Add lr_lambda type check in MultiplicativeLR (#151973)

Fixes #81554

## TestResult

### Before

```python
In [3]: import torch
   ...: class SimpleLinearModel(torch.nn.Module):
   ...:     def __init__(self):
   ...:         super(SimpleLinearModel, self).__init__()
   ...:         self.linear = torch.nn.Linear(10, 1)
   ...:
   ...:     def forward(self, x):
   ...:         return self.linear(x)
   ...:
   ...: net = SimpleLinearModel()
   ...: optimizer = torch.optim.Adam(net.parameters(), lr=0.01)
   ...: scheduler = torch.optim.lr_scheduler.MultiplicativeLR(optimizer, 0.95)
   ...: for i in range(10):
   ...:     print(i, scheduler.get_last_lr())
   ...:     scheduler.step()
TypeError: 'float' object is not callable

### After

```python
   ...: scheduler = torch.optim.lr_scheduler.MultiplicativeLR(optimizer, 0.95)
TypeError: lr_lambda should be a function, but got float
```

Pull Request resolved: https://github.com/pytorch/pytorch/pull/151973
Approved by: https://github.com/janeyx99

test/optim/test_lrscheduler.py

@@ -1837,6 +1837,15 @@ class TestLRScheduler(TestCase):
         )
         self._test(scheduler, targets, epochs)
 
+    def test_multiplicative_lr_with_lr_lambda(self):
+        lr_lambda = 0.95
+        with self.assertRaisesRegex(TypeError, "lr_lambda should be a function"):
+            MultiplicativeLR(self.opt, lr_lambda)
+
+        lr_lambda2 = 0.95
+        with self.assertRaisesRegex(TypeError, "lr_lambda should be a function"):
+            MultiplicativeLR(self.opt, [lr_lambda, lr_lambda2])
+
     @parametrize("T_mult", [1, 2, 4])
     def test_CosineAnnealingWarmRestarts_lr1(self, T_mult):
         iters = 100

torch/optim/lr_scheduler.py

@@ -398,6 +398,11 @@ class MultiplicativeLR(LRScheduler):
                     f"Expected {len(optimizer.param_groups)} lr_lambdas, but got {len(lr_lambda)}"
                 )
             self.lr_lambdas = list(lr_lambda)
+        for lr_lambda in self.lr_lambdas:
+            if not callable(lr_lambda):
+                raise TypeError(
+                    f"lr_lambda should be a function, but got {type(lr_lambda).__name__}"
+                )
         super().__init__(optimizer, last_epoch)
 
     @override