Avoid dual semantics of extend_input_len by computing the candidate on the fly (#27616)

This commit is contained in:
fzyzcjy
2026-06-25 08:17:37 +08:00
committed by GitHub
parent 5d4e63d49e
commit 563c3418a7
3 changed files with 42 additions and 28 deletions
@@ -77,6 +77,8 @@ class TestPrefillAdder(CustomTestCase):
req.rid = str(rid)
req.priority = priority
req.extend_input_len = 0
req.prefix_indices = []
req.full_untruncated_fill_ids = []
req.extend_logprob_start_len = 0
req.output_ids = [0] * output_len
req.sampling_params = SimpleNamespace(max_new_tokens=max_new_tokens)