[Fix ] Fix Spark2.5 hybrid SWA config (#37335)

This commit is contained in:
KnightYao
2026-09-01 20:29:09 -07:00
committed by GitHub
parent 2d9c64394f
commit a58751d54d
+5 -1
View File
@@ -59,5 +59,9 @@ class Spark2_5Config(PretrainedConfig):
"partial_rotary_factor": 1.0,
},
}
self.hybrid_layer_pattern = [
1 if layer_type == "sliding_attention" else 0
for layer_type in self.layer_types
]
self.is_hybrid_swa = any(self.hybrid_layer_pattern)
super().__init__(**kwargs, tie_word_embeddings=tie_word_embeddings)