Implement presence and frequency penalties (#95)

This commit is contained in:
Woosuk Kwon
2023-05-10 23:39:12 -07:00
committed by GitHub
parent 9f88db35da
commit 55f8b0a5de
9 changed files with 215 additions and 82 deletions

View File

@@ -96,7 +96,7 @@ class FastAPIServer:
seqs: List[Sequence] = []
for _ in range(sampling_params.n):
seq_id = next(self.seq_counter)
seq = Sequence(seq_id, token_ids, block_size=self.block_size)
seq = Sequence(seq_id, prompt, token_ids, block_size=self.block_size)
seqs.append(seq)
arrival_time = time.time()