4_4__rpc.log +++ b/benchmark/results-rpc/GLM-4.6-UD-Q4_K_XL-00001-of-00005__rocm6_4_4__rpc.log @@ -1,8 +1,8 @@ ggml_cuda_init: found 1 ROCm devices: Device 0: Radeon 8060S Graphics, gfx1151 (0x1151), VMM: no, Wave Size: 32 -| model | size | params | backend | ngl | fa | mmap | test | t/s | -| ------------------------------ | ---------: | ---------: | ---------- | --: | -: | ---: | --------------: | -------------------: | -| glm4moe 355B.A32B Q4_K - Medium | 189.69 GiB | 356.79 B | ROCm,RPC | 99 | 1 | 0 | pp512 | 95.55 ± 0.26 | -| glm4moe 355B.A32B Q4_K - Medium | 189.69 GiB | 356.79 B | ROCm,RPC | 99 | 1 | 0 | tg128 | 8.78 ± 0.00 | +| model | size | params | backend | ngl | fa | test | t/s | +| ------------------------------ | ---------: | ---------: | ---------- | --: | -: | --------------: | -------------------: | +| glm4moe 355B.A32B Q4_K - Medium | 189.69 GiB | 356.79 B | ROCm,RPC | 99 | 1 | pp512 | 99.22 ± 0.25 | +| glm4moe 355B.A32B Q4_K - Medium | 189.69 GiB | 356.79 B | ROCm,RPC | 99 | 1 | tg128 | 8.55 ± 0.00 | -build: 2656c0d26 (7693) +build: e0c93af2a (7938) diff --git a/benchmark/results-rpc/GLM-4.6-UD-Q4_K_XL-00001-of-00005__rocm7-nightlies__hblt0__longctx32768__rpc.log b/benchmark/results-rpc/GLM-4.6-UD-Q4_K_XL-00001-of-00005__rocm7-nightlies__hblt0__longctx32768__rpc.log index d7afa2a..9fa9c59 100644 --- a/benchmark/results-rpc/GLM-4.6-UD-Q4_K_XL-00001-of-00005__rocm7-nightlies__hblt0__longctx32768__rpc.log +++ b/benchmark/results-rpc/GLM-4.6-UD-Q4_K_XL-00001-of-00005__rocm7-nightlies__hblt0__longctx32768__rpc.log @@ -1,8 +1,8 @@ ggml_cuda_init: found 1 ROCm devices: Device 0: Radeon 8060S Graphics, gfx1151 (0x1151), VMM: no, Wave Size: 32 -| model | size | params | backend | ngl | n_ubatch | fa | mmap | test | t/s | -| ------------------------------ | ---------: | ---------: | ---------- | --: | -------: | -: | ---: | --------------: | -------------------: | -| glm4moe 355B.A32B Q4_K - Medium | 189.69 GiB | 356.79 B | ROCm,RPC | 99 | 2048 | 1 | 0 | pp2048 @ d32768 | 18.59 ± 0.00 | -| glm4moe 355B.A32B Q4_K - Medium | 189.69 GiB | 356.79 B | ROCm,RPC | 99 | 2048 | 1 | 0 | tg32 @ d32768 | 3.63 ± 0.00 | +| model | size | params | backend | ngl | n_ubatch | fa | test | t/s | +| ------------------------------ | ---------: | ---------: | ---------- | --: | -------: | -: | --------------: | -------------------: | +| glm4moe 355B.A32B Q4_K - Medium | 189.69 GiB | 356.79 B | ROCm,RPC | 99 | 2048 | 1 | pp2048 @ d32768 | 18.35 ± 0.00 | +| glm4moe 355B.A32B Q4_K - Medium | 189.69 GiB | 356.79 B | ROCm,RPC | 99 | 2048 | 1 | tg32 @ d32768 | 3.69 ± 0.00 | -build: 2656c0d26 (7693) +build: e0c93af2a (7938) diff --git a/benchmark/results-rpc/GLM-4.6-UD-Q4_K_XL-00001-of-00005__rocm7-nightlies__hblt0__rpc.log b/benchmark/results-rpc/GLM-4.6-UD-Q4_K_XL-00001-of-00005__rocm7-nightlies__hblt0__rpc.log index 9e62d86..36c2a5f 100644 --- a/benchmark/results-rpc/GLM-4.6-UD-Q4_K_XL-00001-of-00005__rocm7-nightlies__hblt0__rpc.log +++ b/benchmark/results-rpc/GLM-4.6-UD-Q4_K_XL-00001-of-00005__rocm7-nightlies__hblt0__rpc.log @@ -1,8 +1,8 @@ ggml_cuda_init: found 1 ROCm devices: Device 0: Radeon 8060S Graphics, gfx1151 (0x1151), VMM: no, Wave Size: 32 -| model | size | params | backend | ngl | fa | mmap | test | t/s | -| ------------------------------ | ---------: | ---------: | ---------- | --: | -: | ---: | --------------: | -------------------: | -| glm4moe 355B.A32B Q4_K - Medium | 189.69 GiB | 356.79 B | ROCm,RPC | 99 | 1 | 0 | pp512 | 103.11 ± 0.08 | -| glm4moe 355B.A32B Q4_K - Medium | 189.69 GiB | 356.79 B | ROCm,RPC | 99 | 1 | 0 | tg128 | 9.11 ± 0.03 | +| model | size | params | backend | ngl | fa | test | t/s | +| ------------------------------ | ---------: | ---------: | ---------- | --: | -: | --------------: | -------------------: | +| glm4moe 355B.A32B Q4_K - Medium | 189.69 GiB | 356.79 B | ROCm,RPC | 99 | 1 | pp512 | 99.58 ± 0.65 | +| glm4moe 355B.A32B Q4_K - Medium | 189.69 GiB | 356.79 B | ROCm,RPC | 99 | 1 | tg128 | 9.04 ± 0.01 | -build: 2656c0d26 (7693) +build: e0c93af2a (7938) diff --git a/benchmark/results-rpc/GLM-4.6-UD-Q4_K_XL-00001-of-00005__rocm7-nightlies__longctx32768__rpc.log b/benchmark/results-rpc/GLM-4.6-UD-Q4_K_XL-00001-of-00005__rocm7-nightlies__longctx32768__rpc.log index e5e59f5..e572b4d 100644 --- a/benchmark/results-rpc/GLM-4.6-UD-Q4_K_XL-00001-of-00005__rocm7-nightlies__longctx32768__rpc.log +++ b/benchmark/results-rpc/GLM-4.6-UD-Q4_K_XL-00001-of-00005__rocm7-nightlies__longctx32768__rpc.log @@ -1,8 +1,8 @@ ggml_cuda_init: found 1 ROCm devices: Device 0: Radeon 8060S Graphics, gfx1151 (0x1151), VMM: no, Wave Size: 32 -| model | size | params | backend | ngl | n_ubatch | fa | mmap | test | t/s | -| ------------------------------ | ---------: | ---------: | ---------- | --: | -------: | -: | ---: | --------------: | -------------------: | -| glm4moe 355B.A32B Q4_K - Medium | 189.69 GiB | 356.79 B | ROCm,RPC | 99 | 2048 | 1 | 0 | pp2048 @ d32768 | 18.03 ± 0.00 | -| glm4moe 355B.A32B Q4_K - Medium | 189.69 GiB | 356.79 B | ROCm,RPC | 99 | 2048 | 1 | 0 | tg32 @ d32768 | 3.63 ± 0.00 | +| model | size | params | backend | ngl | n_ubatch | fa | test | t/s | +| ------------------------------ | ---------: | ---------: | ---------- | --: | -------: | -: | --------------: | -------------------: | +| glm4moe 355B.A32B Q4_K - Medium | 189.69 GiB | 356.79 B | ROCm,RPC | 99 | 2048 | 1 | pp2048 @ d32768 | 17.17 ± 0.00 | +| glm4moe 355B.A32B Q4_K - Medium | 189.69 GiB | 356.79 B | ROCm,RPC | 99 | 2048 | 1 | tg32 @ d32768 | 3.67 ± 0.00 | -build: 2656c0d26 (7693) +build: e0c93af2a (7938) diff --git a/benchmark/results-rpc/GLM-4.6-UD-Q4_K_XL-00001-of-00005__rocm7-nightlies__rpc.log b/benchmark/results-rpc/GLM-4.6-UD-Q4_K_XL-00001-of-00005__rocm7-nightlies__rpc.log index 0abf42d..19521cd 100644 --- a/benchmark/results-rpc/GLM-4.6-UD-Q4_K_XL-00001-of-00005__rocm7-nightlies__rpc.log +++ b/benchmark/results-rpc/GLM-4.6-UD-Q4_K_XL-00001-of-00005__rocm7-nightlies__rpc.log @@ -1,8 +1,8 @@ ggml_cuda_init: found 1 ROCm devices: Device 0: Radeon 8060S Graphics, gfx1151 (0x1151), VMM: no, Wave Size: 32 -| model | size | params | backend | ngl | fa | mmap | test | t/s | -| ------------------------------ | ---------: | ---------: | ---------- | --: | -: | ---: | --------------: | -------------------: | -| glm4moe 355B.A32B Q4_K - Medium | 189.69 GiB | 356.79 B | ROCm,RPC | 99 | 1 | 0 | pp512 | 87.98 ± 0.29 | -| glm4moe 355B.A32B Q4_K - Medium | 189.69 GiB | 356.79 B | ROCm,RPC | 99 | 1 | 0 | tg128 | 9.10 ± 0.02 | +| model | size | params | backend | ngl | fa | test | t/s | +| ------------------------------ | ---------: | ---------: | ---------- | --: | -: | --------------: | -------------------: | +| glm4moe 355B.A32B Q4_K - Medium | 189.69 GiB | 356.79 B | ROCm,RPC | 99 | 1 | pp512 | 72.73 ± 0.53 | +| glm4moe 355B.A32B Q4_K - Medium | 189.69 GiB | 356.79 B | ROCm,RPC | 99 | 1 | tg128 | 9.05 ± 0.00 | -build: 2656c0d26 (7693) +build: e0c93af2a (7938) diff --git a/benchmark/results-rpc/GLM-4.6-UD-Q4_K_XL-00001-of-00005__vulkan_amdvlk__longctx32768__rpc.log b/benchmark/results-rpc/GLM-4.6-UD-Q4_K_XL-00001-of-00005__vulkan_amdvlk__longctx32768__rpc.log index bff85ad..f1c441f 100644 --- a/benchmark/results-rpc/GLM-4.6-UD-Q4_K_XL-00001-of-00005__vulkan_amdvlk__longctx32768__rpc.log +++ b/benchmark/results-rpc/GLM-4.6-UD-Q4_K_XL-00001-of-00005__vulkan_amdvlk__longctx32768__rpc.log @@ -1,19 +1,3 @@ ggml_vulkan: Found 1 Vulkan devices: ggml_vulkan: 0 = Radeon 8060S Graphics (AMD open-source driver) | uma: 1 | fp16: 1 | bf16: 0 | warp size: 64 | shared memory: 32768 | int dot: 1 | matrix cores: KHR_coopmat -| model | size | params | backend | ngl | fa | mmap | test | t/s | -| ------------------------------ | ---------: | ---------: | ---------- | --: | -: | ---: | --------------: | -------------------: | -/opt/llama.cpp/ggml/src/ggml-rpc/ggml-rpc.cpp:724: Remote RPC server crashed or returned malformed response -/lib64/libggml-base.so.0(+0x35a5) [0x7f92f39eb5a5] -/lib64/libggml-base.so.0(ggml_print_backtrace+0x1eb) [0x7f92f39eb96b] -/lib64/libggml-base.so.0(ggml_abort+0x11f) [0x7f92f39ebaef] -/lib64/libggml-rpc.so.0(+0x5b4a) [0x7f92f7010b4a] -/lib64/libggml-base.so.0(+0x171b2) [0x7f92f39ff1b2] -/lib64/libggml-base.so.0(+0x1749f) [0x7f92f39ff49f] -/lib64/libggml-base.so.0(ggml_backend_alloc_ctx_tensors_from_buft+0x19) [0x7f92f3a00509] -/lib64/libllama.so.0(_ZN11llama_model12load_tensorsER18llama_model_loader+0x3c61) [0x7f92f72603c1] -/lib64/libllama.so.0(+0x25568) [0x7f92f71b6568] -/lib64/libllama.so.0(llama_model_load_from_file+0xac) [0x7f92f71b73cc] -/usr/sbin/llama-bench() [0x4077b5] -/lib64/libc.so.6(+0x35b5) [0x7f92f33815b5] -/lib64/libc.so.6(__libc_start_main+0x88) [0x7f92f3381668] -/usr/sbin/llama-bench() [0x409cf5] +Failed to connect to 192.168.100.2:50052 diff --git a/benchmark/results-rpc/GLM-4.6-UD-Q4_K_XL-00001-of-00005__vulkan_amdvlk__rpc.log b/benchmark/results-rpc/GLM-4.6-UD-Q4_K_XL-00001-of-00005__vulkan_amdvlk__rpc.log index 6f7a3a0..f85bb5c 100644 --- a/benchmark/results-rpc/GLM-4.6-UD-Q4_K_XL-00001-of-00005__vulkan_amdvlk__rpc.log +++ b/benchmark/results-rpc/GLM-4.6-UD-Q4_K_XL-00001-of-00005__vulkan_amdvlk__rpc.log @@ -1,19 +1,19 @@ ggml_vulkan: Found 1 Vulkan devices: ggml_vulkan: 0 = Radeon 8060S Graphics (AMD open-source driver) | uma: 1 | fp16: 1 | bf16: 0 | warp size: 64 | shared memory: 32768 | int dot: 1 | matrix cores: KHR_coopmat -| model | size | params | backend | ngl | fa | mmap | test | t/s | -| ------------------------------ | ---------: | ---------: | ---------- | --: | -: | ---: | --------------: | -------------------: | +| model | size | params | backend | ngl | fa | test | t/s | +| ------------------------------ | ---------: | ---------: | ---------- | --: | -: | --------------: | -------------------: | /opt/llama.cpp/ggml/src/ggml-rpc/ggml-rpc.cpp:724: Remote RPC server crashed or returned malformed response -/lib64/libggml-base.so.0(+0x35a5) [0x7f4efadba5a5] -/lib64/libggml-base.so.0(ggml_print_backtrace+0x1eb) [0x7f4efadba96b] -/lib64/libggml-base.so.0(ggml_abort+0x11f) [0x7f4efadbaaef] -/lib64/libggml-rpc.so.0(+0x5b4a) [0x7f4efe3dfb4a] -/lib64/libggml-base.so.0(+0x171b2) [0x7f4efadce1b2] -/lib64/libggml-base.so.0(+0x1749f) [0x7f4efadce49f] -/lib64/libggml-base.so.0(ggml_backend_alloc_ctx_tensors_from_buft+0x19) [0x7f4efadcf509] -/lib64/libllama.so.0(_ZN11llama_model12load_tensorsER18llama_model_loader+0x3c61) [0x7f4efe62f3c1] -/lib64/libllama.so.0(+0x25568) [0x7f4efe585568] -/lib64/libllama.so.0(llama_model_load_from_file+0xac) [0x7f4efe5863cc] -/usr/sbin/llama-bench() [0x4077b5] -/lib64/libc.so.6(+0x35b5) [0x7f4efa7505b5] -/lib64/libc.so.6(__libc_start_main+0x88) [0x7f4efa750668] -/usr/sbin/llama-bench() [0x409cf5] +/lib64/libggml-base.so.0(+0x35a5) [0x7f14eecbd5a5] +/lib64/libggml-base.so.0(ggml_print_backtrace+0x1eb) [0x7f14eecbd96b] +/lib64/libggml-base.so.0(ggml_abort+0x11f) [0x7f14eecbdaef] +/lib64/libggml-rpc.so.0(+0x5b4a) [0x7f14f2311b4a] +/lib64/libggml-base.so.0(+0x174f2) [0x7f14eecd14f2] +/lib64/libggml-base.so.0(+0x177df) [0x7f14eecd17df] +/lib64/libggml-base.so.0(ggml_backend_alloc_ctx_tensors_from_buft+0x19) [0x7f14eecd2849] +/lib64/libllama.so.0(_ZN11llama_model12load_tensorsER18llama_model_loader+0x3c41) [0x7f14f257dbe1] +/lib64/libllama.so.0(+0x279e8) [0x7f14f24c79e8] +/lib64/libllama.so.0(llama_model_load_from_file+0xac) [0x7f14f24c884c] +/usr/sbin/llama-bench() [0x407fbd] +/lib64/libc.so.6(+0x35b5) [0x7f14ee1055b5] +/lib64/libc.so.6(__libc_start_main+0x88) [0x7f14ee105668] +/usr/sbin/llama-bench() [0x40a7b5] diff --git a/benchmark/results-rpc/GLM-4.6-UD-Q4_K_XL-00001-of-00005__vulkan_radv__longctx32768__rpc.log b/benchmark/results-rpc/GLM-4.6-UD-Q4_K_XL-00001-of-00005__vulkan_radv__longctx32768__rpc.log index 13c4edf..906b475 100644 --- a/benchmark/results-rpc/GLM-4.6-UD-Q4_K_XL-00001-of-00005__vulkan_radv__longctx32768__rpc.log +++ b/benchmark/results-rpc/GLM-4.6-UD-Q4_K_XL-00001-of-00005__vulkan_radv__longctx32768__rpc.log @@ -1 +1 @@ -Error: unable to find user kyuz0: no matching entries in passwd file +Error: failed to start container llama-vulkan-radv diff --git a/benchmark/results-rpc/GLM-4.6-UD-Q4_K_XL-00001-of-00005__vulkan_radv__rpc.log b/benchmark/results-rpc/GLM-4.6-UD-Q4_K_XL-00001-of-00005__vulkan_radv__rpc.log index 13c4edf..906b475 100644 --- a/benchmark/results-rpc/GLM-4.6-UD-Q4_K_XL-00001-of-00005__vulkan_radv__rpc.log +++ b/benchmark/results-rpc/GLM-4.6-UD-Q4_K_XL-00001-of-00005__vulkan_radv__rpc.log @@ -1 +1 @@ -Error: unable to find user kyuz0: no matching entries in passwd file +Error: failed to start container llama-vulkan-radv diff --git a/benchmark/results-rpc/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__rocm-7_2__hblt0__longctx32768__rpc.log b/benchmark/results-rpc/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__rocm-7_2__hblt0__longctx32768__rpc.log new file mode 100644 index 0000000..99e7150 --- /dev/null +++ b/benchmark/results-rpc/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__rocm-7_2__hblt0__longctx32768__rpc.log @@ -0,0 +1,8 @@ +ggml_cuda_init: found 1 ROCm devices: + Device 0: Radeon 8060S Graphics, gfx1151 (0x1151), VMM: no, Wave Size: 32 +| model | size | params | backend | ngl | n_ubatch | fa | test | t/s | +| ------------------------------ | ---------: | ---------: | ---------- | --: | -------: | -: | --------------: | -------------------: | +| minimax-m2 230B.A10B Q6_K | 180.94 GiB | 228.69 B | ROCm,RPC | 99 | 2048 | 1 | pp2048 @ d32768 | 58.89 ± 0.00 | +| minimax-m2 230B.A10B Q6_K | 180.94 GiB | 228.69 B | ROCm,RPC | 99 | 2048 | 1 | tg32 @ d32768 | 5.92 ± 0.00 | + +build: e0c93af2a (7938) diff --git a/benchmark/results-rpc/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__rocm-7_2__hblt0__rpc.log b/benchmark/results-rpc/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__rocm-7_2__hblt0__rpc.log new file mode 100644 index 0000000..a5bcaa6 --- /dev/null +++ b/benchmark/results-rpc/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__rocm-7_2__hblt0__rpc.log @@ -0,0 +1,8 @@ +ggml_cuda_init: found 1 ROCm devices: + Device 0: Radeon 8060S Graphics, gfx1151 (0x1151), VMM: no, Wave Size: 32 +| model | size | params | backend | ngl | fa | test | t/s | +| ------------------------------ | ---------: | ---------: | ---------- | --: | -: | --------------: | -------------------: | +| minimax-m2 230B.A10B Q6_K | 180.94 GiB | 228.69 B | ROCm,RPC | 99 | 1 | pp512 | 158.25 ± 0.52 | +| minimax-m2 230B.A10B Q6_K | 180.94 GiB | 228.69 B | ROCm,RPC | 99 | 1 | tg128 | 19.04 ± 0.00 | + +build: e0c93af2a (7938) diff --git a/benchmark/results-rpc/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__rocm-7_2__longctx32768__rpc.log b/benchmark/results-rpc/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__rocm-7_2__longctx32768__rpc.log new file mode 100644 index 0000000..a04603f --- /dev/null +++ b/benchmark/results-rpc/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__rocm-7_2__longctx32768__rpc.log @@ -0,0 +1,8 @@ +ggml_cuda_init: found 1 ROCm devices: + Device 0: Radeon 8060S Graphics, gfx1151 (0x1151), VMM: no, Wave Size: 32 +| model | size | params | backend | ngl | n_ubatch | fa | test | t/s | +| ------------------------------ | ---------: | ---------: | ---------- | --: | -------: | -: | --------------: | -------------------: | +| minimax-m2 230B.A10B Q6_K | 180.94 GiB | 228.69 B | ROCm,RPC | 99 | 2048 | 1 | pp2048 @ d32768 | 59.95 ± 0.00 | +| minimax-m2 230B.A10B Q6_K | 180.94 GiB | 228.69 B | ROCm,RPC | 99 | 2048 | 1 | tg32 @ d32768 | 5.91 ± 0.00 | + +build: e0c93af2a (7938) diff --git a/benchmark/results-rpc/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__rocm-7_2__rpc.log b/benchmark/results-rpc/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__rocm-7_2__rpc.log new file mode 100644 index 0000000..65cdcde --- /dev/null +++ b/benchmark/results-rpc/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__rocm-7_2__rpc.log @@ -0,0 +1,8 @@ +ggml_cuda_init: found 1 ROCm devices: + Device 0: Radeon 8060S Graphics, gfx1151 (0x1151), VMM: no, Wave Size: 32 +| model | size | params | backend | ngl | fa | test | t/s | +| ------------------------------ | ---------: | ---------: | ---------- | --: | -: | --------------: | -------------------: | +| minimax-m2 230B.A10B Q6_K | 180.94 GiB | 228.69 B | ROCm,RPC | 99 | 1 | pp512 | 159.79 ± 0.35 | +| minimax-m2 230B.A10B Q6_K | 180.94 GiB | 228.69 B | ROCm,RPC | 99 | 1 | tg128 | 19.05 ± 0.00 | + +build: e0c93af2a (7938) diff --git a/benchmark/results-rpc/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__rocm6_4_4__hblt0__longctx32768__rpc.log b/benchmark/results-rpc/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__rocm6_4_4__hblt0__longctx32768__rpc.log index ed18edd..d3d7632 100644 --- a/benchmark/results-rpc/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__rocm6_4_4__hblt0__longctx32768__rpc.log +++ b/benchmark/results-rpc/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__rocm6_4_4__hblt0__longctx32768__rpc.log @@ -1,8 +1,8 @@ ggml_cuda_init: found 1 ROCm devices: Device 0: Radeon 8060S Graphics, gfx1151 (0x1151), VMM: no, Wave Size: 32 -| model | size | params | backend | ngl | n_ubatch | fa | mmap | test | t/s | -| ------------------------------ | ---------: | ---------: | ---------- | --: | -------: | -: | ---: | --------------: | -------------------: | -| minimax-m2 230B.A10B Q6_K | 180.94 GiB | 228.69 B | ROCm,RPC | 99 | 2048 | 1 | 0 | pp2048 @ d32768 | 59.80 ± 0.00 | -| minimax-m2 230B.A10B Q6_K | 180.94 GiB | 228.69 B | ROCm,RPC | 99 | 2048 | 1 | 0 | tg32 @ d32768 | 6.45 ± 0.00 | +| model | size | params | backend | ngl | n_ubatch | fa | test | t/s | +| ------------------------------ | ---------: | ---------: | ---------- | --: | -------: | -: | --------------: | -------------------: | +| minimax-m2 230B.A10B Q6_K | 180.94 GiB | 228.69 B | ROCm,RPC | 99 | 2048 | 1 | pp2048 @ d32768 | 64.41 ± 0.00 | +| minimax-m2 230B.A10B Q6_K | 180.94 GiB | 228.69 B | ROCm,RPC | 99 | 2048 | 1 | tg32 @ d32768 | 6.41 ± 0.00 | -build: 2656c0d26 (7693) +build: e0c93af2a (7938) diff --git a/benchmark/results-rpc/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__rocm6_4_4__hblt0__rpc.log b/benchmark/results-rpc/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__rocm6_4_4__hblt0__rpc.log index 7578042..b07a592 100644 --- a/benchmark/results-rpc/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__rocm6_4_4__hblt0__rpc.log +++ b/benchmark/results-rpc/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__rocm6_4_4__hblt0__rpc.log @@ -1,8 +1,8 @@ ggml_cuda_init: found 1 ROCm devices: Device 0: Radeon 8060S Graphics, gfx1151 (0x1151), VMM: no, Wave Size: 32 -| model | size | params | backend | ngl | fa | mmap | test | t/s | -| ------------------------------ | ---------: | ---------: | ---------- | --: | -: | ---: | --------------: | -------------------: | -| minimax-m2 230B.A10B Q6_K | 180.94 GiB | 228.69 B | ROCm,RPC | 99 | 1 | 0 | pp512 | 172.78 ± 2.43 | -| minimax-m2 230B.A10B Q6_K | 180.94 GiB | 228.69 B | ROCm,RPC | 99 | 1 | 0 | tg128 | 18.17 ± 0.05 | +| model | size | params | backend | ngl | fa | test | t/s | +| ------------------------------ | ---------: | ---------: | ---------- | --: | -: | --------------: | -------------------: | +| minimax-m2 230B.A10B Q6_K | 180.94 GiB | 228.69 B | ROCm,RPC | 99 | 1 | pp512 | 169.30 ± 0.95 | +| minimax-m2 230B.A10B Q6_K | 180.94 GiB | 228.69 B | ROCm,RPC | 99 | 1 | tg128 | 18.93 ± 0.01 | -build: 2656c0d26 (7693) +build: e0c93af2a (7938) diff --git a/benchmark/results-rpc/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__rocm6_4_4__longctx32768__rpc.log b/benchmark/results-rpc/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__rocm6_4_4__longctx32768__rpc.log index 3ea4755..d4933d6 100644 --- a/benchmark/results-rpc/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__rocm6_4_4__longctx32768__rpc.log +++ b/benchmark/results-rpc/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__rocm6_4_4__longctx32768__rpc.log @@ -1,8 +1,8 @@ ggml_cuda_init: found 1 ROCm devices: Device 0: Radeon 8060S Graphics, gfx1151 (0x1151), VMM: no, Wave Size: 32 -| model | size | params | backend | ngl | n_ubatch | fa | mmap | test | t/s | -| ------------------------------ | ---------: | ---------: | ---------- | --: | -------: | -: | ---: | --------------: | -------------------: | -| minimax-m2 230B.A10B Q6_K | 180.94 GiB | 228.69 B | ROCm,RPC | 99 | 2048 | 1 | 0 | pp2048 @ d32768 | 59.95 ± 0.00 | -| minimax-m2 230B.A10B Q6_K | 180.94 GiB | 228.69 B | ROCm,RPC | 99 | 2048 | 1 | 0 | tg32 @ d32768 | 6.45 ± 0.00 | +| model | size | params | backend | ngl | n_ubatch | fa | test | t/s | +| ------------------------------ | ---------: | ---------: | ---------- | --: | -------: | -: | --------------: | -------------------: | +| minimax-m2 230B.A10B Q6_K | 180.94 GiB | 228.69 B | ROCm,RPC | 99 | 2048 | 1 | pp2048 @ d32768 | 65.69 ± 0.00 | +| minimax-m2 230B.A10B Q6_K | 180.94 GiB | 228.69 B | ROCm,RPC | 99 | 2048 | 1 | tg32 @ d32768 | 6.47 ± 0.00 | -build: 2656c0d26 (7693) +build: e0c93af2a (7938) diff --git a/benchmark/results-rpc/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__rocm6_4_4__rpc.log b/benchmark/results-rpc/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__rocm6_4_4__rpc.log index 1978468..c216092 100644 --- a/benchmark/results-rpc/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__rocm6_4_4__rpc.log +++ b/benchmark/results-rpc/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__rocm6_4_4__rpc.log @@ -1,8 +1,8 @@ ggml_cuda_init: found 1 ROCm devices: Device 0: Radeon 8060S Graphics, gfx1151 (0x1151), VMM: no, Wave Size: 32 -| model | size | params | backend | ngl | fa | mmap | test | t/s | -| ------------------------------ | ---------: | ---------: | ---------- | --: | -: | ---: | --------------: | -------------------: | -| minimax-m2 230B.A10B Q6_K | 180.94 GiB | 228.69 B | ROCm,RPC | 99 | 1 | 0 | pp512 | 173.98 ± 1.76 | -| minimax-m2 230B.A10B Q6_K | 180.94 GiB | 228.69 B | ROCm,RPC | 99 | 1 | 0 | tg128 | 18.17 ± 0.04 | +| model | size | params | backend | ngl | fa | test | t/s | +| ------------------------------ | ---------: | ---------: | ---------- | --: | -: | --------------: | -------------------: | +| minimax-m2 230B.A10B Q6_K | 180.94 GiB | 228.69 B | ROCm,RPC | 99 | 1 | pp512 | 169.50 ± 1.04 | +| minimax-m2 230B.A10B Q6_K | 180.94 GiB | 228.69 B | ROCm,RPC | 99 | 1 | tg128 | 18.89 ± 0.00 | -build: 2656c0d26 (7693) +build: e0c93af2a (7938) diff --git a/benchmark/results-rpc/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__rocm7-nightlies__hblt0__longctx32768__rpc.log b/benchmark/results-rpc/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__rocm7-nightlies__hblt0__longctx32768__rpc.log index d67ecf9..38e1dda 100644 --- a/benchmark/results-rpc/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__rocm7-nightlies__hblt0__longctx32768__rpc.log +++ b/benchmark/results-rpc/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__rocm7-nightlies__hblt0__longctx32768__rpc.log @@ -1,8 +1,8 @@ ggml_cuda_init: found 1 ROCm devices: Device 0: Radeon 8060S Graphics, gfx1151 (0x1151), VMM: no, Wave Size: 32 -| model | size | params | backend | ngl | n_ubatch | fa | mmap | test | t/s | -| ------------------------------ | ---------: | ---------: | ---------- | --: | -------: | -: | ---: | --------------: | -------------------: | -| minimax-m2 230B.A10B Q6_K | 180.94 GiB | 228.69 B | ROCm,RPC | 99 | 2048 | 1 | 0 | pp2048 @ d32768 | 60.12 ± 0.00 | -| minimax-m2 230B.A10B Q6_K | 180.94 GiB | 228.69 B | ROCm,RPC | 99 | 2048 | 1 | 0 | tg32 @ d32768 | 6.04 ± 0.00 | +| model | size | params | backend | ngl | n_ubatch | fa | test | t/s | +| ------------------------------ | ---------: | ---------: | ---------- | --: | -------: | -: | --------------: | -------------------: | +| minimax-m2 230B.A10B Q6_K | 180.94 GiB | 228.69 B | ROCm,RPC | 99 | 2048 | 1 | pp2048 @ d32768 | 60.88 ± 0.00 | +| minimax-m2 230B.A10B Q6_K | 180.94 GiB | 228.69 B | ROCm,RPC | 99 | 2048 | 1 | tg32 @ d32768 | 6.10 ± 0.00 | -build: 2656c0d26 (7693) +build: e0c93af2a (7938) diff --git a/benchmark/results-rpc/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__rocm7-nightlies__hblt0__rpc.log b/benchmark/results-rpc/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__rocm7-nightlies__hblt0__rpc.log index 733c3da..22e2443 100644 --- a/benchmark/results-rpc/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__rocm7-nightlies__hblt0__rpc.log +++ b/benchmark/results-rpc/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__rocm7-nightlies__hblt0__rpc.log @@ -1,8 +1,8 @@ ggml_cuda_init: found 1 ROCm devices: Device 0: Radeon 8060S Graphics, gfx1151 (0x1151), VMM: no, Wave Size: 32 -| model | size | params | backend | ngl | fa | mmap | test | t/s | -| ------------------------------ | ---------: | ---------: | ---------- | --: | -: | ---: | --------------: | -------------------: | -| minimax-m2 230B.A10B Q6_K | 180.94 GiB | 228.69 B | ROCm,RPC | 99 | 1 | 0 | pp512 | 157.51 ± 1.13 | -| minimax-m2 230B.A10B Q6_K | 180.94 GiB | 228.69 B | ROCm,RPC | 99 | 1 | 0 | tg128 | 18.24 ± 0.10 | +| model | size | params | backend | ngl | fa | test | t/s | +| ------------------------------ | ---------: | ---------: | ---------- | --: | -: | --------------: | -------------------: | +| minimax-m2 230B.A10B Q6_K | 180.94 GiB | 228.69 B | ROCm,RPC | 99 | 1 | pp512 | 171.03 ± 0.56 | +| minimax-m2 230B.A10B Q6_K | 180.94 GiB | 228.69 B | ROCm,RPC | 99 | 1 | tg128 | 18.98 ± 0.02 | -build: 2656c0d26 (7693) +build: e0c93af2a (7938) diff --git a/benchmark/results-rpc/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__rocm7-nightlies__longctx32768__rpc.log b/benchmark/results-rpc/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__rocm7-nightlies__longctx32768__rpc.log index 2b480f3..b3239b5 100644 --- a/benchmark/results-rpc/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__rocm7-nightlies__longctx32768__rpc.log +++ b/benchmark/results-rpc/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__rocm7-nightlies__longctx32768__rpc.log @@ -1,8 +1,8 @@ ggml_cuda_init: found 1 ROCm devices: Device 0: Radeon 8060S Graphics, gfx1151 (0x1151), VMM: no, Wave Size: 32 -| model | size | params | backend | ngl | n_ubatch | fa | mmap | test | t/s | -| ------------------------------ | ---------: | ---------: | ---------- | --: | -------: | -: | ---: | --------------: | -------------------: | -| minimax-m2 230B.A10B Q6_K | 180.94 GiB | 228.69 B | ROCm,RPC | 99 | 2048 | 1 | 0 | pp2048 @ d32768 | 60.47 ± 0.00 | -| minimax-m2 230B.A10B Q6_K | 180.94 GiB | 228.69 B | ROCm,RPC | 99 | 2048 | 1 | 0 | tg32 @ d32768 | 6.04 ± 0.00 | +| model | size | params | backend | ngl | n_ubatch | fa | test | t/s | +| ------------------------------ | ---------: | ---------: | ---------- | --: | -------: | -: | --------------: | -------------------: | +| minimax-m2 230B.A10B Q6_K | 180.94 GiB | 228.69 B | ROCm,RPC | 99 | 2048 | 1 | pp2048 @ d32768 | 60.78 ± 0.00 | +| minimax-m2 230B.A10B Q6_K | 180.94 GiB | 228.69 B | ROCm,RPC | 99 | 2048 | 1 | tg32 @ d32768 | 6.10 ± 0.00 | -build: 2656c0d26 (7693) +build: e0c93af2a (7938) diff --git a/benchmark/results-rpc/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__rocm7-nightlies__rpc.log b/benchmark/results-rpc/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__rocm7-nightlies__rpc.log index 6d1ea52..ce19ab2 100644 --- a/benchmark/results-rpc/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__rocm7-nightlies__rpc.log +++ b/benchmark/results-rpc/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__rocm7-nightlies__rpc.log @@ -1,8 +1,8 @@ ggml_cuda_init: found 1 ROCm devices: Device 0: Radeon 8060S Graphics, gfx1151 (0x1151), VMM: no, Wave Size: 32 -| model | size | params | backend | ngl | fa | mmap | test | t/s | -| ------------------------------ | ---------: | ---------: | ---------- | --: | -: | ---: | --------------: | -------------------: | -| minimax-m2 230B.A10B Q6_K | 180.94 GiB | 228.69 B | ROCm,RPC | 99 | 1 | 0 | pp512 | 162.36 ± 1.16 | -| minimax-m2 230B.A10B Q6_K | 180.94 GiB | 228.69 B | ROCm,RPC | 99 | 1 | 0 | tg128 | 18.23 ± 0.08 | +| model | size | params | backend | ngl | fa | test | t/s | +| ------------------------------ | ---------: | ---------: | ---------- | --: | -: | --------------: | -------------------: | +| minimax-m2 230B.A10B Q6_K | 180.94 GiB | 228.69 B | ROCm,RPC | 99 | 1 | pp512 | 173.91 ± 0.29 | +| minimax-m2 230B.A10B Q6_K | 180.94 GiB | 228.69 B | ROCm,RPC | 99 | 1 | tg128 | 19.05 ± 0.00 | -build: 2656c0d26 (7693) +build: e0c93af2a (7938) diff --git a/benchmark/results-rpc/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__vulkan_amdvlk__longctx32768__rpc.log b/benchmark/results-rpc/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__vulkan_amdvlk__longctx32768__rpc.log index ce783c3..2e20f75 100644 --- a/benchmark/results-rpc/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__vulkan_amdvlk__longctx32768__rpc.log +++ b/benchmark/results-rpc/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__vulkan_amdvlk__longctx32768__rpc.log @@ -1,3 +1,8 @@ ggml_vulkan: Found 1 Vulkan devices: ggml_vulkan: 0 = Radeon 8060S Graphics (AMD open-source driver) | uma: 1 | fp16: 1 | bf16: 0 | warp size: 64 | shared memory: 32768 | int dot: 1 | matrix cores: KHR_coopmat -Failed to connect to 10.0.0.1:50052 +Failed to connect to 192.168.100.2:50052 +| model | size | params | backend | ngl | fa | test | t/s | +| ------------------------------ | ---------: | ---------: | ---------- | --: | -: | --------------: | -------------------: | +ggml_vulkan: Device memory allocation of size 990904320 failed. +ggml_vulkan: vk::Device::allocateMemory: ErrorOutOfHostMemory +main: error: failed to load model '/mnt/storage/MiniMax-M2-GGUF/UD-Q6_K_XL/MiniMax-M2-UD-Q6_K_XL-00001-of-00004.gguf' diff --git a/benchmark/results-rpc/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__vulkan_amdvlk__rpc.log b/benchmark/results-rpc/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__vulkan_amdvlk__rpc.log index 264c27f..d79d665 100644 --- a/benchmark/results-rpc/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__vulkan_amdvlk__rpc.log +++ b/benchmark/results-rpc/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__vulkan_amdvlk__rpc.log @@ -1,19 +1,19 @@ ggml_vulkan: Found 1 Vulkan devices: ggml_vulkan: 0 = Radeon 8060S Graphics (AMD open-source driver) | uma: 1 | fp16: 1 | bf16: 0 | warp size: 64 | shared memory: 32768 | int dot: 1 | matrix cores: KHR_coopmat -| model | size | params | backend | ngl | fa | mmap | test | t/s | -| ------------------------------ | ---------: | ---------: | ---------- | --: | -: | ---: | --------------: | -------------------: | +| model | size | params | backend | ngl | fa | test | t/s | +| ------------------------------ | ---------: | ---------: | ---------- | --: | -: | --------------: | -------------------: | /opt/llama.cpp/ggml/src/ggml-rpc/ggml-rpc.cpp:724: Remote RPC server crashed or returned malformed response -/lib64/libggml-base.so.0(+0x35a5) [0x7f7c046f25a5] -/lib64/libggml-base.so.0(ggml_print_backtrace+0x1eb) [0x7f7c046f296b] -/lib64/libggml-base.so.0(ggml_abort+0x11f) [0x7f7c046f2aef] -/lib64/libggml-rpc.so.0(+0x5b4a) [0x7f7c07d17b4a] -/lib64/libggml-base.so.0(+0x171b2) [0x7f7c047061b2] -/lib64/libggml-base.so.0(+0x1749f) [0x7f7c0470649f] -/lib64/libggml-base.so.0(ggml_backend_alloc_ctx_tensors_from_buft+0x19) [0x7f7c04707509] -/lib64/libllama.so.0(_ZN11llama_model12load_tensorsER18llama_model_loader+0x3c61) [0x7f7c07f673c1] -/lib64/libllama.so.0(+0x25568) [0x7f7c07ebd568] -/lib64/libllama.so.0(llama_model_load_from_file+0xac) [0x7f7c07ebe3cc] -/usr/sbin/llama-bench() [0x4077b5] -/lib64/libc.so.6(+0x35b5) [0x7f7c040885b5] -/lib64/libc.so.6(__libc_start_main+0x88) [0x7f7c04088668] -/usr/sbin/llama-bench() [0x409cf5] +/lib64/libggml-base.so.0(+0x35a5) [0x7f12de48e5a5] +/lib64/libggml-base.so.0(ggml_print_backtrace+0x1eb) [0x7f12de48e96b] +/lib64/libggml-base.so.0(ggml_abort+0x11f) [0x7f12de48eaef] +/lib64/libggml-rpc.so.0(+0x5b4a) [0x7f12e1ae2b4a] +/lib64/libggml-base.so.0(+0x174f2) [0x7f12de4a24f2] +/lib64/libggml-base.so.0(+0x177df) [0x7f12de4a27df] +/lib64/libggml-base.so.0(ggml_backend_alloc_ctx_tensors_from_buft+0x19) [0x7f12de4a3849] +/lib64/libllama.so.0(_ZN11llama_model12load_tensorsER18llama_model_loader+0x3c41) [0x7f12e1d4ebe1] +/lib64/libllama.so.0(+0x279e8) [0x7f12e1c989e8] +/lib64/libllama.so.0(llama_model_load_from_file+0xac) [0x7f12e1c9984c] +/usr/sbin/llama-bench() [0x407fbd] +/lib64/libc.so.6(+0x35b5) [0x7f12dd8d65b5] +/lib64/libc.so.6(__libc_start_main+0x88) [0x7f12dd8d6668] +/usr/sbin/llama-bench() [0x40a7b5] diff --git a/benchmark/results-rpc/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__vulkan_radv__longctx32768__rpc.log b/benchmark/results-rpc/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__vulkan_radv__longctx32768__rpc.log index 381ed94..a98164f 100644 --- a/benchmark/results-rpc/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__vulkan_radv__longctx32768__rpc.log +++ b/benchmark/results-rpc/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__vulkan_radv__longctx32768__rpc.log @@ -1,10 +1,6 @@ ggml_vulkan: Found 1 Vulkan devices: ggml_vulkan: 0 = Radeon 8060S Graphics (RADV GFX1151) (radv) | uma: 1 | fp16: 1 | bf16: 0 | warp size: 64 | shared memory: 65536 | int dot: 1 | matrix cores: KHR_coopmat -Failed to connect to 10.0.0.1:50052 -radv/amdgpu: Failed to allocate a buffer: -radv/amdgpu: size : 990904320 bytes -radv/amdgpu: alignment : 262144 bytes -radv/amdgpu: domains : 4 +Failed to connect to 192.168.100.2:50052 radv/amdgpu: Failed to allocate a buffer: radv/amdgpu: size : 990904320 bytes radv/amdgpu: alignment : 262144 bytes diff --git a/benchmark/results-rpc/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__vulkan_radv__rpc.log b/benchmark/results-rpc/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__vulkan_radv__rpc.log index f1f44b3..d6f5c31 100644 --- a/benchmark/results-rpc/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__vulkan_radv__rpc.log +++ b/benchmark/results-rpc/MiniMax-M2-UD-Q6_K_XL-00001-of-00004__vulkan_radv__rpc.log @@ -1,19 +1,19 @@ ggml_vulkan: Found 1 Vulkan devices: ggml_vulkan: 0 = Radeon 8060S Graphics (RADV GFX1151) (radv) | uma: 1 | fp16: 1 | bf16: 0 | warp size: 64 | shared memory: 65536 | int dot: 1 | matrix cores: KHR_coopmat -| model | size | params | backend | ngl | fa | mmap | test | t/s | -| ------------------------------ | ---------: | ---------: | ---------- | --: | -: | ---: | --------------: | -------------------: | +| model | size | params | backend | ngl | fa | test | t/s | +| ------------------------------ | ---------: | ---------: | ---------- | --: | -: | --------------: | -------------------: | /opt/llama.cpp/ggml/src/ggml-rpc/ggml-rpc.cpp:724: Remote RPC server crashed or returned malformed response -/lib64/libggml-base.so.0(+0x35a5) [0x7fe6965fe5a5] -/lib64/libggml-base.so.0(ggml_print_backtrace+0x1eb) [0x7fe6965fe96b] -/lib64/libggml-base.so.0(ggml_abort+0x11f) [0x7fe6965feaef] -/lib64/libggml-rpc.so.0(+0x5b4a) [0x7fe699c23b4a] -/lib64/libggml-base.so.0(+0x171b2) [0x7fe6966121b2] -/lib64/libggml-base.so.0(+0x1749f) [0x7fe69661249f] -/lib64/libggml-base.so.0(ggml_backend_alloc_ctx_tensors_from_buft+0x19) [0x7fe696613509] -/lib64/libllama.so.0(_ZN11llama_model12load_tensorsER18llama_model_loader+0x3c61) [0x7fe699e733c1] -/lib64/libllama.so.0(+0x25568) [0x7fe699dc9568] -/lib64/libllama.so.0(llama_model_load_from_file+0xac) [0x7fe699dca3cc] -/usr/sbin/llama-bench() [0x4077b5] -/lib64/libc.so.6(+0x35b5) [0x7fe695f945b5] -/lib64/libc.so.6(__libc_start_main+0x88) [0x7fe695f94668] -/usr/sbin/llama-bench() [0x409cf5] +/lib64/libggml-base.so.0(+0x35a5) [0x7f62e4d465a5] +/lib64/libggml-base.so.0(ggml_print_backtrace+0x1eb) [0x7f62e4d4696b] +/lib64/libggml-base.so.0(ggml_abort+0x11f) [0x7f62e4d46aef] +/lib64/libggml-rpc.so.0(+0x5b4a) [0x7f62e839ab4a] +/lib64/libggml-base.so.0(+0x174f2) [0x7f62e4d5a4f2] +/lib64/libggml-base.so.0(+0x177df) [0x7f62e4d5a7df] +/lib64/libggml-base.so.0(ggml_backend_alloc_ctx_tensors_from_buft+0x19) [0x7f62e4d5b849] +/lib64/libllama.so.0(_ZN11llama_model12load_tensorsER18llama_model_loader+0x3c41) [0x7f62e8631be1] +/lib64/libllama.so.0(+0x279e8) [0x7f62e857b9e8] +/lib64/libllama.so.0(llama_model_load_from_file+0xac) [0x7f62e857c84c] +/usr/sbin/llama-bench() [0x407fbd] +/lib64/libc.so.6(+0x35b5) [0x7f62e418e5b5] +/lib64/libc.so.6(__libc_start_main+0x88) [0x7f62e418e668] +/usr/sbin/llama-bench() [0x40a7b5] diff --git a/benchmark/results/09-01-2026/Devstral-2-123B-Instruct-2512-UD-Q4_K_XL-00001-of-00002__rocm-7.2__fa1.log b/benchmark/results/09-01-2026/Devstral-2-123B-Instruct-2512-UD-Q4_K_XL-00001-of-00002__rocm-7_2__fa1.log similarity index 100% rename from benchmark/results/09-01-2026/Devstral-2-123B-Instruct-2512-UD-Q4_K_XL-00001-of-00002__rocm-7.2__fa1.log rename to benchmark/results/09-01-2026/Devstral-2-123B-Instruct-2512-UD-Q4_K_XL-00001-of-00002__rocm-7_2__fa1.log diff --git a/benchmark/results/09-01-2026/Devstral-2-123B-Instruct-2512-UD-Q4_K_XL-00001-of-00002__rocm-7.2__fa1__longctx32768.log b/benchmark/results/09-01-2026/Devstral-2-123B-Instruct-2512-UD-Q4_K_XL-00001-of-00002__rocm-7_2__fa1__longctx32768.log similarity index 100% rename from benchmark/results/09-01-2026/Devstral-2-123B-Instruct-2512-UD-Q4_K_XL-00001-of-00002__rocm-7.2__fa1__longctx32768.log rename to benchmark/results/09-01-2026/Devstral-2-123B-Instruct-2512-UD-Q4_K_XL-00001-of-00002__rocm-7_2__fa1__longctx32768.log diff --git a/benchmark/results/09-01-2026/Devstral-2-123B-Instruct-2512-UD-Q4_K_XL-00001-of-00002__rocm-7.2__hblt0__fa1.log b/benchmark/results/09-01-2026/Devstral-2-123B-Instruct-2512-UD-Q4_K_XL-00001-of-00002__rocm-7_2__hblt0__fa1.log similarity index 100% rename from benchmark/results/09-01-2026/Devstral-2-123B-Instruct-2512-UD-Q4_K_XL-00001-of-00002__rocm-7.2__hblt0__fa1.log rename to benchmark/results/09-01-2026/Devstral-2-123B-Instruct-2512-UD-Q4_K_XL-00001-of-00002__rocm-7_2__hblt0__fa1.log diff --git a/benchmark/results/09-01-2026/Devstral-2-123B-Instruct-2512-UD-Q4_K_XL-00001-of-00002__rocm-7.2__hblt0__fa1__longctx32768.log b/benchmark/results/09-01-2026/Devstral-2-123B-Instruct-2512-UD-Q4_K_XL-00001-of-00002__rocm-7_2__hblt0__fa1__longctx32768.log similarity index 100% rename from benchmark/results/09-01-2026/Devstral-2-123B-Instruct-2512-UD-Q4_K_XL-00001-of-00002__rocm-7.2__hblt0__fa1__longctx32768.log rename to benchmark/results/09-01-2026/Devstral-2-123B-Instruct-2512-UD-Q4_K_XL-00001-of-00002__rocm-7_2__hblt0__fa1__longctx32768.log diff --git a/benchmark/results/09-01-2026/GLM-4.5-Air-UD-Q4_K_XL-00001-of-00002__rocm-7.2__fa1.log b/benchmark/results/09-01-2026/GLM-4.5-Air-UD-Q4_K_XL-00001-of-00002__rocm-7_2__fa1.log similarity index 100% rename from benchmark/results/09-01-2026/GLM-4.5-Air-UD-Q4_K_XL-00001-of-00002__rocm-7.2__fa1.log rename to benchmark/results/09-01-2026/GLM-4.5-Air-UD-Q4_K_XL-00001-of-00002__rocm-7_2__fa1.log diff --git a/benchmark/results/09-01-2026/GLM-4.5-Air-UD-Q4_K_XL-00001-of-00002__rocm-7.2__fa1__longctx32768.log b/benchmark/results/09-01-2026/GLM-4.5-Air-UD-Q4_K_XL-00001-of-00002__rocm-7_2__fa1__longctx32768.log similarity index 100% rename from benchmark/results/09-01-2026/GLM-4.5-Air-UD-Q4_K_XL-00001-of-00002__rocm-7.2__fa1__longctx32768.log rename to benchmark/results/09-01-2026/GLM-4.5-Air-UD-Q4_K_XL-00001-of-00002__rocm-7_2__fa1__longctx32768.log diff --git a/benchmark/results/09-01-2026/GLM-4.5-Air-UD-Q4_K_XL-00001-of-00002__rocm-7.2__hblt0__fa1.log b/benchmark/results/09-01-2026/GLM-4.5-Air-UD-Q4_K_XL-00001-of-00002__rocm-7_2__hblt0__fa1.log similarity index 100% rename from benchmark/results/09-01-2026/GLM-4.5-Air-UD-Q4_K_XL-00001-of-00002__rocm-7.2__hblt0__fa1.log rename to benchmark/results/09-01-2026/GLM-4.5-Air-UD-Q4_K_XL-00001-of-00002__rocm-7_2__hblt0__fa1.log diff --git a/benchmark/results/09-01-2026/GLM-4.5-Air-UD-Q4_K_XL-00001-of-00002__rocm-7.2__hblt0__fa1__longctx32768.log b/benchmark/results/09-01-2026/GLM-4.5-Air-UD-Q4_K_XL-00001-of-00002__rocm-7_2__hblt0__fa1__longctx32768.log similarity index 100% rename from benchmark/results/09-01-2026/GLM-4.5-Air-UD-Q4_K_XL-00001-of-00002__rocm-7.2__hblt0__fa1__longctx32768.log rename to benchmark/results/09-01-2026/GLM-4.5-Air-UD-Q4_K_XL-00001-of-00002__rocm-7_2__hblt0__fa1__longctx32768.log diff --git a/benchmark/results/09-01-2026/GLM-4.5-Air-UD-Q6_K_XL-00001-of-00003__rocm-7.2__fa1.log b/benchmark/results/09-01-2026/GLM-4.5-Air-UD-Q6_K_XL-00001-of-00003__rocm-7_2__fa1.log similarity index 100% rename from benchmark/results/09-01-2026/GLM-4.5-Air-UD-Q6_K_XL-00001-of-00003__rocm-7.2__fa1.log rename to benchmark/results/09-01-2026/GLM-4.5-Air-UD-Q6_K_XL-00001-of-00003__rocm-7_2__fa1.log diff --git a/benchmark/results/09-01-2026/GLM-4.5-Air-UD-Q6_K_XL-00001-of-00003__rocm-7.2__fa1__longctx32768.log b/benchmark/results/09-01-2026/GLM-4.5-Air-UD-Q6_K_XL-00001-of-00003__rocm-7_2__fa1__longctx32768.log similarity index 100% rename from benchmark/results/09-01-2026/GLM-4.5-Air-UD-Q6_K_XL-00001-of-00003__rocm-7.2__fa1__longctx32768.log rename to benchmark/results/09-01-2026/GLM-4.5-Air-UD-Q6_K_XL-00001-of-00003__rocm-7_2__fa1__longctx32768.log diff --git a/benchmark/results/09-01-2026/GLM-4.5-Air-UD-Q6_K_XL-00001-of-00003__rocm-7.2__hblt0__fa1.log b/benchmark/results/09-01-2026/GLM-4.5-Air-UD-Q6_K_XL-00001-of-00003__rocm-7_2__hblt0__fa1.log similarity index 100% rename from benchmark/results/09-01-2026/GLM-4.5-Air-UD-Q6_K_XL-00001-of-00003__rocm-7.2__hblt0__fa1.log rename to benchmark/results/09-01-2026/GLM-4.5-Air-UD-Q6_K_XL-00001-of-00003__rocm-7_2__hblt0__fa1.log diff --git a/benchmark/results/09-01-2026/GLM-4.5-Air-UD-Q6_K_XL-00001-of-00003__rocm-7.2__hblt0__fa1__longctx32768.log b/benchmark/results/09-01-2026/GLM-4.5-Air-UD-Q6_K_XL-00001-of-00003__rocm-7_2__hblt0__fa1__longctx32768.log similarity index 100% rename from benchmark/results/09-01-2026/GLM-4.5-Air-UD-Q6_K_XL-00001-of-00003__rocm-7.2__hblt0__fa1__longctx32768.log rename to benchmark/results/09-01-2026/GLM-4.5-Air-UD-Q6_K_XL-00001-of-00003__rocm-7_2__hblt0__fa1__longctx32768.log diff --git a/benchmark/results/09-01-2026/Llama-3.3-70B-Instruct-UD-Q8_K_XL-00001-of-00002__rocm-7.2__fa1.log b/benchmark/results/09-01-2026/Llama-3.3-70B-Instruct-UD-Q8_K_XL-00001-of-00002__rocm-7_2__fa1.log similarity index 100% rename from benchmark/results/09-01-2026/Llama-3.3-70B-Instruct-UD-Q8_K_XL-00001-of-00002__rocm-7.2__fa1.log rename to benchmark/results/09-01-2026/Llama-3.3-70B-Instruct-UD-Q8_K_XL-00001-of-00002__rocm-7_2__fa1.log diff --git a/benchmark/results/09-01-2026/Llama-3.3-70B-Instruct-UD-Q8_K_XL-00001-of-00002__rocm-7.2__fa1__longctx32768.log b/benchmark/results/09-01-2026/Llama-3.3-70B-Instruct-UD-Q8_K_XL-00001-of-00002__rocm-7_2__fa1__longctx32768.log similarity index 100% rename from benchmark/results/09-01-2026/Llama-3.3-70B-Instruct-UD-Q8_K_XL-00001-of-00002__rocm-7.2__fa1__longctx32768.log rename to benchmark/results/09-01-2026/Llama-3.3-70B-Instruct-UD-Q8_K_XL-00001-of-00002__rocm-7_2__fa1__longctx32768.log diff --git a/benchmark/results/09-01-2026/Llama-3.3-70B-Instruct-UD-Q8_K_XL-00001-of-00002__rocm-7.2__hblt0__fa1.log b/benchmark/results/09-01-2026/Llama-3.3-70B-Instruct-UD-Q8_K_XL-00001-of-00002__rocm-7_2__hblt0__fa1.log similarity index 100% rename from benchmark/results/09-01-2026/Llama-3.3-70B-Instruct-UD-Q8_K_XL-00001-of-00002__rocm-7.2__hblt0__fa1.log rename to benchmark/results/09-01-2026/Llama-3.3-70B-Instruct-UD-Q8_K_XL-00001-of-00002__rocm-7_2__hblt0__fa1.log diff --git a/benchmark/results/09-01-2026/Llama-3.3-70B-Instruct-UD-Q8_K_XL-00001-of-00002__rocm-7.2__hblt0__fa1__longctx32768.log b/benchmark/results/09-01-2026/Llama-3.3-70B-Instruct-UD-Q8_K_XL-00001-of-00002__rocm-7_2__hblt0__fa1__longctx32768.log similarity index 100% rename from benchmark/results/09-01-2026/Llama-3.3-70B-Instruct-UD-Q8_K_XL-00001-of-00002__rocm-7.2__hblt0__fa1__longctx32768.log rename to benchmark/results/09-01-2026/Llama-3.3-70B-Instruct-UD-Q8_K_XL-00001-of-00002__rocm-7_2__hblt0__fa1__longctx32768.log diff --git a/benchmark/results/09-01-2026/Ministral-3-14B-Instruct-2512-BF16__rocm-7.2__fa1.log b/benchmark/results/09-01-2026/Ministral-3-14B-Instruct-2512-BF16__rocm-7_2__fa1.log similarity index 100% rename from benchmark/results/09-01-2026/Ministral-3-14B-Instruct-2512-BF16__rocm-7.2__fa1.log rename to benchmark/results/09-01-2026/Ministral-3-14B-Instruct-2512-BF16__rocm-7_2__fa1.log diff --git a/benchmark/results/09-01-2026/Ministral-3-14B-Instruct-2512-BF16__rocm-7.2__fa1__longctx32768.log b/benchmark/results/09-01-2026/Ministral-3-14B-Instruct-2512-BF16__rocm-7_2__fa1__longctx32768.log similarity index 100% rename from benchmark/results/09-01-2026/Ministral-3-14B-Instruct-2512-BF16__rocm-7.2__fa1__longctx32768.log rename to benchmark/results/09-01-2026/Ministral-3-14B-Instruct-2512-BF16__rocm-7_2__fa1__longctx32768.log diff --git a/benchmark/results/09-01-2026/Ministral-3-14B-Instruct-2512-BF16__rocm-7.2__hblt0__fa1.log b/benchmark/results/09-01-2026/Ministral-3-14B-Instruct-2512-BF16__rocm-7_2__hblt0__fa1.log similarity index 100% rename from benchmark/results/09-01-2026/Ministral-3-14B-Instruct-2512-BF16__rocm-7.2__hblt0__fa1.log rename to benchmark/results/09-01-2026/Ministral-3-14B-Instruct-2512-BF16__rocm-7_2__hblt0__fa1.log diff --git a/benchmark/results/09-01-2026/Ministral-3-14B-Instruct-2512-BF16__rocm-7.2__hblt0__fa1__longctx32768.log b/benchmark/results/09-01-2026/Ministral-3-14B-Instruct-2512-BF16__rocm-7_2__hblt0__fa1__longctx32768.log similarity index 100% rename from benchmark/results/09-01-2026/Ministral-3-14B-Instruct-2512-BF16__rocm-7.2__hblt0__fa1__longctx32768.log rename to benchmark/results/09-01-2026/Ministral-3-14B-Instruct-2512-BF16__rocm-7_2__hblt0__fa1__longctx32768.log diff --git a/benchmark/results/09-01-2026/Nemotron-3-Nano-30B-A3B-UD-Q8_K_XL__rocm-7.2__fa1.log b/benchmark/results/09-01-2026/Nemotron-3-Nano-30B-A3B-UD-Q8_K_XL__rocm-7_2__fa1.log similarity index 100% rename from benchmark/results/09-01-2026/Nemotron-3-Nano-30B-A3B-UD-Q8_K_XL__rocm-7.2__fa1.log rename to benchmark/results/09-01-2026/Nemotron-3-Nano-30B-A3B-UD-Q8_K_XL__rocm-7_2__fa1.log diff --git a/benchmark/results/09-01-2026/Nemotron-3-Nano-30B-A3B-UD-Q8_K_XL__rocm-7.2__fa1__longctx32768.log b/benchmark/results/09-01-2026/Nemotron-3-Nano-30B-A3B-UD-Q8_K_XL__rocm-7_2__fa1__longctx32768.log similarity index 100% rename from benchmark/results/09-01-2026/Nemotron-3-Nano-30B-A3B-UD-Q8_K_XL__rocm-7.2__fa1__longctx32768.log rename to benchmark/results/09-01-2026/Nemotron-3-Nano-30B-A3B-UD-Q8_K_XL__rocm-7_2__fa1__longctx32768.log diff --git a/benchmark/results/09-01-2026/Nemotron-3-Nano-30B-A3B-UD-Q8_K_XL__rocm-7.2__hblt0__fa1.log b/benchmark/results/09-01-2026/Nemotron-3-Nano-30B-A3B-UD-Q8_K_XL__rocm-7_2__hblt0__fa1.log similarity index 100% rename from benchmark/results/09-01-2026/Nemotron-3-Nano-30B-A3B-UD-Q8_K_XL__rocm-7.2__hblt0__fa1.log rename to benchmark/results/09-01-2026/Nemotron-3-Nano-30B-A3B-UD-Q8_K_XL__rocm-7_2__hblt0__fa1.log diff --git a/benchmark/results/09-01-2026/Nemotron-3-Nano-30B-A3B-UD-Q8_K_XL__rocm-7.2__hblt0__fa1__longctx32768.log b/benchmark/results/09-01-2026/Nemotron-3-Nano-30B-A3B-UD-Q8_K_XL__rocm-7_2__hblt0__fa1__longctx32768.log similarity index 100% rename from