From d7dc9b2c78aafd585be673a9e592e56cb35386cb Mon Sep 17 00:00:00 2001
From: Jirka B <j.borovec+github@gmail.com>
Date: Fri, 25 Jul 2025 14:04:27 +0200
Subject: [PATCH 01/25] ci: split common & distributed

---
 .github/workflows/ci-tests.yml | 27 +++++++++++++++++----------
 1 file changed, 17 insertions(+), 10 deletions(-)

diff --git a/.github/workflows/ci-tests.yml b/.github/workflows/ci-tests.yml
index 04f8d027c00..654bc4b6f33 100644
--- a/.github/workflows/ci-tests.yml
+++ b/.github/workflows/ci-tests.yml
@@ -45,19 +45,26 @@ jobs:
           - "2.5.1"
           - "2.6.0"
           - "2.7.1"
+        testing: ["common", "distributed"]
         include:
           # cover additional python and PT combinations
-          - { os: "ubuntu-22.04", python-version: "3.9", pytorch-version: "2.0.1", requires: "oldest" }
-          - { os: "ubuntu-22.04", python-version: "3.12", pytorch-version: "2.6.0" }
-          - { os: "ubuntu-22.04", python-version: "3.12", pytorch-version: "2.7.1" }
+          - { os: "ubuntu-22.04", python-version: "3.9", pytorch-version: "2.0.1", requires: "oldest", testing: "common" }
+          - { os: "ubuntu-22.04", python-version: "3.9", pytorch-version: "2.0.1", requires: "oldest", testing: "distributed" }
+          - { os: "ubuntu-22.04", python-version: "3.12", pytorch-version: "2.7.1", testing: "common" }
+          - { os: "ubuntu-22.04", python-version: "3.12", pytorch-version: "2.7.1", testing: "distributed" }
           # standard mac machine, not the M1
-          - { os: "macOS-13", python-version: "3.10", pytorch-version: "2.0.1" }
+          - { os: "macOS-14", python-version: "3.10", pytorch-version: "2.0.1", testing: "common" }
+          - { os: "macOS-14", python-version: "3.10", pytorch-version: "2.0.1", testing: "distributed" }
           # using the ARM based M1 machine
-          - { os: "macOS-14", python-version: "3.10", pytorch-version: "2.0.1" }
-          - { os: "macOS-14", python-version: "3.12", pytorch-version: "2.7.1" }
+          - { os: "macOS-14", python-version: "3.10", pytorch-version: "2.0.1", testing: "common" }
+          - { os: "macOS-14", python-version: "3.10", pytorch-version: "2.0.1", testing: "distributed" }
+          - { os: "macOS-14", python-version: "3.12", pytorch-version: "2.7.1", testing: "common" }
+          - { os: "macOS-14", python-version: "3.12", pytorch-version: "2.7.1", testing: "distributed" }
           # some windows
-          - { os: "windows-2022", python-version: "3.10", pytorch-version: "2.0.1" }
-          - { os: "windows-2022", python-version: "3.12", pytorch-version: "2.7.1" }
+          - { os: "windows-2022", python-version: "3.10", pytorch-version: "2.0.1", testing: "common" }
+          - { os: "windows-2022", python-version: "3.10", pytorch-version: "2.0.1", testing: "distributed" }
+          - { os: "windows-2022", python-version: "3.12", pytorch-version: "2.7.1", testing: "common" }
+          - { os: "windows-2022", python-version: "3.12", pytorch-version: "2.7.1", testing: "distributed" }
           # Future released version
           #- { os: "ubuntu-22.04", python-version: "3.11", pytorch-version: "2.8.0" }
           #- { os: "macOS-14", python-version: "3.11", pytorch-version: "2.8.0" }
@@ -167,7 +174,7 @@ jobs:
 
       - name: Unittests common
         # skip for PR if there is nothing to test, note that outside PR there is default 'unittests'
-        if: ${{ env.TEST_DIRS != '' }}
+        if: ${{ env.TEST_DIRS != '' && matrix.testing == 'common' }}
         working-directory: ./tests
         run: |
           python -m pytest \
@@ -183,7 +190,7 @@ jobs:
 
       - name: Unittests DDP
         # skip for PR if there is nothing to test, note that outside PR there is default 'unittests'
-        if: ${{ env.TEST_DIRS != '' }}
+        if: ${{ env.TEST_DIRS != '' && matrix.testing == 'distributed' }}
         working-directory: ./tests
         env:
           USE_PYTEST_POOL: "1"

From bf756d3ee93ec5671f3d0ccac14f9ed35c984bb8 Mon Sep 17 00:00:00 2001
From: "pre-commit-ci[bot]"
 <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Date: Fri, 25 Jul 2025 12:06:35 +0000
Subject: [PATCH 02/25] [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci
---
 .github/workflows/ci-tests.yml | 16 ++++++++++++++--
 1 file changed, 14 insertions(+), 2 deletions(-)

diff --git a/.github/workflows/ci-tests.yml b/.github/workflows/ci-tests.yml
index 654bc4b6f33..9dca7e4f4ed 100644
--- a/.github/workflows/ci-tests.yml
+++ b/.github/workflows/ci-tests.yml
@@ -48,8 +48,20 @@ jobs:
         testing: ["common", "distributed"]
         include:
           # cover additional python and PT combinations
-          - { os: "ubuntu-22.04", python-version: "3.9", pytorch-version: "2.0.1", requires: "oldest", testing: "common" }
-          - { os: "ubuntu-22.04", python-version: "3.9", pytorch-version: "2.0.1", requires: "oldest", testing: "distributed" }
+          - {
+              os: "ubuntu-22.04",
+              python-version: "3.9",
+              pytorch-version: "2.0.1",
+              requires: "oldest",
+              testing: "common",
+            }
+          - {
+              os: "ubuntu-22.04",
+              python-version: "3.9",
+              pytorch-version: "2.0.1",
+              requires: "oldest",
+              testing: "distributed",
+            }
           - { os: "ubuntu-22.04", python-version: "3.12", pytorch-version: "2.7.1", testing: "common" }
           - { os: "ubuntu-22.04", python-version: "3.12", pytorch-version: "2.7.1", testing: "distributed" }
           # standard mac machine, not the M1

From dfeee3d261cfa40f85bc8f530f5039f309a875db Mon Sep 17 00:00:00 2001
From: Nicki Skafte <skaftenicki@gmail.com>
Date: Mon, 4 Aug 2025 08:19:01 +0200
Subject: [PATCH 03/25] try random port instead

---
 tests/unittests/conftest.py | 9 +++------
 1 file changed, 3 insertions(+), 6 deletions(-)

diff --git a/tests/unittests/conftest.py b/tests/unittests/conftest.py
index 65c086cb9ba..be221996de8 100644
--- a/tests/unittests/conftest.py
+++ b/tests/unittests/conftest.py
@@ -56,14 +56,11 @@ def setup_ddp(rank, world_size):
         world_size: the number of processes
 
     """
-    global CURRENT_PORT
+    import random
 
+    port = random.randint(10000, 20000)  # noqa: S311
     os.environ["MASTER_ADDR"] = "localhost"
-    os.environ["MASTER_PORT"] = str(CURRENT_PORT)
-
-    CURRENT_PORT += 1
-    if CURRENT_PORT > MAX_PORT:
-        CURRENT_PORT = START_PORT
+    os.environ["MASTER_PORT"] = str(port)
 
     if torch.distributed.group.WORLD is not None:  # if already initialized, destroy the process group
         torch.distributed.destroy_process_group()

