diff --git a/.github/workflows/integration-cloud.yml b/.github/workflows/integration-cloud.yml
index 3f028465..6e9b3cac 100644
--- a/.github/workflows/integration-cloud.yml
+++ b/.github/workflows/integration-cloud.yml
@@ -5,6 +5,8 @@ on:
   push:
     branches:
       - master
+  pull_request:
+    paths: [".github/workflows/integration-cloud.yml"]
   schedule:
     - cron: "0 11 * * 0"
 
@@ -17,16 +19,18 @@ jobs:
     runs-on: ubuntu-latest
     steps:
       - uses: actions/checkout@v4
+
       - name: Set up Python
         uses: actions/setup-python@v5
+        id: setup-python
         with:
-          python-version: "3.9"
+          python-version: "3.10"
           cache: "pip"
-          cache-dependency-path: |
-            pyproject.toml
+      - run: echo '${{ steps.setup-python.outputs.cache-hit }}' # true if cache-hit occurred on the primary key
       - name: Install dependencies
         run: |
           pip install -e .[dev]
+
       - name: Run integration tests
         env:
           VESPA_TEAM_API_KEY: ${{ secrets.VESPA_TEAM_API_KEY }}
@@ -35,42 +39,44 @@ jobs:
 
   integration-cloud-token:
     runs-on: ubuntu-latest
-    needs: integration-cloud
     steps:
       - uses: actions/checkout@v4
+
       - name: Set up Python
         uses: actions/setup-python@v5
+        id: setup-python
         with:
-          python-version: "3.9"
+          python-version: "3.10"
           cache: "pip"
-          cache-dependency-path: |
-            pyproject.toml
+      - run: echo '${{ steps.setup-python.outputs.cache-hit }}' # true if cache-hit occurred on the primary key
       - name: Install dependencies
         run: |
           pip install -e .[dev]
+
       - name: Run integration tests
         env:
           VESPA_TEAM_API_KEY: ${{ secrets.VESPA_TEAM_API_KEY }}
           VESPA_CLOUD_SECRET_TOKEN: ${{ secrets.VESPA_CLOUD_SECRET_TOKEN }}
-          VESPA_CLIENT_TOKEN_ID: ${{ secrets.VESPA_CLIENT_TOKEN_ID}}
+          VESPA_CLIENT_TOKEN_ID: ${{ secrets.VESPA_CLIENT_TOKEN_ID }}
         run: |
           pytest tests/integration/test_integration_vespa_cloud_token.py -s -v
 
