Set weights_only in tests to avoid warnings in PyTorch 2.4

tests/tests_fabric/strategies/test_deepspeed_integration.py

@@ -312,11 +312,11 @@ def _assert_saved_model_is_equal(fabric, model, checkpoint_path):
             single_ckpt_path = checkpoint_path / "single_model.pt"
             # the tag is hardcoded in DeepSpeedStrategy
             convert_zero_checkpoint_to_fp32_state_dict(checkpoint_path, single_ckpt_path, tag="checkpoint")
-            state_dict = torch.load(single_ckpt_path)
+            state_dict = torch.load(single_ckpt_path, weights_only=False)
         else:
             # 'checkpoint' is the tag, hardcoded in DeepSpeedStrategy
             single_ckpt_path = checkpoint_path / "checkpoint" / "mp_rank_00_model_states.pt"
-            state_dict = torch.load(single_ckpt_path)["module"]
+            state_dict = torch.load(single_ckpt_path, weights_only=False)["module"]
 
         model = model.cpu()
 

tests/tests_fabric/strategies/test_fsdp_integration.py

@@ -191,7 +191,7 @@ def test_save_full_state_dict(tmp_path):
     state = {"model": trainer.model, "optimizer": trainer.optimizer, "steps": 1}
     fabric.save(checkpoint_path, state)
 
-    checkpoint = torch.load(checkpoint_path)
+    checkpoint = torch.load(checkpoint_path, weights_only=True)
     assert checkpoint["steps"] == 1
     loaded_state_dict = checkpoint["model"]
 
@@ -248,7 +248,7 @@ def test_save_full_state_dict(tmp_path):
     # get optimizer state after loading
     normal_checkpoint_path = Path(fabric.broadcast(str(tmp_path / "normal-checkpoint.pt")))
     fabric.save(normal_checkpoint_path, {"model": trainer.model, "optimizer": trainer.optimizer, "steps": 2})
-    optimizer_state_after = torch.load(normal_checkpoint_path)["optimizer"]
+    optimizer_state_after = torch.load(normal_checkpoint_path, weights_only=True)["optimizer"]
     optimizer_state_after = FullyShardedDataParallel.rekey_optim_state_dict(
         optimizer_state_after, optim_state_key_type=OptimStateKeyType.PARAM_NAME, model=trainer.model
     )
@@ -330,7 +330,7 @@ def test_load_full_state_dict_into_sharded_model(tmp_path):
     # Create a raw state-dict checkpoint to test `Fabric.load_raw` too
     raw_checkpoint_path = checkpoint_path.with_name("model-state-dict")
     if fabric.global_rank == 0:
-        checkpoint = torch.load(checkpoint_path)
+        checkpoint = torch.load(checkpoint_path, weights_only=True)
         torch.save(checkpoint["model"], raw_checkpoint_path)
     fabric.barrier()
 
@@ -485,7 +485,7 @@ def test_save_filter(tmp_path):
 
     checkpoint_path = tmp_path / "full.pth"
     fabric.save(checkpoint_path, state, filter=filter)
-    checkpoint = torch.load(checkpoint_path)["model"]
+    checkpoint = torch.load(checkpoint_path, weights_only=True)["model"]
     assert set(checkpoint) == {"bias"}
     assert type(checkpoint["bias"]) is torch.Tensor
 

tests/tests_fabric/strategies/test_model_parallel_integration.py

@@ -321,7 +321,7 @@ def test_save_full_state_dict(tmp_path):
     state = {"model": model, "optimizer": optimizer, "steps": 1}
     fabric.save(checkpoint_path, state)
 
-    checkpoint = torch.load(checkpoint_path)
+    checkpoint = torch.load(checkpoint_path, weights_only=True)
     assert checkpoint["steps"] == 1
     loaded_state_dict = checkpoint["model"]
 
@@ -369,7 +369,7 @@ def test_save_full_state_dict(tmp_path):
     normal_checkpoint_path = Path(fabric.broadcast(str(tmp_path / "normal-checkpoint.pt")))
     fabric.save(normal_checkpoint_path, {"model": model, "optimizer": optimizer, "steps": 2})
 