From c17511db63fbcb5152e5881c28a0c82e6aabf45a Mon Sep 17 00:00:00 2001
From: Nicki Skafte <skaftenicki@gmail.com>
Date: Mon, 4 Aug 2025 09:13:18 +0200
Subject: [PATCH 04/25] try instead a free socket approach

---
 tests/unittests/conftest.py                 | 17 ++++++++++-------
 tests/unittests/image/test_ms_ssim.py       |  4 ++--
 tests/unittests/image/test_ssim.py          |  4 ++--
 tests/unittests/utilities/test_utilities.py | 14 --------------
 4 files changed, 14 insertions(+), 25 deletions(-)

diff --git a/tests/unittests/conftest.py b/tests/unittests/conftest.py
index be221996de8..746a3e66c26 100644
--- a/tests/unittests/conftest.py
+++ b/tests/unittests/conftest.py
@@ -13,6 +13,7 @@
 # limitations under the License.
 import contextlib
 import os
+import socket
 import sys
 
 import pytest
@@ -30,9 +31,6 @@
 EXTRA_DIM = 3
 THRESHOLD = 0.5
 
-MAX_PORT = 8100
-START_PORT = 8088
-CURRENT_PORT = START_PORT
 USE_PYTEST_POOL = os.getenv("USE_PYTEST_POOL", "0") == "1"
 
 
@@ -44,6 +42,14 @@ def use_deterministic_algorithms():
     torch.use_deterministic_algorithms(False)
 
 
+def get_free_port():
+    """Find an available free port on localhost."""
+    with socket.socket(socket.AF_INET, socket.SOCK_STREAM) as s:
+        s.bind(("localhost", 0))  # Bind to a free port provided by the OS
+        s.setsockopt(socket.SOL_SOCKET, socket.SO_REUSEADDR, 1)
+        return s.getsockname()[1]
+
+
 def setup_ddp(rank, world_size):
     """Initialize ddp environment.
 
@@ -56,11 +62,8 @@ def setup_ddp(rank, world_size):
         world_size: the number of processes
 
     """
-    import random
-
-    port = random.randint(10000, 20000)  # noqa: S311
     os.environ["MASTER_ADDR"] = "localhost"
-    os.environ["MASTER_PORT"] = str(port)
+    os.environ["MASTER_PORT"] = str(get_free_port())
 
     if torch.distributed.group.WORLD is not None:  # if already initialized, destroy the process group
         torch.distributed.destroy_process_group()
diff --git a/tests/unittests/image/test_ms_ssim.py b/tests/unittests/image/test_ms_ssim.py
index a5ba60bf2fd..c7501ff1c4e 100644
--- a/tests/unittests/image/test_ms_ssim.py
+++ b/tests/unittests/image/test_ms_ssim.py
@@ -24,8 +24,8 @@
 from unittests import NUM_BATCHES, _Input
 from unittests._helpers import seed_all
 from unittests._helpers.testers import MetricTester
+from unittests.conftest import get_free_port
 from unittests.image import cleanup_ddp, setup_ddp
-from unittests.utilities.test_utilities import find_free_port
 
 seed_all(42)
 
@@ -137,7 +137,7 @@ def test_ms_ssim_reduction_none_ddp():
 
     """
     world_size = 2
-    free_port = find_free_port()
+    free_port = get_free_port()
     if free_port == -1:
         pytest.skip("No free port available for DDP test.")
     mp.spawn(_run_ms_ssim_ddp, args=(world_size, free_port), nprocs=world_size, join=True)
diff --git a/tests/unittests/image/test_ssim.py b/tests/unittests/image/test_ssim.py
index 6bfc642bd2b..a3e0e1ba299 100644
--- a/tests/unittests/image/test_ssim.py
+++ b/tests/unittests/image/test_ssim.py
@@ -27,8 +27,8 @@
 from unittests import NUM_BATCHES, _Input
 from unittests._helpers import seed_all
 from unittests._helpers.testers import MetricTester
+from unittests.conftest import get_free_port
 from unittests.image import cleanup_ddp, setup_ddp
-from unittests.utilities.test_utilities import find_free_port
 
 seed_all(42)
 
@@ -392,7 +392,7 @@ def test_ssim_reduction_none_ddp():
 
     """
     world_size = 2
-    free_port = find_free_port()
+    free_port = get_free_port()
     if free_port == -1:
         pytest.skip("No free port available for DDP test.")
     mp.spawn(_run_ssim_ddp, args=(world_size, free_port), nprocs=world_size, join=True)
diff --git a/tests/unittests/utilities/test_utilities.py b/tests/unittests/utilities/test_utilities.py
index e34bbda2662..6f9f5d62843 100644
--- a/tests/unittests/utilities/test_utilities.py
+++ b/tests/unittests/utilities/test_utilities.py
@@ -11,7 +11,6 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
-import socket
 import sys
 
 import numpy as np
@@ -19,7 +18,6 @@
 import torch
 from lightning_utilities.test.warning import no_warning_call
 from torch import tensor
-from unittests.conftest import MAX_PORT, START_PORT
 
 from torchmetrics.regression import MeanSquaredError, PearsonCorrCoef
 from torchmetrics.utilities import check_forward_full_state_property, rank_zero_debug, rank_zero_info, rank_zero_warn
@@ -241,15 +239,3 @@ def test_half_precision_top_k_cpu_raises_error():
     x = torch.randn(100, 10, dtype=torch.half)
     with pytest.raises(RuntimeError, match="\"topk_cpu\" not implemented for 'Half'"):
         torch.topk(x, k=3, dim=1)
-
-
-def find_free_port(start=START_PORT, end=MAX_PORT):
-    """Returns an available localhost port in the given range or returns -1 if no port available."""
-    for port in range(start, end + 1):
-        with socket.socket(socket.AF_INET, socket.SOCK_STREAM) as s:
-            try:
-                s.bind(("localhost", port))
-                return port
-            except OSError:
-                continue
-    return -1

From e2af42077fc4a60b2a985ce4d01deb2089960ac1 Mon Sep 17 00:00:00 2001
From: Nicki Skafte <skaftenicki@gmail.com>
Date: Mon, 4 Aug 2025 12:57:12 +0200
Subject: [PATCH 05/25] fix ports not syncing

---
 tests/unittests/bases/test_ddp.py     | 13 +++++++------
 tests/unittests/conftest.py           | 14 +++++++++++---
 tests/unittests/image/__init__.py     | 17 -----------------
 tests/unittests/image/test_ms_ssim.py |  3 +--
 tests/unittests/image/test_ssim.py    |  3 +--
 5 files changed, 20 insertions(+), 30 deletions(-)

diff --git a/tests/unittests/bases/test_ddp.py b/tests/unittests/bases/test_ddp.py
index ad29b0f4c6e..1eecad747d9 100644
--- a/tests/unittests/bases/test_ddp.py
+++ b/tests/unittests/bases/test_ddp.py
@@ -27,7 +27,7 @@
 from unittests import NUM_PROCESSES, USE_PYTEST_POOL
 from unittests._helpers import seed_all
 from unittests._helpers.testers import DummyListMetric, DummyMetric, DummyMetricSum
-from unittests.conftest import setup_ddp
+from unittests.conftest import get_free_port, setup_ddp
 
 seed_all(42)
 
