[Speculative decoding 6/9] Integrate speculative decoding with LLMEngine (#3894)
This commit is contained in:
@@ -104,7 +104,6 @@ class BlockTable:
|
||||
token_ids (List[int]): The sequence of token IDs to be appended.
|
||||
"""
|
||||
assert self._is_allocated
|
||||
assert token_ids, "can't append empty token ids"
|
||||
|
||||
self.ensure_num_empty_slots(num_empty_slots=len(token_ids) +
|
||||
num_lookahead_slots)
|
||||
|
||||
Reference in New Issue
Block a user