TST: Skip AQLM test that is incompatible with torch 2.5 (huggingface#…

…2187)

See: Vahe1994/AQLM#139

It is unclear if/when AQLM will fix this and if there might not be other
issues with torch 2.5.

tests/test_gpu_examples.py

@@ -2731,6 +2731,8 @@ def _check_inference_finite(self, model, batch):
         model.train(training)
 
     @pytest.mark.single_gpu_tests
+    # see https://github.com/Vahe1994/AQLM/pull/139
+    @pytest.mark.xfail(reason="AQLM does not work with PyTorch 2.5 (yet)", strict=True, raises=AttributeError)
     def test_causal_lm_training_aqlm(self):
         r"""
         Test the CausalLM training on a single GPU device. The test would simply fail if the adapters are not set