benchmark/results/09-01-2026/Nemotron-3-Nano-30B-A3B-UD-Q8_K_XL__rocm-7.2__hblt0__fa1__longctx32768.log rename to benchmark/results/09-01-2026/Nemotron-3-Nano-30B-A3B-UD-Q8_K_XL__rocm-7_2__hblt0__fa1__longctx32768.log diff --git a/benchmark/results/09-01-2026/Qwen3-235B-A22B-Instruct-2507-UD-Q3_K_XL-00001-of-00003__rocm-7.2__fa1.log b/benchmark/results/09-01-2026/Qwen3-235B-A22B-Instruct-2507-UD-Q3_K_XL-00001-of-00003__rocm-7_2__fa1.log similarity index 100% rename from benchmark/results/09-01-2026/Qwen3-235B-A22B-Instruct-2507-UD-Q3_K_XL-00001-of-00003__rocm-7.2__fa1.log rename to benchmark/results/09-01-2026/Qwen3-235B-A22B-Instruct-2507-UD-Q3_K_XL-00001-of-00003__rocm-7_2__fa1.log diff --git a/benchmark/results/09-01-2026/Qwen3-235B-A22B-Instruct-2507-UD-Q3_K_XL-00001-of-00003__rocm-7.2__fa1__longctx32768.log b/benchmark/results/09-01-2026/Qwen3-235B-A22B-Instruct-2507-UD-Q3_K_XL-00001-of-00003__rocm-7_2__fa1__longctx32768.log similarity index 100% rename from benchmark/results/09-01-2026/Qwen3-235B-A22B-Instruct-2507-UD-Q3_K_XL-00001-of-00003__rocm-7.2__fa1__longctx32768.log rename to benchmark/results/09-01-2026/Qwen3-235B-A22B-Instruct-2507-UD-Q3_K_XL-00001-of-00003__rocm-7_2__fa1__longctx32768.log diff --git a/benchmark/results/09-01-2026/Qwen3-235B-A22B-Instruct-2507-UD-Q3_K_XL-00001-of-00003__rocm-7.2__hblt0__fa1.log b/benchmark/results/09-01-2026/Qwen3-235B-A22B-Instruct-2507-UD-Q3_K_XL-00001-of-00003__rocm-7_2__hblt0__fa1.log similarity index 100% rename from benchmark/results/09-01-2026/Qwen3-235B-A22B-Instruct-2507-UD-Q3_K_XL-00001-of-00003__rocm-7.2__hblt0__fa1.log rename to benchmark/results/09-01-2026/Qwen3-235B-A22B-Instruct-2507-UD-Q3_K_XL-00001-of-00003__rocm-7_2__hblt0__fa1.log diff --git a/benchmark/results/09-01-2026/Qwen3-235B-A22B-Instruct-2507-UD-Q3_K_XL-00001-of-00003__rocm-7.2__hblt0__fa1__longctx32768.log b/benchmark/results/09-01-2026/Qwen3-235B-A22B-Instruct-2507-UD-Q3_K_XL-00001-of-00003__rocm-7_2__hblt0__fa1__longctx32768.log similarity index 100% rename from benchmark/results/09-01-2026/Qwen3-235B-A22B-Instruct-2507-UD-Q3_K_XL-00001-of-00003__rocm-7.2__hblt0__fa1__longctx32768.log rename to benchmark/results/09-01-2026/Qwen3-235B-A22B-Instruct-2507-UD-Q3_K_XL-00001-of-00003__rocm-7_2__hblt0__fa1__longctx32768.log diff --git a/benchmark/results/09-01-2026/Qwen3-30B-A3B-BF16-00001-of-00002__rocm-7.2__fa1.log b/benchmark/results/09-01-2026/Qwen3-30B-A3B-BF16-00001-of-00002__rocm-7_2__fa1.log similarity index 100% rename from benchmark/results/09-01-2026/Qwen3-30B-A3B-BF16-00001-of-00002__rocm-7.2__fa1.log rename to benchmark/results/09-01-2026/Qwen3-30B-A3B-BF16-00001-of-00002__rocm-7_2__fa1.log diff --git a/benchmark/results/09-01-2026/Qwen3-30B-A3B-BF16-00001-of-00002__rocm-7.2__fa1__longctx32768.log b/benchmark/results/09-01-2026/Qwen3-30B-A3B-BF16-00001-of-00002__rocm-7_2__fa1__longctx32768.log similarity index 100% rename from benchmark/results/09-01-2026/Qwen3-30B-A3B-BF16-00001-of-00002__rocm-7.2__fa1__longctx32768.log rename to benchmark/results/09-01-2026/Qwen3-30B-A3B-BF16-00001-of-00002__rocm-7_2__fa1__longctx32768.log diff --git a/benchmark/results/09-01-2026/Qwen3-30B-A3B-BF16-00001-of-00002__rocm-7.2__hblt0__fa1.log b/benchmark/results/09-01-2026/Qwen3-30B-A3B-BF16-00001-of-00002__rocm-7_2__hblt0__fa1.log similarity index 100% rename from benchmark/results/09-01-2026/Qwen3-30B-A3B-BF16-00001-of-00002__rocm-7.2__hblt0__fa1.log rename to benchmark/results/09-01-2026/Qwen3-30B-A3B-BF16-00001-of-00002__rocm-7_2__hblt0__fa1.log diff --git a/benchmark/results/09-01-2026/Qwen3-30B-A3B-BF16-00001-of-00002__rocm-7.2__hblt0__fa1__longctx32768.log b/benchmark/results/09-01-2026/Qwen3-30B-A3B-BF16-00001-of-00002__rocm-7_2__hblt0__fa1__longctx32768.log similarity index 100% rename from benchmark/results/09-01-2026/Qwen3-30B-A3B-BF16-00001-of-00002__rocm-7.2__hblt0__fa1__longctx32768.log rename to benchmark/results/09-01-2026/Qwen3-30B-A3B-BF16-00001-of-00002__rocm-7_2__hblt0__fa1__longctx32768.log diff --git a/benchmark/results/09-01-2026/Qwen3-30B-A3B-Instruct-2507-UD-Q6_K_XL__rocm-7.2__fa1.log b/benchmark/results/09-01-2026/Qwen3-30B-A3B-Instruct-2507-UD-Q6_K_XL__rocm-7_2__fa1.log similarity index 100% rename from benchmark/results/09-01-2026/Qwen3-30B-A3B-Instruct-2507-UD-Q6_K_XL__rocm-7.2__fa1.log rename to benchmark/results/09-01-2026/Qwen3-30B-A3B-Instruct-2507-UD-Q6_K_XL__rocm-7_2__fa1.log diff --git a/benchmark/results/09-01-2026/Qwen3-30B-A3B-Instruct-2507-UD-Q6_K_XL__rocm-7.2__fa1__longctx32768.log b/benchmark/results/09-01-2026/Qwen3-30B-A3B-Instruct-2507-UD-Q6_K_XL__rocm-7_2__fa1__longctx32768.log similarity index 100% rename from benchmark/results/09-01-2026/Qwen3-30B-A3B-Instruct-2507-UD-Q6_K_XL__rocm-7.2__fa1__longctx32768.log rename to benchmark/results/09-01-2026/Qwen3-30B-A3B-Instruct-2507-UD-Q6_K_XL__rocm-7_2__fa1__longctx32768.log diff --git a/benchmark/results/09-01-2026/Qwen3-30B-A3B-Instruct-2507-UD-Q6_K_XL__rocm-7.2__hblt0__fa1.log b/benchmark/results/09-01-2026/Qwen3-30B-A3B-Instruct-2507-UD-Q6_K_XL__rocm-7_2__hblt0__fa1.log similarity index 100% rename from benchmark/results/09-01-2026/Qwen3-30B-A3B-Instruct-2507-UD-Q6_K_XL__rocm-7.2__hblt0__fa1.log rename to benchmark/results/09-01-2026/Qwen3-30B-A3B-Instruct-2507-UD-Q6_K_XL__rocm-7_2__hblt0__fa1.log diff --git a/benchmark/results/09-01-2026/Qwen3-30B-A3B-Instruct-2507-UD-Q6_K_XL__rocm-7.2__hblt0__fa1__longctx32768.log b/benchmark/results/09-01-2026/Qwen3-30B-A3B-Instruct-2507-UD-Q6_K_XL__rocm-7_2__hblt0__fa1__longctx32768.log similarity index 100% rename from benchmark/results/09-01-2026/Qwen3-30B-A3B-Instruct-2507-UD-Q6_K_XL__rocm-7.2__hblt0__fa1__longctx32768.log rename to benchmark/results/09-01-2026/Qwen3-30B-A3B-Instruct-2507-UD-Q6_K_XL__rocm-7_2__hblt0__fa1__longctx32768.log diff --git a/benchmark/results/09-01-2026/Qwen3-Coder-30B-A3B-Instruct-Q4_K_M__rocm-7.2__fa1.log b/benchmark/results/09-01-2026/Qwen3-Coder-30B-A3B-Instruct-Q4_K_M__rocm-7_2__fa1.log similarity index 100% rename from benchmark/results/09-01-2026/Qwen3-Coder-30B-A3B-Instruct-Q4_K_M__rocm-7.2__fa1.log rename to benchmark/results/09-01-2026/Qwen3-Coder-30B-A3B-Instruct-Q4_K_M__rocm-7_2__fa1.log diff --git a/benchmark/results/09-01-2026/Qwen3-Coder-30B-A3B-Instruct-Q4_K_M__rocm-7.2__fa1__longctx32768.log b/benchmark/results/09-01-2026/Qwen3-Coder-30B-A3B-Instruct-Q4_K_M__rocm-7_2__fa1__longctx32768.log similarity index 100% rename from benchmark/results/09-01-2026/Qwen3-Coder-30B-A3B-Instruct-Q4_K_M__rocm-7.2__fa1__longctx32768.log rename to benchmark/results/09-01-2026/Qwen3-Coder-30B-A3B-Instruct-Q4_K_M__rocm-7_2__fa1__longctx32768.log diff --git a/benchmark/results/09-01-2026/Qwen3-Coder-30B-A3B-Instruct-Q4_K_M__rocm-7.2__hblt0__fa1.log b/benchmark/results/09-01-2026/Qwen3-Coder-30B-A3B-Instruct-Q4_K_M__rocm-7_2__hblt0__fa1.log similarity index 100% rename from benchmark/results/09-01-2026/Qwen3-Coder-30B-A3B-Instruct-Q4_K_M__rocm-7.2__hblt0__fa1.log rename to benchmark/results/09-01-2026/Qwen3-Coder-30B-A3B-Instruct-Q4_K_M__rocm-7_2__hblt0__fa1.log diff --git a/benchmark/results/09-01-2026/Qwen3-Coder-30B-A3B-Instruct-Q4_K_M__rocm-7.2__hblt0__fa1__longctx32768.log b/benchmark/results/09-01-2026/Qwen3-Coder-30B-A3B-Instruct-Q4_K_M__rocm-7_2__hblt0__fa1__longctx32768.log similarity index 100% rename from benchmark/results/09-01-2026/Qwen3-Coder-30B-A3B-Instruct-Q4_K_M__rocm-7.2__hblt0__fa1__longctx32768.log rename to benchmark/results/09-01-2026/Qwen3-Coder-30B-A3B-Instruct-Q4_K_M__rocm-7_2__hblt0__fa1__longctx32768.log diff --git a/benchmark/results/09-01-2026/Qwen3-Next-80B-A3B-Instruct-UD-Q8_K_XL-00001-of-00002__rocm-7.2__fa1.log b/benchmark/results/09-01-2026/Qwen3-Next-80B-A3B-Instruct-UD-Q8_K_XL-00001-of-00002__rocm-7_2__fa1.log similarity index 100% rename from benchmark/results/09-01-2026/Qwen3-Next-80B-A3B-Instruct-UD-Q8_K_XL-00001-of-00002__rocm-7.2__fa1.log rename to benchmark/results/09-01-2026/Qwen3-Next-80B-A3B-Instruct-UD-Q8_K_XL-00001-of-00002__rocm-7_2__fa1.log diff --git a/benchmark/results/09-01-2026/Qwen3-Next-80B-A3B-Instruct-UD-Q8_K_XL-00001-of-00002__rocm-7.2__fa1__longctx32768.log b/benchmark/results/09-01-2026/Qwen3-Next-80B-A3B-Instruct-UD-Q8_K_XL-00001-of-00002__rocm-7_2__fa1__longctx32768.log similarity index 100% rename from benchmark/results/09-01-2026/Qwen3-Next-80B-A3B-Instruct-UD-Q8_K_XL-00001-of-00002__rocm-7.2__fa1__longctx32768.log rename to benchmark/results/09-01-2026/Qwen3-Next-80B-A3B-Instruct-UD-Q8_K_XL-00001-of-00002__rocm-7_2__fa1__longctx32768.log diff --git a/benchmark/results/09-01-2026/Qwen3-Next-80B-A3B-Instruct-UD-Q8_K_XL-00001-of-00002__rocm-7.2__hblt0__fa1.log b/benchmark/results/09-01-2026/Qwen3-Next-80B-A3B-Instruct-UD-Q8_K_XL-00001-of-00002__rocm-7_2__hblt0__fa1.log similarity index 100% rename from benchmark/results/09-01-2026/Qwen3-Next-80B-A3B-Instruct-UD-Q8_K_XL-00001-of-00002__rocm-7.2__hblt0__fa1.log rename to benchmark/results/09-01-2026/Qwen3-Next-80B-A3B-Instruct-UD-Q8_K_XL-00001-of-00002__rocm-7_2__hblt0__fa1.log diff --git a/benchmark/results/09-01-2026/Qwen3-Next-80B-A3B-Instruct-UD-Q8_K_XL-00001-of-00002__rocm-7.2__hblt0__fa1__longctx32768.log b/benchmark/results/09-01-2026/Qwen3-Next-80B-A3B-Instruct-UD-Q8_K_XL-00001-of-00002__rocm-7_2__hblt0__fa1__longctx32768.log similarity index 100% rename from benchmark/results/09-01-2026/Qwen3-Next-80B-A3B-Instruct-UD-Q8_K_XL-00001-of-00002__rocm-7.2__hblt0__fa1__longctx32768.log rename to benchmark/results/09-01-2026/Qwen3-Next-80B-A3B-Instruct-UD-Q8_K_XL-00001-of-00002__rocm-7_2__hblt0__fa1__longctx32768.log diff --git a/benchmark/results/09-01-2026/gemma-3-12b-it-UD-Q8_K_XL__rocm-7.2__fa1.log b/benchmark/results/09-01-2026/gemma-3-12b-it-UD-Q8_K_XL__rocm-7_2__fa1.log similarity index 100% rename from benchmark/results/09-01-2026/gemma-3-12b-it-UD-Q8_K_XL__rocm-7.2__fa1.log rename to benchmark/results/09-01-2026/gemma-3-12b-it-UD-Q8_K_XL__rocm-7_2__fa1.log diff --git a/benchmark/results/09-01-2026/gemma-3-12b-it-UD-Q8_K_XL__rocm-7.2__fa1__longctx32768.log b/benchmark/results/09-01-2026/gemma-3-12b-it-UD-Q8_K_XL__rocm-7_2__fa1__longctx32768.log similarity index 100% rename from benchmark/results/09-01-2026/gemma-3-12b-it-UD-Q8_K_XL__rocm-7.2__fa1__longctx32768.log rename to benchmark/results/09-01-2026/gemma-3-12b-it-UD-Q8_K_XL__rocm-7_2__fa1__longctx32768.log diff --git a/benchmark/results/09-01-2026/gemma-3-12b-it-UD-Q8_K_XL__rocm-7.2__hblt0__fa1.log b/benchmark/results/09-01-2026/gemma-3-12b-it-UD-Q8_K_XL__rocm-7_2__hblt0__fa1.log similarity index 100% rename from benchmark/results/09-01-2026/gemma-3-12b-it-UD-Q8_K_XL__rocm-7.2__hblt0__fa1.log rename to benchmark/results/09-01-2026/gemma-3-12b-it-UD-Q8_K_XL__rocm-7_2__hblt0__fa1.log diff --git a/benchmark/results/09-01-2026/gemma-3-12b-it-UD-Q8_K_XL__rocm-7.2__hblt0__fa1__longctx32768.log b/benchmark/results/09-01-2026/gemma-3-12b-it-UD-Q8_K_XL__rocm-7_2__hblt0__fa1__longctx32768.log similarity index 100% rename from benchmark/results/09-01-2026/gemma-3-12b-it-UD-Q8_K_XL__rocm-7.2__hblt0__fa1__longctx32768.log rename to benchmark/results/09-01-2026/gemma-3-12b-it-UD-Q8_K_XL__rocm-7_2__hblt0__fa1__longctx32768.log diff --git a/benchmark/results/09-01-2026/gemma-3-27b-it-BF16-00001-of-00002__rocm-7.2__fa1.log b/benchmark/results/09-01-2026/gemma-3-27b-it-BF16-00001-of-00002__rocm-7_2__fa1.log similarity index 100% rename from benchmark/results/09-01-2026/gemma-3-27b-it-BF16-00001-of-00002__rocm-7.2__fa1.log rename to benchmark/results/09-01-2026/gemma-3-27b-it-BF16-00001-of-00002__rocm-7_2__fa1.log diff --git a/benchmark/results/09-01-2026/gemma-3-27b-it-BF16-00001-of-00002__rocm-7.2__fa1__longctx32768.log b/benchmark/results/09-01-2026/gemma-3-27b-it-BF16-00001-of-00002__rocm-7_2__fa1__longctx32768.log similarity index 100% rename from benchmark/results/09-01-2026/gemma-3-27b-it-BF16-00001-of-00002__rocm-7.2__fa1__longctx32768.log rename to