Skip to content

Commit 26eacc1

Browse files
Naveen Sudafacebook-github-bot
Naveen Suda
authored andcommitted
fix head_dim in metadata
Summary: After #8846, the models exported correctly, but failed to run on-device with segmentation error. This diff fixes that error. Reviewed By: sxu, billmguo Differential Revision: D70538475
1 parent b2bee8c commit 26eacc1

File tree

1 file changed

+1
-1
lines changed

1 file changed

+1
-1
lines changed

examples/qualcomm/oss_scripts/llama/model/static_llama.py

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -461,7 +461,7 @@ def get_metadata(self):
461461
"get_bos_id": 1,
462462
"get_eos_id": 2,
463463
"get_dim": self.dim,
464-
"get_head_dim": self.dim // self.n_heads,
464+
"get_head_dim": self.head_dim,
465465
"get_max_batch_size": self.max_batch_size,
466466
"get_max_seq_len": self.max_seq_len,
467467
"get_n_bos": 1,

0 commit comments

Comments
 (0)