zhangyiss/mlx-examples
1
0
Fork 0
mirror of https://github.com/ml-explore/mlx-examples.git synced 2025-09-01 04:14:38 +08:00
Code Issues Actions Packages Projects Releases Wiki Activity

Length masking for batch inputs (#1173)

Browse Source 
* length masking

* add mask to mlx_lm model interface

* remove lengths

* fix test:

* comment + fix
This commit is contained in:
Alex Barron
2024-12-18 19:43:52 -08:00
committed by GitHub
parent db109184b7
commit d4ef909d4a
34 changed files with 191 additions and 72 deletions
Download Patch File Download Diff File Expand all files Collapse all files

8
llms/mlx_lm/models/deepseek_v2.py
View File

@@ -370,9 +370,12 @@ class DeepseekV2Model(nn.Module):
self,
x: mx.array,
cache: Optional[Any] = None,
mask: Optional[mx.array] = None,
) -> mx.array:
h = self.embed_tokens(x)
mask = create_attention_mask(h, cache)
if mask is None:
mask = create_attention_mask(h, cache)
if cache is None:
cache = [None] * len(self.layers)
@@ -395,8 +398,9 @@ class Model(nn.Module):
self,
inputs: mx.array,
cache: Optional[Any] = None,
mask: Optional[mx.array] = None,
):
out = self.model(inputs, cache)
out = self.model(inputs, cache, mask)
return self.lm_head(out)
def sanitize(self, weights):