Fix test failures: remove cuda parameter from test GradScaler instantiations and revert to correct PyTorch version requirements

Co-authored-by: vfdev-5 <[email protected]>

Author: Copilot

examples/references/classification/imagenet/main.py

@@ -8,7 +8,7 @@
 try:
     from torch.amp import autocast, GradScaler
 except ImportError:
-    raise RuntimeError("Please, use recent PyTorch version, e.g. >=2.3.1")
+    raise RuntimeError("Please, use recent PyTorch version, e.g. >=1.12.0")
 
 import dataflow as data
 import utils

examples/references/segmentation/pascal_voc2012/main.py

@@ -8,7 +8,7 @@
 try:
     from torch.amp import autocast, GradScaler
 except ImportError:
-    raise RuntimeError("Please, use recent PyTorch version, e.g. >=2.3.1")
+    raise RuntimeError("Please, use recent PyTorch version, e.g. >=1.12.0")
 
 import dataflow as data
 import utils

ignite/engine/__init__.py

@@ -187,7 +187,7 @@ def supervised_training_step_amp(
     try:
         from torch.amp import autocast, GradScaler
     except ImportError:
-        raise ImportError("Please install torch>=2.3.1 to use amp_mode='amp'.")
+        raise ImportError("Please install torch>=1.12.0 to use amp_mode='amp'.")
 
     if gradient_accumulation_steps <= 0:
         raise ValueError(
@@ -412,7 +412,7 @@ def _check_arg(
             try:
                 from torch.amp import GradScaler
             except ImportError:
-                raise ImportError("Please install torch>=2.3.1 to use scaler argument.")
+                raise ImportError("Please install torch>=1.6.0 to use scaler argument.")
             scaler = GradScaler(enabled=True)
 
     if on_tpu:

tests/ignite/engine/test_create_supervised.py

@@ -170,7 +170,7 @@ def _():
 @pytest.mark.skipif(Version(torch.__version__) < Version("1.12.0"), reason="Skip if < 1.12.0")
 def test_create_supervised_training_scalar_assignment():
     with mock.patch("ignite.engine._check_arg") as check_arg_mock:
-        check_arg_mock.return_value = None, torch.amp.GradScaler('cuda', enabled=False)
+        check_arg_mock.return_value = None, torch.amp.GradScaler(enabled=False)
         trainer, _ = _default_create_supervised_trainer(model_device="cpu", trainer_device="cpu", scaler=True)
         assert hasattr(trainer.state, "scaler")
         assert isinstance(trainer.state.scaler, torch.amp.GradScaler)
@@ -462,7 +462,7 @@ def test_create_supervised_trainer_amp_error(mock_torch_cuda_amp_module):
 
 @pytest.mark.skipif(Version(torch.__version__) < Version("1.12.0"), reason="Skip if < 1.12.0")
 def test_create_supervised_trainer_scaler_not_amp():
-    scaler = torch.amp.GradScaler('cuda', enabled=torch.cuda.is_available())
+    scaler = torch.amp.GradScaler(enabled=torch.cuda.is_available())
 
     with pytest.raises(ValueError, match=f"scaler argument is {scaler}, but amp_mode is None."):
         _test_create_supervised_trainer(amp_mode=None, scaler=scaler)
@@ -540,7 +540,7 @@ def test_create_supervised_trainer_on_cuda_amp_scaler():
     _test_create_mocked_supervised_trainer(
         model_device=model_device, trainer_device=trainer_device, amp_mode="amp", scaler=True
     )
-    scaler = torch.amp.GradScaler('cuda', enabled=torch.cuda.is_available())
+    scaler = torch.amp.GradScaler(enabled=torch.cuda.is_available())
     _test_create_supervised_trainer(
         gradient_accumulation_steps=1,
         model_device=model_device,