@@ -106,9 +106,9 @@ def test_ddp(process):
     pytest.pool.map(process, range(NUM_PROCESSES))
 
 
-def _test_ddp_gather_all_autograd_same_shape(rank: int, worldsize: int = NUM_PROCESSES) -> None:
+def _test_ddp_gather_all_autograd_same_shape(rank: int, worldsize: int, port: int) -> None:
     """Test that ddp gather preserves local rank's autograd graph for same-shaped tensors across ranks."""
-    setup_ddp(rank, worldsize)
+    setup_ddp(rank, worldsize, port)
     x = (rank + 1) * torch.ones(10, requires_grad=True)
 
     # random linear transformation, it should really not matter what we do here
@@ -121,9 +121,9 @@ def _test_ddp_gather_all_autograd_same_shape(rank: int, worldsize: int = NUM_PRO
     assert torch.allclose(grad, a * torch.ones_like(x))
 
 
-def _test_ddp_gather_all_autograd_different_shape(rank: int, worldsize: int = NUM_PROCESSES) -> None:
+def _test_ddp_gather_all_autograd_different_shape(rank: int, worldsize: int, port: int) -> None:
     """Test that ddp gather preserves local rank's autograd graph for differently-shaped tensors across ranks."""
-    setup_ddp(rank, worldsize)
+    setup_ddp(rank, worldsize, port)
     x = (rank + 1) * torch.ones(rank + 1, 2 - rank, requires_grad=True)
 
     # random linear transformation, it should really not matter what we do here
@@ -144,7 +144,8 @@ def _test_ddp_gather_all_autograd_different_shape(rank: int, worldsize: int = NU
 )
 def test_ddp_autograd(process):
     """Test ddp functions for autograd compatibility."""
-    pytest.pool.map(process, range(NUM_PROCESSES))
+    port = get_free_port()
+    pytest.pool.starmap(process, [(rank, NUM_PROCESSES, port) for rank in range(NUM_PROCESSES)])
 
 
 def _test_non_contiguous_tensors(rank):
diff --git a/tests/unittests/conftest.py b/tests/unittests/conftest.py
index 746a3e66c26..dde15ca3a90 100644
--- a/tests/unittests/conftest.py
+++ b/tests/unittests/conftest.py
@@ -50,7 +50,7 @@ def get_free_port():
         return s.getsockname()[1]
 
 
-def setup_ddp(rank, world_size):
+def setup_ddp(rank, world_size, master_port):
     """Initialize ddp environment.
 
     If a particular test relies on the order of the processes in the pool to be [0, 1, 2, ...], then this function
@@ -60,10 +60,11 @@ def setup_ddp(rank, world_size):
     Args:
         rank: the rank of the process
         world_size: the number of processes
+        master_port: the port to use for the master process
 
     """
     os.environ["MASTER_ADDR"] = "localhost"
-    os.environ["MASTER_PORT"] = str(get_free_port())
+    os.environ["MASTER_PORT"] = str(master_port)
 
     if torch.distributed.group.WORLD is not None:  # if already initialized, destroy the process group
         torch.distributed.destroy_process_group()
@@ -72,12 +73,19 @@ def setup_ddp(rank, world_size):
         torch.distributed.init_process_group("gloo", rank=rank, world_size=world_size)
 
 
+def cleanup_ddp():
+    """Clean up the DDP process group if initialized."""
+    if torch.distributed.is_initialized():
+        torch.distributed.destroy_process_group()
+
+
 def pytest_sessionstart():
     """Global initialization of multiprocessing pool; runs before any test."""
     if not USE_PYTEST_POOL:
         return
+    port = get_free_port()
     pool = Pool(processes=NUM_PROCESSES)
-    pool.starmap(setup_ddp, [(rank, NUM_PROCESSES) for rank in range(NUM_PROCESSES)])
+    pool.starmap(setup_ddp, [(rank, NUM_PROCESSES, port) for rank in range(NUM_PROCESSES)])
     pytest.pool = pool
 
 
diff --git a/tests/unittests/image/__init__.py b/tests/unittests/image/__init__.py
index 8eea7d284b8..13ab22f7140 100644
--- a/tests/unittests/image/__init__.py
+++ b/tests/unittests/image/__init__.py
@@ -13,23 +13,6 @@
 # limitations under the License.
 import os
 
-import torch
-import torch.distributed as dist
-
 from unittests import _PATH_ALL_TESTS
 
 _SAMPLE_IMAGE = os.path.join(_PATH_ALL_TESTS, "_data", "image", "i01_01_5.bmp")
-
-
-def setup_ddp(rank: int, world_size: int, free_port: int):
-    """Set up DDP with a free port and assign CUDA device to the given rank."""
-    os.environ["MASTER_ADDR"] = "localhost"
-    os.environ["MASTER_PORT"] = str(free_port)
-    dist.init_process_group(backend="nccl", rank=rank, world_size=world_size)
-    torch.cuda.set_device(rank)
-
-
-def cleanup_ddp():
-    """Clean up the DDP process group if initialized."""
-    if dist.is_initialized():
-        dist.destroy_process_group()
diff --git a/tests/unittests/image/test_ms_ssim.py b/tests/unittests/image/test_ms_ssim.py
index c7501ff1c4e..8995d8a6b2c 100644
--- a/tests/unittests/image/test_ms_ssim.py
+++ b/tests/unittests/image/test_ms_ssim.py
@@ -24,8 +24,7 @@
 from unittests import NUM_BATCHES, _Input
 from unittests._helpers import seed_all
 from unittests._helpers.testers import MetricTester
-from unittests.conftest import get_free_port
-from unittests.image import cleanup_ddp, setup_ddp
+from unittests.conftest import cleanup_ddp, get_free_port, setup_ddp
 
 seed_all(42)
 
diff --git a/tests/unittests/image/test_ssim.py b/tests/unittests/image/test_ssim.py
index a3e0e1ba299..352b33c5fe1 100644
--- a/tests/unittests/image/test_ssim.py
+++ b/tests/unittests/image/test_ssim.py
@@ -27,8 +27,7 @@
 from unittests import NUM_BATCHES, _Input
 from unittests._helpers import seed_all
 from unittests._helpers.testers import MetricTester
-from unittests.conftest import get_free_port
-from unittests.image import cleanup_ddp, setup_ddp
+from unittests.conftest import cleanup_ddp, get_free_port, setup_ddp
 
 seed_all(42)
 

From 6c92ed981b5ba70ea375fd470a736360af673ded Mon Sep 17 00:00:00 2001
From: Nicki Skafte <skaftenicki@gmail.com>
Date: Tue, 5 Aug 2025 07:02:27 +0200
Subject: [PATCH 06/25] try limit number of parallel jobs for debugging

---
 .github/workflows/ci-tests.yml | 1 +
 1 file changed, 1 insertion(+)

diff --git a/.github/workflows/ci-tests.yml b/.github/workflows/ci-tests.yml
index 9dca7e4f4ed..cfff7987d7e 100644
--- a/.github/workflows/ci-tests.yml
+++ b/.github/workflows/ci-tests.yml
@@ -33,6 +33,7 @@ jobs:
     needs: check-diff
     strategy:
       fail-fast: false
+      max-parallel: 2
       matrix:
         os: ["ubuntu-22.04"]
         python-version: ["3.10"]

