[Bugfix] Fix LLM priority normalization for single-string prompts (#40011)

Signed-off-by: daiyu1111 <2356690121@qq.com>
This commit is contained in:
daiyu1111
2026-04-16 07:56:06 -07:00
committed by GitHub
parent 4e8c3f1c19
commit a302a8fd1b
2 changed files with 7 additions and 1 deletions
+6
View File
@@ -91,6 +91,12 @@ def test_multiple_priority(llm: LLM):
outputs = llm.generate(PROMPTS, sampling_params=None, priority=[])
def test_single_prompt_priority(llm: LLM):
# Single string prompts should be normalized to one request.
outputs = llm.generate(PROMPTS[0], sampling_params=None, priority=[0])
assert len(outputs) == 1
def test_max_model_len():
max_model_len = 20
llm = LLM(
+1 -1
View File
@@ -1575,7 +1575,7 @@ class LLM:
seq_prompts = prompt_to_seq(prompts)
seq_params = self._params_to_seq(params, len(seq_prompts))
seq_lora_requests = self._lora_request_to_seq(lora_request, len(seq_prompts))
seq_priority = self._priority_to_seq(priority, len(prompts))
seq_priority = self._priority_to_seq(priority, len(seq_prompts))
return self._render_and_add_requests(
prompts=(