add readme

README.md

@@ -0,0 +1 @@
+This is a build of the [SageAttention](https://github.com/thu-ml/SageAttention) compatible with kernels library

torch-ext/sage_attention/core.py

@@ -116,7 +116,7 @@ def sageattn(
     - The tensors `q`, `k`, and `v` must have the dtype ``torch.float16`` or ``torch.bfloat16``
     - All tensors must be on the same cuda device.
     """
-
+    print("########tensor_layout", tensor_layout)
     arch = get_cuda_arch_versions()[q.device.index]
     if arch == "sm80":
         return sageattn_qk_int8_pv_fp16_cuda(