Multiple refactors of DeepSeek V32 and context parallel (#16305)

2026-01-03 02:21:22 +08:00
parent 1cfd2b2ded
commit f07e76b229
7 changed files with 193 additions and 38 deletions
@@ -0,0 +1,172 @@
+"""
+Usage:
+cd test/src
+python3 -m unittest test_deepseek_v32_cp_single_node.TestDeepseekV32CP.test_a_gsm8k
+"""
+
+import unittest
+from types import SimpleNamespace
+
+import requests
+
+from sglang.srt.utils import kill_process_tree
+from sglang.test.few_shot_gsm8k import run_eval as run_eval_few_shot_gsm8k
+from sglang.test.test_utils import (
+    DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
+    DEFAULT_URL_FOR_TEST,
+    CustomTestCase,
+    is_in_ci,
+    popen_launch_server,
+    write_github_step_summary,
+)
+
+FULL_DEEPSEEK_V32_MODEL_PATH = "deepseek-ai/DeepSeek-V3.2-Exp"
+
+
+class TestDeepseekV32CP(CustomTestCase):
+    @classmethod
+    def setUpClass(cls):
+        cls.model = FULL_DEEPSEEK_V32_MODEL_PATH
+        cls.base_url = DEFAULT_URL_FOR_TEST
+        other_args = [
+            "--trust-remote-code",
+            "--tp",
+            "8",
+            "--dp",
+            "2",
+            "--enable-dp-attention",
+            "--speculative-algorithm",
+            "EAGLE",
+            "--speculative-num-steps",
+            "3",
+            "--speculative-eagle-topk",
+            "1",
+            "--speculative-num-draft-tokens",
+            "4",
+            "--mem-frac",
+            "0.7",
+            "--cuda-graph-max-bs",
+            "32",
+            "--max-running-requests",
+            "32",
+            "--enable-nsa-prefill-context-parallel",
+            "--nsa-prefill-cp-mode",
+            "in-seq-split",
+        ]
+        cls.process = popen_launch_server(
+            cls.model,
+            cls.base_url,
+            timeout=DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
+            other_args=other_args,
+        )
+
+    @classmethod
+    def tearDownClass(cls):
+        kill_process_tree(cls.process.pid)
+
+    def test_a_gsm8k(
+        self,
+    ):  # Append an "a" to make this test run first (alphabetically) to warm up the server
+        requests.get(self.base_url + "/flush_cache")
+
+        args = SimpleNamespace(
+            num_shots=8,
+            data_path=None,
+            num_questions=200,
+            parallel=32,
+            max_new_tokens=512,
+            host="http://127.0.0.1",
+            port=int(self.base_url.split(":")[-1]),
+        )
+        metrics = run_eval_few_shot_gsm8k(args)
+        print(f"{metrics=}")
+
+        server_info = requests.get(self.base_url + "/get_server_info")
+        avg_spec_accept_length = server_info.json()["internal_states"][0][
+            "avg_spec_accept_length"
+        ]
+        print(f"{avg_spec_accept_length=}")
+
+        if is_in_ci():
+            write_github_step_summary(
+                f"### test_gsm8k (deepseek-v32 nsa-cp)\n"
+                f'{metrics["accuracy"]=:.3f}\n'
+                f"{avg_spec_accept_length=:.2f}\n"
+            )
+            self.assertGreater(metrics["accuracy"], 0.935)
+            self.assertGreater(avg_spec_accept_length, 2.7)
+
+
+class TestDeepseekV32CPMode1(CustomTestCase):
+    @classmethod
+    def setUpClass(cls):
+        cls.model = FULL_DEEPSEEK_V32_MODEL_PATH
+        cls.base_url = DEFAULT_URL_FOR_TEST
+        other_args = [
+            "--trust-remote-code",
+            "--tp",
+            "8",
+            "--speculative-algorithm",
+            "EAGLE",
+            "--speculative-num-steps",
+            "3",
+            "--speculative-eagle-topk",
+            "1",
+            "--speculative-num-draft-tokens",
+            "4",
+            "--mem-frac",
+            "0.7",
+            "--cuda-graph-max-bs",
+            "32",
+            "--max-running-requests",
+            "32",
+            "--enable-nsa-prefill-context-parallel",
+            "--nsa-prefill-cp-mode",
+            "round-robin-split",
+        ]
+        cls.process = popen_launch_server(
+            cls.model,
+            cls.base_url,
+            timeout=DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
+            other_args=other_args,
+        )
+
+    @classmethod
+    def tearDownClass(cls):
+        kill_process_tree(cls.process.pid)
+
+    def test_a_gsm8k(
+        self,
+    ):  # Append an "a" to make this test run first (alphabetically) to warm up the server
+        requests.get(self.base_url + "/flush_cache")
+
+        args = SimpleNamespace(
+            num_shots=8,
+            data_path=None,
+            num_questions=200,
+            parallel=32,
+            max_new_tokens=512,
+            host="http://127.0.0.1",
+            port=int(self.base_url.split(":")[-1]),
+        )
+        metrics = run_eval_few_shot_gsm8k(args)
+        print(f"{metrics=}")
+
+        server_info = requests.get(self.base_url + "/get_server_info")
+        avg_spec_accept_length = server_info.json()["internal_states"][0][
+            "avg_spec_accept_length"
+        ]
+        print(f"{avg_spec_accept_length=}")
+
+        if is_in_ci():
+            write_github_step_summary(
+                f"### test_gsm8k (deepseek-v32 nsa-cp)\n"
+                f'{metrics["accuracy"]=:.3f}\n'
+                f"{avg_spec_accept_length=:.2f}\n"
+            )
+            self.assertGreater(metrics["accuracy"], 0.935)
+            self.assertGreater(avg_spec_accept_length, 2.7)
+
+
+if __name__ == "__main__":
+    unittest.main()