ci: migrate RL tests to test/registered/rl/ (#16417)

2026-01-05 22:13:13 -08:00
parent 6ffe1fc02f
commit 861a35fb6e
10 changed files with 76 additions and 17 deletions
--- a/.github/workflows/pr-test-amd.yml
+++ b/.github/workflows/pr-test-amd.yml
@@ -224,6 +224,45 @@ jobs:
        run: |
          bash scripts/ci/amd_ci_exec.sh -w "/sglang-checkout/test" python3 run_suite.py --hw amd --suite stage-b-test-small-1-gpu --auto-partition-id ${{ matrix.part }} --auto-partition-size 4

+  stage-b-test-large-2-gpu-amd:
+    needs: [check-changes, stage-a-test-1-amd]
+    if: |
+      always() &&
+      (
+        (inputs.target_stage == 'stage-b-test-large-2-gpu-amd') ||
+        (
+          !inputs.target_stage &&
+          (!failure() && !cancelled()) &&
+          ((needs.check-changes.outputs.main_package == 'true') || (needs.check-changes.outputs.sgl_kernel == 'true'))
+        )
+      )
+    strategy:
+      fail-fast: false
+      matrix:
+        runner: [linux-mi325-gpu-2]
+    runs-on: ${{matrix.runner}}
+    steps:
+      - name: Checkout code
+        uses: actions/checkout@v4
+        with:
+          ref: ${{ inputs.pr_head_sha || inputs.ref || github.sha }}
+
+      - name: Ensure VRAM is clear
+        run: bash scripts/ensure_vram_clear.sh rocm
+
+      - name: Start CI container
+        run: bash scripts/ci/amd_ci_start_container.sh
+        env:
+          GITHUB_WORKSPACE: ${{ github.workspace }}
+
+      - name: Install dependencies
+        run: bash scripts/ci/amd_ci_install_dependency.sh
+
+      - name: Run test
+        timeout-minutes: 30
+        run: |
+          bash scripts/ci/amd_ci_exec.sh -w "/sglang-checkout/test" python3 run_suite.py --hw amd --suite stage-b-test-large-2-gpu-amd
+
  multimodal-gen-test-1-gpu-amd:
    needs: [check-changes]
    if: needs.check-changes.outputs.multimodal_gen == 'true'
@@ -884,6 +923,7 @@ jobs:

        stage-a-test-1-amd,
        stage-b-test-small-1-gpu-amd,
+        stage-b-test-large-2-gpu-amd,
        unit-test-backend-1-gpu-amd,
        unit-test-backend-2-gpu-amd,
        unit-test-backend-8-gpu-amd,
--- a/scripts/ci/slash_command_handler.py
+++ b/scripts/ci/slash_command_handler.py
@@ -178,6 +178,7 @@ def handle_rerun_stage(
        "sgl-kernel-unit-test-amd",
        "stage-a-test-1-amd",
        "stage-b-test-small-1-gpu-amd",
+        "stage-b-test-large-2-gpu-amd",
        "unit-test-backend-1-gpu-amd",
        "unit-test-backend-2-gpu-amd",
        "unit-test-backend-8-gpu-amd",
--- a/test/registered/rl/test_fp32_lm_head.py
+++ b/test/registered/rl/test_fp32_lm_head.py
@@ -1,3 +1,8 @@
+from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci
+
+register_cuda_ci(est_time=9, suite="stage-b-test-small-1-gpu")
+register_amd_ci(est_time=15, suite="stage-b-test-small-1-gpu")
+
 import unittest
 from types import SimpleNamespace
 from unittest.mock import patch
--- a/test/registered/rl/test_patch_torch.py
+++ b/test/registered/rl/test_patch_torch.py
@@ -1,3 +1,10 @@
+from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci
+
+register_cuda_ci(est_time=19, suite="stage-b-test-large-2-gpu")
+register_amd_ci(
+    est_time=19, suite="stage-b-test-large-2-gpu-amd", disabled="see #11127"
+)
+
 import os
 import traceback
 import unittest
--- a/test/registered/rl/test_return_routed_experts.py
+++ b/test/registered/rl/test_return_routed_experts.py
@@ -1,3 +1,7 @@
+from sglang.test.ci.ci_register import register_cuda_ci
+
+register_cuda_ci(est_time=180, suite="stage-c-test-large-4-gpu")
+
 import asyncio
 import logging
 import unittest
--- a/test/registered/rl/test_update_weights_from_disk.py
+++ b/test/registered/rl/test_update_weights_from_disk.py
@@ -1,3 +1,8 @@
+from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci
+
+register_cuda_ci(est_time=210, suite="stage-b-test-small-1-gpu", disabled="see #14021")
+register_amd_ci(est_time=210, suite="stage-b-test-small-1-gpu", disabled="see #14021")
+
 import json
 import random
 import time
--- a/test/registered/rl/test_update_weights_from_distributed.py
+++ b/test/registered/rl/test_update_weights_from_distributed.py
@@ -1,3 +1,8 @@
+from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci
+
+register_cuda_ci(est_time=103, suite="stage-b-test-large-2-gpu")
+register_amd_ci(est_time=103, suite="stage-b-test-large-2-gpu-amd")
+
 """Test distributed weight updates.