   integration-cloud-vector-search:
     runs-on: ubuntu-latest
-    needs: integration-cloud-token
     steps:
       - uses: actions/checkout@v4
+
       - name: Set up Python
+        id: setup-python
         uses: actions/setup-python@v5
         with:
-          python-version: "3.9"
+          python-version: "3.10"
           cache: "pip"
-          cache-dependency-path: |
-            pyproject.toml
+      - run: echo '${{ steps.setup-python.outputs.cache-hit }}' # true if cache-hit occurred on the primary key
       - name: Install dependencies
         run: |
           pip install -e .[dev]
+
       - name: Run integration tests
         env:
           VESPA_TEAM_API_KEY: ${{ secrets.VESPA_TEAM_API_KEY }}
diff --git a/docs/sphinx/source/examples/colpali-benchmark-vqa-vlm_Vespa-cloud.ipynb b/docs/sphinx/source/examples/colpali-benchmark-vqa-vlm_Vespa-cloud.ipynb
index 7ac2da7a..483229b8 100644
--- a/docs/sphinx/source/examples/colpali-benchmark-vqa-vlm_Vespa-cloud.ipynb
+++ b/docs/sphinx/source/examples/colpali-benchmark-vqa-vlm_Vespa-cloud.ipynb
@@ -59,7 +59,8 @@
     "from colpali_engine.utils.colpali_processing_utils import (\n",
     "    process_images,\n",
     "    process_queries,\n",
-    ")"
+    ")\n",
+    "from torch.amp import autocast"
    ]
   },
   {
@@ -88,13 +89,13 @@
    "source": [
     "if torch.cuda.is_available():\n",
     "    device = torch.device(\"cuda\")\n",
-    "    type = torch.bfloat16\n",
+    "    dtype = torch.bfloat16\n",
     "elif torch.backends.mps.is_available():\n",
     "    device = torch.device(\"mps\")\n",
-    "    type = torch.float32\n",
+    "    dtype = torch.float32\n",
     "else:\n",
     "    device = torch.device(\"cpu\")\n",
-    "    type = torch.float32"
+    "    dtype = torch.float32"
    ]
   },
   {
@@ -286,7 +287,7 @@
    "source": [
     "model_name = \"vidore/colpali-v1.2\"\n",
     "model = ColPali.from_pretrained(\n",
-    "    \"vidore/colpaligemma-3b-pt-448-base\", torch_dtype=type\n",
+    "    \"vidore/colpaligemma-3b-pt-448-base\", torch_dtype=dtype\n",
     ").eval()\n",
     "model.load_adapter(model_name)\n",
     "model = model.eval()\n",
@@ -356,9 +357,10 @@
     "embeddings = []\n",
     "for batch_doc in tqdm(dataloader):\n",
     "    with torch.no_grad():\n",
-    "        batch_doc = {k: v.to(model.device) for k, v in batch_doc.items()}\n",
-    "        embeddings_doc = model(**batch_doc)\n",
-    "        embeddings.extend(list(torch.unbind(embeddings_doc.to(\"cpu\"))))"
+    "        with autocast(device_type=device.type):\n",
+    "            batch_doc = {k: v.to(model.device) for k, v in batch_doc.items()}\n",
+    "            embeddings_doc = model(**batch_doc)\n",
+    "            embeddings.extend(list(torch.unbind(embeddings_doc.to(\"cpu\"))))"
    ]
   },
   {
@@ -392,9 +394,10 @@
     "query_embeddings = []\n",
     "for batch_query in tqdm(dataloader):\n",
     "    with torch.no_grad():\n",
-    "        batch_query = {k: v.to(model.device) for k, v in batch_query.items()}\n",
-    "        embeddings_query = model(**batch_query)\n",
-    "        query_embeddings.extend(list(torch.unbind(embeddings_query.to(\"cpu\"))))"
+    "        with autocast(device_type=device.type):\n",
+    "            batch_query = {k: v.to(model.device) for k, v in batch_query.items()}\n",
+    "            embeddings_query = model(**batch_query)\n",
+    "            query_embeddings.extend(list(torch.unbind(embeddings_query.to(\"cpu\"))))"
    ]
   },
   {
@@ -6238,4 +6241,4 @@
  },
  "nbformat": 4,
  "nbformat_minor": 0
-}
\ No newline at end of file
+}
diff --git a/docs/sphinx/source/examples/colpali-document-retrieval-vision-language-models-cloud.ipynb b/docs/sphinx/source/examples/colpali-document-retrieval-vision-language-models-cloud.ipynb
index 124450c1..bc5b6c3b 100644
--- a/docs/sphinx/source/examples/colpali-document-retrieval-vision-language-models-cloud.ipynb
+++ b/docs/sphinx/source/examples/colpali-document-retrieval-vision-language-models-cloud.ipynb
@@ -22,8 +22,8 @@
     "\n",
     "[ColPali: Efficient Document Retrieval with Vision Language Models Manuel Faysse, Hugues Sibille, Tony Wu, Bilel Omrani, Gautier Viaud, Céline Hudelot, Pierre Colombo](https://arxiv.org/abs/2407.01449v2)\n",
     "\n",
-    "ColPail is a combination of [ColBERT](https://blog.vespa.ai/announcing-colbert-embedder-in-vespa/) \n",
-    "and [PailGemma](https://huggingface.co/blog/paligemma):\n",
+    "ColPali is a combination of [ColBERT](https://blog.vespa.ai/announcing-colbert-embedder-in-vespa/) \n",
+    "and [PaliGemma](https://huggingface.co/blog/paligemma):\n",
     "\n",
     ">ColPali is enabled by the latest advances in Vision Language Models, notably the PaliGemma model from the Google Zürich team, and leverages multi-vector retrieval through late interaction mechanisms as proposed in ColBERT by Omar Khattab.\n",
     "\n",
@@ -61,7 +61,7 @@
     "\n",
     "Let us get started. \n",
     "\n",
-    "[![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/vespa-engine/pyvespa/blob/master/docs/sphinx/source/examples/colpali-document-retrieval-vision-language-models.ipynb)\n",
+    "[![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/vespa-engine/pyvespa/blob/master/docs/sphinx/source/examples/colpali-document-retrieval-vision-language-models-cloud.ipynb)\n",
     "\n",
     "\n",
     "Install dependencies: \n",
@@ -100,7 +100,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 3,
+   "execution_count": null,
    "metadata": {
     "id": "qKFOvdo5nCVl"
    },
@@ -118,7 +118,8 @@
     "    process_images,\n",
     "    process_queries,\n",
     ")\n",
-    "from colpali_engine.utils.image_utils import scale_image, get_base64_image"
+    "from colpali_engine.utils.image_utils import scale_image, get_base64_image\n",
+    "from torch.amp import autocast"
    ]
   },
   {
@@ -142,19 +143,19 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 4,
+   "execution_count": null,
    "metadata": {},
    "outputs": [],
    "source": [
     "if torch.cuda.is_available():\n",
     "    device = torch.device(\"cuda\")\n",
-    "    type = torch.bfloat16\n",
+    "    dtype = torch.bfloat16\n",
     "elif torch.backends.mps.is_available():\n",