From 86f33bd0fd68553f865826a157d94f11a67a470c Mon Sep 17 00:00:00 2001
From: "pre-commit-ci[bot]"
 <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Date: Wed, 6 Aug 2025 15:07:13 +0000
Subject: [PATCH 07/25] [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci
---
 tests/unittests/utilities/test_utilities.py | 1 -
 1 file changed, 1 deletion(-)

diff --git a/tests/unittests/utilities/test_utilities.py b/tests/unittests/utilities/test_utilities.py
index dc1d4563a52..00ffa53b3a1 100644
--- a/tests/unittests/utilities/test_utilities.py
+++ b/tests/unittests/utilities/test_utilities.py
@@ -19,7 +19,6 @@
 from lightning_utilities.test.warning import no_warning_call
 from torch import tensor
 from unittests._helpers import _IS_WINDOWS
-from unittests.conftest import MAX_PORT, START_PORT
 
 from torchmetrics.regression import MeanSquaredError, PearsonCorrCoef
 from torchmetrics.utilities import check_forward_full_state_property, rank_zero_debug, rank_zero_info, rank_zero_warn

From 0e502a64c262baf10ddbc185a9d2d97d4ff7234e Mon Sep 17 00:00:00 2001
From: Jirka B <j.borovec+github@gmail.com>
Date: Wed, 6 Aug 2025 21:37:55 +0200
Subject: [PATCH 08/25] pytest

---
 .github/workflows/ci-tests.yml | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/.github/workflows/ci-tests.yml b/.github/workflows/ci-tests.yml
index cfff7987d7e..b75a3d32930 100644
--- a/.github/workflows/ci-tests.yml
+++ b/.github/workflows/ci-tests.yml
@@ -190,7 +190,7 @@ jobs:
         if: ${{ env.TEST_DIRS != '' && matrix.testing == 'common' }}
         working-directory: ./tests
         run: |
-          python -m pytest \
+          pytest \
             $TEST_DIRS \
             --cov=torchmetrics \
             --durations=50 \
@@ -208,7 +208,7 @@ jobs:
         env:
           USE_PYTEST_POOL: "1"
         run: |
-          python -m pytest -v \
+          pytest -v \
             $TEST_DIRS \
             --cov=torchmetrics \
             --durations=50 \

From 913cea6ed97d37296005623422b591b00c79fef0 Mon Sep 17 00:00:00 2001
From: Jirka B <j.borovec+github@gmail.com>
Date: Wed, 6 Aug 2025 21:54:47 +0200
Subject: [PATCH 09/25] gpu

---
 .azure/gpu-unittests.yml       | 25 ++++++++++++++++---------
 .github/workflows/ci-tests.yml |  1 -
 2 files changed, 16 insertions(+), 10 deletions(-)

diff --git a/.azure/gpu-unittests.yml b/.azure/gpu-unittests.yml
index 12b45a0ed53..f6f4dae5648 100644
--- a/.azure/gpu-unittests.yml
+++ b/.azure/gpu-unittests.yml
@@ -24,11 +24,18 @@ jobs:
   - job: unitest_GPU
     strategy:
       matrix:
-        "PyTorch | 2.0 oldest":
+        "PyTorch | 2.0 oldest | commonn":
           # Torch does not have build wheels with old Torch versions for newer CUDA
           docker-image: "ubuntu22.04-cuda12.1.1-py3.9-torch2.0"
           torch-ver: "2.0"
-        "PyTorch | 2.X stable":
+        "PyTorch | 2.0 oldest | DDP":
+          # Torch does not have build wheels with old Torch versions for newer CUDA
+          docker-image: "ubuntu22.04-cuda12.1.1-py3.9-torch2.0"
+          torch-ver: "2.0"
+        "PyTorch | 2.X stable | commonn":
+          docker-image: "ubuntu22.04-cuda12.6.3-py3.11-torch2.7"
+          torch-ver: "2.7"
+        "PyTorch | 2.X stable | DDP":
           docker-image: "ubuntu22.04-cuda12.6.3-py3.11-torch2.7"
           torch-ver: "2.7"
         #"PyTorch | 2.X future":
@@ -123,7 +130,7 @@ jobs:
 
       - bash: |
           python .github/assistant.py set-oldest-versions
-        condition: eq(variables['torch-ver'], '2.0')
+        condition: contains(variables['Agent.JobName'], 'oldest')
         displayName: "Setting oldest versions"
 
       - bash: |
@@ -185,27 +192,27 @@ jobs:
 
       - bash: |
           du -h --max-depth=1 .
-          python -m pytest $(TEST_DIRS) \
+          pytest $(TEST_DIRS) \
             -m "not DDP" --numprocesses=6 --dist=loadfile \
             --cov=torchmetrics --timeout=240 --durations=100 \
             --reruns 3 --reruns-delay 1
         workingDirectory: "tests/"
         # skip for PR if there is nothing to test, note that outside PR there is default 'unittests'
-        condition: and(succeeded(), ne(variables['TEST_DIRS'], ''))
+        condition: and(succeeded(), ne(variables['TEST_DIRS'], ''), contains(variables['Agent.JobName'], 'commonn'))
         timeoutInMinutes: "120"
-        displayName: "UnitTesting common"
+        displayName: "Testing common"
 
       - bash: |
-          python -m pytest $(TEST_DIRS) -v \
+          pytest $(TEST_DIRS) -v \
             --cov=torchmetrics -m "DDP" \
             --timeout=240 --durations=100
         env:
           USE_PYTEST_POOL: "1"
         workingDirectory: "tests/"
         # skip for PR if there is nothing to test, note that outside PR there is default 'unittests'
-        condition: and(succeeded(), ne(variables['TEST_DIRS'], ''))
+        condition: and(succeeded(), ne(variables['TEST_DIRS'], ''), contains(variables['Agent.JobName'], 'DDP'))
         timeoutInMinutes: "120"
-        displayName: "UnitTesting DDP"
+        displayName: "Testing DDP"
 
       - bash: |
           du -h --max-depth=1 tests/
diff --git a/.github/workflows/ci-tests.yml b/.github/workflows/ci-tests.yml
index b75a3d32930..e8e14dd789a 100644
--- a/.github/workflows/ci-tests.yml
+++ b/.github/workflows/ci-tests.yml
@@ -33,7 +33,6 @@ jobs:
     needs: check-diff
     strategy:
       fail-fast: false
-      max-parallel: 2
       matrix:
         os: ["ubuntu-22.04"]
         python-version: ["3.10"]

From 5637d3b836f3888f845b5f70a44446292b3bcbc3 Mon Sep 17 00:00:00 2001
From: Jirka B <j.borovec+github@gmail.com>
Date: Wed, 6 Aug 2025 22:40:28 +0200
Subject: [PATCH 10/25] doctest

