[AMD] Add AMD CI registration (1-gpu unit test) to nightly CI. (#16941)

2026-01-13 23:41:45 -08:00
parent 030496eb06
commit b025cff441
39 changed files with 3701 additions and 2180 deletions
@@ -5,12 +5,15 @@ import unittest
 import torch

 from sglang.srt.batch_invariant_ops import batch_invariant_ops
-from sglang.test.ci.ci_register import register_cuda_ci
-
-register_cuda_ci(est_time=10, suite="nightly-1-gpu", nightly=True)
 from sglang.srt.batch_invariant_ops.batch_invariant_ops import set_batch_invariant_mode
+from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci
 from sglang.test.test_utils import CustomTestCase

+# Note: MI300 (gfx942) has 64KB shared memory limit but kernel needs 66KB
+# MI35x (gfx950/CDNA4) may have different limits - testing on MI35x only
+register_cuda_ci(est_time=10, suite="nightly-1-gpu", nightly=True)
+register_amd_ci(est_time=10, suite="nightly-amd-1-gpu-mi35x", nightly=True)
+
 device_type = getattr(torch.accelerator.current_accelerator(), "type", "cpu")
 torch.set_default_device(device_type)

@@ -13,6 +13,7 @@ from sglang.test.test_utils import (
    popen_launch_server,
 )

+# Note: AMD registration removed - test_cpp_radix_cache fails on AMD due to C++ radix tree issues
 register_cuda_ci(est_time=60, suite="nightly-1-gpu", nightly=True)