[Frontend] Add LLM.reward specific to reward models (#21720)

Signed-off-by: wang.yuqi <noooop@126.com>
This commit is contained in:
wang.yuqi
2025-07-30 11:56:03 +08:00
committed by GitHub
parent 1b0a155534
commit 65f311ce59
7 changed files with 174 additions and 35 deletions

View File

@@ -95,7 +95,7 @@ def test_prm_models(
monkeypatch.setenv("VLLM_USE_TRITON_FLASH_ATTN", "False")
with vllm_runner(model, max_model_len=1024, dtype=dtype) as vllm_model:
vllm_outputs = vllm_model.encode(math_step_prompts)
vllm_outputs = vllm_model.reward(math_step_prompts)
with hf_runner(model, dtype=dtype, auto_cls=AutoModel) as hf_model:
hf_model = step_reward_patch_hf_model(hf_model)