---
 .azure/gpu-unittests.yml | 8 ++++++++
 1 file changed, 8 insertions(+)

diff --git a/.azure/gpu-unittests.yml b/.azure/gpu-unittests.yml
index f6f4dae5648..7812224a9e7 100644
--- a/.azure/gpu-unittests.yml
+++ b/.azure/gpu-unittests.yml
@@ -28,10 +28,17 @@ jobs:
           # Torch does not have build wheels with old Torch versions for newer CUDA
           docker-image: "ubuntu22.04-cuda12.1.1-py3.9-torch2.0"
           torch-ver: "2.0"
+        "PyTorch | 2.0 oldest | doctest":
+          # Torch does not have build wheels with old Torch versions for newer CUDA
+          docker-image: "ubuntu22.04-cuda12.1.1-py3.9-torch2.0"
+          torch-ver: "2.0"
         "PyTorch | 2.0 oldest | DDP":
           # Torch does not have build wheels with old Torch versions for newer CUDA
           docker-image: "ubuntu22.04-cuda12.1.1-py3.9-torch2.0"
           torch-ver: "2.0"
+        "PyTorch | 2.X stable | doctest":
+          docker-image: "ubuntu22.04-cuda12.6.3-py3.11-torch2.7"
+          torch-ver: "2.7"
         "PyTorch | 2.X stable | commonn":
           docker-image: "ubuntu22.04-cuda12.6.3-py3.11-torch2.7"
           torch-ver: "2.7"
@@ -155,6 +162,7 @@ jobs:
             --timeout=240 --durations=50 \
             --reruns 2 --reruns-delay 1
           #  --numprocesses=5 --dist=loadfile
+        condition: contains(variables['Agent.JobName'], 'doctest')
         env:
           DOCTEST_DOWNLOAD_TIMEOUT: "180"
           SKIP_SLOW_DOCTEST: "1"

From a0ad4885ef90081d37eaee1c192c565abd906cda Mon Sep 17 00:00:00 2001
From: Jirka B <j.borovec+github@gmail.com>
Date: Thu, 7 Aug 2025 19:40:08 +0200
Subject: [PATCH 11/25] rev GPU

---
 .azure/gpu-integrations.yml |  4 ++--
 .azure/gpu-unittests.yml    | 37 +++++++++++--------------------------
 2 files changed, 13 insertions(+), 28 deletions(-)

diff --git a/.azure/gpu-integrations.yml b/.azure/gpu-integrations.yml
index fa3bc22c5d7..9d956bc434d 100644
--- a/.azure/gpu-integrations.yml
+++ b/.azure/gpu-integrations.yml
@@ -22,8 +22,8 @@ jobs:
           torch-ver: "2.0"
           requires: "oldest"
         "torch | 2.x":
-          docker-image: "pytorch/pytorch:2.7.1-cuda12.6-cudnn9-runtime"
-          torch-ver: "2.7"
+          docker-image: "pytorch/pytorch:2.8.0-cuda12.6-cudnn9-runtime"
+          torch-ver: "2.8"
     # how long to run the job before automatically cancelling
     timeoutInMinutes: "40"
     # how much time to give 'run always even if cancelled tasks' before stopping them
diff --git a/.azure/gpu-unittests.yml b/.azure/gpu-unittests.yml
index 7812224a9e7..8d287268df1 100644
--- a/.azure/gpu-unittests.yml
+++ b/.azure/gpu-unittests.yml
@@ -24,27 +24,13 @@ jobs:
   - job: unitest_GPU
     strategy:
       matrix:
-        "PyTorch | 2.0 oldest | commonn":
+        "PyTorch | 2.0 oldest":
           # Torch does not have build wheels with old Torch versions for newer CUDA
           docker-image: "ubuntu22.04-cuda12.1.1-py3.9-torch2.0"
           torch-ver: "2.0"
-        "PyTorch | 2.0 oldest | doctest":
-          # Torch does not have build wheels with old Torch versions for newer CUDA
-          docker-image: "ubuntu22.04-cuda12.1.1-py3.9-torch2.0"
-          torch-ver: "2.0"
-        "PyTorch | 2.0 oldest | DDP":
-          # Torch does not have build wheels with old Torch versions for newer CUDA
-          docker-image: "ubuntu22.04-cuda12.1.1-py3.9-torch2.0"
-          torch-ver: "2.0"
-        "PyTorch | 2.X stable | doctest":
-          docker-image: "ubuntu22.04-cuda12.6.3-py3.11-torch2.7"
-          torch-ver: "2.7"
-        "PyTorch | 2.X stable | commonn":
-          docker-image: "ubuntu22.04-cuda12.6.3-py3.11-torch2.7"
-          torch-ver: "2.7"
-        "PyTorch | 2.X stable | DDP":
-          docker-image: "ubuntu22.04-cuda12.6.3-py3.11-torch2.7"
-          torch-ver: "2.7"
+        "PyTorch | 2.X stable":
+          docker-image: "ubuntu24.04-cuda12.6.3-py3.12-torch2.8"
+          torch-ver: "2.8"
         #"PyTorch | 2.X future":
         #  docker-image: "ubuntu22.04-cuda12.6.3-py3.11-torch2.8"
         #  torch-ver: "2.8"
@@ -137,7 +123,7 @@ jobs:
 
       - bash: |
           python .github/assistant.py set-oldest-versions
-        condition: contains(variables['Agent.JobName'], 'oldest')
+        condition: eq(variables['torch-ver'], '2.0')
         displayName: "Setting oldest versions"
 
       - bash: |
@@ -162,7 +148,6 @@ jobs:
             --timeout=240 --durations=50 \
             --reruns 2 --reruns-delay 1
           #  --numprocesses=5 --dist=loadfile
-        condition: contains(variables['Agent.JobName'], 'doctest')
         env:
           DOCTEST_DOWNLOAD_TIMEOUT: "180"
           SKIP_SLOW_DOCTEST: "1"
@@ -200,27 +185,27 @@ jobs:
 
       - bash: |
           du -h --max-depth=1 .
-          pytest $(TEST_DIRS) \
+          python -m pytest $(TEST_DIRS) \
             -m "not DDP" --numprocesses=6 --dist=loadfile \
             --cov=torchmetrics --timeout=240 --durations=100 \
             --reruns 3 --reruns-delay 1
         workingDirectory: "tests/"
         # skip for PR if there is nothing to test, note that outside PR there is default 'unittests'
-        condition: and(succeeded(), ne(variables['TEST_DIRS'], ''), contains(variables['Agent.JobName'], 'commonn'))
+        condition: and(succeeded(), ne(variables['TEST_DIRS'], ''))
         timeoutInMinutes: "120"
-        displayName: "Testing common"
+        displayName: "UnitTesting common"
 
       - bash: |
-          pytest $(TEST_DIRS) -v \
+          python -m pytest $(TEST_DIRS) -v \
             --cov=torchmetrics -m "DDP" \
             --timeout=240 --durations=100
         env:
           USE_PYTEST_POOL: "1"
         workingDirectory: "tests/"
         # skip for PR if there is nothing to test, note that outside PR there is default 'unittests'
-        condition: and(succeeded(), ne(variables['TEST_DIRS'], ''), contains(variables['Agent.JobName'], 'DDP'))
+        condition: and(succeeded(), ne(variables['TEST_DIRS'], ''))
         timeoutInMinutes: "120"
-        displayName: "Testing DDP"
+        displayName: "UnitTesting DDP"
 
       - bash: |
           du -h --max-depth=1 tests/

From d10a688bc66f7a6ad39d04e0869fa4a3fbd4c9d4 Mon Sep 17 00:00:00 2001
From: Jirka B <j.borovec+github@gmail.com>
Date: Mon, 11 Aug 2025 19:39:48 +0200
Subject: [PATCH 12/25] get_free_port

---
 tests/unittests/conftest.py | 21 +++++++++++++++------
 1 file changed, 15 insertions(+), 6 deletions(-)

diff --git a/tests/unittests/conftest.py b/tests/unittests/conftest.py
index dde15ca3a90..c94474d0fbd 100644
--- a/tests/unittests/conftest.py
+++ b/tests/unittests/conftest.py
@@ -42,12 +42,21 @@ def use_deterministic_algorithms():
     torch.use_deterministic_algorithms(False)
 
