[AMD] add testcases for Qwen3 235b Instruct 2507 models (#18805)

Co-authored-by: bingxche <Bingxu.Chen@amd.com>

.github/workflows/nightly-test-amd-rocm720.yml

@@ -32,9 +32,11 @@ on:
           - 'nightly-8-gpu-deepseek-v32-rocm720'
           - 'nightly-8-gpu-deepseek-v32-mtp-rocm720'
           - 'nightly-8-gpu-kimi-k25-rocm720'
+          - 'nightly-8-gpu-qwen3-235b-rocm720'
           - 'nightly-8-gpu-glm5-rocm720'
           # MI35x ROCm 7.2 jobs
           - 'nightly-test-1-gpu-mi35x-rocm720'
+          - 'nightly-8-gpu-mi35x-qwen3-235b-mxfp4-rocm720'
           - 'nightly-accuracy-8-gpu-mi35x-rocm720'
           - 'nightly-8-gpu-mi35x-grok1-int4-rocm720'
           - 'nightly-8-gpu-mi35x-grok2-rocm720'
@@ -496,6 +498,36 @@ jobs:
           echo "$(<github_summary.md )" >> $GITHUB_STEP_SUMMARY || true
           exit ${TEST_EXIT_CODE:-0}
 
+  # 8-GPU Qwen3-235B (Accuracy + Performance) ROCm 7.2
+  nightly-8-gpu-qwen3-235b-rocm720:
+    if: (github.repository == 'sgl-project/sglang' || github.event_name == 'pull_request') && (inputs.job_filter == '' || inputs.job_filter == 'all' || inputs.job_filter == 'nightly-8-gpu-qwen3-235b-rocm720')
+    runs-on: linux-mi325-gpu-8
+    steps:
+      - name: Checkout code
+        uses: actions/checkout@v4
+        with:
+          ref: ${{ inputs.ref || github.ref }}
+
+      - name: Setup docker (ROCm 7.2)
+        run: |
+          touch github_summary.md
+          bash scripts/ci/amd/amd_ci_start_container.sh --rocm-version rocm720
+        env:
+          GITHUB_WORKSPACE: ${{ github.workspace }}
+
+      - name: Install dependencies
+        run: bash scripts/ci/amd/amd_ci_install_dependency.sh --skip-aiter-build --skip-test-time-deps
+
+      - name: Accuracy Test + Performance Test ROCm 7.2 (8-GPU Qwen3)
+        timeout-minutes: 120
+        run: |
+          > github_summary.md  # Clear summary file
+          bash scripts/ci/amd/amd_ci_exec.sh -w /sglang-checkout/test \
+            -e GITHUB_STEP_SUMMARY="/sglang-checkout/github_summary.md" \
+            python3 run_suite.py --hw amd --suite nightly-8-gpu-qwen3-235b --nightly --timeout-per-file 3600 || TEST_EXIT_CODE=$?
+          echo "$(<github_summary.md )" >> $GITHUB_STEP_SUMMARY || true
+          exit ${TEST_EXIT_CODE:-0}
+
   # 8-GPU GLM-5 (Accuracy) ROCm 7.2
   nightly-8-gpu-glm5-rocm720:
     if: (github.repository == 'sgl-project/sglang' || github.event_name == 'pull_request') && (inputs.job_filter == '' || inputs.job_filter == 'all' || inputs.job_filter == 'nightly-8-gpu-glm5-rocm720')
@@ -860,6 +892,39 @@ jobs:
           echo "$(<github_summary.md )" >> $GITHUB_STEP_SUMMARY || true
           exit ${TEST_EXIT_CODE:-0}
 