     "    device = torch.device(\"mps\")\n",
-    "    type = torch.float32\n",
+    "    dtype = torch.float32\n",
     "else:\n",
     "    device = torch.device(\"cpu\")\n",
-    "    type = torch.float32"
+    "    dtype = torch.float32"
    ]
   },
   {
@@ -339,7 +340,7 @@
    "source": [
     "model_name = \"vidore/colpali-v1.2\"\n",
     "model = ColPali.from_pretrained(\n",
-    "    \"vidore/colpaligemma-3b-pt-448-base\", torch_dtype=type\n",
+    "    \"vidore/colpaligemma-3b-pt-448-base\", torch_dtype=dtype\n",
     ").eval()\n",
     "model.load_adapter(model_name)\n",
     "model = model.eval()\n",
@@ -497,7 +498,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 10,
+   "execution_count": null,
    "metadata": {
     "colab": {
      "base_uri": "https://localhost:8080/"
@@ -526,9 +527,10 @@
     "    )\n",
     "    for batch_doc in tqdm(dataloader):\n",
     "        with torch.no_grad():\n",
-    "            batch_doc = {k: v.to(model.device) for k, v in batch_doc.items()}\n",
-    "            embeddings_doc = model(**batch_doc)\n",
-    "            page_embeddings.extend(list(torch.unbind(embeddings_doc.to(\"cpu\"))))\n",
+    "            with autocast(device_type=device.type):\n",
+    "                batch_doc = {k: v.to(model.device) for k, v in batch_doc.items()}\n",
+    "                embeddings_doc = model(**batch_doc)\n",
+    "                page_embeddings.extend(list(torch.unbind(embeddings_doc.cpu())))\n",
     "    pdf[\"embeddings\"] = page_embeddings"
    ]
   },
@@ -927,7 +929,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 26,
+   "execution_count": null,
    "metadata": {
     "id": "NxeDd3mcYDpL"
    },
@@ -942,9 +944,10 @@
     "qs = []\n",
     "for batch_query in dataloader:\n",
     "    with torch.no_grad():\n",
-    "        batch_query = {k: v.to(model.device) for k, v in batch_query.items()}\n",
-    "        embeddings_query = model(**batch_query)\n",
-    "        qs.extend(list(torch.unbind(embeddings_query.to(\"cpu\"))))"
+    "        with autocast(device_type=device.type):\n",
+    "            batch_query = {k: v.to(model.device) for k, v in batch_query.items()}\n",
+    "            embeddings_query = model(**batch_query)\n",
+    "            qs.extend(list(torch.unbind(embeddings_query.cpu())))"
    ]
   },
   {
diff --git a/docs/sphinx/source/examples/simplified-retrieval-with-colpali-vlm_Vespa-cloud.ipynb b/docs/sphinx/source/examples/simplified-retrieval-with-colpali-vlm_Vespa-cloud.ipynb
index 2bb9fd57..2867a9ef 100644
--- a/docs/sphinx/source/examples/simplified-retrieval-with-colpali-vlm_Vespa-cloud.ipynb
+++ b/docs/sphinx/source/examples/simplified-retrieval-with-colpali-vlm_Vespa-cloud.ipynb
@@ -103,7 +103,8 @@
     "    process_images,\n",
     "    process_queries,\n",
     ")\n",
-    "from colpali_engine.utils.image_utils import scale_image, get_base64_image"
+    "from colpali_engine.utils.image_utils import scale_image, get_base64_image\n",
+    "from torch.amp import autocast"
    ]
   },
   {
@@ -133,13 +134,13 @@
    "source": [
     "if torch.cuda.is_available():\n",
     "    device = torch.device(\"cuda\")\n",
-    "    type = torch.bfloat16\n",
+    "    dtype = torch.bfloat16\n",
     "elif torch.backends.mps.is_available():\n",
     "    device = torch.device(\"mps\")\n",
-    "    type = torch.float32\n",
+    "    dtype = torch.float32\n",
     "else:\n",
     "    device = torch.device(\"cpu\")\n",
-    "    type = torch.float32"
+    "    dtype = torch.float32"
    ]
   },
   {
@@ -331,7 +332,7 @@
    "source": [
     "model_name = \"vidore/colpali-v1.2\"\n",
     "model = ColPali.from_pretrained(\n",
-    "    \"vidore/colpaligemma-3b-pt-448-base\", torch_dtype=type\n",
+    "    \"vidore/colpaligemma-3b-pt-448-base\", torch_dtype=dtype\n",
     ").eval()\n",
     "model.load_adapter(model_name)\n",
     "model = model.eval()\n",
@@ -432,7 +433,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 6,
+   "execution_count": null,
    "metadata": {
     "id": "YaDInfmT3Tbu"
    },
@@ -440,6 +441,7 @@
    "source": [
     "for pdf in sample_pdfs:\n",
     "    page_images, page_texts = get_pdf_images(pdf[\"url\"])\n",
+    "\n",
     "    pdf[\"images\"] = page_images\n",
     "    pdf[\"texts\"] = page_texts"
    ]
@@ -590,9 +592,10 @@
     "    )\n",
     "    for batch_doc in tqdm(dataloader):\n",
     "        with torch.no_grad():\n",
-    "            batch_doc = {k: v.to(model.device) for k, v in batch_doc.items()}\n",
-    "            embeddings_doc = model(**batch_doc)\n",
-    "            page_embeddings.extend(list(torch.unbind(embeddings_doc.to(\"cpu\"))))\n",
+    "            with autocast(device_type=device.type):\n",
+    "                batch_doc = {k: v.to(model.device) for k, v in batch_doc.items()}\n",
+    "                embeddings_doc = model(**batch_doc)\n",
+    "                page_embeddings.extend(list(torch.unbind(embeddings_doc.to(\"cpu\"))))\n",
     "    pdf[\"embeddings\"] = page_embeddings"
    ]
   },
@@ -974,9 +977,10 @@
     "qs = []\n",
     "for batch_query in dataloader:\n",
     "    with torch.no_grad():\n",
-    "        batch_query = {k: v.to(model.device) for k, v in batch_query.items()}\n",
-    "        embeddings_query = model(**batch_query)\n",
-    "        qs.extend(list(torch.unbind(embeddings_query.to(\"cpu\"))))"
+    "        with autocast(device_type=device.type):\n",
+    "            batch_query = {k: v.to(model.device) for k, v in batch_query.items()}\n",
+    "            embeddings_query = model(**batch_query)\n",
+    "            qs.extend(list(torch.unbind(embeddings_query.to(\"cpu\"))))"
    ]
   },
   {
diff --git a/tests/integration/test_integration_vespa_cloud_vector_search.py b/tests/integration/test_integration_vespa_cloud_vector_search.py
index 38ee5510..3797eb42 100644
--- a/tests/integration/test_integration_vespa_cloud_vector_search.py
+++ b/tests/integration/test_integration_vespa_cloud_vector_search.py
@@ -21,7 +21,6 @@
     ContainerCluster,
     Nodes,
     DeploymentConfiguration,
-    EmptyDeploymentConfiguration,
     Validation,
     ValidationID,
 )
@@ -93,7 +92,7 @@ def test_vector_indexing_and_query(self):
 