-    optimizer_state_after = torch.load(normal_checkpoint_path)["optimizer"]
+    optimizer_state_after = torch.load(normal_checkpoint_path, weights_only=True)["optimizer"]
     assert set(optimizer_state_after.keys()) == set(optimizer_state_before.keys()) == {"state", "param_groups"}
     assert torch.equal(
         optimizer_state_after["state"][0]["exp_avg"],
@@ -433,7 +433,7 @@ def test_load_full_state_dict_into_sharded_model(tmp_path):
     # Create a raw state-dict checkpoint to test `Fabric.load_raw` too
     raw_checkpoint_path = checkpoint_path.with_name("model-state-dict")
     if fabric.global_rank == 0:
-        checkpoint = torch.load(checkpoint_path)
+        checkpoint = torch.load(checkpoint_path, weights_only=True)
         torch.save(checkpoint["model"], raw_checkpoint_path)
     fabric.barrier()
 
@@ -519,7 +519,7 @@ def test_save_filter(tmp_path):
 
     checkpoint_path = tmp_path / "full.pth"
     fabric.save(checkpoint_path, state, filter=filter)
-    checkpoint = torch.load(checkpoint_path)["model"]
+    checkpoint = torch.load(checkpoint_path, weights_only=True)["model"]
     assert set(checkpoint) == {"w1.bias", "w2.bias", "w3.bias"}
     assert type(checkpoint["w1.bias"]) is torch.Tensor
 

tests/tests_pytorch/accelerators/test_xla.py

@@ -67,7 +67,7 @@ def test_resume_training_on_cpu(tmp_path):
     model_path = trainer.checkpoint_callback.best_model_path
 
     # Verify saved Tensors are on CPU
-    ckpt = torch.load(model_path)
+    ckpt = torch.load(model_path, weights_only=True)
     weight_tensor = list(ckpt["state_dict"].values())[0]
     assert weight_tensor.device == torch.device("cpu")
 

tests/tests_pytorch/callbacks/test_early_stopping.py

@@ -15,13 +15,15 @@
 import math
 import os
 import pickle
+from contextlib import nullcontext
 from typing import List, Optional
 from unittest import mock
 from unittest.mock import Mock
 
 import cloudpickle
 import pytest
 import torch
+from lightning.fabric.utilities.imports import _TORCH_GREATER_EQUAL_2_4
 from lightning.pytorch import Trainer, seed_everything
 from lightning.pytorch.callbacks import EarlyStopping, ModelCheckpoint
 from lightning.pytorch.demos.boring_classes import BoringModel
@@ -82,7 +84,7 @@ def test_resume_early_stopping_from_checkpoint(tmp_path):
 
     checkpoint_filepath = checkpoint_callback.kth_best_model_path
     # ensure state is persisted properly
-    checkpoint = torch.load(checkpoint_filepath)
+    checkpoint = torch.load(checkpoint_filepath, weights_only=True)
     # the checkpoint saves "epoch + 1"
     early_stop_callback_state = early_stop_callback.saved_states[checkpoint["epoch"]]
     assert len(early_stop_callback.saved_states) == 4
@@ -191,11 +193,13 @@ def test_pickling():
     early_stopping = EarlyStopping(monitor="foo")
 
     early_stopping_pickled = pickle.dumps(early_stopping)
-    early_stopping_loaded = pickle.loads(early_stopping_pickled)
+    with pytest.warns(FutureWarning, match="`weights_only=False`") if _TORCH_GREATER_EQUAL_2_4 else nullcontext():
+        early_stopping_loaded = pickle.loads(early_stopping_pickled)
     assert vars(early_stopping) == vars(early_stopping_loaded)
 
     early_stopping_pickled = cloudpickle.dumps(early_stopping)
-    early_stopping_loaded = cloudpickle.loads(early_stopping_pickled)
+    with pytest.warns(FutureWarning, match="`weights_only=False`") if _TORCH_GREATER_EQUAL_2_4 else nullcontext():
+        early_stopping_loaded = cloudpickle.loads(early_stopping_pickled)
     assert vars(early_stopping) == vars(early_stopping_loaded)
 
 

tests/tests_pytorch/callbacks/test_pruning.py

@@ -273,7 +273,7 @@ def test_multiple_pruning_callbacks(tmp_path, caplog, make_pruning_permanent: bo
     filepath = str(tmp_path / "foo.ckpt")
     trainer.save_checkpoint(filepath)
 