+  # MI35x 8-GPU Qwen3-235B-MXFP4 (Accuracy + Performance) ROCm 7.2
+  nightly-8-gpu-mi35x-qwen3-235b-mxfp4-rocm720:
+    if: (github.repository == 'sgl-project/sglang' || github.event_name == 'pull_request') && (inputs.job_filter == '' || inputs.job_filter == 'all' || inputs.job_filter == 'nightly-8-gpu-mi35x-qwen3-235b-mxfp4-rocm720')
+    runs-on: linux-mi35x-gpu-8
+    steps:
+      - name: Checkout code
+        uses: actions/checkout@v4
+        with:
+          ref: ${{ inputs.ref || github.ref }}
+
+      - name: Setup docker (ROCm 7.2)
+        run: |
+          touch github_summary.md
+          bash scripts/ci/amd/amd_ci_start_container.sh --rocm-version rocm720
+        env:
+          GITHUB_WORKSPACE: ${{ github.workspace }}
+
+      - name: Install dependencies
+        run: |
+          bash scripts/ci/amd/amd_ci_install_dependency.sh --skip-aiter-build --skip-test-time-deps
+          # Install tabulate for run_suite.py (missing in MI35x container)
+          bash scripts/ci/amd/amd_ci_exec.sh pip install tabulate
+
+      - name: Accuracy Test + Performance Test MI35x ROCm 7.2 (8-GPU Qwen3-235B-MXFP4)
+        timeout-minutes: 120
+        run: |
+          > github_summary.md  # Clear summary file
+          bash scripts/ci/amd/amd_ci_exec.sh -w /sglang-checkout/test \
+            -e GITHUB_STEP_SUMMARY="/sglang-checkout/github_summary.md" \
+            python3 run_suite.py --hw amd --suite nightly-8-gpu-mi35x-qwen3-235b-mxfp4 --nightly --timeout-per-file 3600 || TEST_EXIT_CODE=$?
+          echo "$(<github_summary.md )" >> $GITHUB_STEP_SUMMARY || true
+          exit ${TEST_EXIT_CODE:-0}
+
   nightly-8-gpu-mi35x-glm5-rocm720:
     if: (github.repository == 'sgl-project/sglang' || github.event_name == 'pull_request') && (inputs.job_filter == '' || inputs.job_filter == 'all' || inputs.job_filter == 'nightly-8-gpu-mi35x-glm5-rocm720')
     runs-on: linux-mi35x-gpu-8
@@ -946,6 +1011,7 @@ jobs:
       - nightly-8-gpu-deepseek-v32-rocm720
       - nightly-8-gpu-deepseek-v32-mtp-rocm720
       - nightly-8-gpu-kimi-k25-rocm720
+      - nightly-8-gpu-qwen3-235b-rocm720
       - nightly-8-gpu-glm5-rocm720
       # MI35x ROCm 7.2 jobs
       - nightly-test-1-gpu-mi35x-rocm720
@@ -958,6 +1024,7 @@ jobs:
       - nightly-perf-8-gpu-mi35x-deepseek-v32-basic-rocm720
       - nightly-perf-8-gpu-mi35x-deepseek-v32-mtp-rocm720
       - nightly-8-gpu-mi35x-kimi-k25-rocm720
+      - nightly-8-gpu-mi35x-qwen3-235b-mxfp4-rocm720
       - nightly-8-gpu-mi35x-glm5-rocm720
     runs-on: ubuntu-latest
     steps:

.github/workflows/nightly-test-amd.yml

@@ -32,9 +32,11 @@ on:
           - 'nightly-8-gpu-deepseek-v32'
           - 'nightly-8-gpu-deepseek-v32-mtp'
           - 'nightly-8-gpu-kimi-k25'
+          - 'nightly-8-gpu-qwen3-235b'
           - 'nightly-8-gpu-glm5'
           # MI35x jobs
           - 'nightly-test-1-gpu-mi35x'
+          - 'nightly-8-gpu-mi35x-qwen3-235b-mxfp4'
           - 'nightly-8-gpu-mi35x-kimi-k25'
           - 'nightly-8-gpu-mi35x-glm5'
           - 'nightly-accuracy-8-gpu-mi35x'
@@ -496,6 +498,35 @@ jobs:
           echo "$(<github_summary.md )" >> $GITHUB_STEP_SUMMARY || true
           exit ${TEST_EXIT_CODE:-0}
 
