Skip to content

feat(agentx): DSv4-Pro FP4 MTP DPattn+EP8 on FlyDSL mega-MoE (conc 32) / 新增 DSv4-Pro FP4 MTP DPattn+EP8 FlyDSL mega-MoE 配置（并发 32） #2503

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Closed
jiacao-amd wants to merge 7 commits into from
Closed

feat(agentx): DSv4-Pro FP4 MTP DPattn+EP8 on FlyDSL mega-MoE (conc 32) / 新增 DSv4-Pro FP4 MTP DPattn+EP8 FlyDSL mega-MoE 配置（并发 32） #2503

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
7 commits
Select commit Hold shift + click to select a range
d8be7ba
feat(agentx): dsv4-fp4-mi355x-vllm-agentic-mtp DPattn+EP8 on FlyDSL m…
jiacao-amd Aug 5, 2026
3c6e77c
chore(agentx): backfill perf-changelog pr-link to #2503 / 回填 perf-cha…
jiacao-amd Aug 5, 2026
10077e4
fix(agentx): drop --moe-backend flydsl, DSv4-Pro MXFP4 rejects it at …
jiacao-amd Aug 5, 2026
2fe1f2b
feat(agentx): capture torch-profiler traces during AgentX replay
jiacao-amd Aug 5, 2026
10c6808
chore(agentx): add DEP-only config for trace collection
jiacao-amd Aug 5, 2026
3464ea3
fix(agentx): configure profiler via --profiler-config, not the remove…
jiacao-amd Aug 5, 2026
6e9f17c
chore(agentx): add DPattn-only (EP off) config for the DEP comparison
jiacao-amd Aug 5, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
18 changes: 18 additions & 0 deletions .github/workflows/benchmark-tmpl.yml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -139,6 +139,21 @@ on:
required: false
type: boolean
default: false
agentx-trace:
description: "Capture a short torch-profiler trace mid-replay (AgentX recipes that implement it)"
required: false
type: boolean
default: false
agentx-trace-delay-s:
description: "Seconds into the AgentX replay before the trace window opens"
required: false
type: string
default: ''
agentx-trace-window-s:
description: "Length of the trace window in seconds"
required: false
type: string
default: ''
eval-limit:
description: "Eval instance count: empty/full = whole split (default); N = first-N smoke slice"
required: false
Expand Down Expand Up @@ -186,6 +201,9 @@ env:
DURATION: ${{ inputs.duration }}
REQUIRE_POWER: ${{ inputs.require-power && '1' || '0' }}
AIPERF_EXPERIMENTAL_FAST: ${{ inputs.agentx-fast && '1' || '0' }}
AGENTX_TRACE: ${{ inputs.agentx-trace && '1' || '0' }}
AGENTX_TRACE_DELAY_S: ${{ inputs.agentx-trace-delay-s }}
AGENTX_TRACE_WINDOW_S: ${{ inputs.agentx-trace-window-s }}
EVAL_LIMIT: ${{ inputs.eval-limit }}
SWEBENCH_GEN_MODE: ${{ inputs.swebench-gen-mode }}
AIPERF_FAILED_REQUEST_THRESHOLD: '0.10'
Expand Down
33 changes: 33 additions & 0 deletions .github/workflows/e2e-tests.yml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -40,6 +40,21 @@ on:
required: false
type: boolean
default: false
agentx-trace:
description: "Capture a short torch-profiler trace mid-replay (AgentX recipes that implement it)."
required: false
type: boolean
default: false
agentx-trace-delay-s:
description: "Seconds into the AgentX replay before the trace window opens."
required: false
type: string
default: ""
agentx-trace-window-s:
description: "Length of the trace window in seconds."
required: false
type: string
default: ""
eval-limit:
description: "Eval instance count: empty/full = whole split (default); N = first-N smoke slice"
required: false
Expand Down Expand Up @@ -120,6 +135,21 @@ on:
required: false
type: boolean
default: false
agentx-trace:
description: "Capture a short torch-profiler trace mid-replay (AgentX recipes that implement it)."
required: false
type: boolean
default: false
agentx-trace-delay-s:
description: "Seconds into the AgentX replay before the trace window opens."
required: false
type: string
default: ""
agentx-trace-window-s:
description: "Length of the trace window in seconds."
required: false
type: string
default: ""
eval-limit:
description: "Eval instance count: empty/full = whole split (default); N = first-N smoke slice"
required: false
Expand Down Expand Up @@ -412,6 +442,9 @@ jobs:
total-cpu-dram-gb: ${{ matrix.config.total-cpu-dram-gb }}
duration: ${{ inputs.agentx-fast && '1200' || (inputs.duration-override != '' && inputs.duration-override || matrix.config.duration) }}
agentx-fast: ${{ inputs.agentx-fast }}
agentx-trace: ${{ inputs.agentx-trace }}
agentx-trace-delay-s: ${{ inputs.agentx-trace-delay-s }}
agentx-trace-window-s: ${{ inputs.agentx-trace-window-s }}
isl: '0'
osl: '0'
max-model-len: '0'
Expand Down
Loading