Unify memory management across (overlap, non-overlap) x (page>=1) x (spec, non-spec, spec v2) x (retract, finished) (#12224)

test/srt/hicache/test_hicache_storage_mooncake_backend.py

@@ -4,18 +4,14 @@ Usage:
     python3.10 -m pytest test/srt/hicache/test_hicache_storage_mooncake_backend.py -v
 """
 
-import json
 import os
 import subprocess
 import time
 import unittest
-from types import SimpleNamespace
 
 import requests
 from test_hicache_storage_file_backend import HiCacheStorageBaseMixin
 
-from sglang.test.few_shot_gsm8k import run_eval as run_eval_few_shot_gsm8k
-from sglang.test.test_disaggregation_utils import get_rdma_devices_args
 from sglang.test.test_utils import (
     DEFAULT_MLA_MODEL_NAME_FOR_TEST,
     CustomTestCase,

test/srt/test_eagle_infer_beta.py

@@ -68,6 +68,7 @@ class TestEagleServerBase(CustomTestCase, MatchedStopMixin):
 
     def test_radix_attention(self):
         run_radix_attention_test(self.base_url)
+        assert self.process.poll() is None
 
     def test_gsm8k(self):
         args = SimpleNamespace(
@@ -84,6 +85,11 @@ class TestEagleServerBase(CustomTestCase, MatchedStopMixin):
         self.assertGreater(
             metrics["accuracy"], 0.23
         )  # 0.3333 for 60 questions; 0.234 for 1319 questions
+        assert self.process.poll() is None
+
+
+class TestEagleServerPage(TestEagleServerBase):
+    other_launch_args = ["--page-size", "64"]
 
 
 if __name__ == "__main__":