From 52d42686560a9e8f441f9b9780c8890c37d2802d Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Sigbj=C3=B8rn=20Skj=C3=A6ret?=
 <sigbjorn.skjaeret@huggingface.co>
Date: Thu, 10 Sep 2026 18:11:32 +0200
Subject: [PATCH] ci : add self-hosted-gpu-cuda and server-sanitize to hf-jobs
 (#28693)

---
 .github/workflows/build-self-hosted.yml | 34 ++++++++++++++++-
 .github/workflows/server-sanitize.yml   | 50 ++++++++++++++++---------
 2 files changed, 64 insertions(+), 20 deletions(-)

diff --git a/.github/workflows/build-self-hosted.yml b/.github/workflows/build-self-hosted.yml
index ccfe2a6046..fda4879e21 100644
--- a/.github/workflows/build-self-hosted.yml
+++ b/.github/workflows/build-self-hosted.yml
@@ -58,18 +58,48 @@ env:
 
 jobs:
   gpu-cuda:
-    runs-on: [self-hosted, Linux, NVIDIA]
+    runs-on: "hf-jobs-t4-small:cuda13"
 
     steps:
       - name: Clone
         id: checkout
         uses: actions/checkout@v6
 
+      - name: Install dependencies
+        run: |
+          sudo apt update
+          sudo apt install -y cmake libssl-dev time unzip wget python3 python3-venv python3-pip
+
+      - name: ccache
+        uses: ggml-org/ccache-action@v1.2.24
+        with:
+          restore: false
+          save: false
+
+      - name: ccache-buckets-restore
+        uses: ./.github/actions/ccache-buckets
+        with:
+          key: self-hosted-gpu-cuda
+          folder: llama.cpp
+          hf_bucket: ggml-org/cache
+
       - name: Test
         id: ggml-ci
         run: |
           nvidia-smi
-          GG_BUILD_CUDA=1 bash ./ci/run.sh ~/results/llama.cpp ~/mnt/llama.cpp
+          GG_BUILD_CUDA=1 CUDACXX=/usr/local/cuda/bin/nvcc bash ./ci/run.sh ~/results/llama.cpp ~/mnt/llama.cpp
+
+      - name: ccache-buckets-save
+        if: ${{ github.event_name == 'push' && github.ref == 'refs/heads/master' }}
+        uses: ./.github/actions/ccache-buckets
+        env:
+          HF_TOKEN: ${{ secrets.HF_TOKEN_CACHE_OUTPUT }}
+        with:
+          key: self-hosted-gpu-cuda
+          folder: llama.cpp
+          evict-old-files: 1d
+          hf_bucket: ggml-org/cache
+          save: true
 
   gpu-rocm:
     runs-on: [self-hosted, Linux, AMD]
diff --git a/.github/workflows/server-sanitize.yml b/.github/workflows/server-sanitize.yml
index 77549ee871..52e175f834 100644
--- a/.github/workflows/server-sanitize.yml
+++ b/.github/workflows/server-sanitize.yml
@@ -32,6 +32,8 @@ on:
     ]
 
 env:
+  # note: this is dud token to avoid rate limiting (https://github.com/ggml-org/llama.cpp/pull/25706#issuecomment-4979941302)
+  HF_TOKEN: ${{ secrets.HF_TOKEN_CI }}
   LLAMA_ARG_LOG_COLORS: 1
   LLAMA_ARG_LOG_PREFIX: 1
   LLAMA_ARG_LOG_TIMESTAMPS: 1
@@ -43,7 +45,7 @@ concurrency:
 
 jobs:
   server:
-    runs-on: [self-hosted, CPU, Linux, llama-server]
+    runs-on: hf-jobs-cpu-upgrade
 
     strategy:
       matrix:
@@ -52,20 +54,6 @@ jobs:
       fail-fast: false
 
     steps:
-      #- name: Dependencies
-      #  id: depends
-      #  run: |
-      #    sudo apt-get update
-      #    sudo apt-get -y install \
-      #      build-essential \
-      #      xxd \
-      #      git \
-      #      cmake \
-      #      curl \
-      #      wget \
-      #      language-pack-en \
-      #      libssl-dev
-
       - name: Clone
         id: checkout
         uses: actions/checkout@v6
@@ -73,6 +61,24 @@ jobs:
           fetch-depth: 0
           ref: ${{ github.event.inputs.sha || github.event.pull_request.head.sha || github.sha || github.head_ref || github.ref_name }}
 
+      - name: Install dependencies
+        run: |
+          sudo apt update
+          sudo apt install -y build-essential cmake python3-full
+
+      - name: ccache
+        uses: ggml-org/ccache-action@v1.2.24
+        with:
+          restore: false
+          save: false
+
+      - name: ccache-buckets-restore
+        uses: ./.github/actions/ccache-buckets
+        with:
+          key: server-sanitize
+          folder: llama.cpp
+          hf_bucket: ggml-org/cache
+
       - name: Build
         id: cmake_build
         run: |
@@ -87,9 +93,17 @@ jobs:
             -DLLAMA_SANITIZE_UNDEFINED=${{ matrix.sanitizer == 'UNDEFINED' }}
           cmake --build build --config ${{ matrix.build_type }} -j $(nproc) --target llama-server
 
-      - name: Python setup
-        id: setup_python
-        uses: actions/setup-python@v7
+      - name: ccache-buckets-save
+        if: ${{ github.event_name == 'push' && github.ref == 'refs/heads/master' }}
+        uses: ./.github/actions/ccache-buckets
+        env:
+          HF_TOKEN: ${{ secrets.HF_TOKEN_CACHE_OUTPUT }}
+        with:
+          key: server-sanitize
+          folder: llama.cpp
+          evict-old-files: 1d
+          hf_bucket: ggml-org/cache
+          save: true
 
       - name: Install Python dependencies
         run: |