drop nll_loss shapes forbidden by PT 2.9 (#2508)

Author: t-vi

thunder/tests/opinfos.py

@@ -9529,7 +9529,6 @@ def nll_loss_sample_generator(op, device, dtype, requires_grad, **kwargs):
         if target.ndim != 0:
             # sprinkle ignore_index in the target, verify correctness, see issue 1744.
             target = torch.where(make(target_shape, low=0.0, high=1.0, requires_grad=False) > 0.3, target, ignore_index)
-
         yield SampleInput(
             a,
             target=target,
@@ -9539,13 +9538,16 @@ def nll_loss_sample_generator(op, device, dtype, requires_grad, **kwargs):
         )
 
     # Test empty input and target tensor short-circuit
-    for reduction_str, ignore_index in itertools.product(reduction_options, ignore_index_options):
-        yield SampleInput(
-            torch.tensor([], device=device, dtype=dtype),
-            torch.tensor([], device=device, dtype=torch.long),
-            ignore_index=ignore_index,
-            reduction=reduction_str,
-        )
+    # PyTorch disallows these inputs as of
+    # https://github.com/pytorch/pytorch/pull/161412
+    # so we drop this shape
+    # for reduction_str, ignore_index in itertools.product(reduction_options, ignore_index_options):
+    #     yield SampleInput(
+    #         torch.tensor([], device=device, dtype=dtype),
+    #         torch.tensor([], device=device, dtype=torch.long),
+    #         ignore_index=ignore_index,
+    #         reduction=reduction_str,
+    #     )
 
 
 def nll_loss_error_generator(op, device, dtype=torch.float32, **kwargs):