+  nightly-8-gpu-qwen3-235b:
+    if: (github.repository == 'sgl-project/sglang' || github.event_name == 'pull_request') && (inputs.job_filter == '' || inputs.job_filter == 'all' || inputs.job_filter == 'nightly-8-gpu-qwen3-235b')
+    runs-on: linux-mi325-gpu-8
+    steps:
+      - name: Checkout code
+        uses: actions/checkout@v4
+        with:
+          ref: ${{ inputs.ref || github.ref }}
+
+      - name: Setup docker
+        run: |
+          touch github_summary.md
+          bash scripts/ci/amd/amd_ci_start_container.sh
+        env:
+          GITHUB_WORKSPACE: ${{ github.workspace }}
+
+      - name: Install dependencies
+        run: bash scripts/ci/amd/amd_ci_install_dependency.sh
+
+      - name: Accuracy Test + Performance Test (8-GPU Qwen3)
+        timeout-minutes: 120
+        run: |
+          > github_summary.md  # Clear summary file
+          bash scripts/ci/amd/amd_ci_exec.sh -w /sglang-checkout/test \
+            -e GITHUB_STEP_SUMMARY="/sglang-checkout/github_summary.md" \
+            python3 run_suite.py --hw amd --suite nightly-8-gpu-qwen3-235b --nightly --timeout-per-file 3600 || TEST_EXIT_CODE=$?
+          echo "$(<github_summary.md )" >> $GITHUB_STEP_SUMMARY || true
+          exit ${TEST_EXIT_CODE:-0}
+
   nightly-8-gpu-glm5:
     if: (github.repository == 'sgl-project/sglang' || github.event_name == 'pull_request') && (inputs.job_filter == '' || inputs.job_filter == 'all' || inputs.job_filter == 'nightly-8-gpu-glm5')
     runs-on: linux-mi325-gpu-8
@@ -861,6 +892,39 @@ jobs:
           echo "$(<github_summary.md )" >> $GITHUB_STEP_SUMMARY || true
           exit ${TEST_EXIT_CODE:-0}
 
+  # MI35x 8-GPU Qwen3-235B-MXFP4 (Accuracy + Performance)
+  nightly-8-gpu-mi35x-qwen3-235b-mxfp4:
+    if: (github.repository == 'sgl-project/sglang' || github.event_name == 'pull_request') && (inputs.job_filter == '' || inputs.job_filter == 'all' || inputs.job_filter == 'nightly-8-gpu-mi35x-qwen3-235b-mxfp4')
+    runs-on: linux-mi35x-gpu-8
+    steps:
+      - name: Checkout code
+        uses: actions/checkout@v4
+        with:
+          ref: ${{ inputs.ref || github.ref }}
+
+      - name: Setup docker
+        run: |
+          touch github_summary.md
+          bash scripts/ci/amd/amd_ci_start_container.sh
+        env:
+          GITHUB_WORKSPACE: ${{ github.workspace }}
+
+      - name: Install dependencies
+        run: |
+          bash scripts/ci/amd/amd_ci_install_dependency.sh
+          # Install tabulate for run_suite.py (missing in MI35x container)
+          bash scripts/ci/amd/amd_ci_exec.sh pip install tabulate
+
+      - name: Accuracy Test + Performance Test MI35x (8-GPU Qwen3-235B-MXFP4)
+        timeout-minutes: 120
+        run: |
+          > github_summary.md  # Clear summary file
+          bash scripts/ci/amd/amd_ci_exec.sh -w /sglang-checkout/test \
+            -e GITHUB_STEP_SUMMARY="/sglang-checkout/github_summary.md" \
+            python3 run_suite.py --hw amd --suite nightly-8-gpu-mi35x-qwen3-235b-mxfp4 --nightly --timeout-per-file 3600 || TEST_EXIT_CODE=$?
+          echo "$(<github_summary.md )" >> $GITHUB_STEP_SUMMARY || true
+          exit ${TEST_EXIT_CODE:-0}
+
   nightly-8-gpu-mi35x-glm5:
     if: (github.repository == 'sgl-project/sglang' || github.event_name == 'pull_request') && (inputs.job_filter == '' || inputs.job_filter == 'all' || inputs.job_filter == 'nightly-8-gpu-mi35x-glm5')
     runs-on: linux-mi35x-gpu-8
@@ -947,6 +1011,7 @@ jobs:
       - nightly-8-gpu-deepseek-v32
       - nightly-8-gpu-deepseek-v32-mtp
       - nightly-8-gpu-kimi-k25
+      - nightly-8-gpu-qwen3-235b
       - nightly-8-gpu-glm5
       # MI35x jobs
       - nightly-test-1-gpu-mi35x
@@ -957,6 +1022,7 @@ jobs:
       - nightly-accuracy-8-gpu-mi35x-deepseek-v32
       - nightly-accuracy-8-gpu-mi35x-deepseek-v32-mtp
       - nightly-8-gpu-mi35x-kimi-k25
+      - nightly-8-gpu-mi35x-qwen3-235b-mxfp4
       - nightly-8-gpu-mi35x-glm5
       # MI35x perf jobs excluded from check - perf failures don't block CI
       # - nightly-perf-8-gpu-mi35x-deepseek-v32-basic

