From 3df6e2f9681a5f926d3b73286d19de9fa76b44a2 Mon Sep 17 00:00:00 2001
From: iridiumine <42236072+iridiumine@users.noreply.github.com>
Date: Mon, 15 Jun 2026 19:57:49 +0800
Subject: [PATCH] [NPU] Add MiMo-V2-Flash manual testcases (#28223)

---
 .../sglang/test/ascend/test_ascend_utils.py   |  1 +
 .../llm_models/test_npu_mimo_v2_flash.py      | 39 +++++++++++++++++++
 2 files changed, 40 insertions(+)
 create mode 100644 test/manual/ascend/llm_models/test_npu_mimo_v2_flash.py

diff --git a/python/sglang/test/ascend/test_ascend_utils.py b/python/sglang/test/ascend/test_ascend_utils.py
index c711d6dad..ac503f4d1 100644
--- a/python/sglang/test/ascend/test_ascend_utils.py
+++ b/python/sglang/test/ascend/test_ascend_utils.py
@@ -108,6 +108,7 @@ META_LLAMA_3_1_8B_INSTRUCT = os.path.join(
     MODEL_WEIGHTS_DIR, "LLM-Research/Meta-Llama-3.1-8B-Instruct"
 )
 MIMO_7B_RL_WEIGHTS_PATH = os.path.join(MODEL_WEIGHTS_DIR, "XiaomiMiMo/MiMo-7B-RL")
+MIMO_V2_FLASH_WEIGHTS_PATH = os.path.join(MODEL_WEIGHTS_DIR, "XiaomiMiMo/MiMo-V2-Flash")
 MINICPM3_4B_WEIGHTS_PATH = os.path.join(MODEL_WEIGHTS_DIR, "OpenBMB/MiniCPM3-4B")
 MISTRAL_7B_INSTRUCT_V0_2_WEIGHTS_PATH = os.path.join(
     MODEL_WEIGHTS_DIR, "mistralai/Mistral-7B-Instruct-v0.2"
diff --git a/test/manual/ascend/llm_models/test_npu_mimo_v2_flash.py b/test/manual/ascend/llm_models/test_npu_mimo_v2_flash.py
new file mode 100644
index 000000000..337e30fa1
--- /dev/null
+++ b/test/manual/ascend/llm_models/test_npu_mimo_v2_flash.py
@@ -0,0 +1,39 @@
+import unittest
+
+from sglang.test.ascend.gsm8k_ascend_mixin import GSM8KAscendMixin
+from sglang.test.ascend.test_ascend_utils import MIMO_V2_FLASH_WEIGHTS_PATH
+from sglang.test.test_utils import CustomTestCase
+
+
+class TestMiMoV2FlashGraphWithMTP(GSM8KAscendMixin, CustomTestCase):
+    """Testcase: Verify the inference accuracy of MiMo-V2-Flash on GSM8K with cuda graph and MTP (speculative decoding).
+
+    [Test Category] Model
+    [Test Target] XiaomiMiMo/MiMo-V2-Flash
+    [Test Config] Prefill+Decode, cuda graph enabled, EAGLE speculative decoding
+    """
+
+    model = MIMO_V2_FLASH_WEIGHTS_PATH
+    accuracy = 0.9
+    other_args = [
+        "--trust-remote-code",
+        "--mem-fraction-static",
+        "0.8",
+        "--attention-backend",
+        "ascend",
+        "--tp-size",
+        "16",
+        "--speculative-algorithm",
+        "EAGLE",
+        "--speculative-num-steps",
+        "3",
+        "--speculative-eagle-topk",
+        "1",
+        "--speculative-num-draft-tokens",
+        "4",
+        "--enable-multi-layer-eagle",
+    ]
+
+
+if __name__ == "__main__":
+    unittest.main()