From 6f075bd5cef43d83bfa32ea0a7846a6541f015dd Mon Sep 17 00:00:00 2001 From: zhenzew Date: Mon, 26 Oct 2026 08:18:34 -0500 Subject: [PATCH 1/5] ci(oga): apply microsoft/onnxruntime-genai#2714 and its dependencies Qwen3.8-Flash-Next needs the runtime in microsoft/onnxruntime-genai#2714, which is stacked on #2713 and on main changes made after v0.17.0. Add the 35 main PRs those diffs need, #2713, #2714 and #2782 (IndexerCache fix on top of #2714). Drop 2705, superseded by #2672 which is now in the list. Exclude Markdown from the applied diffs and retry .diff fetches on Windows. --- .github/workflows/linux-build.yml | 7 ++++--- .github/workflows/windows-deps.yml | 10 ++++++---- 2 files changed, 10 insertions(+), 7 deletions(-) diff --git a/.github/workflows/linux-build.yml b/.github/workflows/linux-build.yml index e00ef323e4..214172e2a5 100644 --- a/.github/workflows/linux-build.yml +++ b/.github/workflows/linux-build.yml @@ -35,7 +35,7 @@ env: # Space-separated microsoft/onnxruntime-genai PR numbers fetched from # pull/.diff and applied on top of v. Part of the OGA cache # key. Remove a PR once it lands in the pinned OGA release. - OGA_PR_PATCHES: "2685 2691 2705 2712" + OGA_PR_PATCHES: "2685 2691 2712 2600 2473 2524 2624 2627 2634 2642 2644 2584 2605 2631 2643 2645 2656 2660 2657 2654 2678 2677 2679 2686 2696 2672 2664 2594 2745 2746 2545 2503 2742 2606 2724 2648 2582 2700 2713 2714 2782" # The GPU arch the artifact targets. CI runners have no GPU, so it is passed # explicitly (build.py auto-detects from /sys on a real GPU host). HIP_ARCHITECTURES: gfx11-generic @@ -365,7 +365,8 @@ jobs: # and fails the step. Tests are excluded because OGA does not build # them and their fixtures do conflict; OGA's own CI workflows are # excluded because the build does not use them and some no longer - # exist at the pinned tag. + # exist at the pinned tag. Markdown is excluded because the build + # does not read it and doc edits across the listed PRs conflict. for pr in $prs; do url="https://github.com/microsoft/onnxruntime-genai/pull/${pr}.diff" echo "Fetching $url" @@ -374,7 +375,7 @@ jobs: curl -fsSL --retry 6 --retry-delay 10 --retry-all-errors \ "$url" -o "/tmp/oga-pr-${pr}.diff" echo "Applying PR #${pr}" - git apply --whitespace=nowarn --3way --exclude='test/*' \ + git apply --whitespace=n--exclude='*.md' owarn --3way --exclude='test/*' \ --exclude='.github/*' "/tmp/oga-pr-${pr}.diff" \ || { echo "PR #${pr} failed to apply"; exit 1; } done diff --git a/.github/workflows/windows-deps.yml b/.github/workflows/windows-deps.yml index 962210f87f..9ec6b23bc2 100644 --- a/.github/workflows/windows-deps.yml +++ b/.github/workflows/windows-deps.yml @@ -37,7 +37,7 @@ env: # Space-separated microsoft/onnxruntime-genai PR numbers fetched from # pull/.diff and applied on top of v. Part of the OGA cache # key. Remove a PR once it lands in the pinned OGA release. - OGA_PR_PATCHES: "2685 2691 2705 2712 2763" + OGA_PR_PATCHES: "2685 2691 2712 2763 2600 2473 2524 2624 2627 2634 2642 2644 2584 2605 2631 2643 2645 2656 2660 2657 2654 2678 2677 2679 2686 2696 2672 2664 2594 2745 2746 2545 2503 2742 2606 2724 2648 2582 