a8nova
/

OpenELM-270M-Instruct

Text Generation

Model card Files Files and versions Community

a8nova commited on May 19

Commit

f4b93d3

•

1 Parent(s): 6296b1d

Update modeling_openelm.py

Files changed (1) hide show

modeling_openelm.py +3 -1

modeling_openelm.py CHANGED Viewed

@@ -779,7 +779,9 @@ class OpenELMModel(OpenELMPreTrainedModel):
                 :, None, None, :
             ].eq(0.0)
             causal_mask = causal_mask.clone()
-            causal_mask[..., :mask_length] = causal_mask[..., :mask_length].masked_fill(...)
             #causal_mask[..., :mask_length] = causal_mask[..., :mask_length].masked_fill(
             #    padding_mask, min_dtype
             #)

                 :, None, None, :
             ].eq(0.0)
             causal_mask = causal_mask.clone()
+            causal_mask[..., :mask_length] = causal_mask[..., :mask_length].masked_fill(
+                padding_mask, min_dtype
+            )
             #causal_mask[..., :mask_length] = causal_mask[..., :mask_length].masked_fill(
             #    padding_mask, min_dtype
             #)