Skip to content

Commit 5c18fbd

Browse files
committed
ci: use ubuntu-latest-large runner for GGUF publish job
Switch the publish job from ubuntu-latest to ubuntu-latest-large, a GitHub- hosted 'Larger Runner' SKU with ~2x disk (~150 GB SSD), 4 cores, 16 GB RAM. This is documented at: https://docs.github.com/en/actions/using-github-hosted-runners/ Previously we relied on the jlumbroso/free-disk-space@main action to wipe preinstalled toolchains before the giant variants would fit on the stock runner. With the larger SKU the giant checkpoint + venv fits without any pre-cleanup, so the free-disk-space step is no longer needed (and was already removed by the preceding 'port to bash script' refactor). The validate job stays on ubuntu-latest because it only converts the smallest variant and doesn't need the extra disk; keeping it on the default SKU avoids burning paid larger-runner minutes for the smoke test. If the org doesn't have larger runners provisioned, the job will queue indefinitely; switch back to ubuntu-latest if that happens.
1 parent aa326d2 commit 5c18fbd

1 file changed

Lines changed: 1 addition & 1 deletion

File tree

‎.github/workflows/convert-and-publish-gguf.yml‎

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -39,7 +39,7 @@ jobs:
3939
name: Convert ${{ matrix.variant }}
4040
needs: validate
4141
if: ${{ !cancelled() && needs.validate.result == 'success' }}
42-
runs-on: ubuntu-latest
42+
runs-on: ubuntu-latest-large
4343
timeout-minutes: 60
4444
strategy:
4545
fail-fast: false

0 commit comments

Comments
 (0)