[Qwen3-next] support mamba radix cache for overlap scheduler (#14792)

2025-12-14 18:54:16 -08:00
parent 36e7c8c59f
commit e61dabf5e4
30 changed files with 1414 additions and 204 deletions
@@ -146,7 +146,7 @@ suites = {
        TestFile("test_eagle_dp_attention.py", 200),
    ],
    "per-commit-4-gpu": [
-        TestFile("models/test_qwen3_next_models.py", 472),
+        TestFile("models/test_qwen3_next_models.py", 590),
        TestFile("test_gpt_oss_4gpu.py", 300),
        TestFile("test_local_attn.py", 411),
        TestFile("test_multi_instance_release_memory_occupation.py", 64),