2700 2713 2714 2782" ACTIONS_ALLOW_USE_UNSECURE_NODE_VERSION: "true" AMDGPU_EP_REF: "f3dccce2a142567da050da1032617fb39db4086a" # Space-separated onnxruntime/onnxruntime-ep-amdgpu PR numbers applied on top @@ -605,15 +605,17 @@ jobs: # and fails the step. Tests are excluded because OGA does not build # them and their fixtures do conflict; OGA's own CI workflows are # excluded because the build does not use them and some no longer - # exist at the pinned tag. + # exist at the pinned tag. Markdown is excluded because the build + # does not read it and doc edits across the listed PRs conflict. foreach ($pr in $prs -split '\s+') { if (-not $pr) { continue } $url = "https://github.com/microsoft/onnxruntime-genai/pull/$pr.diff" $tmp = Join-Path $env:RUNNER_TEMP "oga-pr-$pr.diff" Write-Host "Fetching $url" - Invoke-WebRequest -Uri $url -OutFile $tmp -UseBasicParsing + # github.com throttles raw .diff fetches with HTTP 429. + Invoke-WebRequest -Uri $url -OutFile $tmp -UseBasicParsing -MaximumRetryCount 6 -RetryIntervalSec 10 Write-Host "Applying PR #$pr" - git apply --whitespace=nowarn --3way --exclude='test/*' --exclude='.github/*' $tmp + git apply --whitespace=nowarn --3way --exclude='test/*' --exclude='.github/*' --exclude='*.md' $tmp if ($LASTEXITCODE -ne 0) { throw "PR #$pr failed to apply (exit $LASTEXITCODE)" } From f9f913e765d9ef64e9bb11e79924fbfb9ef78ff9 Mon Sep 17 00:00:00 2001 From: zhenzew Date: Mon, 26 Oct 2026 22:28:27 -0500 Subject: [PATCH 2/5] ci(oga): order OGA_PR_PATCHES by OGA main merge order --- .github/workflows/linux-build.yml | 2 +- .github/workflows/windows-deps.yml | 2 +- 2 files changed, 2 insertions(+), 2 deletions(-) diff --git a/.github/workflows/linux-build.yml b/.github/workflows/linux-build.yml index 214172e2a5..5b42b43266 100644 --- a/.github/workflows/linux-build.yml +++ b/.github/workflows/linux-build.yml @@ -35,7 +35,7 @@ env: # Space-separated microsoft/onnxruntime-genai PR numbers fetched from # pull/.diff and applied on top of v. Part of the OGA cache # key. Remove a PR once it lands in the pinned OGA release. - OGA_PR_PATCHES: "2685 2691 2712 2600 2473 2524 2624 2627 2634 2642 2644 2584 2605 2631 2643 2645 2656 2660 2657 2654 2678 2677 2679 2686 2696 2672 2664 2594 2745 2746 2545 2503 2742 2606 2724 2648 2582 2700 2713 2714 2782" + OGA_PR_PATCHES: "2600 2473 2524 2624 2627 2634 2642 2644 2584 2605 2631 2643 2645 2656 2660 2657 2654 2678 2677 2679 2686 2685 2696 2691 2672 2712 2664 2594 2745 2746 2545 2503 2742 2606 2724 2648 2582 2700 2713 2714 2782" # The GPU arch the artifact targets. CI runners have no GPU, so it is passed # explicitly (build.py auto-detects from /sys on a real GPU host). HIP_ARCHITECTURES: gfx11-generic diff --git a/.github/workflows/windows-deps.yml b/.github/workflows/windows-deps.yml index 9ec6b23bc2..1a707f0dd8 100644 --- a/.github/workflows/windows-deps.yml +++ b/.github/workflows/windows-deps.yml @@ -37,7 +37,7 @@ env: # Space-separated microsoft/onnxruntime-genai PR numbers fetched from # pull/.diff and applied on top of v. Part of the OGA cache # key. Remove a PR once it lands in the pinned OGA release. - OGA_PR_PATCHES: "2685 