test/registered/amd/test_qwen3_instruct.py

@@ -0,0 +1,92 @@
+import os
+import unittest
+from types import SimpleNamespace
+
+import requests
+
+from sglang.srt.utils import kill_process_tree
+from sglang.test.ci.ci_register import register_amd_ci
+from sglang.test.few_shot_gsm8k import run_eval as run_eval_few_shot_gsm8k
+from sglang.test.send_one import BenchArgs, send_one_prompt
+from sglang.test.test_utils import (
+    DEFAULT_URL_FOR_TEST,
+    CustomTestCase,
+    is_in_amd_ci,
+    is_in_ci,
+    popen_launch_server,
+    write_github_step_summary,
+)
+
+register_amd_ci(est_time=3600, suite="nightly-8-gpu-qwen3-235b", nightly=True)
+
+QWEN3_MODEL_PATH = "Qwen/Qwen3-235B-A22B-Instruct-2507"
+SERVER_LAUNCH_TIMEOUT = 3600
+
+
+class TestQwen3Instruct2507(CustomTestCase):
+    @classmethod
+    def setUpClass(cls):
+        cls.model = QWEN3_MODEL_PATH
+        cls.base_url = DEFAULT_URL_FOR_TEST
+        other_args = [
+            "--tp",
+            "8",
+            "--trust-remote-code",
+            "--attention-backend",
+            "aiter",
+        ]
+        env = os.environ.copy()
+        env["SGLANG_USE_AITER"] = "1"
+        cls.process = popen_launch_server(
+            cls.model,
+            cls.base_url,
+            timeout=SERVER_LAUNCH_TIMEOUT,
+            other_args=other_args,
+            env=env,
+        )
+
+    @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=1319,
+            parallel=1319,
+            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=}")
+
+        if is_in_ci():
+            write_github_step_summary(
+                f"### test_gsm8k ({self.model})\n" f'{metrics["accuracy"]=:.3f}\n'
+            )
+            self.assertGreater(metrics["accuracy"], 0.95)
+
+    def test_bs_1_speed(self):
+        args = BenchArgs(port=int(self.base_url.split(":")[-1]), max_new_tokens=2048)
+        _, speed = send_one_prompt(args)
+
+        print(f"{speed=:.2f}")
+
+        if is_in_ci():
+            write_github_step_summary(
+                f"### test_bs_1_speed ({self.model})\n" f"{speed=:.2f} token/s\n"
+            )
+            if is_in_amd_ci():
+                self.assertGreater(speed, 50)
+            else:
+                self.assertGreater(speed, 75)
+
+
+if __name__ == "__main__":
+    unittest.main()

