Use qr_2wave pipeline for fp8

example/ck_tile/01_fmha/codegen/ops/fmha_fwd.py

@@ -467,7 +467,10 @@ def get_fwd_blobs(kernel_filter : Optional[str], receipt, mask_impl) -> Tuple[Fm
         elif dtype in ['fp8', 'bf8']:
             # no need lse/dropout kernels
             for mask, bias in itertools.product(get_mask_map(mask_impl).keys(), BIAS_MAP.keys()):
-                pipelines.append(FmhaFwdPipeline('qr', 'col', 'f', 'f', 'f', 'f', bias, 'f', 'f', squant, mask))
+                if receipt == 3:
+                    pipelines.append(FmhaFwdPipeline('qr_2wave', 'col', 'f', 'f', 'f', 'f', bias, 'f', 'f', squant, mask))
+                else:
+                    pipelines.append(FmhaFwdPipeline('qr', 'col', 'f', 'f', 'f', 'f', bias, 'f', 'f', squant, mask))
         else:
             assert False
         return pipelines