2691 2712 2763 2600 2473 2524 2624 2627 2634 2642 2644 2584 2605 2631 2643 2645 2656 2660 2657 2654 2678 2677 2679 2686 2696 2672 2664 2594 2745 2746 2545 2503 2742 2606 2724 2648 2582 2700 2713 2714 2782" + OGA_PR_PATCHES: "2600 2473 2524 2624 2627 2634 2642 2644 2584 2605 2631 2643 2645 2656 2660 2657 2654 2678 2677 2679 2686 2685 2696 2691 2672 2712 2664 2594 2745 2746 2545 2503 2742 2606 2724 2648 2582 2700 2763 2713 2714 2782" ACTIONS_ALLOW_USE_UNSECURE_NODE_VERSION: "true" AMDGPU_EP_REF: "f3dccce2a142567da050da1032617fb39db4086a" # Space-separated onnxruntime/onnxruntime-ep-amdgpu PR numbers applied on top From 6882e6421b3cc8101d6f25b8d7e6df22f3da54d5 Mon Sep 17 00:00:00 2001 From: zhenzew Date: Mon, 26 Oct 2026 22:37:11 -0500 Subject: [PATCH 3/5] ci(oga): replace 2763 with microsoft/onnxruntime-genai#2792 --- .github/workflows/windows-deps.yml | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/.github/workflows/windows-deps.yml b/.github/workflows/windows-deps.yml index 1a707f0dd8..574827b954 100644 --- a/.github/workflows/windows-deps.yml +++ b/.github/workflows/windows-deps.yml @@ -37,7 +37,7 @@ env: # Space-separated microsoft/onnxruntime-genai PR numbers fetched from # pull/.diff and applied on top of v. Part of the OGA cache # key. Remove a PR once it lands in the pinned OGA release. - OGA_PR_PATCHES: "2600 2473 2524 2624 2627 2634 2642 2644 2584 2605 2631 2643 2645 2656 2660 2657 2654 2678 2677 2679 2686 2685 2696 2691 2672 2712 2664 2594 2745 2746 2545 2503 2742 2606 2724 2648 2582 2700 2763 2713 2714 2782" + OGA_PR_PATCHES: "2600 2473 2524 2624 2627 2634 2642 2644 2584 2605 2631 2643 2645 2656 2660 2657 2654 2678 2677 2679 2686 2685 2696 2691 2672 2712 2664 2594 2745 2746 2545 2503 2742 2606 2724 2648 2582 2700 2792 2713 2714 2782" ACTIONS_ALLOW_USE_UNSECURE_NODE_VERSION: "true" AMDGPU_EP_REF: "f3dccce2a142567da050da1032617fb39db4086a" # Space-separated onnxruntime/onnxruntime-ep-amdgpu PR numbers applied on top From 734b6e147e99b04344cf3a7aabdd873ff832b77c Mon Sep 17 00:00:00 2001 From: zhenzew Date: Mon, 26 Oct 2026 22:56:42 -0500 Subject: [PATCH 4/5] ci(oga): fix the Markdown exclude in the Linux OGA patch step --- .github/workflows/linux-build.yml | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/.github/workflows/linux-build.yml b/.github/workflows/linux-build.yml index 5b42b43266..ce97fc90c6 100644 --- a/.github/workflows/linux-build.yml +++ b/.github/workflows/linux-build.yml @@ -375,8 +375,8 @@ jobs: curl -fsSL --retry 6 --retry-delay 10 --retry-all-errors \ "$url" -o "/tmp/oga-pr-${pr}.diff" echo "Applying PR #${pr}" - git apply --whitespace=n--exclude='*.md' owarn --3way --exclude='test/*' \ - --exclude='.github/*' "/tmp/oga-pr-${pr}.diff" \ + git apply --whitespace=nowarn --3way --exclude='test/*' \ + --exclude='.github/*' --exclude='*.md' "/tmp/oga-pr-${pr}.diff" \ || { echo "PR #${pr} failed to apply"; exit 1; } done From 097397468b35425d9aa5dc9c885c2fdf35e692f4 Mon Sep 17 00:00:00 2001 From: "Wang, Zhenze" Date: Wed, 28 Oct 2026 02:32:29 -0400 Subject: [PATCH 5/5] ci(oga): replace the Qwen3.8 patch chain with one v0.17.0 draft (#1351) ## Summary Replace