benchmark/results/09-01-2026/gemma-3-27b-it-BF16-00001-of-00002__rocm-7_2__fa1__longctx32768.log diff --git a/benchmark/results/09-01-2026/gemma-3-27b-it-BF16-00001-of-00002__rocm-7.2__hblt0__fa1.log b/benchmark/results/09-01-2026/gemma-3-27b-it-BF16-00001-of-00002__rocm-7_2__hblt0__fa1.log similarity index 100% rename from benchmark/results/09-01-2026/gemma-3-27b-it-BF16-00001-of-00002__rocm-7.2__hblt0__fa1.log rename to benchmark/results/09-01-2026/gemma-3-27b-it-BF16-00001-of-00002__rocm-7_2__hblt0__fa1.log diff --git a/benchmark/results/09-01-2026/gemma-3-27b-it-BF16-00001-of-00002__rocm-7.2__hblt0__fa1__longctx32768.log b/benchmark/results/09-01-2026/gemma-3-27b-it-BF16-00001-of-00002__rocm-7_2__hblt0__fa1__longctx32768.log similarity index 100% rename from benchmark/results/09-01-2026/gemma-3-27b-it-BF16-00001-of-00002__rocm-7.2__hblt0__fa1__longctx32768.log rename to benchmark/results/09-01-2026/gemma-3-27b-it-BF16-00001-of-00002__rocm-7_2__hblt0__fa1__longctx32768.log diff --git a/benchmark/results/09-01-2026/gemma-3-4b-it-Q3_K_S__rocm-7.2__fa1.log b/benchmark/results/09-01-2026/gemma-3-4b-it-Q3_K_S__rocm-7_2__fa1.log similarity index 100% rename from benchmark/results/09-01-2026/gemma-3-4b-it-Q3_K_S__rocm-7.2__fa1.log rename to benchmark/results/09-01-2026/gemma-3-4b-it-Q3_K_S__rocm-7_2__fa1.log diff --git a/benchmark/results/09-01-2026/gemma-3-4b-it-Q3_K_S__rocm-7.2__fa1__longctx32768.log b/benchmark/results/09-01-2026/gemma-3-4b-it-Q3_K_S__rocm-7_2__fa1__longctx32768.log similarity index 100% rename from benchmark/results/09-01-2026/gemma-3-4b-it-Q3_K_S__rocm-7.2__fa1__longctx32768.log rename to benchmark/results/09-01-2026/gemma-3-4b-it-Q3_K_S__rocm-7_2__fa1__longctx32768.log diff --git a/benchmark/results/09-01-2026/gemma-3-4b-it-Q3_K_S__rocm-7.2__hblt0__fa1.log b/benchmark/results/09-01-2026/gemma-3-4b-it-Q3_K_S__rocm-7_2__hblt0__fa1.log similarity index 100% rename from benchmark/results/09-01-2026/gemma-3-4b-it-Q3_K_S__rocm-7.2__hblt0__fa1.log rename to benchmark/results/09-01-2026/gemma-3-4b-it-Q3_K_S__rocm-7_2__hblt0__fa1.log diff --git a/benchmark/results/09-01-2026/gemma-3-4b-it-Q3_K_S__rocm-7.2__hblt0__fa1__longctx32768.log b/benchmark/results/09-01-2026/gemma-3-4b-it-Q3_K_S__rocm-7_2__hblt0__fa1__longctx32768.log similarity index 100% rename from benchmark/results/09-01-2026/gemma-3-4b-it-Q3_K_S__rocm-7.2__hblt0__fa1__longctx32768.log rename to benchmark/results/09-01-2026/gemma-3-4b-it-Q3_K_S__rocm-7_2__hblt0__fa1__longctx32768.log diff --git a/benchmark/results/09-01-2026/gpt-oss-120b-mxfp4-00001-of-00003__rocm-7.2__fa1.log b/benchmark/results/09-01-2026/gpt-oss-120b-mxfp4-00001-of-00003__rocm-7_2__fa1.log similarity index 100% rename from benchmark/results/09-01-2026/gpt-oss-120b-mxfp4-00001-of-00003__rocm-7.2__fa1.log rename to benchmark/results/09-01-2026/gpt-oss-120b-mxfp4-00001-of-00003__rocm-7_2__fa1.log diff --git a/benchmark/results/09-01-2026/gpt-oss-120b-mxfp4-00001-of-00003__rocm-7.2__fa1__longctx32768.log b/benchmark/results/09-01-2026/gpt-oss-120b-mxfp4-00001-of-00003__rocm-7_2__fa1__longctx32768.log similarity index 100% rename from benchmark/results/09-01-2026/gpt-oss-120b-mxfp4-00001-of-00003__rocm-7.2__fa1__longctx32768.log rename to benchmark/results/09-01-2026/gpt-oss-120b-mxfp4-00001-of-00003__rocm-7_2__fa1__longctx32768.log diff --git a/benchmark/results/09-01-2026/gpt-oss-120b-mxfp4-00001-of-00003__rocm-7.2__hblt0__fa1.log b/benchmark/results/09-01-2026/gpt-oss-120b-mxfp4-00001-of-00003__rocm-7_2__hblt0__fa1.log similarity index 100% rename from benchmark/results/09-01-2026/gpt-oss-120b-mxfp4-00001-of-00003__rocm-7.2__hblt0__fa1.log rename to benchmark/results/09-01-2026/gpt-oss-120b-mxfp4-00001-of-00003__rocm-7_2__hblt0__fa1.log diff --git a/benchmark/results/09-01-2026/gpt-oss-120b-mxfp4-00001-of-00003__rocm-7.2__hblt0__fa1__longctx32768.log b/benchmark/results/09-01-2026/gpt-oss-120b-mxfp4-00001-of-00003__rocm-7_2__hblt0__fa1__longctx32768.log similarity index 100% rename from benchmark/results/09-01-2026/gpt-oss-120b-mxfp4-00001-of-00003__rocm-7.2__hblt0__fa1__longctx32768.log rename to benchmark/results/09-01-2026/gpt-oss-120b-mxfp4-00001-of-00003__rocm-7_2__hblt0__fa1__longctx32768.log diff --git a/benchmark/results/09-01-2026/gpt-oss-20b-mxfp4__rocm-7.2__fa1.log b/benchmark/results/09-01-2026/gpt-oss-20b-mxfp4__rocm-7_2__fa1.log similarity index 100% rename from benchmark/results/09-01-2026/gpt-oss-20b-mxfp4__rocm-7.2__fa1.log rename to benchmark/results/09-01-2026/gpt-oss-20b-mxfp4__rocm-7_2__fa1.log diff --git a/benchmark/results/09-01-2026/gpt-oss-20b-mxfp4__rocm-7.2__fa1__longctx32768.log b/benchmark/results/09-01-2026/gpt-oss-20b-mxfp4__rocm-7_2__fa1__longctx32768.log similarity index 100% rename from benchmark/results/09-01-2026/gpt-oss-20b-mxfp4__rocm-7.2__fa1__longctx32768.log rename to benchmark/results/09-01-2026/gpt-oss-20b-mxfp4__rocm-7_2__fa1__longctx32768.log diff --git a/benchmark/results/09-01-2026/gpt-oss-20b-mxfp4__rocm-7.2__hblt0__fa1.log b/benchmark/results/09-01-2026/gpt-oss-20b-mxfp4__rocm-7_2__hblt0__fa1.log similarity index 100% rename from benchmark/results/09-01-2026/gpt-oss-20b-mxfp4__rocm-7.2__hblt0__fa1.log rename to benchmark/results/09-01-2026/gpt-oss-20b-mxfp4__rocm-7_2__hblt0__fa1.log diff --git a/benchmark/results/09-01-2026/gpt-oss-20b-mxfp4__rocm-7.2__hblt0__fa1__longctx32768.log b/benchmark/results/09-01-2026/gpt-oss-20b-mxfp4__rocm-7_2__hblt0__fa1__longctx32768.log similarity index 100% rename from benchmark/results/09-01-2026/gpt-oss-20b-mxfp4__rocm-7.2__hblt0__fa1__longctx32768.log rename to benchmark/results/09-01-2026/gpt-oss-20b-mxfp4__rocm-7_2__hblt0__fa1__longctx32768.log diff --git a/benchmark/results/09-01-2026/llama-2-7b.Q4_0__rocm-7.2__fa1.log b/benchmark/results/09-01-2026/llama-2-7b.Q4_0__rocm-7_2__fa1.log similarity index 100% rename from benchmark/results/09-01-2026/llama-2-7b.Q4_0__rocm-7.2__fa1.log rename to benchmark/results/09-01-2026/llama-2-7b.Q4_0__rocm-7_2__fa1.log diff --git a/benchmark/results/09-01-2026/llama-2-7b.Q4_0__rocm-7.2__fa1__longctx32768.log b/benchmark/results/09-01-2026/llama-2-7b.Q4_0__rocm-7_2__fa1__longctx32768.log similarity index 100% rename from benchmark/results/09-01-2026/llama-2-7b.Q4_0__rocm-7.2__fa1__longctx32768.log rename to benchmark/results/09-01-2026/llama-2-7b.Q4_0__rocm-7_2__fa1__longctx32768.log diff --git a/benchmark/results/09-01-2026/llama-2-7b.Q4_0__rocm-7.2__hblt0__fa1.log b/benchmark/results/09-01-2026/llama-2-7b.Q4_0__rocm-7_2__hblt0__fa1.log similarity index 100% rename from benchmark/results/09-01-2026/llama-2-7b.Q4_0__rocm-7.2__hblt0__fa1.log rename to benchmark/results/09-01-2026/llama-2-7b.Q4_0__rocm-7_2__hblt0__fa1.log diff --git a/benchmark/results/09-01-2026/llama-2-7b.Q4_0__rocm-7.2__hblt0__fa1__longctx32768.log b/benchmark/results/09-01-2026/llama-2-7b.Q4_0__rocm-7_2__hblt0__fa1__longctx32768.log similarity index 100% rename from benchmark/results/09-01-2026/llama-2-7b.Q4_0__rocm-7.2__hblt0__fa1__longctx32768.log rename to benchmark/results/09-01-2026/llama-2-7b.Q4_0__rocm-7_2__hblt0__fa1__longctx32768.log diff --git a/benchmark/results/Devstral-2-123B-Instruct-2512-UD-Q4_K_XL-00001-of-00002__rocm-7.2__fa1.log b/benchmark/results/Devstral-2-123B-Instruct-2512-UD-Q4_K_XL-00001-of-00002__rocm-7_2__fa1.log similarity index 100% rename from benchmark/results/Devstral-2-123B-Instruct-2512-UD-Q4_K_XL-00001-of-00002__rocm-7.2__fa1.log rename to benchmark/results/Devstral-2-123B-Instruct-2512-UD-Q4_K_XL-00001-of-00002__rocm-7_2__fa1.log diff --git a/benchmark/results/Devstral-2-123B-Instruct-2512-UD-Q4_K_XL-00001-of-00002__rocm-7.2__fa1__longctx32768.log b/benchmark/results/Devstral-2-123B-Instruct-2512-UD-Q4_K_XL-00001-of-00002__rocm-7_2__fa1__longctx32768.log similarity index 100% rename from benchmark/results/Devstral-2-123B-Instruct-2512-UD-Q4_K_XL-00001-of-00002__rocm-7.2__fa1__longctx32768.log rename to benchmark/results/Devstral-2-123B-Instruct-2512-UD-Q4_K_XL-00001-of-00002__rocm-7_2__fa1__longctx32768.log diff --git a/benchmark/results/Devstral-2-123B-Instruct-2512-UD-Q4_K_XL-00001-of-00002__rocm-7.2__hblt0__fa1.log b/benchmark/results/Devstral-2-123B-Instruct-2512-UD-Q4_K_XL-00001-of-00002__rocm-7_2__hblt0__fa1.log similarity index 100% rename from benchmark/results/Devstral-2-123B-Instruct-2512-UD-Q4_K_XL-00001-of-00002__rocm-7.2__hblt0__fa1.log rename to benchmark/results/Devstral-2-123B-Instruct-2512-UD-Q4_K_XL-00001-of-00002__rocm-7_2__hblt0__fa1.log diff --git a/benchmark/results/Devstral-2-123B-Instruct-2512-UD-Q4_K_XL-00001-of-00002__rocm-7.2__hblt0__fa1__longctx32768.log b/benchmark/results/Devstral-2-123B-Instruct-2512-UD-Q4_K_XL-00001-of-00002__rocm-7_2__hblt0__fa1__longctx32768.log similarity index 100% rename from benchmark/results/Devstral-2-123B-Instruct-2512-UD-Q4_K_XL-00001-of-00002__rocm-7.2__hblt0__fa1__longctx32768.log rename to benchmark/results/Devstral-2-123B-Instruct-2512-UD-Q4_K_XL-00001-of-00002__rocm-7_2__hblt0__fa1__longctx32768.log diff --git a/benchmark/results/GLM-4.7-Flash-BF16-00001-of-00002__rocm-7.2__fa1.log b/benchmark/results/GLM-4.7-Flash-BF16-00001-of-00002__rocm-7_2__fa1.log similarity index 100% rename from benchmark/results/GLM-4.7-Flash-BF16-00001-of-00002__rocm-7.2__fa1.log rename to benchmark/results/GLM-4.7-Flash-BF16-00001-of-00002__rocm-7_2__fa1.log diff --git a/benchmark/results/GLM-4.7-Flash-BF16-00001-of-00002__rocm-7.2__fa1__longctx32768.log b/benchmark/results/GLM-4.7-Flash-BF16-00001-of-00002__rocm-7_2__fa1__longctx32768.log similarity index 100% rename from benchmark/results/GLM-4.7-Flash-BF16-00001-of-00002__rocm-7.2__fa1__longctx32768.log rename to benchmark/results/GLM-4.7-Flash-BF16-00001-of-00002__rocm-7_2__fa1__longctx32768.log diff --git a/benchmark/results/GLM-4.7-Flash-BF16-00001-of-00002__rocm-7.2__hblt0__fa1.log b/benchmark/results/GLM-4.7-Flash-BF16-00001-of-00002__rocm-7_2__hblt0__fa1.log similarity index 100% rename from benchmark/results/GLM-4.7-Flash-BF16-00001-of-00002__rocm-7.2__hblt0__fa1.log rename to benchmark/results/GLM-4.7-Flash-BF16-00001-of-00002__rocm-7_2__hblt0__fa1.log diff --git a/benchmark/results/GLM-4.7-Flash-BF16-00001-of-00002__rocm-7.2__hblt0__fa1__longctx32768.log b/benchmark/results/GLM-4.7-Flash-BF16-00001-of-00002__rocm-7_2__hblt0__fa1__longctx32768.log similarity index 100% rename from benchmark/results/GLM-4.7-Flash-BF16-00001-of-00002__rocm-7.2__hblt0__fa1__longctx32768.log rename to benchmark/results/GLM-4.7-Flash-BF16-00001-of-00002__rocm-7_2__hblt0__fa1__longctx32768.log diff --git a/benchmark/results/GLM-4.7-Flash-UD-Q8_K_XL__rocm-7.2__fa1.log b/benchmark/results/GLM-4.7-Flash-UD-Q8_K_XL__rocm-7_2__fa1.log similarity index 100% rename from benchmark/results/GLM-4.7-Flash-UD-Q8_K_XL__rocm-7.2__fa1.log rename to benchmark/results/GLM-4.7-Flash-UD-Q8_K_XL__rocm-7_2__fa1.log diff --git a/benchmark/results/GLM-4.7-Flash-UD-Q8_K_XL__rocm-7.2__fa1__longctx32768.log b/benchmark/results/GLM-4.7-Flash-UD-Q8_K_XL__rocm-7_2__fa1__longctx32768.log similarity index 100% rename from benchmark/results/GLM-4.7-Flash-UD-Q8_K_XL__rocm-7.2__fa1__longctx32768.log rename to benchmark/results/GLM-4.7-Flash-UD-Q8_K_XL__rocm-7_2__fa1__longctx32768.log diff --git a/benchmark/results/GLM-4.7-Flash-UD-Q8_K_XL__rocm-7.2__hblt0__fa1.log b/benchmark/results/GLM-4.7-Flash-UD-Q8_K_XL__rocm-7_2__hblt0__fa1.log similarity index 100% rename from benchmark/results/GLM-4.7-Flash-UD-Q8_K_XL__rocm-7.2__hblt0__fa1.log rename to benchmark/results/GLM-4.7-Flash-UD-Q8_K_XL__rocm-7_2__hblt0__fa1.log diff --git a/benchmark/results/GLM-4.7-Flash-UD-Q8_K_XL__rocm-7.2__hblt0__fa1__longctx32768.log b/benchmark/results/GLM-4.7-Flash-UD-Q8_K_XL__rocm-7_2__hblt0__fa1__longctx32768.log similarity index 100% rename from benchmark/results/GLM-4.7-Flash-UD-Q8_K_XL__rocm-7.2__hblt0__fa1__longctx32768.log rename to