Disable register_da8w4_concat_linear_cpu_pass (#2623)

Summary: at

Differential Revision: D79130500

Author: henrylhtsang

test/quantization/test_da8w4_cpu.py

@@ -123,6 +123,7 @@ def test_8da4w_cpu(self, dtype, x_dim, bias, bs, sym_quant_a):
     @common_utils.parametrize("x_dim", [2, 3])
     @common_utils.parametrize("bias", [True, False])
     def test_8da4w_concat_linear_cpu(self, x_dim, bias):
+        self.skipTest("Disabled for now")
         N, K = 64, 128
 
         class Mod(torch.nn.Module):

torchao/dtypes/uintx/dyn_int8_act_int4_wei_cpu_layout.py

@@ -317,6 +317,6 @@ def _linear_int8_act_int4_weight_cpu_impl(input_tensor, weight_tensor, bias):
 
 
 # Register the concat linear fusion pass
-from ...prototype.inductor.fx_passes import register_da8w4_concat_linear_cpu_pass
+# from ...prototype.inductor.fx_passes import register_da8w4_concat_linear_cpu_pass
 
-register_da8w4_concat_linear_cpu_pass()
+# register_da8w4_concat_linear_cpu_pass()