mirror of
https://github.com/vllm-project/vllm.git
synced 2026-06-06 00:16:14 +00:00
[Test][BugFix] Fix double-BOS in PD+specdec acceptance test (#44234)
Signed-off-by: Nick Hill <nickhill123@gmail.com>
This commit is contained in:
@@ -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]
|
||||
|
||||
Reference in New Issue
Block a user