 
-def get_free_port():
-    """Find an available free port on localhost."""
-    with socket.socket(socket.AF_INET, socket.SOCK_STREAM) as s:
-        s.bind(("localhost", 0))  # Bind to a free port provided by the OS
-        s.setsockopt(socket.SOL_SOCKET, socket.SO_REUSEADDR, 1)
-        return s.getsockname()[1]
+def get_free_port(max_tries=10):
+    """Find an available free port on localhost, retrying if necessary."""
+    for _ in range(max_tries):
+        with socket.socket(socket.AF_INET, socket.SOCK_STREAM) as s:
+            s.bind(("localhost", 0))
+            s.setsockopt(socket.SOL_SOCKET, socket.SO_REUSEADDR, 1)
+            port = s.getsockname()[1]
+        # Try to bind again to check if port is still free
+        try:
+            with socket.socket(socket.AF_INET, socket.SOCK_STREAM) as test_sock:
+                test_sock.bind(("localhost", port))
+                return port
+        except OSError:
+            continue
+    raise RuntimeError("Could not find a free port after several attempts")
 
 
 def setup_ddp(rank, world_size, master_port):

From 1c93193ff6f433add9ec7a0352278423d440f554 Mon Sep 17 00:00:00 2001
From: Jirka B <j.borovec+github@gmail.com>
Date: Mon, 11 Aug 2025 19:52:00 +0200
Subject: [PATCH 13/25] get_free_port

---
 tests/unittests/conftest.py | 24 +++++++++---------------
 1 file changed, 9 insertions(+), 15 deletions(-)

diff --git a/tests/unittests/conftest.py b/tests/unittests/conftest.py
index c94474d0fbd..c592658b54d 100644
--- a/tests/unittests/conftest.py
+++ b/tests/unittests/conftest.py
@@ -42,21 +42,15 @@ def use_deterministic_algorithms():
     torch.use_deterministic_algorithms(False)
 
 
-def get_free_port(max_tries=10):
-    """Find an available free port on localhost, retrying if necessary."""
-    for _ in range(max_tries):
-        with socket.socket(socket.AF_INET, socket.SOCK_STREAM) as s:
-            s.bind(("localhost", 0))
-            s.setsockopt(socket.SOL_SOCKET, socket.SO_REUSEADDR, 1)
-            port = s.getsockname()[1]
-        # Try to bind again to check if port is still free
-        try:
-            with socket.socket(socket.AF_INET, socket.SOCK_STREAM) as test_sock:
-                test_sock.bind(("localhost", port))
-                return port
-        except OSError:
-            continue
-    raise RuntimeError("Could not find a free port after several attempts")
+def get_free_port():
+    """Find an available free port on localhost and keep it reserved."""
+    with socket.socket(socket.AF_INET, socket.SOCK_STREAM) as s:
+        s.bind(("localhost", 0))  # Bind to a free port provided by the OS
+        s.setsockopt(socket.SOL_SOCKET, socket.SO_REUSEADDR, 1)
+        port = s.getsockname()[1]
+        # Keep socket open longer to prevent immediate reuse
+        s.listen(1)
+        return port
 
 
 def setup_ddp(rank, world_size, master_port):

From 6f9bec7d5ec721dce5e93da7577a8f8384b474c6 Mon Sep 17 00:00:00 2001
From: Jirka B <j.borovec+github@gmail.com>
Date: Tue, 12 Aug 2025 00:40:23 +0200
Subject: [PATCH 14/25] get_free_port

---
 tests/unittests/conftest.py | 28 ++++++++++++++++++++++------
 1 file changed, 22 insertions(+), 6 deletions(-)

diff --git a/tests/unittests/conftest.py b/tests/unittests/conftest.py
index c592658b54d..aff46720797 100644
--- a/tests/unittests/conftest.py
+++ b/tests/unittests/conftest.py
@@ -43,14 +43,30 @@ def use_deterministic_algorithms():
 
 
 def get_free_port():
-    """Find an available free port on localhost and keep it reserved."""
+    """Find an available free port on localhost with better reservation."""
+    import time
+    import random
+
+    # Try multiple times with different base ports to avoid conflicts
+    for _ in range(10):
+        try:
+            with socket.socket(socket.AF_INET, socket.SOCK_STREAM) as s:
+                # Use a random port in a higher range to avoid common conflicts
+                base_port = random.randint(20000, 30000)
+                s.bind(("localhost", base_port))
+                s.setsockopt(socket.SOL_SOCKET, socket.SO_REUSEADDR, 1)
+                port = s.getsockname()[1]
+                # Brief delay to reduce race conditions
+                time.sleep(0.1)
+                return port
+        except OSError:
+            continue
+
+    # Fallback to original method
     with socket.socket(socket.AF_INET, socket.SOCK_STREAM) as s:
-        s.bind(("localhost", 0))  # Bind to a free port provided by the OS
+        s.bind(("localhost", 0))
         s.setsockopt(socket.SOL_SOCKET, socket.SO_REUSEADDR, 1)
-        port = s.getsockname()[1]
-        # Keep socket open longer to prevent immediate reuse
-        s.listen(1)
-        return port
+        return s.getsockname()[1]
 
 
 def setup_ddp(rank, world_size, master_port):

From 5f81e754a7aa1f436643ce9f6cd1a88676e5d76d Mon Sep 17 00:00:00 2001
From: "pre-commit-ci[bot]"
 <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Date: Mon, 11 Aug 2025 22:41:49 +0000
Subject: [PATCH 15/25] [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci
---
 tests/unittests/conftest.py | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/tests/unittests/conftest.py b/tests/unittests/conftest.py
index aff46720797..790ad91de90 100644
--- a/tests/unittests/conftest.py
+++ b/tests/unittests/conftest.py
@@ -44,8 +44,8 @@ def use_deterministic_algorithms():
 
 def get_free_port():
     """Find an available free port on localhost with better reservation."""
-    import time
     import random
+    import time
 
     # Try multiple times with different base ports to avoid conflicts
     for _ in range(10):

From 4b3c37e63957589be04bc2af893bd322e84f0150 Mon Sep 17 00:00:00 2001
From: Jirka B <j.borovec+github@gmail.com>
Date: Tue, 12 Aug 2025 10:06:49 +0200
Subject: [PATCH 16/25] get_free_port()

---
 tests/unittests/conftest.py | 33 ++++++++-------------------------
 1 file changed, 8 insertions(+), 25 deletions(-)

diff --git a/tests/unittests/conftest.py b/tests/unittests/conftest.py
index 790ad91de90..7afa62e9433 100644
--- a/tests/unittests/conftest.py
+++ b/tests/unittests/conftest.py
@@ -42,31 +42,15 @@ def use_deterministic_algorithms():
     torch.use_deterministic_algorithms(False)
 
