Skip int4 QAT tests for nightly for now

int4 tinygemm quantization is currently broken in master and
being fixed in #517. Let's
skip these tests for now until that is fixed.

Author: andrewor14

test/quantization/test_qat.py

@@ -366,6 +366,8 @@ def _assert_close_4w(self, val, ref):
 
     @unittest.skipIf(not TORCH_VERSION_AFTER_2_4, "skipping when torch version is 2.4 or lower")
     @unittest.skipIf(not _CUDA_IS_AVAILABLE, "skipping when cuda is not available")
+    # TODO: remove once we fix int4 error: https://github.com/pytorch/ao/pull/517
+    @unittest.skipIf(TORCH_VERSION_AFTER_2_5, "int4 doesn't work for 2.5+ right now")
     def test_qat_4w_primitives(self):
         n_bit = 4
         group_size = 32
@@ -411,6 +413,8 @@ def test_qat_4w_primitives(self):
 
     @unittest.skipIf(not TORCH_VERSION_AFTER_2_4, "skipping when torch version is 2.4 or lower")
     @unittest.skipIf(not _CUDA_IS_AVAILABLE, "skipping when cuda is not available")
+    # TODO: remove once we fix int4 error: https://github.com/pytorch/ao/pull/517
+    @unittest.skipIf(TORCH_VERSION_AFTER_2_5, "int4 doesn't work for 2.5+ right now")
     def test_qat_4w_linear(self):
         from torchao.quantization.prototype.qat import Int4WeightOnlyQATLinear
         from torchao.quantization.GPTQ import WeightOnlyInt4Linear
@@ -439,6 +443,8 @@ def test_qat_4w_linear(self):
 
     @unittest.skipIf(not TORCH_VERSION_AFTER_2_4, "skipping when torch version is 2.4 or lower")
     @unittest.skipIf(not _CUDA_IS_AVAILABLE, "skipping when cuda is not available")
+    # TODO: remove once we fix int4 error: https://github.com/pytorch/ao/pull/517
+    @unittest.skipIf(TORCH_VERSION_AFTER_2_5, "int4 doesn't work for 2.5+ right now")
     def test_qat_4w_quantizer(self):
         from torchao.quantization.prototype.qat import Int4WeightOnlyQATQuantizer
         from torchao.quantization.GPTQ import Int4WeightOnlyQuantizer