 This test suite simulates a distributed training environment to ensure
--- a/test/registered/rl/test_update_weights_from_tensor.py
+++ b/test/registered/rl/test_update_weights_from_tensor.py
@@ -1,3 +1,7 @@
+from sglang.test.ci.ci_register import register_cuda_ci
+
+register_cuda_ci(est_time=195, suite="stage-b-test-small-1-gpu")
+
 import gc
 import json
 import random
--- a/test/run_suite.py
+++ b/test/run_suite.py
@@ -18,7 +18,11 @@ HW_MAPPING = {
 # Per-commit test suites (run on every PR)
 PER_COMMIT_SUITES = {
    HWBackend.CPU: ["default", "stage-a-cpu-only"],
-    HWBackend.AMD: ["stage-a-test-1", "stage-b-test-small-1-gpu"],
+    HWBackend.AMD: [
+        "stage-a-test-1",
+        "stage-b-test-small-1-gpu",
+        "stage-b-test-large-2-gpu-amd",
+    ],
    HWBackend.CUDA: [
        "stage-a-test-1",
        "stage-b-test-small-1-gpu",
--- a/test/srt/run_suite.py
+++ b/test/srt/run_suite.py
@@ -28,10 +28,6 @@ suites = {
        TestFile("openai_server/validation/test_openai_server_ignore_eos.py", 6),
        TestFile("openai_server/validation/test_request_length_validation.py", 38),
        TestFile("ops/test_repeat_interleave.py", 60),
-        # quant tests moved to test/registered/quant/
-        TestFile("rl/test_fp32_lm_head.py", 9),
-        # TestFile("rl/test_update_weights_from_disk.py", 210),  # Temporarily disabled, see https://github.com/sgl-project/sglang/pull/13998
-        TestFile("rl/test_update_weights_from_tensor.py", 195),
        TestFile("dllm/test_llada2_mini.py", 520),
        TestFile("test_abort.py", 131),
        TestFile("test_chunked_prefill.py", 312),
@@ -84,12 +80,10 @@ suites = {
        TestFile("hicache/test_hicache_storage_mooncake_backend.py", 300),
        TestFile("models/test_kimi_linear_models.py", 90),
        TestFile("models/test_nvidia_nemotron_nano_v2.py", 132),
-        TestFile("rl/test_update_weights_from_distributed.py", 103),
        TestFile("test_data_parallelism.py", 73),
        TestFile("test_disaggregation_basic.py", 400),
        TestFile("test_dp_attention.py", 350),
        TestFile("test_load_weights_from_remote_instance.py", 72),
-        TestFile("test_patch_torch.py", 19),
    ],
    "per-commit-4-gpu": [
        TestFile("models/test_qwen3_next_models.py", 650),
@@ -97,7 +91,6 @@ suites = {
        TestFile("test_multi_instance_release_memory_occupation.py", 64),
        TestFile("test_pp_single_node.py", 500),
        TestFile("test_epd_disaggregation.py", 150),
-        TestFile("rl/test_return_routed_experts.py", 300),
    ],
    "per-commit-8-gpu-h200": [
        TestFile("test_deepseek_v3_basic.py", 275),
@@ -146,9 +139,6 @@ suites = {
    "__not_in_ci__": [
        TestFile("test_release_memory_occupation.py", 200),  # Temporarily disabled
        TestFile("models/test_dummy_grok_models.py"),
-        TestFile(
-            "rl/test_update_weights_from_disk.py"
-        ),  # Temporarily disabled, see https://github.com/sgl-project/sglang/pull/13998
        TestFile("test_bench_one_batch.py"),
        TestFile("test_bench_serving.py"),
        TestFile("test_eval_accuracy_large.py"),
@@ -191,9 +181,6 @@ suite_amd = {
        TestFile("openai_server/validation/test_openai_server_ignore_eos.py", 85),
        TestFile("openai_server/validation/test_request_length_validation.py", 31),
        TestFile("ops/test_repeat_interleave.py", 75),
-        # quant tests moved to test/registered/quant/
-        TestFile("rl/test_fp32_lm_head.py", 15),
-        # TestFile("rl/test_update_weights_from_disk.py", 210),  # Temporarily disabled, see https://github.com/sgl-project/sglang/pull/13998
        TestFile("rotary_embedding/test_mrope.py", 15),
        TestFile("test_abort.py", 51),
        TestFile("test_bench_typebaseddispatcher.py", 10),
@@ -242,11 +229,8 @@ suite_amd = {
        TestFile("test_gpt_oss_1gpu.py", 750),
    ],
    "per-commit-2-gpu-amd": [
-        # TestFile("lora/test_lora_tp.py", 116), # Disabled temporarily, see https://github.com/sgl-project/sglang/issues/13107. Moved to test/registered/lora/
-        TestFile("rl/test_update_weights_from_distributed.py", 103),
        TestFile("test_data_parallelism.py", 73),
        TestFile("test_load_weights_from_remote_instance.py", 72),
-        # TestFile("test_patch_torch.py", 19), # Disabled temporarily, see https://github.com/sgl-project/sglang/issues/11127
    ],
    "per-commit-4-gpu-amd": [
        TestFile("test_pp_single_node.py", 150),