saymrwulf/onnxruntime
1
0
Fork 0
mirror of https://github.com/saymrwulf/onnxruntime.git synced 2026-07-22 19:23:30 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

Bugfix: Add initializer to model in AttentionMask directly (#17719)

Browse source
This commit is contained in:
Jambay Kinley 2023-09-27 14:53:37 -07:00 committed by GitHub
parent 8621bdcd44
commit 1f4a3529dd
No known key found for this signature in database
GPG key ID: 4AEE18F83AFDEB23
1 changed files with 9 additions and 1 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

10
onnxruntime/python/tools/transformers/fusion_attention.py
View file

@ -78,7 +78,15 @@ class AttentionMask:
# ReduceSum-13: axes is moved from attribute to input
axes_name = "ort_const_1_reduce_sum_axes"
if self.model.get_initializer(axes_name) is None:
self.add_initializer(name=axes_name, data_type=TensorProto.INT64, dims=[1], vals=[1], raw=False)
self.model.add_initializer(
helper.make_tensor(
name=axes_name,
data_type=TensorProto.INT64,
dims=[1],
vals=[1],
raw=False,
)
)
mask_index_node = helper.make_node(
"ReduceSum",
inputs=[input_name, axes_name],