 
-def get_free_port():
-    """Find an available free port on localhost with better reservation."""
-    import random
-    import time
-
-    # Try multiple times with different base ports to avoid conflicts
-    for _ in range(10):
-        try:
-            with socket.socket(socket.AF_INET, socket.SOCK_STREAM) as s:
-                # Use a random port in a higher range to avoid common conflicts
-                base_port = random.randint(20000, 30000)
-                s.bind(("localhost", base_port))
-                s.setsockopt(socket.SOL_SOCKET, socket.SO_REUSEADDR, 1)
-                port = s.getsockname()[1]
-                # Brief delay to reduce race conditions
-                time.sleep(0.1)
-                return port
-        except OSError:
-            continue
-
-    # Fallback to original method
+def get_free_port() -> int:
+    """Find an available free port on localhost and keep it reserved."""
     with socket.socket(socket.AF_INET, socket.SOCK_STREAM) as s:
-        s.bind(("localhost", 0))
+        s.bind(("localhost", 0))  # Bind to a free port provided by the OS
         s.setsockopt(socket.SOL_SOCKET, socket.SO_REUSEADDR, 1)
-        return s.getsockname()[1]
+        port = s.getsockname()[1]
+        # Keep socket open longer to prevent immediate reuse
+        s.listen(1)
+        return port
 
 
 def setup_ddp(rank, world_size, master_port):
@@ -102,9 +86,8 @@ def pytest_sessionstart():
     """Global initialization of multiprocessing pool; runs before any test."""
     if not USE_PYTEST_POOL:
         return
-    port = get_free_port()
     pool = Pool(processes=NUM_PROCESSES)
-    pool.starmap(setup_ddp, [(rank, NUM_PROCESSES, port) for rank in range(NUM_PROCESSES)])
+    pool.starmap(setup_ddp, [(rank, NUM_PROCESSES, get_free_port()) for rank in range(NUM_PROCESSES)])
     pytest.pool = pool
 
 

From b5f9ae37f5e482f6d6e38339b9206f35eb1aeacf Mon Sep 17 00:00:00 2001
From: Jirka B <j.borovec+github@gmail.com>
Date: Tue, 12 Aug 2025 10:58:17 +0200
Subject: [PATCH 17/25] settimeout

---
 tests/unittests/conftest.py | 1 +
 1 file changed, 1 insertion(+)

diff --git a/tests/unittests/conftest.py b/tests/unittests/conftest.py
index 7afa62e9433..597bb3538f2 100644
--- a/tests/unittests/conftest.py
+++ b/tests/unittests/conftest.py
@@ -45,6 +45,7 @@ def use_deterministic_algorithms():
 def get_free_port() -> int:
     """Find an available free port on localhost and keep it reserved."""
     with socket.socket(socket.AF_INET, socket.SOCK_STREAM) as s:
+        s.settimeout(5)  # Set a timeout of 5 seconds
         s.bind(("localhost", 0))  # Bind to a free port provided by the OS
         s.setsockopt(socket.SOL_SOCKET, socket.SO_REUSEADDR, 1)
         port = s.getsockname()[1]

From 666cb7fb134d18657667d548a962975a2a058c24 Mon Sep 17 00:00:00 2001
From: Jirka B <j.borovec+github@gmail.com>
Date: Tue, 12 Aug 2025 11:00:37 +0200
Subject: [PATCH 18/25] setup_ddp

---
 tests/unittests/conftest.py | 7 +++----
 1 file changed, 3 insertions(+), 4 deletions(-)

diff --git a/tests/unittests/conftest.py b/tests/unittests/conftest.py
index 597bb3538f2..34104cee3fc 100644
--- a/tests/unittests/conftest.py
+++ b/tests/unittests/conftest.py
@@ -54,7 +54,7 @@ def get_free_port() -> int:
         return port
 
 
-def setup_ddp(rank, world_size, master_port):
+def setup_ddp(rank: int, world_size: int) -> None:
     """Initialize ddp environment.
 
     If a particular test relies on the order of the processes in the pool to be [0, 1, 2, ...], then this function
@@ -64,11 +64,10 @@ def setup_ddp(rank, world_size, master_port):
     Args:
         rank: the rank of the process
         world_size: the number of processes
-        master_port: the port to use for the master process
 
     """
     os.environ["MASTER_ADDR"] = "localhost"
-    os.environ["MASTER_PORT"] = str(master_port)
+    os.environ["MASTER_PORT"] = str(get_free_port())
 
     if torch.distributed.group.WORLD is not None:  # if already initialized, destroy the process group
         torch.distributed.destroy_process_group()
@@ -88,7 +87,7 @@ def pytest_sessionstart():
     if not USE_PYTEST_POOL:
         return
     pool = Pool(processes=NUM_PROCESSES)
-    pool.starmap(setup_ddp, [(rank, NUM_PROCESSES, get_free_port()) for rank in range(NUM_PROCESSES)])
+    pool.starmap(setup_ddp, [(rank, NUM_PROCESSES) for rank in range(NUM_PROCESSES)])
     pytest.pool = pool
 
 

From 12253e60f387060db99af8e6fd58766cc72aaa0a Mon Sep 17 00:00:00 2001
From: Jirka B <j.borovec+github@gmail.com>
Date: Tue, 12 Aug 2025 11:21:10 +0200
Subject: [PATCH 19/25] timeout-minutes: 70

---
 .github/workflows/ci-tests.yml | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/.github/workflows/ci-tests.yml b/.github/workflows/ci-tests.yml
index a44ddb5e7d8..5075cebb459 100644
--- a/.github/workflows/ci-tests.yml
+++ b/.github/workflows/ci-tests.yml
@@ -91,7 +91,7 @@ jobs:
 
     # Timeout: https://stackoverflow.com/a/59076067/4521646
     # seems that macOS jobs take much more than orger OS
-    timeout-minutes: 120
+    timeout-minutes: 70
 
     steps:
       - uses: actions/checkout@v4

From e73e1d3078bd2abfb9ec07d17fe8b543f84c5059 Mon Sep 17 00:00:00 2001
From: Jirka B <j.borovec+github@gmail.com>
Date: Tue, 12 Aug 2025 11:24:56 +0200
Subject: [PATCH 20/25] enumerate

---
 tests/unittests/conftest.py | 7 ++++---
 1 file changed, 4 insertions(+), 3 deletions(-)

diff --git a/tests/unittests/conftest.py b/tests/unittests/conftest.py
index 34104cee3fc..c17e084d148 100644
--- a/tests/unittests/conftest.py
+++ b/tests/unittests/conftest.py
@@ -54,7 +54,7 @@ def get_free_port() -> int:
         return port
 
 
-def setup_ddp(rank: int, world_size: int) -> None:
+def setup_ddp(rank: int, world_size: int, port: int) -> None:
     """Initialize ddp environment.
 
     If a particular test relies on the order of the processes in the pool to be [0, 1, 2, ...], then this function
@@ -67,7 +67,7 @@ def setup_ddp(rank: int, world_size: int) -> None:
 
     """
     os.environ["MASTER_ADDR"] = "localhost"
-    os.environ["MASTER_PORT"] = str(get_free_port())
+    os.environ["MASTER_PORT"] = str(port)
 
     if torch.distributed.group.WORLD is not None:  # if already initialized, destroy the process group
         torch.distributed.destroy_process_group()
@@ -86,8 +86,9 @@ def pytest_sessionstart():
     """Global initialization of multiprocessing pool; runs before any test."""
     if not USE_PYTEST_POOL:
         return
+    port = get_free_port()
     pool = Pool(processes=NUM_PROCESSES)
-    pool.starmap(setup_ddp, [(rank, NUM_PROCESSES) for rank in range(NUM_PROCESSES)])
+    pool.starmap(setup_ddp, [(rank, NUM_PROCESSES, port + i) for i, rank in enumerate(range(NUM_PROCESSES))])
     pytest.pool = pool
 
 