-    model.load_state_dict(torch.load(filepath), strict=False)
+    model.load_state_dict(torch.load(filepath, weights_only=True), strict=False)
     has_pruning = hasattr(model.layer.mlp_1, "weight_orig")
     assert not has_pruning if make_pruning_permanent else has_pruning
 

tests/tests_pytorch/checkpointing/test_legacy_checkpoints.py

@@ -78,7 +78,7 @@ def load_model():
         from lightning.pytorch.utilities.migration import pl_legacy_patch
 
         with pl_legacy_patch():
-            _ = torch.load(path_ckpt)
+            _ = torch.load(path_ckpt, weights_only=False)
 
     with patch("sys.path", [PATH_LEGACY] + sys.path):
         t1 = ThreadExceptionHandler(target=load_model)

tests/tests_pytorch/checkpointing/test_model_checkpoint.py

@@ -17,6 +17,7 @@
 import re
 import time
 from argparse import Namespace
+from contextlib import nullcontext
 from datetime import timedelta
 from inspect import signature
 from pathlib import Path
@@ -31,6 +32,7 @@
 import yaml
 from jsonargparse import ArgumentParser
 from lightning.fabric.utilities.cloud_io import _load as pl_load
+from lightning.fabric.utilities.imports import _TORCH_GREATER_EQUAL_2_4
 from lightning.pytorch import Trainer, seed_everything
 from lightning.pytorch.callbacks import ModelCheckpoint
 from lightning.pytorch.demos.boring_classes import BoringModel
@@ -350,11 +352,13 @@ def test_pickling(tmp_path):
     ckpt = ModelCheckpoint(dirpath=tmp_path)
 
     ckpt_pickled = pickle.dumps(ckpt)
-    ckpt_loaded = pickle.loads(ckpt_pickled)
+    with pytest.warns(FutureWarning, match="`weights_only=False`") if _TORCH_GREATER_EQUAL_2_4 else nullcontext():
+        ckpt_loaded = pickle.loads(ckpt_pickled)
     assert vars(ckpt) == vars(ckpt_loaded)
 
     ckpt_pickled = cloudpickle.dumps(ckpt)
-    ckpt_loaded = cloudpickle.loads(ckpt_pickled)
+    with pytest.warns(FutureWarning, match="`weights_only=False`") if _TORCH_GREATER_EQUAL_2_4 else nullcontext():
+        ckpt_loaded = cloudpickle.loads(ckpt_pickled)
     assert vars(ckpt) == vars(ckpt_loaded)
 
 
@@ -920,8 +924,8 @@ def test_model_checkpoint_save_last_checkpoint_contents(tmp_path):
     assert os.path.isfile(path_last_epoch)
     assert os.path.isfile(path_last)
 
-    ckpt_last_epoch = torch.load(path_last_epoch)
-    ckpt_last = torch.load(path_last)
+    ckpt_last_epoch = torch.load(path_last_epoch, weights_only=True)
+    ckpt_last = torch.load(path_last, weights_only=True)
 
     assert ckpt_last_epoch["epoch"] == ckpt_last["epoch"]
     assert ckpt_last_epoch["global_step"] == ckpt_last["global_step"]
@@ -1168,7 +1172,7 @@ def training_step(self, *args):
     )
     trainer.fit(TestModel())
     assert model_checkpoint.current_score == 0.3
-    ckpts = [torch.load(ckpt) for ckpt in tmp_path.iterdir()]
+    ckpts = [torch.load(ckpt, weights_only=True) for ckpt in tmp_path.iterdir()]
     ckpts = [
         ckpt["callbacks"][
             "ModelCheckpoint{'monitor': 'foo', 'mode': 'min', 'every_n_train_steps': 0, 'every_n_epochs': 1,"
@@ -1452,7 +1456,7 @@ def test_save_last_saves_correct_last_model_path(tmp_path):
     expected = "foo=1-last.ckpt"
     assert os.listdir(tmp_path) == [expected]
     full_path = tmp_path / expected
-    ckpt = torch.load(full_path)
+    ckpt = torch.load(full_path, weights_only=True)
     assert ckpt["callbacks"][mc.state_key]["last_model_path"] == str(full_path)
 
 
@@ -1484,7 +1488,7 @@ def test_none_monitor_saves_correct_best_model_path(tmp_path):
     expected = "epoch=0-step=0.ckpt"
     assert os.listdir(tmp_path) == [expected]
     full_path = str(tmp_path / expected)
-    ckpt = torch.load(full_path)
+    ckpt = torch.load(full_path, weights_only=True)
     assert ckpt["callbacks"][mc.state_key]["best_model_path"] == full_path
 
