diff options
| author | Kohaku-Blueleaf <59680068+KohakuBlueleaf@users.noreply.github.com> | 2023-08-14 13:34:51 +0800 |
|---|---|---|
| committer | Kohaku-Blueleaf <59680068+KohakuBlueleaf@users.noreply.github.com> | 2023-08-14 13:34:51 +0800 |
| commit | e7c03ccdcefb2a80129703931ef1f8455708945b (patch) | |
| tree | 6769bdb3a32bdf99cf25422206ff03457bdfa86c /modules/sub_quadratic_attention.py | |
| parent | d9cc27cb29926c9cc5dce331da8fbaf996cf4973 (diff) | |
| parent | 007ecfbb29771aa7cdcf0263ab1811bc75fa5446 (diff) | |
Merge branch 'dev' into extra-norm-module
Diffstat (limited to 'modules/sub_quadratic_attention.py')
| -rw-r--r-- | modules/sub_quadratic_attention.py | 4 |
1 files changed, 2 insertions, 2 deletions
diff --git a/modules/sub_quadratic_attention.py b/modules/sub_quadratic_attention.py index 497568eb..ae4ee4bb 100644 --- a/modules/sub_quadratic_attention.py +++ b/modules/sub_quadratic_attention.py @@ -58,7 +58,7 @@ def _summarize_chunk( scale: float, ) -> AttnChunk: attn_weights = torch.baddbmm( - torch.empty(1, 1, 1, device=query.device, dtype=query.dtype), + torch.zeros(1, 1, 1, device=query.device, dtype=query.dtype), query, key.transpose(1,2), alpha=scale, @@ -121,7 +121,7 @@ def _get_attention_scores_no_kv_chunking( scale: float, ) -> Tensor: attn_scores = torch.baddbmm( - torch.empty(1, 1, 1, device=query.device, dtype=query.dtype), + torch.zeros(1, 1, 1, device=query.device, dtype=query.dtype), query, key.transpose(1,2), alpha=scale, |