Fix mistral attention on Metal (#2699)

Co-authored-by: Luka Zakrajsek <luka.zakrajsek@soniox.com>
This commit is contained in:
Luka Zakrajšek
2025-01-04 16:11:20 +01:00
committed by GitHub
parent cbaa0ad46f
commit 57f41da13b

View File

@ -262,7 +262,8 @@ impl Attention {
.contiguous()?;
let value_states = value_states
.reshape((b_sz, q_len, self.num_kv_heads, self.head_dim))?
.transpose(1, 2)?;
.transpose(1, 2)?
.contiguous()?;
let (query_states, key_states) =
self.rotary_emb