From 21b7546a05e225f45aa2436dc3f583e73faa2b7a Mon Sep 17 00:00:00 2001 From: "Chuan (Richard) Li" Date: Fri, 13 Feb 2026 09:49:03 -0800 Subject: [PATCH 1/6] Update search-space configurations in amd-master.yaml revise the search space config --- .github/configs/amd-master.yaml | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) diff --git a/.github/configs/amd-master.yaml b/.github/configs/amd-master.yaml index 0dba92a59..9fbd35a99 100644 --- a/.github/configs/amd-master.yaml +++ b/.github/configs/amd-master.yaml @@ -33,18 +33,18 @@ dsr1-fp4-mi355x-atom: - isl: 1024 osl: 1024 search-space: - - { tp: 4, ep: 1, conc-start: 32, conc-end: 128 } - - { tp: 8, ep: 1, conc-start: 4, conc-end: 32 } + - { tp: 4, ep: 1, conc-start: 4, conc-end: 128 } + - { tp: 8, ep: 1, conc-start: 4, conc-end: 128 } - isl: 1024 osl: 8192 search-space: - - { tp: 4, ep: 1, conc-start: 128, conc-end: 128 } + - { tp: 4, ep: 1, conc-start: 4, conc-end: 128 } - { tp: 8, ep: 1, conc-start: 4, conc-end: 128 } - isl: 8192 osl: 1024 search-space: - { tp: 4, ep: 1, conc-start: 4, conc-end: 128 } - - { tp: 8, ep: 1, conc-start: 4, conc-end: 4 } + - { tp: 8, ep: 1, conc-start: 4, conc-end: 128 } dsr1-fp4-mi355x-atom-mtp: image: rocm/atom:rocm7.2.0-ubuntu24.04-pytorch2.9-atom0.1.1 From 958a967e29e4eab6f8ea97ea6215874a4e44daf4 Mon Sep 17 00:00:00 2001 From: "Chuan (Richard) Li" Date: Fri, 13 Feb 2026 10:06:17 -0800 Subject: [PATCH 2/6] Update conc-end values in amd-master.yaml Updated to fullscan --- .github/configs/amd-master.yaml | 14 +++++++------- 1 file changed, 7 insertions(+), 7 deletions(-) diff --git a/.github/configs/amd-master.yaml b/.github/configs/amd-master.yaml index 9fbd35a99..b83d83c8e 100644 --- a/.github/configs/amd-master.yaml +++ b/.github/configs/amd-master.yaml @@ -33,18 +33,18 @@ dsr1-fp4-mi355x-atom: - isl: 1024 osl: 1024 search-space: - - { tp: 4, ep: 1, conc-start: 4, conc-end: 128 } - - { tp: 8, ep: 1, conc-start: 4, conc-end: 128 } + - { tp: 4, ep: 1, conc-start: 4, conc-end: 256 } + - { tp: 8, ep: 1, conc-start: 4, conc-end: 256 } - isl: 1024 osl: 8192 search-space: - - { tp: 4, ep: 1, conc-start: 4, conc-end: 128 } - - { tp: 8, ep: 1, conc-start: 4, conc-end: 128 } + - { tp: 4, ep: 1, conc-start: 4, conc-end: 256 } + - { tp: 8, ep: 1, conc-start: 4, conc-end: 256 } - isl: 8192 osl: 1024 search-space: - - { tp: 4, ep: 1, conc-start: 4, conc-end: 128 } - - { tp: 8, ep: 1, conc-start: 4, conc-end: 128 } + - { tp: 4, ep: 1, conc-start: 4, conc-end: 256 } + - { tp: 8, ep: 1, conc-start: 4, conc-end: 256 } dsr1-fp4-mi355x-atom-mtp: image: rocm/atom:rocm7.2.0-ubuntu24.04-pytorch2.9-atom0.1.1 @@ -64,7 +64,7 @@ dsr1-fp4-mi355x-atom-mtp: - isl: 1024 osl: 8192 search-space: - - { tp: 8, conc-start: 256, conc-end: 256, spec-decoding: mtp } + - { tp: 8, conc-start: 4, conc-end: 256, spec-decoding: mtp } - isl: 8192 osl: 1024 search-space: From 861c9180f5c1b4992bb5b8afd0dc02bed0c41bf4 Mon Sep 17 00:00:00 2001 From: "Chuan (Richard) Li" Date: Thu, 19 Feb 2026 22:57:35 -0800 Subject: [PATCH 3/6] Update amd-master.yaml Reduce the search space to depends less on the compute as most of the pareto frontier fall under TP=8. --- .github/configs/amd-master.yaml | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) diff --git a/.github/configs/amd-master.yaml b/.github/configs/amd-master.yaml index b83d83c8e..50a4c74d5 100644 --- a/.github/configs/amd-master.yaml +++ b/.github/configs/amd-master.yaml @@ -33,17 +33,17 @@ dsr1-fp4-mi355x-atom: - isl: 1024 osl: 1024 search-space: - - { tp: 4, ep: 1, conc-start: 4, conc-end: 256 } + # - { tp: 4, ep: 1, conc-start: 4, conc-end: 256 } - { tp: 8, ep: 1, conc-start: 4, conc-end: 256 } - isl: 1024 osl: 8192 search-space: - - { tp: 4, ep: 1, conc-start: 4, conc-end: 256 } + # - { tp: 4, ep: 1, conc-start: 4, conc-end: 256 } - { tp: 8, ep: 1, conc-start: 4, conc-end: 256 } - isl: 8192 osl: 1024 search-space: - - { tp: 4, ep: 1, conc-start: 4, conc-end: 256 } + # - { tp: 4, ep: 1, conc-start: 4, conc-end: 256 } - { tp: 8, ep: 1, conc-start: 4, conc-end: 256 } dsr1-fp4-mi355x-atom-mtp: @@ -59,7 +59,7 @@ dsr1-fp4-mi355x-atom-mtp: - isl: 1024 osl: 1024 search-space: - - { tp: 4, conc-start: 4, conc-end: 256, spec-decoding: mtp } + # - { tp: 4, conc-start: 4, conc-end: 256, spec-decoding: mtp } - { tp: 8, conc-start: 4, conc-end: 256, spec-decoding: mtp } - isl: 1024 osl: 8192 From 11b94484080e73bea50aa98b51c209ac2f8a5492 Mon Sep 17 00:00:00 2001 From: Cam Quilici Date: Mon, 23 Feb 2026 15:07:33 -0600 Subject: [PATCH 4/6] add perf changelog --- perf-changelog.yaml | 10 ++++++++++ 1 file changed, 10 insertions(+) diff --git a/perf-changelog.yaml b/perf-changelog.yaml index 344d4129f..548b021ae 100644 --- a/perf-changelog.yaml +++ b/perf-changelog.yaml @@ -695,3 +695,13 @@ description: - "Add more configs for MI355X FP8 Disagg" pr-link: https://github.com/SemiAnalysisAI/InferenceX/pull/770 + +- config-keys: + - dsr1-fp4-mi355x-atom + - dsr1-fp4-mi355x-atom-mtp + description: + - "Update search-space configurations for DSR1 FP4 MI355X ATOM and ATOM-MTP" + - "Comment out TP=4 configs, consolidate to TP=8 only" + - "Extend concurrency range to conc-end: 256 across all sequence lengths (1k1k, 1k8k, 8k1k)" + - "Fix MTP 1k8k conc-start from 256 to 4 to enable full concurrency sweep" + pr-link: https://github.com/SemiAnalysisAI/InferenceX/pull/699 From 9f353901715b28c9b4f76472c220456b5b4913bd Mon Sep 17 00:00:00 2001 From: "Chuan (Richard) Li" Date: Tue, 24 Feb 2026 21:42:16 -0800 Subject: [PATCH 5/6] Update amd-master.yaml Revise the changes to make sure we get the pareto's frontier. --- .github/configs/amd-master.yaml | 22 ++++++++++++---------- 1 file changed, 12 insertions(+), 10 deletions(-) diff --git a/.github/configs/amd-master.yaml b/.github/configs/amd-master.yaml index 516d15351..7d39132d6 100644 --- a/.github/configs/amd-master.yaml +++ b/.github/configs/amd-master.yaml @@ -33,18 +33,18 @@ dsr1-fp4-mi355x-atom: - isl: 1024 osl: 1024 search-space: - # - { tp: 4, ep: 1, conc-start: 4, conc-end: 256 } - - { tp: 8, ep: 1, conc-start: 4, conc-end: 256 } + - { tp: 4, ep: 1, conc-start: 32, conc-end: 256 } + - { tp: 8, ep: 1, conc-start: 4, conc-end: 32 } - isl: 1024 osl: 8192 search-space: - # - { tp: 4, ep: 1, conc-start: 4, conc-end: 256 } - - { tp: 8, ep: 1, conc-start: 4, conc-end: 256 } + - { tp: 4, ep: 1, conc-start: 32, conc-end: 256 } + - { tp: 8, ep: 1, conc-start: 4, conc-end: 32 } - isl: 8192 osl: 1024 