 

tests/tests_pytorch/checkpointing/test_torch_saving.py

@@ -30,7 +30,7 @@ def test_model_torch_save(tmp_path):
     # Ensure these do not fail
     torch.save(trainer.model, temp_path)
     torch.save(trainer, temp_path)
-    trainer = torch.load(temp_path)
+    trainer = torch.load(temp_path, weights_only=False)
 
 
 @RunIf(skip_windows=True)

tests/tests_pytorch/core/test_datamodules.py

@@ -208,7 +208,7 @@ def load_state_dict(self, state_dict: Dict[str, Any]) -> None:
     # fit model
     trainer.fit(model, datamodule=dm)
     checkpoint_path = list(trainer.checkpoint_callback.best_k_models.keys())[0]
-    checkpoint = torch.load(checkpoint_path)
+    checkpoint = torch.load(checkpoint_path, weights_only=True)
     assert dm.__class__.__qualname__ in checkpoint
     assert checkpoint[dm.__class__.__qualname__] == {"my": "state_dict"}
 

tests/tests_pytorch/core/test_metric_result_integration.py

@@ -19,6 +19,7 @@
 import lightning.pytorch as pl
 import pytest
 import torch
+from lightning.fabric.utilities.imports import _TORCH_GREATER_EQUAL_2_4
 from lightning.fabric.utilities.warnings import PossibleUserWarning
 from lightning.pytorch import Trainer
 from lightning.pytorch.callbacks import OnExceptionCheckpoint
@@ -253,11 +254,12 @@ def lightning_log(fx, *args, **kwargs):
         }
 
         # make sure can be pickled
-        pickle.loads(pickle.dumps(result))
+        with pytest.warns(FutureWarning, match="`weights_only=False`") if _TORCH_GREATER_EQUAL_2_4 else nullcontext():
+            pickle.loads(pickle.dumps(result))
         # make sure can be torch.loaded
         filepath = str(tmp_path / "result")
         torch.save(result, filepath)
-        torch.load(filepath)
+        torch.load(filepath, weights_only=False)
 
         # assert metric state reset to default values
         result.reset()

tests/tests_pytorch/core/test_saving.py

@@ -39,7 +39,7 @@ def test_load_from_checkpoint_map_location_automatic(accelerator, tmp_path, monk
     create_boring_checkpoint(tmp_path, BoringModel(), accelerator=accelerator)
 
     # The checkpoint contains tensors with storage tag on the accelerator
-    checkpoint = torch.load(f"{tmp_path}/checkpoint.ckpt")
+    checkpoint = torch.load(f"{tmp_path}/checkpoint.ckpt", weights_only=True)
     assert checkpoint["state_dict"]["layer.weight"].device.type.startswith(accelerator)
 
     # Pretend that the accelerator is not available
@@ -113,7 +113,7 @@ def test_load_from_checkpoint_warn_on_empty_state_dict(tmp_path):
     """Test that checkpoints can be loaded with an empty state dict and that the appropriate warning is raised."""
     create_boring_checkpoint(tmp_path, BoringModel(), accelerator="cpu")
     # Now edit so the state_dict is empty
-    checkpoint = torch.load(tmp_path / "checkpoint.ckpt")
+    checkpoint = torch.load(tmp_path / "checkpoint.ckpt", weights_only=True)
     checkpoint["state_dict"] = {}
     torch.save(checkpoint, tmp_path / "checkpoint.ckpt")
 

tests/tests_pytorch/helpers/datasets.py

@@ -106,7 +106,7 @@ def _try_load(path_data, trials: int = 30, delta: float = 1.0):
         assert os.path.isfile(path_data), f"missing file: {path_data}"
         for _ in range(trials):
             try:
-                res = torch.load(path_data)
+                res = torch.load(path_data, weights_only=True)
             # todo: specify the possible exception
             except Exception as ex:
                 exception = ex

tests/tests_pytorch/helpers/test_datasets.py

@@ -12,10 +12,12 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 import pickle
+from contextlib import nullcontext
 
 import cloudpickle
 import pytest
 import torch
+from lightning.fabric.utilities.imports import _TORCH_GREATER_EQUAL_2_4
 
 from tests_pytorch import _PATH_DATASETS
 from tests_pytorch.helpers.datasets import MNIST, AverageDataset, TrialMNIST
@@ -42,9 +44,9 @@ def test_pickling_dataset_mnist(dataset_cls, args):
     mnist = dataset_cls(**args)
 
