[Misc] Enhance attention selector (#4751)
This commit is contained in:
@@ -235,7 +235,6 @@ class EmbeddingModelRunner(ModelRunner):
|
||||
num_decode_tokens=num_decode_tokens,
|
||||
prefill_metadata=prefill_attn_metadata,
|
||||
decode_metadata=decode_attn_metadata,
|
||||
kv_cache_dtype=self.kv_cache_dtype,
|
||||
)
|
||||
|
||||
return (input_tokens, input_positions, attn_metadata, pooling_metadata,
|
||||
|
||||
Reference in New Issue
Block a user