[Test][BugFix] Fix double-BOS in PD+specdec acceptance test (#44234)

Signed-off-by: Nick Hill <nickhill123@gmail.com>
This commit is contained in:
Nick Hill
2026-06-01 14:31:12 -07:00
committed by GitHub
parent 6f8b40a23f
commit e4cbc4385d
@@ -158,6 +158,10 @@ def test_spec_decode_acceptance_length():
max_tokens=DEFAULT_OUTPUT_LEN,
temperature=0.0,
top_p=1.0,
# Prompts are already chat-templated (contain BOS); avoid the
# completions API prepending a second BOS, which would lower
# acceptance ~5% vs the add_special_tokens=False standalone baselines.
extra_body={"add_special_tokens": False},
)
if i < 3:
text = resp.choices[0].text.strip()[:100]