Update machine image for GPU CI

.cirun.yml

@@ -2,7 +2,8 @@ runners:
   - name: aws-gpu-runner
     cloud: aws
     instance_type: g4dn.xlarge
-    machine_image: ami-0678adbdcb4c3a662
+    machine_image: ami-067a4ba2816407ee9
+    region: eu-north-1
     preemptible: false
     workflow: .github/workflows/test-gpu.yml
     labels:

.github/workflows/test-gpu.yml

@@ -3,7 +3,11 @@ name: AWS GPU
 on:
   push:
     branches: [main]
-  workflow_dispatch:
+  pull_request:
+    types:
+      - labeled
+      - opened
+      - synchronize
 
 # Cancel the job if new commits are pushed
 # https://stackoverflow.com/questions/66335225/how-to-cancel-previous-runs-in-the-pr-when-you-push-new-commitsupdate-the-curre
@@ -13,6 +17,8 @@ concurrency:
 
 jobs:
   test:
+    name: GPU Tests
+    if: ${{ contains(github.event.pull_request.labels.*.name, 'run-gpu-ci') }}
     runs-on: "cirun-aws-gpu--${{ github.run_id }}"
     defaults:
       run:

docs/contributing.md

@@ -5,3 +5,10 @@ AnnData follows the development practices outlined in the [Scanpy contribution g
 ```{eval-rst}
 .. include:: _key_contributors.rst
 ```
+
+## CI
+
+### GPU CI
+
+To test GPU specific code we have a paid self-hosted runner to run the gpu specific tests on.
+This CI runs by default on the main branch, but for PRs requires the `run-gpu-ci` label to prevent unneccesary runs.

docs/release-notes/0.10.0.md

@@ -6,7 +6,7 @@
 **GPU Support**
 
 * Support for dense and sparse cupy arrays {pr}`1066` {user}`ivirshup`
-* anndata now has GPU enabled CI. Made possibly by a grant from [CZI's EOSS program](https://chanzuckerberg.com/eoss/) and managed via [Cirun](https://Cirun.io) {pr}`1066` {user}`Zethson` {user}`ivirshup`
+* anndata now has GPU enabled CI. Made possibly by a grant from [CZI's EOSS program](https://chanzuckerberg.com/eoss/) and managed via [Cirun](https://Cirun.io) {pr}`1066` {pr}`1084` {user}`Zethson` {user}`ivirshup`
 
 **Improved errors and warnings**