search-space: - # - { tp: 4, ep: 1, conc-start: 4, conc-end: 256 } - - { tp: 8, ep: 1, conc-start: 4, conc-end: 256 } + - { tp: 4, ep: 1, conc-start: 4, conc-end: 256 } + - { tp: 8, ep: 1, conc-start: 4, conc-end: 32 } dsr1-fp4-mi355x-atom-mtp: image: rocm/atom:rocm7.2.0-ubuntu24.04-pytorch2.9-atom0.1.1 @@ -59,16 +59,18 @@ dsr1-fp4-mi355x-atom-mtp: - isl: 1024 osl: 1024 search-space: - # - { tp: 4, conc-start: 4, conc-end: 256, spec-decoding: mtp } - - { tp: 8, conc-start: 4, conc-end: 256, spec-decoding: mtp } + - { tp: 4, conc-start: 32, conc-end: 256, spec-decoding: mtp } + - { tp: 8, conc-start: 4, conc-end: 32, spec-decoding: mtp } - isl: 1024 osl: 8192 search-space: - - { tp: 8, conc-start: 4, conc-end: 256, spec-decoding: mtp } + - { tp: 4, conc-start: 32, conc-end: 256, spec-decoding: mtp } + - { tp: 8, conc-start: 4, conc-end: 32, spec-decoding: mtp } - isl: 8192 osl: 1024 search-space: - - { tp: 8, conc-start: 4, conc-end: 256, spec-decoding: mtp } + - { tp: 4, conc-start: 32, conc-end: 256, spec-decoding: mtp } + - { tp: 8, conc-start: 4, conc-end: 32, spec-decoding: mtp } dsr1-fp8-mi300x-sglang: image: lmsysorg/sglang:v0.5.8-rocm700-mi30x From a18adc3eac0a70bb60d34f809096d9c618da4eb6 Mon Sep 17 00:00:00 2001 From: "Chuan (Richard) Li" Date: Tue, 24 Feb 2026 21:45:50 -0800 Subject: [PATCH 6/6] Update amd-master.yaml update the search space --- .github/configs/amd-master.yaml | 18 +++++++++--------- 1 file changed, 9 insertions(+), 9 deletions(-) diff --git a/.github/configs/amd-master.yaml b/.github/configs/amd-master.yaml index 7d39132d6..7590eb2d1 100644 --- a/.github/configs/amd-master.yaml +++ b/.github/configs/amd-master.yaml @@ -38,13 +38,13 @@ dsr1-fp4-mi355x-atom: - isl: 1024 osl: 8192 search-space: - - { tp: 4, ep: 1, conc-start: 32, conc-end: 256 } - - { tp: 8, ep: 1, conc-start: 4, conc-end: 32 } + - { tp: 4, ep: 1, conc-start: 128, conc-end: 256 } + - { tp: 8, ep: 1, conc-start: 4, conc-end: 128 } - isl: 8192 osl: 1024 search-space: - { tp: 4, ep: 1, conc-start: 4, conc-end: 256 } - - { tp: 8, ep: 1, conc-start: 4, conc-end: 32 } + - { tp: 8, ep: 1, conc-start: 4, conc-end: 4 } dsr1-fp4-mi355x-atom-mtp: image: rocm/atom:rocm7.2.0-ubuntu24.04-pytorch2.9-atom0.1.1 @@ -59,18 +59,18 @@ dsr1-fp4-mi355x-atom-mtp: - isl: 1024 osl: 1024 search-space: - - { tp: 4, conc-start: 32, conc-end: 256, spec-decoding: mtp } - - { tp: 8, conc-start: 4, conc-end: 32, spec-decoding: mtp } + - { tp: 4, conc-start: 4, conc-end: 256, spec-decoding: mtp } + - { tp: 8, conc-start: 4, conc-end: 256, spec-decoding: mtp } - isl: 1024 osl: 8192 search-space: - - { tp: 4, conc-start: 32, conc-end: 256, spec-decoding: mtp } - - { tp: 8, conc-start: 4, conc-end: 32, spec-decoding: mtp } + # - { tp: 4, conc-start: 4, conc-end: 256, spec-decoding: mtp } + - { tp: 8, conc-start: 4, conc-end: 256, spec-decoding: mtp } - isl: 8192 osl: 1024 search-space: - - { tp: 4, conc-start: 32, conc-end: 256, spec-decoding: mtp } - - { tp: 8, conc-start: 4, conc-end: 32, spec-decoding: mtp } + #- { tp: 4, conc-start: 32, conc-end: 256, spec-decoding: mtp } + - { tp: 8, conc-start: 4, conc-end: 256, spec-decoding: mtp } dsr1-fp8-mi300x-sglang: image: lmsysorg/sglang:v0.5.8-rocm700-mi30x