From 5e4cfbe0e19c46a3cc02234400ae6485dd51364a Mon Sep 17 00:00:00 2001
From: "Claude Sonnet 4.5" <claude@anthropic.com>
Date: Wed, 15 Jul 2026 20:22:04 +0200
Subject: [PATCH] ci: fix XPU image build (disk space + torch 2.13) and
 idempotent release

The v0.4.0 XPU Docker job failed for two reasons; also harden the release step:

- build-xpu ran the runner out of disk mid-extract of the torch-xpu wheel
  (os error 28). Add a "Free disk space" step reclaiming ~25 GB of unused
  preinstalled toolchains (dotnet/android/ghc/boost/tool-cache) + a docker
  image prune before the build.
- Dockerfile.xpu still pinned torch==2.11.0; bump to torch==2.13.0 +
  torchvision==0.28.0 (verified on the PyTorch XPU index) to match the
  lockfile. Drop the explicit torchaudio (not a project dep) and triton-xpu
  pin (torch pulls the matching pytorch-triton-xpu itself on the Linux base).
- python-publish "Create GitHub Release" now upserts: if the release already
  exists (e.g. cut manually) it uploads the dists with --clobber instead of
  failing the whole workflow.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
---
 .github/workflows/docker-publish.yml |  9 +++++++++
 .github/workflows/python-publish.yml | 14 +++++++++-----
 docker/Dockerfile.xpu                |  6 ++++--
 3 files changed, 22 insertions(+), 7 deletions(-)

diff --git a/.github/workflows/docker-publish.yml b/.github/workflows/docker-publish.yml
index 85f3a1e..974488d 100644
--- a/.github/workflows/docker-publish.yml
+++ b/.github/workflows/docker-publish.yml
@@ -100,6 +100,15 @@ jobs:
     steps:
       - uses: actions/checkout@v6
 
+      # The XPU image bundles the large torch-xpu wheel (SYCL/MKL/OneCCL);
+      # the stock runner runs out of disk mid-extract (os error 28). Reclaim
+      # ~25 GB of preinstalled toolchains we don't need before building.
+      - name: Free disk space
+        run: |
+          sudo rm -rf /usr/share/dotnet /opt/ghc /usr/local/lib/android \
+            /usr/local/share/boost "$AGENT_TOOLSDIRECTORY"
+          sudo docker image prune -af
+
       - uses: docker/setup-buildx-action@v4
 
       - name: Log in to GHCR
diff --git a/.github/workflows/python-publish.yml b/.github/workflows/python-publish.yml
index bd69c90..e13c2c6 100644
--- a/.github/workflows/python-publish.yml
+++ b/.github/workflows/python-publish.yml
@@ -76,14 +76,18 @@ jobs:
       - name: Extract version from tag
         id: version
         run: echo "VERSION=${GITHUB_REF#refs/tags/v}" >> $GITHUB_OUTPUT
-      - name: Create GitHub Release
+      - name: Create or update GitHub Release
         env:
           GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
         run: |
-          gh release create ${{ github.ref_name }} \
-            --title "Release ${{ steps.version.outputs.VERSION }}" \
-            --generate-notes \
-            dist/*
+          if gh release view ${{ github.ref_name }} >/dev/null 2>&1; then
+            gh release upload ${{ github.ref_name }} dist/* --clobber
+          else
+            gh release create ${{ github.ref_name }} \
+              --title "Release ${{ steps.version.outputs.VERSION }}" \
+              --generate-notes \
+              dist/*
+          fi
 
       - name: Trigger docker-publish workflow
         env:
diff --git a/docker/Dockerfile.xpu b/docker/Dockerfile.xpu
index eba2bc5..8b84c74 100644
--- a/docker/Dockerfile.xpu
+++ b/docker/Dockerfile.xpu
@@ -45,9 +45,11 @@ USER 1000
 RUN uv sync --no-dev --extra all
 
 # Replace PyPI torch with the XPU build (bundles SYCL runtime, MKL, OneCCL).
-# triton-xpu is Linux-only; Linux-only base image makes the marker moot.
+# torch pulls its matching pytorch-triton-xpu itself; the Linux base resolves
+# that dependency's sys_platform marker, so no explicit triton pin is needed.
+# torchaudio is not a project dependency — dropped to save image space.
 RUN uv pip install --index-url https://download.pytorch.org/whl/xpu \
-        torch==2.11.0 torchvision==0.26.0 torchaudio==2.11.0 triton-xpu==3.7.0
+        torch==2.13.0 torchvision==0.28.0
 
 ENV VIRTUAL_ENV=/app/.venv
 ENV PATH="$VIRTUAL_ENV/bin:$PATH"