benchmark/results/GLM-4.7-Flash-UD-Q8_K_XL__rocm-7_2__hblt0__fa1__longctx32768.log diff --git a/benchmark/results/Llama-3.3-70B-Instruct-UD-Q8_K_XL-00001-of-00002__rocm-7.2__fa1.log b/benchmark/results/Llama-3.3-70B-Instruct-UD-Q8_K_XL-00001-of-00002__rocm-7_2__fa1.log similarity index 100% rename from benchmark/results/Llama-3.3-70B-Instruct-UD-Q8_K_XL-00001-of-00002__rocm-7.2__fa1.log rename to benchmark/results/Llama-3.3-70B-Instruct-UD-Q8_K_XL-00001-of-00002__rocm-7_2__fa1.log diff --git a/benchmark/results/Llama-3.3-70B-Instruct-UD-Q8_K_XL-00001-of-00002__rocm-7.2__fa1__longctx32768.log b/benchmark/results/Llama-3.3-70B-Instruct-UD-Q8_K_XL-00001-of-00002__rocm-7_2__fa1__longctx32768.log similarity index 100% rename from benchmark/results/Llama-3.3-70B-Instruct-UD-Q8_K_XL-00001-of-00002__rocm-7.2__fa1__longctx32768.log rename to benchmark/results/Llama-3.3-70B-Instruct-UD-Q8_K_XL-00001-of-00002__rocm-7_2__fa1__longctx32768.log diff --git a/benchmark/results/Llama-3.3-70B-Instruct-UD-Q8_K_XL-00001-of-00002__rocm-7.2__hblt0__fa1.log b/benchmark/results/Llama-3.3-70B-Instruct-UD-Q8_K_XL-00001-of-00002__rocm-7_2__hblt0__fa1.log similarity index 100% rename from benchmark/results/Llama-3.3-70B-Instruct-UD-Q8_K_XL-00001-of-00002__rocm-7.2__hblt0__fa1.log rename to benchmark/results/Llama-3.3-70B-Instruct-UD-Q8_K_XL-00001-of-00002__rocm-7_2__hblt0__fa1.log diff --git a/benchmark/results/Llama-3.3-70B-Instruct-UD-Q8_K_XL-00001-of-00002__rocm-7.2__hblt0__fa1__longctx32768.log b/benchmark/results/Llama-3.3-70B-Instruct-UD-Q8_K_XL-00001-of-00002__rocm-7_2__hblt0__fa1__longctx32768.log similarity index 100% rename from benchmark/results/Llama-3.3-70B-Instruct-UD-Q8_K_XL-00001-of-00002__rocm-7.2__hblt0__fa1__longctx32768.log rename to benchmark/results/Llama-3.3-70B-Instruct-UD-Q8_K_XL-00001-of-00002__rocm-7_2__hblt0__fa1__longctx32768.log diff --git a/benchmark/results/Ministral-3-14B-Instruct-2512-BF16__rocm-7.2__fa1.log b/benchmark/results/Ministral-3-14B-Instruct-2512-BF16__rocm-7_2__fa1.log similarity index 100% rename from benchmark/results/Ministral-3-14B-Instruct-2512-BF16__rocm-7.2__fa1.log rename to benchmark/results/Ministral-3-14B-Instruct-2512-BF16__rocm-7_2__fa1.log diff --git a/benchmark/results/Ministral-3-14B-Instruct-2512-BF16__rocm-7.2__fa1__longctx32768.log b/benchmark/results/Ministral-3-14B-Instruct-2512-BF16__rocm-7_2__fa1__longctx32768.log similarity index 100% rename from benchmark/results/Ministral-3-14B-Instruct-2512-BF16__rocm-7.2__fa1__longctx32768.log rename to benchmark/results/Ministral-3-14B-Instruct-2512-BF16__rocm-7_2__fa1__longctx32768.log diff --git a/benchmark/results/Ministral-3-14B-Instruct-2512-BF16__rocm-7.2__hblt0__fa1.log b/benchmark/results/Ministral-3-14B-Instruct-2512-BF16__rocm-7_2__hblt0__fa1.log similarity index 100% rename from benchmark/results/Ministral-3-14B-Instruct-2512-BF16__rocm-7.2__hblt0__fa1.log rename to benchmark/results/Ministral-3-14B-Instruct-2512-BF16__rocm-7_2__hblt0__fa1.log diff --git a/benchmark/results/Ministral-3-14B-Instruct-2512-BF16__rocm-7.2__hblt0__fa1__longctx32768.log b/benchmark/results/Ministral-3-14B-Instruct-2512-BF16__rocm-7_2__hblt0__fa1__longctx32768.log similarity index 100% rename from benchmark/results/Ministral-3-14B-Instruct-2512-BF16__rocm-7.2__hblt0__fa1__longctx32768.log rename to benchmark/results/Ministral-3-14B-Instruct-2512-BF16__rocm-7_2__hblt0__fa1__longctx32768.log diff --git a/benchmark/results/Nemotron-3-Nano-30B-A3B-UD-Q8_K_XL__rocm-7.2__fa1.log b/benchmark/results/Nemotron-3-Nano-30B-A3B-UD-Q8_K_XL__rocm-7_2__fa1.log similarity index 100% rename from benchmark/results/Nemotron-3-Nano-30B-A3B-UD-Q8_K_XL__rocm-7.2__fa1.log rename to benchmark/results/Nemotron-3-Nano-30B-A3B-UD-Q8_K_XL__rocm-7_2__fa1.log diff --git a/benchmark/results/Nemotron-3-Nano-30B-A3B-UD-Q8_K_XL__rocm-7.2__fa1__longctx32768.log b/benchmark/results/Nemotron-3-Nano-30B-A3B-UD-Q8_K_XL__rocm-7_2__fa1__longctx32768.log similarity index 100% rename from benchmark/results/Nemotron-3-Nano-30B-A3B-UD-Q8_K_XL__rocm-7.2__fa1__longctx32768.log rename to benchmark/results/Nemotron-3-Nano-30B-A3B-UD-Q8_K_XL__rocm-7_2__fa1__longctx32768.log diff --git a/benchmark/results/Nemotron-3-Nano-30B-A3B-UD-Q8_K_XL__rocm-7.2__hblt0__fa1.log b/benchmark/results/Nemotron-3-Nano-30B-A3B-UD-Q8_K_XL__rocm-7_2__hblt0__fa1.log similarity index 100% rename from benchmark/results/Nemotron-3-Nano-30B-A3B-UD-Q8_K_XL__rocm-7.2__hblt0__fa1.log rename to benchmark/results/Nemotron-3-Nano-30B-A3B-UD-Q8_K_XL__rocm-7_2__hblt0__fa1.log diff --git a/benchmark/results/Nemotron-3-Nano-30B-A3B-UD-Q8_K_XL__rocm-7.2__hblt0__fa1__longctx32768.log b/benchmark/results/Nemotron-3-Nano-30B-A3B-UD-Q8_K_XL__rocm-7_2__hblt0__fa1__longctx32768.log similarity index 100% rename from benchmark/results/Nemotron-3-Nano-30B-A3B-UD-Q8_K_XL__rocm-7.2__hblt0__fa1__longctx32768.log rename to benchmark/results/Nemotron-3-Nano-30B-A3B-UD-Q8_K_XL__rocm-7_2__hblt0__fa1__longctx32768.log diff --git a/benchmark/results/Qwen3-235B-A22B-Instruct-2507-UD-Q3_K_XL-00001-of-00003__rocm-7.2__fa1.log b/benchmark/results/Qwen3-235B-A22B-Instruct-2507-UD-Q3_K_XL-00001-of-00003__rocm-7_2__fa1.log similarity index 100% rename from benchmark/results/Qwen3-235B-A22B-Instruct-2507-UD-Q3_K_XL-00001-of-00003__rocm-7.2__fa1.log rename to benchmark/results/Qwen3-235B-A22B-Instruct-2507-UD-Q3_K_XL-00001-of-00003__rocm-7_2__fa1.log diff --git a/benchmark/results/Qwen3-235B-A22B-Instruct-2507-UD-Q3_K_XL-00001-of-00003__rocm-7.2__fa1__longctx32768.log b/benchmark/results/Qwen3-235B-A22B-Instruct-2507-UD-Q3_K_XL-00001-of-00003__rocm-7_2__fa1__longctx32768.log similarity index 100% rename from benchmark/results/Qwen3-235B-A22B-Instruct-2507-UD-Q3_K_XL-00001-of-00003__rocm-7.2__fa1__longctx32768.log rename to benchmark/results/Qwen3-235B-A22B-Instruct-2507-UD-Q3_K_XL-00001-of-00003__rocm-7_2__fa1__longctx32768.log diff --git a/benchmark/results/Qwen3-235B-A22B-Instruct-2507-UD-Q3_K_XL-00001-of-00003__rocm-7.2__hblt0__fa1.log b/benchmark/results/Qwen3-235B-A22B-Instruct-2507-UD-Q3_K_XL-00001-of-00003__rocm-7_2__hblt0__fa1.log similarity index 100% rename from benchmark/results/Qwen3-235B-A22B-Instruct-2507-UD-Q3_K_XL-00001-of-00003__rocm-7.2__hblt0__fa1.log rename to benchmark/results/Qwen3-235B-A22B-Instruct-2507-UD-Q3_K_XL-00001-of-00003__rocm-7_2__hblt0__fa1.log diff --git a/benchmark/results/Qwen3-235B-A22B-Instruct-2507-UD-Q3_K_XL-00001-of-00003__rocm-7.2__hblt0__fa1__longctx32768.log b/benchmark/results/Qwen3-235B-A22B-Instruct-2507-UD-Q3_K_XL-00001-of-00003__rocm-7_2__hblt0__fa1__longctx32768.log similarity index 100% rename from benchmark/results/Qwen3-235B-A22B-Instruct-2507-UD-Q3_K_XL-00001-of-00003__rocm-7.2__hblt0__fa1__longctx32768.log rename to benchmark/results/Qwen3-235B-A22B-Instruct-2507-UD-Q3_K_XL-00001-of-00003__rocm-7_2__hblt0__fa1__longctx32768.log diff --git a/benchmark/results/Qwen3-30B-A3B-BF16-00001-of-00002__rocm-7.2__fa1.log b/benchmark/results/Qwen3-30B-A3B-BF16-00001-of-00002__rocm-7_2__fa1.log similarity index 100% rename from benchmark/results/Qwen3-30B-A3B-BF16-00001-of-00002__rocm-7.2__fa1.log rename to benchmark/results/Qwen3-30B-A3B-BF16-00001-of-00002__rocm-7_2__fa1.log diff --git a/benchmark/results/Qwen3-30B-A3B-BF16-00001-of-00002__rocm-7.2__fa1__longctx32768.log b/benchmark/results/Qwen3-30B-A3B-BF16-00001-of-00002__rocm-7_2__fa1__longctx32768.log similarity index 100% rename from benchmark/results/Qwen3-30B-A3B-BF16-00001-of-00002__rocm-7.2__fa1__longctx32768.log rename to benchmark/results/Qwen3-30B-A3B-BF16-00001-of-00002__rocm-7_2__fa1__longctx32768.log diff --git a/benchmark/results/Qwen3-30B-A3B-BF16-00001-of-00002__rocm-7.2__hblt0__fa1.log b/benchmark/results/Qwen3-30B-A3B-BF16-00001-of-00002__rocm-7_2__hblt0__fa1.log similarity index 100% rename from benchmark/results/Qwen3-30B-A3B-BF16-00001-of-00002__rocm-7.2__hblt0__fa1.log rename to benchmark/results/Qwen3-30B-A3B-BF16-00001-of-00002__rocm-7_2__hblt0__fa1.log diff --git a/benchmark/results/Qwen3-30B-A3B-BF16-00001-of-00002__rocm-7.2__hblt0__fa1__longctx32768.log b/benchmark/results/Qwen3-30B-A3B-BF16-00001-of-00002__rocm-7_2__hblt0__fa1__longctx32768.log similarity index 100% rename from benchmark/results/Qwen3-30B-A3B-BF16-00001-of-00002__rocm-7.2__hblt0__fa1__longctx32768.log rename to benchmark/results/Qwen3-30B-A3B-BF16-00001-of-00002__rocm-7_2__hblt0__fa1__longctx32768.log diff --git a/benchmark/results/Qwen3-30B-A3B-Instruct-2507-UD-Q6_K_XL__rocm-7.2__fa1.log b/benchmark/results/Qwen3-30B-A3B-Instruct-2507-UD-Q6_K_XL__rocm-7_2__fa1.log similarity index 100% rename from benchmark/results/Qwen3-30B-A3B-Instruct-2507-UD-Q6_K_XL__rocm-7.2__fa1.log rename to benchmark/results/Qwen3-30B-A3B-Instruct-2507-UD-Q6_K_XL__rocm-7_2__fa1.log diff --git a/benchmark/results/Qwen3-30B-A3B-Instruct-2507-UD-Q6_K_XL__rocm-7.2__fa1__longctx32768.log b/benchmark/results/Qwen3-30B-A3B-Instruct-2507-UD-Q6_K_XL__rocm-7_2__fa1__longctx32768.log similarity index 100% rename from benchmark/results/Qwen3-30B-A3B-Instruct-2507-UD-Q6_K_XL__rocm-7.2__fa1__longctx32768.log rename to benchmark/results/Qwen3-30B-A3B-Instruct-2507-UD-Q6_K_XL__rocm-7_2__fa1__longctx32768.log diff --git a/benchmark/results/Qwen3-30B-A3B-Instruct-2507-UD-Q6_K_XL__rocm-7.2__hblt0__fa1.log b/benchmark/results/Qwen3-30B-A3B-Instruct-2507-UD-Q6_K_XL__rocm-7_2__hblt0__fa1.log similarity index 100% rename from benchmark/results/Qwen3-30B-A3B-Instruct-2507-UD-Q6_K_XL__rocm-7.2__hblt0__fa1.log rename to benchmark/results/Qwen3-30B-A3B-Instruct-2507-UD-Q6_K_XL__rocm-7_2__hblt0__fa1.log diff --git a/benchmark/results/Qwen3-30B-A3B-Instruct-2507-UD-Q6_K_XL__rocm-7.2__hblt0__fa1__longctx32768.log b/benchmark/results/Qwen3-30B-A3B-Instruct-2507-UD-Q6_K_XL__rocm-7_2__hblt0__fa1__longctx32768.log similarity index 100% rename from benchmark/results/Qwen3-30B-A3B-Instruct-2507-UD-Q6_K_XL__rocm-7.2__hblt0__fa1__longctx32768.log rename to benchmark/results/Qwen3-30B-A3B-Instruct-2507-UD-Q6_K_XL__rocm-7_2__hblt0__fa1__longctx32768.log diff --git a/benchmark/results/Qwen3-Coder-30B-A3B-Instruct-Q4_K_M__rocm-7.2__fa1.log b/benchmark/results/Qwen3-Coder-30B-A3B-Instruct-Q4_K_M__rocm-7_2__fa1.log similarity index 100% rename from benchmark/results/Qwen3-Coder-30B-A3B-Instruct-Q4_K_M__rocm-7.2__fa1.log rename to benchmark/results/Qwen3-Coder-30B-A3B-Instruct-Q4_K_M__rocm-7_2__fa1.log diff --git a/benchmark/results/Qwen3-Coder-30B-A3B-Instruct-Q4_K_M__rocm-7.2__fa1__longctx32768.log b/benchmark/results/Qwen3-Coder-30B-A3B-Instruct-Q4_K_M__rocm-7_2__fa1__longctx32768.log similarity index 100% rename from benchmark/results/Qwen3-Coder-30B-A3B-Instruct-Q4_K_M__rocm-7.2__fa1__longctx32768.log rename to benchmark/results/Qwen3-Coder-30B-A3B-Instruct-Q4_K_M__rocm-7_2__fa1__longctx32768.log diff --git a/benchmark/results/Qwen3-Coder-30B-A3B-Instruct-Q4_K_M__rocm-7.2__hblt0__fa1.log b/benchmark/results/Qwen3-Coder-30B-A3B-Instruct-Q4_K_M__rocm-7_2__hblt0__fa1.log similarity index 100% rename from benchmark/results/Qwen3-Coder-30B-A3B-Instruct-Q4_K_M__rocm-7.2__hblt0__fa1.log rename to benchmark/results/Qwen3-Coder-30B-A3B-Instruct-Q4_K_M__rocm-7_2__hblt0__fa1.log diff --git a/benchmark/results/Qwen3-Coder-30B-A3B-Instruct-Q4_K_M__rocm-7.2__hblt0__fa1__longctx32768.log b/benchmark/results/Qwen3-Coder-30B-A3B-Instruct-Q4_K_M__rocm-7_2__hblt0__fa1__longctx32768.log similarity index 100% rename from benchmark/results/Qwen3-Coder-30B-A3B-Instruct-Q4_K_M__rocm-7.2__hblt0__fa1__longctx32768.log rename to benchmark/results/Qwen3-Coder-30B-A3B-Instruct-Q4_K_M__rocm-7_2__hblt0__fa1__longctx32768.log diff --git a/benchmark/results/Qwen3-Next-80B-A3B-Instruct-UD-Q8_K_XL-00001-of-00002__rocm-7.2__fa1.log b/benchmark/results/Qwen3-Next-80B-A3B-Instruct-UD-Q8_K_XL-00001-of-00002__rocm-7_2__fa1.log similarity index 100% rename from benchmark/results/Qwen3-Next-80B-A3B-Instruct-UD-Q8_K_XL-00001-of-00002__rocm-7.2__fa1.log