[CI] Reorganize stage-b 1-GPU tests for 5090 compatibility (#16826)

test/registered/vlm/test_evs.py

@@ -8,7 +8,6 @@ from sglang.test.test_utils import run_doctests
 
 register_cuda_ci(est_time=20, suite="stage-b-test-small-1-gpu")
 register_amd_ci(est_time=20, suite="stage-b-test-small-1-gpu-amd")
-register_cuda_ci(est_time=7, suite="stage-b-test-small-1-gpu-5090")
 
 
 def test_resolve_evs_config():

test/registered/vlm/test_vision_chunked_prefill.py

@@ -1,6 +1,6 @@
 from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci
 
-register_cuda_ci(est_time=150, suite="stage-b-test-small-1-gpu")
+register_cuda_ci(est_time=150, suite="stage-b-test-large-1-gpu")
 register_amd_ci(est_time=270, suite="stage-b-test-small-1-gpu-amd")
 """
 Usage:

test/registered/vlm/test_vision_openai_server_a.py

@@ -1,7 +1,3 @@
-from sglang.test.ci.ci_register import register_cuda_ci
-
-register_cuda_ci(est_time=957, suite="stage-b-test-small-1-gpu")
-
 """
 Usage:
 python3 -m unittest test_vision_openai_server.TestOpenAIVisionServer.test_mixed_batch
@@ -12,6 +8,7 @@ import unittest
 
 import openai
 
+from sglang.test.ci.ci_register import register_cuda_ci
 from sglang.test.vlm_utils import *
 from sglang.test.vlm_utils import (
     AudioOpenAITestMixin,
@@ -22,6 +19,8 @@ from sglang.test.vlm_utils import (
     VideoOpenAITestMixin,
 )
 
+register_cuda_ci(est_time=957, suite="stage-b-test-large-1-gpu")
+
 
 class TestLlavaServer(ImageOpenAITestMixin):
     model = "lmms-lab/llava-onevision-qwen2-0.5b-ov"

test/registered/vlm/test_vlm_input_format.py

@@ -1,7 +1,3 @@
-from sglang.test.ci.ci_register import register_cuda_ci
-
-register_cuda_ci(est_time=447, suite="stage-b-test-small-1-gpu")
-
 import json
 import unittest
 from io import BytesIO
@@ -22,6 +18,8 @@ from transformers import (
     Qwen2_5_VLForConditionalGeneration,
 )
 
+from sglang.test.ci.ci_register import register_cuda_ci
+
 if not hasattr(_hf_activations, "PytorchGELUTanh"):
 
     class PytorchGELUTanh(torch.nn.Module):
@@ -38,6 +36,8 @@ from sglang import Engine
 from sglang.srt.entrypoints.openai.protocol import ChatCompletionRequest
 from sglang.srt.parser.conversation import generate_chat_conv
 
+register_cuda_ci(est_time=447, suite="stage-b-test-large-1-gpu")
+
 IMAGE_MAN_IRONING_URL = "https://raw.githubusercontent.com/sgl-project/sgl-test-files/refs/heads/main/images/man_ironing_on_back_of_suv.png"
 IMAGE_SGL_LOGO_URL = "https://raw.githubusercontent.com/sgl-project/sgl-test-files/refs/heads/main/images/sgl_logo.png"