From b8eae2c52680209970754a0fd554dafd7a94171b Mon Sep 17 00:00:00 2001
From: Jirka B <j.borovec+github@gmail.com>
Date: Tue, 12 Aug 2025 11:36:27 +0200
Subject: [PATCH 21/25] get_free_port

---
 tests/unittests/conftest.py | 3 ---
 1 file changed, 3 deletions(-)

diff --git a/tests/unittests/conftest.py b/tests/unittests/conftest.py
index c17e084d148..fdec272f9ae 100644
--- a/tests/unittests/conftest.py
+++ b/tests/unittests/conftest.py
@@ -45,12 +45,9 @@ def use_deterministic_algorithms():
 def get_free_port() -> int:
     """Find an available free port on localhost and keep it reserved."""
     with socket.socket(socket.AF_INET, socket.SOCK_STREAM) as s:
-        s.settimeout(5)  # Set a timeout of 5 seconds
         s.bind(("localhost", 0))  # Bind to a free port provided by the OS
         s.setsockopt(socket.SOL_SOCKET, socket.SO_REUSEADDR, 1)
         port = s.getsockname()[1]
-        # Keep socket open longer to prevent immediate reuse
-        s.listen(1)
         return port
 
 

From 2021ca51e577dfad4d11b34dbc01433c9dab71b0 Mon Sep 17 00:00:00 2001
From: Jirka B <j.borovec+github@gmail.com>
Date: Tue, 12 Aug 2025 11:38:30 +0200
Subject: [PATCH 22/25] get_free_port

---
 tests/unittests/conftest.py | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/tests/unittests/conftest.py b/tests/unittests/conftest.py
index fdec272f9ae..5ee2d99ca1b 100644
--- a/tests/unittests/conftest.py
+++ b/tests/unittests/conftest.py
@@ -48,7 +48,7 @@ def get_free_port() -> int:
         s.bind(("localhost", 0))  # Bind to a free port provided by the OS
         s.setsockopt(socket.SOL_SOCKET, socket.SO_REUSEADDR, 1)
         port = s.getsockname()[1]
-        return port
+        return int(port)
 
 
 def setup_ddp(rank: int, world_size: int, port: int) -> None:
@@ -61,6 +61,7 @@ def setup_ddp(rank: int, world_size: int, port: int) -> None:
     Args:
         rank: the rank of the process
         world_size: the number of processes
+        port: the port to use for communication
 
     """
     os.environ["MASTER_ADDR"] = "localhost"

From f7fd132dead578676701578773ec779c2d17b7ee Mon Sep 17 00:00:00 2001
From: Jirka B <j.borovec+github@gmail.com>
Date: Tue, 12 Aug 2025 23:10:05 +0200
Subject: [PATCH 23/25] master port

---
 tests/unittests/conftest.py | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/tests/unittests/conftest.py b/tests/unittests/conftest.py
index 5ee2d99ca1b..dd1ac85ee66 100644
--- a/tests/unittests/conftest.py
+++ b/tests/unittests/conftest.py
@@ -86,7 +86,7 @@ def pytest_sessionstart():
         return
     port = get_free_port()
     pool = Pool(processes=NUM_PROCESSES)
-    pool.starmap(setup_ddp, [(rank, NUM_PROCESSES, port + i) for i, rank in enumerate(range(NUM_PROCESSES))])
+    pool.starmap(setup_ddp, [(rank, NUM_PROCESSES, port) for rank in range(NUM_PROCESSES)])
     pytest.pool = pool
 
 

From 154a4c6045e1202f357c4c01b0c160c04102fb92 Mon Sep 17 00:00:00 2001
From: Jirka B <j.borovec+github@gmail.com>
Date: Wed, 13 Aug 2025 10:21:06 +0200
Subject: [PATCH 24/25] no rerun

---
 .github/workflows/ci-tests.yml | 4 ----
 1 file changed, 4 deletions(-)

diff --git a/.github/workflows/ci-tests.yml b/.github/workflows/ci-tests.yml
index 5075cebb459..94ad52165fc 100644
--- a/.github/workflows/ci-tests.yml
+++ b/.github/workflows/ci-tests.yml
@@ -194,8 +194,6 @@ jobs:
             $TEST_DIRS \
             --cov=torchmetrics \
             --durations=50 \
-            --reruns 3 \
-            --reruns-delay 1 \
             -m "not DDP" \
             -n auto \
             --dist=load \
@@ -213,8 +211,6 @@ jobs:
             --cov=torchmetrics \
             --durations=50 \
             -m DDP \
-            --reruns 3 \
-            --reruns-delay 1 \
             ${{ env.UNITTEST_TIMEOUT }}
 
       - name: Statistics

From 774fdd3820e08684f0d829f29be6f034a1e393bf Mon Sep 17 00:00:00 2001
From: "pre-commit-ci[bot]"
 <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Date: Wed, 8 Oct 2025 10:01:04 +0000
Subject: [PATCH 25/25] [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci
---
 README.md | 12 +++++++-----
 1 file changed, 7 insertions(+), 5 deletions(-)

diff --git a/README.md b/README.md
index fd00bcaa6bd..c77320656fd 100644
--- a/README.md
+++ b/README.md
@@ -39,13 +39,15 @@ ______________________________________________________________________
 </div>
 
 # Looking for GPUs?
-Over 340,000 developers use [Lightning Cloud](https://lightning.ai/?utm_source=tm_readme&utm_medium=referral&utm_campaign=tm_readme) - purpose-built for PyTorch and PyTorch Lightning. 
-- [GPUs](https://lightning.ai/pricing?utm_source=tm_readme&utm_medium=referral&utm_campaign=tm_readme) from $0.19.   
-- [Clusters](https://lightning.ai/clusters?utm_source=tm_readme&utm_medium=referral&utm_campaign=tm_readme): frontier-grade training/inference clusters.   
+
+Over 340,000 developers use [Lightning Cloud](https://lightning.ai/?utm_source=tm_readme&utm_medium=referral&utm_campaign=tm_readme) - purpose-built for PyTorch and PyTorch Lightning.
+
+- [GPUs](https://lightning.ai/pricing?utm_source=tm_readme&utm_medium=referral&utm_campaign=tm_readme) from $0.19.
+- [Clusters](https://lightning.ai/clusters?utm_source=tm_readme&utm_medium=referral&utm_campaign=tm_readme): frontier-grade training/inference clusters.
 - [AI Studio (vibe train)](https://lightning.ai/studios?utm_source=tm_readme&utm_medium=referral&utm_campaign=tm_readme): workspaces where AI helps you debug, tune and vibe train.
-- [AI Studio (vibe deploy)](https://lightning.ai/studios?utm_source=tm_readme&utm_medium=referral&utm_campaign=tm_readme): workspaces where AI helps you optimize, and deploy models.     
+- [AI Studio (vibe deploy)](https://lightning.ai/studios?utm_source=tm_readme&utm_medium=referral&utm_campaign=tm_readme): workspaces where AI helps you optimize, and deploy models.
 - [Notebooks](https://lightning.ai/notebooks?utm_source=tm_readme&utm_medium=referral&utm_campaign=tm_readme): Persistent GPU workspaces where AI helps you code and analyze.
-- [Inference](https://lightning.ai/deploy?utm_source=tm_readme&utm_medium=referral&utm_campaign=tm_readme): Deploy models as inference APIs.   
+- [Inference](https://lightning.ai/deploy?utm_source=tm_readme&utm_medium=referral&utm_campaign=tm_readme): Deploy models as inference APIs.
 
 # Installation