spec: STANDALONE skips hidden_states end-to-end (Optional schema + None-safe consumers) (#25037)

Co-authored-by: Qiaolin Yu <qy254@cornell.edu>
This commit is contained in:
Liangsheng Yin
2026-05-12 12:27:21 -07:00
committed by GitHub
co-authored by Qiaolin Yu
parent e86fb42736
commit 538832c8b7
12 changed files with 280 additions and 79 deletions
@@ -6,6 +6,7 @@ import requests
from sglang.srt.environ import envs
from sglang.srt.utils import kill_process_tree
from sglang.test.ci.ci_register import register_cuda_ci
from sglang.test.kits.radix_cache_server_kit import run_radix_attention_test
from sglang.test.run_eval import run_eval
from sglang.test.test_utils import (
DEFAULT_DRAFT_MODEL_STANDALONE,
@@ -205,6 +206,10 @@ class TestStandaloneV2SpeculativeDecodingTriton(
def get_server_args(cls):
return DEFAULT_SERVER_ARGS_V2 + ["--attention-backend", "triton"]
def test_radix_attention(self):
run_radix_attention_test(self.base_url)
assert self.process.poll() is None
class TestStandaloneV2SpeculativeDecodingFlashinfer(
TestStandaloneV2SpeculativeDecodingBase