[AMD] retrigger dsv4-fp4-mi355x-atom benchmark sweep (#1817)

* [AMD] dsv4-fp4-mi355x-atom: enable DPA TBO at high concurrency, update image to atom0.1.4

- Enable --enable-tbo for ISL=1024/OSL=1024 at CONC>=1024 and ISL=8192/OSL=1024 at CONC>=256
- Update image to atom0.1.4_20260612
- Update ISL=8192 search-space to start at conc=4 and use DPA from conc=128

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

* [AMD] perf-changelog: dsv4-fp4-mi355x-atom DPA TBO + image atom0.1.4

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

* [AMD] perf-changelog: add PR link #1717

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

* [AMD] dsv4_fp4_mi355x_atom.sh: disable prefix caching

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

* [AMD] dsv4-fp4-mi355x-atom: add max-model-len, eval context, extend conc range

- Pass --max-model-len to server using SERVE_MAX_MODEL_LEN
- Add EVAL_ONLY path: compute eval context length via compute_eval_context_length
- Extend conc-end to 8192 (isl=1024) and 4096 (isl=8192) in amd-master.yaml

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

* [AMD] dsv4-fp4-mi355x-atom: narrow eval to single conc=1024 point, disable max-model-len

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

* [AMD] dsv4_fp4_mi355x_atom.sh: add cudagraph-capture-sizes and max-num-seqs

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

* [AMD] dsv4-fp4-mi355x-atom: bump to nightly image, expand search space, enable max-model-len

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

* [AMD] set GPU_MAX_HW_QUEUES=5 in dsv4_fp4_mi355x_atom.sh

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

* [AMD] dsv4-fp4-mi355x-atom: disable TBO, add TP4 rows for isl=8192, cap conc ranges

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

* [AMD] dsv4_fp4_mi355x_atom.sh: quote SERVER_LOG variable

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

* [AMD] dsv4_fp4_mi355x_atom.sh: comment out dense cudagraph sizes

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

* [AMD] dsv4_fp4_mi355x_atom.sh: fix --hf-overrides JSON escaping

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

* [AMD] dsv4_fp4_mi355x_atom.sh: comment out dense cudagraph sizes

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

* [AMD] dsv4-fp4-mi355x-atom: expand search space, restore isl=1024 rows

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

* [AMD] perf-changelog: update dsv4-fp4-mi355x-atom image and search-space description

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

* [AMD] dsv4_fp4_mi355x_atom.sh: restore sparse cudagraph capture sizes

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

* [AMD] perf-changelog: revert dsv4-fp4-mi355x-atom image/search-space, remove stale entries

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

* [AMD] perf-changelog: add dsv4-fp4-mi355x-sglang entry for PR #1762

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

* update dsv4-fp4-mi355x-atom: bump image, enable TBO conditionally, fix mem frac

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

* expand dsv4-fp4-mi355x-atom search space: restore ISL1024 scenarios, add TP4/TP8 conc lists for ISL8192

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

* Update perf-changelog.yaml

* Update perf-changelog.yaml

* Update perf-changelog.yaml

* Update perf-changelog.yaml

* update perf-changelog: move dsv4-fp4-mi355x-atom entry to end

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

* narrow dsv4-fp4-mi355x-atom to DPA conc=256-2048 ISL8192, fix TBO branch override

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

* restore full dsv4-fp4-mi355x-atom search space: ISL1024 + ISL8192 TP4/TP8/DPA

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

* chore: retrigger dsv4 atom benchmark sweep

---------

Co-authored-by: seungrokj <seungrok.jung@amd.com>
Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
Co-authored-by: seungrokj <144636725+seungrokj@users.noreply.github.com>

Author: 3 people

perf-changelog.yaml

@@ -3935,3 +3935,11 @@
     - "Update ISL=8192 search-space: TP8-only from conc=4-64, DPA from conc=128-1024 (previously conc=1-64 and DPA conc=64-512)"
     - "Update Applied TBO on high concurrencies"
   pr-link: https://github.com/SemiAnalysisAI/InferenceX/pull/1717
+  
+- config-keys:
+    - dsv4-fp4-mi355x-atom
+  description:
+    - "Update image to rocm/atom:rocm7.2.4_ubuntu24.04_py3.12_pytorch_release_2.10.0_atom0.1.4_20260612"
+    - "Update ISL=8192 search-space: TP8-only from conc=4-64, DPA from conc=128-1024 (previously conc=1-64 and DPA conc=64-512)"
+    - "Update Applied TBO on high concurrencies"
+  pr-link: https://github.com/SemiAnalysisAI/InferenceX/pull/1717