     mnist_pickled = pickle.dumps(mnist)
-    pickle.loads(mnist_pickled)
-    # assert vars(mnist) == vars(mnist_loaded)
+    with pytest.warns(FutureWarning, match="`weights_only=False`") if _TORCH_GREATER_EQUAL_2_4 else nullcontext():
+        pickle.loads(mnist_pickled)
 
     mnist_pickled = cloudpickle.dumps(mnist)
-    cloudpickle.loads(mnist_pickled)
-    # assert vars(mnist) == vars(mnist_loaded)
+    with pytest.warns(FutureWarning, match="`weights_only=False`") if _TORCH_GREATER_EQUAL_2_4 else nullcontext():
+        cloudpickle.loads(mnist_pickled)

tests/tests_pytorch/loggers/test_all.py

@@ -14,11 +14,13 @@
 import inspect
 import os
 import pickle
+from contextlib import nullcontext
 from unittest import mock
 from unittest.mock import ANY, Mock
 
 import pytest
 import torch
+from lightning.fabric.utilities.imports import _TORCH_GREATER_EQUAL_2_4
 from lightning.pytorch import Callback, Trainer
 from lightning.pytorch.demos.boring_classes import BoringModel
 from lightning.pytorch.loggers import (
@@ -182,7 +184,8 @@ def _test_loggers_pickle(tmp_path, monkeypatch, logger_class):
     trainer = Trainer(max_epochs=1, logger=logger)
     pkl_bytes = pickle.dumps(trainer)
 
-    trainer2 = pickle.loads(pkl_bytes)
+    with pytest.warns(FutureWarning, match="`weights_only=False`") if _TORCH_GREATER_EQUAL_2_4 else nullcontext():
+        trainer2 = pickle.loads(pkl_bytes)
     trainer2.logger.log_metrics({"acc": 1.0})
 
     # make sure we restored properly

tests/tests_pytorch/loggers/test_logger.py

@@ -13,13 +13,15 @@
 # limitations under the License.
 import pickle
 from argparse import Namespace
+from contextlib import nullcontext
 from copy import deepcopy
 from typing import Any, Dict, Optional
 from unittest.mock import patch
 
 import numpy as np
 import pytest
 import torch
+from lightning.fabric.utilities.imports import _TORCH_GREATER_EQUAL_2_4
 from lightning.fabric.utilities.logger import _convert_params, _sanitize_params
 from lightning.pytorch import Trainer
 from lightning.pytorch.demos.boring_classes import BoringDataModule, BoringModel
@@ -122,7 +124,8 @@ def test_multiple_loggers_pickle(tmp_path):
 
     trainer = Trainer(logger=[logger1, logger2])
     pkl_bytes = pickle.dumps(trainer)
-    trainer2 = pickle.loads(pkl_bytes)
+    with pytest.warns(FutureWarning, match="`weights_only=False`") if _TORCH_GREATER_EQUAL_2_4 else nullcontext():
+        trainer2 = pickle.loads(pkl_bytes)
     for logger in trainer2.loggers:
         logger.log_metrics({"acc": 1.0}, 0)
 

tests/tests_pytorch/loggers/test_wandb.py

@@ -13,11 +13,13 @@
 # limitations under the License.
 import os
 import pickle
+from contextlib import nullcontext
 from pathlib import Path
 from unittest import mock
 
 import pytest
 import yaml
+from lightning.fabric.utilities.imports import _TORCH_GREATER_EQUAL_2_4
 from lightning.pytorch import Trainer
 from lightning.pytorch.callbacks import ModelCheckpoint
 from lightning.pytorch.cli import LightningCLI
@@ -160,7 +162,8 @@ def name(self):
     assert trainer.logger.experiment, "missing experiment"
     assert trainer.log_dir == logger.save_dir
     pkl_bytes = pickle.dumps(trainer)
-    trainer2 = pickle.loads(pkl_bytes)
+    with pytest.warns(FutureWarning, match="`weights_only=False`") if _TORCH_GREATER_EQUAL_2_4 else nullcontext():
+        trainer2 = pickle.loads(pkl_bytes)
 
     assert os.environ["WANDB_MODE"] == "dryrun"
     assert trainer2.logger.__class__.__name__ == WandbLogger.__name__