rename to benchmark/results/Qwen3-Next-80B-A3B-Instruct-UD-Q8_K_XL-00001-of-00002__rocm-7_2__fa1.log diff --git a/benchmark/results/Qwen3-Next-80B-A3B-Instruct-UD-Q8_K_XL-00001-of-00002__rocm-7.2__fa1__longctx32768.log b/benchmark/results/Qwen3-Next-80B-A3B-Instruct-UD-Q8_K_XL-00001-of-00002__rocm-7_2__fa1__longctx32768.log similarity index 100% rename from benchmark/results/Qwen3-Next-80B-A3B-Instruct-UD-Q8_K_XL-00001-of-00002__rocm-7.2__fa1__longctx32768.log rename to benchmark/results/Qwen3-Next-80B-A3B-Instruct-UD-Q8_K_XL-00001-of-00002__rocm-7_2__fa1__longctx32768.log diff --git a/benchmark/results/Qwen3-Next-80B-A3B-Instruct-UD-Q8_K_XL-00001-of-00002__rocm-7.2__hblt0__fa1.log b/benchmark/results/Qwen3-Next-80B-A3B-Instruct-UD-Q8_K_XL-00001-of-00002__rocm-7_2__hblt0__fa1.log similarity index 100% rename from benchmark/results/Qwen3-Next-80B-A3B-Instruct-UD-Q8_K_XL-00001-of-00002__rocm-7.2__hblt0__fa1.log rename to benchmark/results/Qwen3-Next-80B-A3B-Instruct-UD-Q8_K_XL-00001-of-00002__rocm-7_2__hblt0__fa1.log diff --git a/benchmark/results/Qwen3-Next-80B-A3B-Instruct-UD-Q8_K_XL-00001-of-00002__rocm-7.2__hblt0__fa1__longctx32768.log b/benchmark/results/Qwen3-Next-80B-A3B-Instruct-UD-Q8_K_XL-00001-of-00002__rocm-7_2__hblt0__fa1__longctx32768.log similarity index 100% rename from benchmark/results/Qwen3-Next-80B-A3B-Instruct-UD-Q8_K_XL-00001-of-00002__rocm-7.2__hblt0__fa1__longctx32768.log rename to benchmark/results/Qwen3-Next-80B-A3B-Instruct-UD-Q8_K_XL-00001-of-00002__rocm-7_2__hblt0__fa1__longctx32768.log diff --git a/benchmark/results/gemma-3-12b-it-UD-Q8_K_XL__rocm-7.2__fa1.log b/benchmark/results/gemma-3-12b-it-UD-Q8_K_XL__rocm-7_2__fa1.log similarity index 100% rename from benchmark/results/gemma-3-12b-it-UD-Q8_K_XL__rocm-7.2__fa1.log rename to benchmark/results/gemma-3-12b-it-UD-Q8_K_XL__rocm-7_2__fa1.log diff --git a/benchmark/results/gemma-3-12b-it-UD-Q8_K_XL__rocm-7.2__fa1__longctx32768.log b/benchmark/results/gemma-3-12b-it-UD-Q8_K_XL__rocm-7_2__fa1__longctx32768.log similarity index 100% rename from benchmark/results/gemma-3-12b-it-UD-Q8_K_XL__rocm-7.2__fa1__longctx32768.log rename to benchmark/results/gemma-3-12b-it-UD-Q8_K_XL__rocm-7_2__fa1__longctx32768.log diff --git a/benchmark/results/gemma-3-12b-it-UD-Q8_K_XL__rocm-7.2__hblt0__fa1.log b/benchmark/results/gemma-3-12b-it-UD-Q8_K_XL__rocm-7_2__hblt0__fa1.log similarity index 100% rename from benchmark/results/gemma-3-12b-it-UD-Q8_K_XL__rocm-7.2__hblt0__fa1.log rename to benchmark/results/gemma-3-12b-it-UD-Q8_K_XL__rocm-7_2__hblt0__fa1.log diff --git a/benchmark/results/gemma-3-12b-it-UD-Q8_K_XL__rocm-7.2__hblt0__fa1__longctx32768.log b/benchmark/results/gemma-3-12b-it-UD-Q8_K_XL__rocm-7_2__hblt0__fa1__longctx32768.log similarity index 100% rename from benchmark/results/gemma-3-12b-it-UD-Q8_K_XL__rocm-7.2__hblt0__fa1__longctx32768.log rename to benchmark/results/gemma-3-12b-it-UD-Q8_K_XL__rocm-7_2__hblt0__fa1__longctx32768.log diff --git a/benchmark/results/gemma-3-27b-it-BF16-00001-of-00002__rocm-7.2__fa1.log b/benchmark/results/gemma-3-27b-it-BF16-00001-of-00002__rocm-7_2__fa1.log similarity index 100% rename from benchmark/results/gemma-3-27b-it-BF16-00001-of-00002__rocm-7.2__fa1.log rename to benchmark/results/gemma-3-27b-it-BF16-00001-of-00002__rocm-7_2__fa1.log diff --git a/benchmark/results/gemma-3-27b-it-BF16-00001-of-00002__rocm-7.2__fa1__longctx32768.log b/benchmark/results/gemma-3-27b-it-BF16-00001-of-00002__rocm-7_2__fa1__longctx32768.log similarity index 100% rename from benchmark/results/gemma-3-27b-it-BF16-00001-of-00002__rocm-7.2__fa1__longctx32768.log rename to benchmark/results/gemma-3-27b-it-BF16-00001-of-00002__rocm-7_2__fa1__longctx32768.log diff --git a/benchmark/results/gemma-3-27b-it-BF16-00001-of-00002__rocm-7.2__hblt0__fa1.log b/benchmark/results/gemma-3-27b-it-BF16-00001-of-00002__rocm-7_2__hblt0__fa1.log similarity index 100% rename from benchmark/results/gemma-3-27b-it-BF16-00001-of-00002__rocm-7.2__hblt0__fa1.log rename to benchmark/results/gemma-3-27b-it-BF16-00001-of-00002__rocm-7_2__hblt0__fa1.log diff --git a/benchmark/results/gemma-3-27b-it-BF16-00001-of-00002__rocm-7.2__hblt0__fa1__longctx32768.log b/benchmark/results/gemma-3-27b-it-BF16-00001-of-00002__rocm-7_2__hblt0__fa1__longctx32768.log similarity index 100% rename from benchmark/results/gemma-3-27b-it-BF16-00001-of-00002__rocm-7.2__hblt0__fa1__longctx32768.log rename to benchmark/results/gemma-3-27b-it-BF16-00001-of-00002__rocm-7_2__hblt0__fa1__longctx32768.log diff --git a/benchmark/results/gemma-3-4b-it-Q3_K_S__rocm-7.2__fa1.log b/benchmark/results/gemma-3-4b-it-Q3_K_S__rocm-7_2__fa1.log similarity index 100% rename from benchmark/results/gemma-3-4b-it-Q3_K_S__rocm-7.2__fa1.log rename to benchmark/results/gemma-3-4b-it-Q3_K_S__rocm-7_2__fa1.log diff --git a/benchmark/results/gemma-3-4b-it-Q3_K_S__rocm-7.2__fa1__longctx32768.log b/benchmark/results/gemma-3-4b-it-Q3_K_S__rocm-7_2__fa1__longctx32768.log similarity index 100% rename from benchmark/results/gemma-3-4b-it-Q3_K_S__rocm-7.2__fa1__longctx32768.log rename to benchmark/results/gemma-3-4b-it-Q3_K_S__rocm-7_2__fa1__longctx32768.log diff --git a/benchmark/results/gemma-3-4b-it-Q3_K_S__rocm-7.2__hblt0__fa1.log b/benchmark/results/gemma-3-4b-it-Q3_K_S__rocm-7_2__hblt0__fa1.log similarity index 100% rename from benchmark/results/gemma-3-4b-it-Q3_K_S__rocm-7.2__hblt0__fa1.log rename to benchmark/results/gemma-3-4b-it-Q3_K_S__rocm-7_2__hblt0__fa1.log diff --git a/benchmark/results/gemma-3-4b-it-Q3_K_S__rocm-7.2__hblt0__fa1__longctx32768.log b/benchmark/results/gemma-3-4b-it-Q3_K_S__rocm-7_2__hblt0__fa1__longctx32768.log similarity index 100% rename from benchmark/results/gemma-3-4b-it-Q3_K_S__rocm-7.2__hblt0__fa1__longctx32768.log rename to benchmark/results/gemma-3-4b-it-Q3_K_S__rocm-7_2__hblt0__fa1__longctx32768.log diff --git a/benchmark/results/gpt-oss-120b-mxfp4-00001-of-00003__rocm-7.2__fa1.log b/benchmark/results/gpt-oss-120b-mxfp4-00001-of-00003__rocm-7_2__fa1.log similarity index 100% rename from benchmark/results/gpt-oss-120b-mxfp4-00001-of-00003__rocm-7.2__fa1.log rename to benchmark/results/gpt-oss-120b-mxfp4-00001-of-00003__rocm-7_2__fa1.log diff --git a/benchmark/results/gpt-oss-120b-mxfp4-00001-of-00003__rocm-7.2__fa1__longctx32768.log b/benchmark/results/gpt-oss-120b-mxfp4-00001-of-00003__rocm-7_2__fa1__longctx32768.log similarity index 100% rename from benchmark/results/gpt-oss-120b-mxfp4-00001-of-00003__rocm-7.2__fa1__longctx32768.log rename to benchmark/results/gpt-oss-120b-mxfp4-00001-of-00003__rocm-7_2__fa1__longctx32768.log diff --git a/benchmark/results/gpt-oss-120b-mxfp4-00001-of-00003__rocm-7.2__hblt0__fa1.log b/benchmark/results/gpt-oss-120b-mxfp4-00001-of-00003__rocm-7_2__hblt0__fa1.log similarity index 100% rename from benchmark/results/gpt-oss-120b-mxfp4-00001-of-00003__rocm-7.2__hblt0__fa1.log rename to benchmark/results/gpt-oss-120b-mxfp4-00001-of-00003__rocm-7_2__hblt0__fa1.log diff --git a/benchmark/results/gpt-oss-120b-mxfp4-00001-of-00003__rocm-7.2__hblt0__fa1__longctx32768.log b/benchmark/results/gpt-oss-120b-mxfp4-00001-of-00003__rocm-7_2__hblt0__fa1__longctx32768.log similarity index 100% rename from benchmark/results/gpt-oss-120b-mxfp4-00001-of-00003__rocm-7.2__hblt0__fa1__longctx32768.log rename to benchmark/results/gpt-oss-120b-mxfp4-00001-of-00003__rocm-7_2__hblt0__fa1__longctx32768.log diff --git a/benchmark/results/gpt-oss-20b-mxfp4__rocm-7.2__fa1.log b/benchmark/results/gpt-oss-20b-mxfp4__rocm-7_2__fa1.log similarity index 100% rename from benchmark/results/gpt-oss-20b-mxfp4__rocm-7.2__fa1.log rename to benchmark/results/gpt-oss-20b-mxfp4__rocm-7_2__fa1.log diff --git a/benchmark/results/gpt-oss-20b-mxfp4__rocm-7.2__fa1__longctx32768.log b/benchmark/results/gpt-oss-20b-mxfp4__rocm-7_2__fa1__longctx32768.log similarity index 100% rename from benchmark/results/gpt-oss-20b-mxfp4__rocm-7.2__fa1__longctx32768.log rename to benchmark/results/gpt-oss-20b-mxfp4__rocm-7_2__fa1__longctx32768.log diff --git a/benchmark/results/gpt-oss-20b-mxfp4__rocm-7.2__hblt0__fa1.log b/benchmark/results/gpt-oss-20b-mxfp4__rocm-7_2__hblt0__fa1.log similarity index 100% rename from benchmark/results/gpt-oss-20b-mxfp4__rocm-7.2__hblt0__fa1.log rename to benchmark/results/gpt-oss-20b-mxfp4__rocm-7_2__hblt0__fa1.log diff --git a/benchmark/results/gpt-oss-20b-mxfp4__rocm-7.2__hblt0__fa1__longctx32768.log b/benchmark/results/gpt-oss-20b-mxfp4__rocm-7_2__hblt0__fa1__longctx32768.log similarity index 100% rename from benchmark/results/gpt-oss-20b-mxfp4__rocm-7.2__hblt0__fa1__longctx32768.log rename to benchmark/results/gpt-oss-20b-mxfp4__rocm-7_2__hblt0__fa1__longctx32768.log diff --git a/benchmark/results/llama-2-7b.Q4_0__rocm-7.2__fa1.log b/benchmark/results/llama-2-7b.Q4_0__rocm-7_2__fa1.log similarity index 100% rename from benchmark/results/llama-2-7b.Q4_0__rocm-7.2__fa1.log rename to benchmark/results/llama-2-7b.Q4_0__rocm-7_2__fa1.log diff --git a/benchmark/results/llama-2-7b.Q4_0__rocm-7.2__fa1__longctx32768.log b/benchmark/results/llama-2-7b.Q4_0__rocm-7_2__fa1__longctx32768.log similarity index 100% rename from benchmark/results/llama-2-7b.Q4_0__rocm-7.2__fa1__longctx32768.log rename to benchmark/results/llama-2-7b.Q4_0__rocm-7_2__fa1__longctx32768.log diff --git a/benchmark/results/llama-2-7b.Q4_0__rocm-7.2__hblt0__fa1.log b/benchmark/results/llama-2-7b.Q4_0__rocm-7_2__hblt0__fa1.log similarity index 100% rename from benchmark/results/llama-2-7b.Q4_0__rocm-7.2__hblt0__fa1.log rename to benchmark/results/llama-2-7b.Q4_0__rocm-7_2__hblt0__fa1.log diff --git a/benchmark/results/llama-2-7b.Q4_0__rocm-7.2__hblt0__fa1__longctx32768.log b/benchmark/results/llama-2-7b.Q4_0__rocm-7_2__hblt0__fa1__longctx32768.log similarity index 100% rename from benchmark/results/llama-2-7b.Q4_0__rocm-7.2__hblt0__fa1__longctx32768.log rename to benchmark/results/llama-2-7b.Q4_0__rocm-7_2__hblt0__fa1__longctx32768.log diff --git a/benchmark/run_benchmarks.sh b/benchmark/run_benchmarks.sh index ddd86ea..71ede28 100755 --- a/benchmark/run_benchmarks.sh +++ b/benchmark/run_benchmarks.sh @@ -62,8 +62,7 @@ echo declare -A CMDS=( [rocm6_4_4]="toolbox run -c llama-rocm-6.4.4 -- /usr/local/bin/llama-bench" - - [rocm-7.2]="toolbox run -c llama-rocm-7.2 -- /usr/local/bin/llama-bench" + [rocm-7_2]="toolbox run -c llama-rocm-7.2 -- /usr/local/bin/llama-bench" [rocm7-nightlies]="toolbox run -c llama-rocm7-nightlies -- /usr/local/bin/llama-bench" [vulkan_amdvlk]="toolbox run -c llama-vulkan-amdvlk -- /usr/sbin/llama-bench" [vulkan_radv]="toolbox run -c llama-vulkan-radv -- /usr/sbin/llama-bench" diff --git a/benchmark/run_rpc_benchmarks.sh b/benchmark/run_rpc_benchmarks.sh index 2751f2f..4a37f99 100755 --- a/benchmark/run_rpc_benchmarks.sh +++ b/benchmark/run_rpc_benchmarks.sh @@ -28,7 +28,7 @@ fi declare -A TOOLBOX_IMAGES=( [rocm6_4_4]="llama-rocm-6.4.4" - [rocm-7.2]="llama-rocm-7.2" + [rocm-7_2]="llama-rocm-7.2" [rocm7-nightlies]="llama-rocm7-nightlies" [vulkan_amdvlk]="llama-vulkan-amdvlk" [vulkan_radv]="llama-vulkan-radv" @@ -36,8 +36,7 @@ declare -A TOOLBOX_IMAGES=( declare -A CLIENT_CMDS=( [rocm6_4_4]="toolbox run -c llama-rocm-6.4.4 -- /usr/local/bin/llama-bench" - - [rocm-7.2]="toolbox run -c llama-rocm-7.2 -- /usr/local/bin/llama-bench" + [rocm-7_2]="toolbox run -c llama-rocm-7.2 -- /usr/local/bin/llama-bench" [rocm7-nightlies]="toolbox run -c llama-rocm7-nightlies -- /usr/local/bin/llama-bench" [vulkan_amdvlk]="toolbox run -c llama-vulkan-amdvlk -- /usr/sbin/llama-bench" [vulkan_radv]="toolbox run -c llama-vulkan-radv -- /usr/sbin/llama-bench" @@ -45,8 +44,7 @@ declare -A CLIENT_CMDS=( ENVIRONMENTS=( rocm6_4_4 - - rocm-7.2 + rocm-7_2 rocm7-nightlies vulkan_amdvlk vulkan_radv diff --git a/docs/index.html b/docs/index.html index a0ab727..27145f1 100644 --- a/docs/index.html +++ b/docs/index.html @@ -108,9 +108,10 @@