MasterJH5574 commented on code in PR #18859:
URL: https://github.com/apache/tvm/pull/18859#discussion_r2868393437
##########
python/tvm/relax/frontend/nn/llm/position_embedding.py:
##########
@@ -814,8 +814,8 @@ def fused_rope_longrope_scaling( # pylint:
disable=too-many-locals
var_position_map, (seq_len,), "int32",
elem_offset=position_map_elem_offset
)
# long factors is the first half, short factors is the second half
- long_factors = T.Buffer((rotary_dim // 2,), "float32",
data=ext_factors.data)
- short_factors = T.Buffer(
+ long_factors = T.decl_buffer((rotary_dim // 2,), "float32",
data=ext_factors.data)
+ short_factors = T.decl_buffer(
(rotary_dim // 2,),
"float32",
data=ext_factors.data,
Review Comment:
Planning to deduplicate later on. Not in this PR.
--
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]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
