[GitHub] [tvm] masahi commented on a diff in pull request #14649: [Unity][BYOC] Fuse attention pattern with `strided_slice`

["masahi (via GitHub)" <gi...@apache.org>]

masahi commented on code in PR #14649:
URL: https://github.com/apache/tvm/pull/14649#discussion_r1169333092


##########
tests/python/relax/test_codegen_cutlass.py:
##########
@@ -676,10 +676,23 @@ def get_relax_stacked_attention_module(qkv, b, s, n, h, h_v, bias=None, qk_scale
             if bias is not None:
                 bias = R.arg("bias", R.Tensor(bias.shape, dtype))
             with R.dataflow() as frame:
-                qkv_tuple = R.split(qkv, [n * h, n * h * 2], axis=2)
-                q = R.reshape(qkv_tuple[0], [b, s, n, h])
-                k = R.reshape(qkv_tuple[1], [b, s, n, h])
-                v = R.reshape(qkv_tuple[2], [b, s, n, h_v])
+                if op == "split":
+                    qkv_tuple = R.split(qkv, [n * h, n * h * 2], axis=2)
+                    q = R.reshape(qkv_tuple[0], [b, s, n, h])
+                    k = R.reshape(qkv_tuple[1], [b, s, n, h])
+                    v = R.reshape(qkv_tuple[2], [b, s, n, h_v])
+                elif op == "strided_slice":
+                    qkv_tuple = R.split(qkv, [n * h, n * h * 2], axis=2)

Review Comment:
   `qkv_tuple` not used (I was confused)



-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: commits-unsubscribe@tvm.apache.org

For queries about this service, please contact Infrastructure at:
users@infra.apache.org