test/registered/amd/test_qwen3_instruct_fp8.py

@@ -0,0 +1,92 @@
+import os
+import unittest
+from types import SimpleNamespace
+
+import requests
+
+from sglang.srt.utils import kill_process_tree
+from sglang.test.ci.ci_register import register_amd_ci
+from sglang.test.few_shot_gsm8k import run_eval as run_eval_few_shot_gsm8k
+from sglang.test.send_one import BenchArgs, send_one_prompt
+from sglang.test.test_utils import (
+    DEFAULT_URL_FOR_TEST,
+    CustomTestCase,
+    is_in_amd_ci,
+    is_in_ci,
+    popen_launch_server,
+    write_github_step_summary,
+)
+
+register_amd_ci(est_time=3600, suite="nightly-8-gpu-qwen3-235b", nightly=True)
+
+QWEN3_MODEL_PATH = "Qwen/Qwen3-235B-A22B-Instruct-2507-FP8"
+SERVER_LAUNCH_TIMEOUT = 3600
+
+
+class TestQwen3Instruct2507FP8(CustomTestCase):
+    @classmethod
+    def setUpClass(cls):
+        cls.model = QWEN3_MODEL_PATH
+        cls.base_url = DEFAULT_URL_FOR_TEST
+        other_args = [
+            "--tp",
+            "4",
+            "--trust-remote-code",
+            "--attention-backend",
+            "aiter",
+        ]
+        env = os.environ.copy()
+        env["SGLANG_USE_AITER"] = "1"
+        cls.process = popen_launch_server(
+            cls.model,
+            cls.base_url,
+            timeout=SERVER_LAUNCH_TIMEOUT,
+            other_args=other_args,
+            env=env,
+        )
+
+    @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=1319,
+            parallel=1319,
+            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=}")
+
+        if is_in_ci():
+            write_github_step_summary(
+                f"### test_gsm8k ({self.model})\n" f'{metrics["accuracy"]=:.3f}\n'
+            )
+            self.assertGreater(metrics["accuracy"], 0.95)
+
+    def test_bs_1_speed(self):
+        args = BenchArgs(port=int(self.base_url.split(":")[-1]), max_new_tokens=2048)
+        _, speed = send_one_prompt(args)
+
+        print(f"{speed=:.2f}")
+
+        if is_in_ci():
+            write_github_step_summary(
+                f"### test_bs_1_speed ({self.model})\n" f"{speed=:.2f} token/s\n"
+            )
+            if is_in_amd_ci():
+                self.assertGreater(speed, 40)
+            else:
+                self.assertGreater(speed, 60)
+
+
+if __name__ == "__main__":
+    unittest.main()

test/registered/amd/test_qwen3_instruct_mxfp4.py

@@ -0,0 +1,96 @@
+import os
+import unittest
+from types import SimpleNamespace
+
+import requests
+
+from sglang.srt.utils import kill_process_tree
+from sglang.test.ci.ci_register import register_amd_ci
+from sglang.test.few_shot_gsm8k import run_eval as run_eval_few_shot_gsm8k
+from sglang.test.send_one import BenchArgs, send_one_prompt
+from sglang.test.test_utils import (
+    DEFAULT_URL_FOR_TEST,
+    CustomTestCase,
+    is_in_amd_ci,
+    is_in_ci,
+    popen_launch_server,
+    write_github_step_summary,
+)
+
+register_amd_ci(
+    est_time=3600, suite="nightly-8-gpu-mi35x-qwen3-235b-mxfp4", nightly=True
+)
+
+QWEN3_MODEL_PATH = "amd/Qwen3-235B-A22B-Instruct-2507-mxfp4"
+SERVER_LAUNCH_TIMEOUT = 3600
+
+
+class TestQwen3Instruct2507MXFP4(CustomTestCase):
+    @classmethod
+    def setUpClass(cls):
+        cls.model = QWEN3_MODEL_PATH
+        cls.base_url = DEFAULT_URL_FOR_TEST
+        other_args = [
+            "--tp",
+            "4",
+            "--ep",
+            "2",
+            "--trust-remote-code",
+            "--attention-backend",
+            "aiter",
+        ]
+        env = os.environ.copy()
+        env["SGLANG_USE_AITER"] = "1"
+        cls.process = popen_launch_server(
+            cls.model,
+            cls.base_url,
+            timeout=SERVER_LAUNCH_TIMEOUT,
+            other_args=other_args,
+            env=env,
+        )
+
+    @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=1319,
+            parallel=1319,
+            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=}")
+
+        if is_in_ci():
+            write_github_step_summary(
+                f"### test_gsm8k ({self.model})\n" f'{metrics["accuracy"]=:.3f}\n'
+            )
+            self.assertGreater(metrics["accuracy"], 0.93)
+
+    def test_bs_1_speed(self):
+        args = BenchArgs(port=int(self.base_url.split(":")[-1]), max_new_tokens=2048)
+        _, speed = send_one_prompt(args)
+
+        print(f"{speed=:.2f}")
+
+        if is_in_ci():
+            write_github_step_summary(
+                f"### test_bs_1_speed ({self.model})\n" f"{speed=:.2f} token/s\n"
+            )
+            if is_in_amd_ci():
+                self.assertGreater(speed, 60)
+            else:
+                self.assertGreater(speed, 90)
+
+
+if __name__ == "__main__":
+    unittest.main()