sunggg commented on code in PR #14150:
URL: https://github.com/apache/tvm/pull/14150#discussion_r1120986730
##########
include/tvm/relax/attrs/nn.h:
##########
@@ -184,6 +184,15 @@ struct DropoutAttrs : public tvm::AttrsNode<DropoutAttrs> {
}
}; // struct DropoutAttrs
+/*! \brief Attributes used in fuse multi head attention operator */
+struct AttentionAttrs : public tvm::AttrsNode<AttentionAttrs> {
+ DataType out_dtype;
Review Comment:
I'm wondering why we need this as attribute. Can we use struct_info instead?
##########
python/tvm/contrib/cutlass/attention_operation.py:
##########
@@ -0,0 +1,95 @@
+# Licensed to the Apache Software Foundation (ASF) under one
+# or more contributor license agreements. See the NOTICE file
+# distributed with this work for additional information
+# regarding copyright ownership. The ASF licenses this file
+# to you under the Apache License, Version 2.0 (the
+# "License"); you may not use this file except in compliance
+# with the License. You may obtain a copy of the License at
+#
+# http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing,
+# software distributed under the License is distributed on an
+# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
+# KIND, either express or implied. See the License for the
+# specific language governing permissions and limitations
+# under the License.
+# pylint: disable=invalid-name, unused-wildcard-import, wildcard-import
+"""Generator for CUTLASS attention kernels."""
+from .library import *
+
+
+def instantiate_attention_template(attrs, func_args):
Review Comment:
Correct me if i'm wrong, but I heard there are various ways to implement
attention.
Is there any specific form (maybe something standard) of attention we
target?
Or can we also support a set of variants?
--
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: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