         from datasets import load_dataset
 
-        sample_size = 1000
+        sample_size = 100
         # streaming=True pages the data from S3. This is needed to avoid memory issues when loading the dataset.
         dataset = load_dataset(
             "KShivendu/dbpedia-entities-openai-1M", split="train", streaming=True
@@ -109,6 +108,10 @@ def test_vector_indexing_and_query(self):
         docs = list(
             pyvespa_feed_format
         )  # we have enough memory to page everything into memory with list()
+        # seems like we sometimes can get more than sample_size docs
+        if len(docs) > sample_size:
+            docs = docs[:sample_size]
+        self.assertEqual(len(docs), sample_size)
         ok = 0
         callbacks = 0
         start_time = time.time()
@@ -127,9 +130,6 @@ def callback(response: VespaResponse, id: str):
             schema="vector",
             namespace="benchmark",
             callback=callback,
-            max_workers=48,
-            max_connections=48,
-            max_queue_size=4000,
         )
         self.assertEqual(ok, sample_size)
         duration = time.time() - start
@@ -164,9 +164,7 @@ def callback(response: VespaResponse, id: str):
         ok = 0
         callbacks = 0
         start_time = time.time()
-        dataset = load_dataset(
-            "KShivendu/dbpedia-entities-openai-1M", split="train", streaming=True
-        ).take(100)
+
         feed_with_wrong_field = dataset.map(
             lambda x: {
                 "id": x["_id"],
@@ -174,21 +172,20 @@ def callback(response: VespaResponse, id: str):
             }
         )
         faulty_docs = list(feed_with_wrong_field)
+        if len(faulty_docs) > sample_size:
+            faulty_docs = faulty_docs[:sample_size]
+        self.assertEqual(len(faulty_docs), sample_size)
         self.app.feed_iterable(
             iter=faulty_docs,
             schema="vector",
             namespace="benchmark",
             callback=callback,
-            max_workers=48,
-            max_connections=48,
         )
         self.assertEqual(ok, 0)
         self.assertEqual(callbacks, 100)
 