the 42-PR OGA patch list used for Qwen3.8-Flash-Next with the existing patches plus two: microsoft/onnxruntime-genai#2792 (onnxruntime-extensions pin with `Qwen3_5Tokenizer`) and microsoft/onnxruntime-genai#2801 (Qwen3.8 runtime from #2714 plus the IndexerCache fix #2782, ported onto v0.17.0). CI vehicle stacked on #1335; not for merge. ## Related issue or design Stacked on #1335. ## Why The previous list applied #2713/#2714 together with 35 OGA main PRs they depend on, which moved a large part of OGA past v0.17.0. #2801 ports only the Generator-path runtime onto v0.17.0, so the OGA used by CI stays close to the release. ## What - `windows-deps.yml`, `linux-build.yml`: `OGA_PR_PATCHES` = `2685 2691 2705 2712 2792 2801`. Linux now also gets the extensions pin. ## Test plan - [x] Local: v0.17.0 + `pull/.diff` of 2685 2691 2705 2712 2792 2801 applied with the CI `git apply --3way` command; Windows DML Release build; Qwen3.8-Flash-Next image description correct, 1024-token prefill / decode on par with the previous chain build. - [ ] Windows Build and Linux Build green. - [ ] Jenkins daily_hip_ep_genai on this PR: Qwen3.8 VLM output correct. ## Notes for reviewers The OGA cache key only contains PR numbers; updating #2801 does not invalidate it. --- .github/workflows/linux-build.yml | 2 +- .github/workflows/windows-deps.yml | 2 +- 2 files changed, 2 insertions(+), 2 deletions(-) diff --git a/.github/workflows/linux-build.yml b/.github/workflows/linux-build.yml index ce97fc90c6..f20508f6ec 100644 --- a/.github/workflows/linux-build.yml +++ b/.github/workflows/linux-build.yml @@ -35,7 +35,7 @@ env: # Space-separated microsoft/onnxruntime-genai PR numbers fetched from # pull/.diff and applied on top of v. Part of the OGA cache # key. Remove a PR once it lands in the pinned OGA release. - OGA_PR_PATCHES: "2600 2473 2524 2624 2627 2634 2642 2644 2584 2605 2631 2643 2645 2656 2660 2657 2654 2678 2677 2679 2686 2685 2696 2691 2672 2712 2664 2594 2745 2746 2545 2503 2742 2606 2724 2648 2582 2700 2713 2714 2782" + OGA_PR_PATCHES: "2685 2691 2705 2712 2792 2801" # The GPU arch the artifact targets. CI runners have no GPU, so it is passed # explicitly (build.py auto-detects from /sys on a real GPU host). HIP_ARCHITECTURES: gfx11-generic diff --git a/.github/workflows/windows-deps.yml b/.github/workflows/windows-deps.yml index 574827b954..c2cda159d8 100644 --- a/.github/workflows/windows-deps.yml +++ b/.github/workflows/windows-deps.yml @@ -37,7 +37,7 @@ env: # Space-separated microsoft/onnxruntime-genai PR numbers fetched from # pull/.diff and applied on top of v. Part of the OGA cache # key. Remove a PR once it lands in the pinned OGA release. - OGA_PR_PATCHES: "2600 2473 2524 2624 2627 2634 2642 2644 2584 2605 2631 2643 2645 2656 2660 2657 2654 2678 2677 2679 2686 2685 2696 2691 2672 2712 2664 2594 2745 2746 2545 2503 2742 2606 2724 2648 2582 2700 2792 2713 2714 2782" + OGA_PR_PATCHES: "2685 2691 2705 2712 2792 2801" ACTIONS_ALLOW_USE_UNSECURE_NODE_VERSION: "true" AMDGPU_EP_REF: "f3dccce2a142567da050da1032617fb39db4086a" # Space-separated onnxruntime/onnxruntime-ep-amdgpu PR numbers applied on top