         ok = 0
-        dataset = load_dataset(
-            "KShivendu/dbpedia-entities-openai-1M", split="train", streaming=True
-        ).take(sample_size)
+
         # Run update - assign all docs with a meta field
 
         updates = dataset.map(lambda x: {"id": x["_id"], "fields": {"meta": "stuff"}})
@@ -239,7 +236,7 @@ def tearDown(self) -> None:
 
 
 class TestProdDeploymentFromDisk(unittest.TestCase):
-    def setUp(self) -> None:
+    def test_setup(self) -> None:
         self.app_package = create_vector_ada_application_package()
         prod_region = "aws-us-east-1c"
         self.app_package.clusters = [
@@ -302,32 +299,33 @@ def test_application_status(self):
     def test_vector_indexing_and_query(self):
         super().test_vector_indexing_and_query()
 
-    @unittest.skip("Do not run when not waiting for deployment.")
-    def tearDown(self) -> None:
-        self.app.delete_all_docs(
-            content_cluster_name="vector_content",
-            schema="vector",
-            namespace="benchmark",
-        )
-        time.sleep(5)
-        with self.app.syncio() as sync_session:
-            response: VespaResponse = sync_session.query(
-                {"yql": "select id from sources * where true", "hits": 10}
-            )
-            self.assertEqual(response.get_status_code(), 200)
-            self.assertEqual(len(response.hits), 0)
-            print(response.get_json())
+    # DO NOT skip tearDown-method, as test will not exit.
+    # @unittest.skip("Do not run when not waiting for deployment.")
+    # def tearDown(self) -> None:
+    #     self.app.delete_all_docs(
+    #         content_cluster_name="vector_content",
+    #         schema="vector",
+    #         namespace="benchmark",
+    #     )
+    #     time.sleep(5)
+    #     with self.app.syncio() as sync_session:
+    #         response: VespaResponse = sync_session.query(
+    #             {"yql": "select id from sources * where true", "hits": 10}
+    #         )
+    #         self.assertEqual(response.get_status_code(), 200)
+    #         self.assertEqual(len(response.hits), 0)
+    #         print(response.get_json())
 
-        # Deployment is deleted by deploying with an empty deployment.xml file.
-        self.app_package.deployment_config = EmptyDeploymentConfiguration()
+    #     # Deployment is deleted by deploying with an empty deployment.xml file.
+    #     self.app_package.deployment_config = EmptyDeploymentConfiguration()
 
-        # Vespa won't push the deleted deployment.xml file unless we add a validation override
-        tomorrow = datetime.now() + timedelta(days=1)
-        formatted_date = tomorrow.strftime("%Y-%m-%d")
-        self.app_package.validations = [
-            Validation(ValidationID("deployment-removal"), formatted_date)
-        ]
-        self.app_package.to_files(self.application_root)
-        # This will delete the deployment
-        self.vespa_cloud._start_prod_deployment(self.application_root)
-        shutil.rmtree(self.application_root, ignore_errors=True)
+    #     # Vespa won't push the deleted deployment.xml file unless we add a validation override
+    #     tomorrow = datetime.now() + timedelta(days=1)
+    #     formatted_date = tomorrow.strftime("%Y-%m-%d")
+    #     self.app_package.validations = [
+    #         Validation(ValidationID("deployment-removal"), formatted_date)
+    #     ]
+    #     self.app_package.to_files(self.application_root)
+    #     # This will delete the deployment
+    #     self.vespa_cloud._start_prod_deployment(self.application_root)
+    #     shutil.rmtree(self.application_root, ignore_errors=True)