Remove smoke wording from tests and comments (#23355)

.claude/skills/sglang-auto-benchmark/SKILL.md

@@ -276,7 +276,7 @@ python3 -m sglang.auto_benchmark validate \
 
 - Tier 1
   - Fastest and smallest sweep.
-  - Best for smoke tests, config validation, and quickly checking whether a model can run at all.
+  - Best for quick checks, config validation, and confirming whether a model can run at all.
   - Uses a very small subset of the search space and mainly does one-at-a-time changes on top of the baseline.
   - Lowest search cost, but also the easiest to miss a better configuration.
 - Tier 2

.github/workflows/pr-test-sgl-kernel.yml

@@ -183,9 +183,9 @@ jobs:
           cd sgl-kernel
           pytest tests/
 
-  # Adding a single CUDA13 smoke test to verify that the kernel builds and runs
+  # Adding a single CUDA13 build-and-run check for the kernel
   # TODO: Add back this test when it can pass on CI
-  # cuda13-kernel-smoke-test:
+  # cuda13-kernel-build-check:
   #   if: inputs.sgl_kernel == 'true'
   #   runs-on: x64-cu13-kernel-tests
   #   steps:

benchmark/bench_rope/benchmark_rope_index.py

@@ -269,7 +269,7 @@ def benchmark_rope_index(
         seed=seed,
     )
 
-    # Smoke test
+    # Validate output shapes before benchmarking.
     has_mm = (image_grid_thw is not None) or (video_grid_thw is not None)
     if has_mm:
         pos, delta = MRotaryEmbedding.get_rope_index_glm4v(

docs_new/src/snippets/autoregressive/deepseek-v4-deployment.jsx

@@ -593,7 +593,7 @@ export const DeepSeekV4Deployment = () => {
         // headroom for DeepEP buffer + mooncake KV recv + CG private pool.
         // Cookbook defaults (mem-frac 0.874, cg_max_bs 512, max-running 256)
         // OOM during CG capture. mem-frac sweep at 0.83 / 0.87 / 0.89 / 0.91
-        // all pass static smoke; 0.9 picked as the default — leaves
+        // all pass static validation; 0.9 picked as the default — leaves
         // ~14 GB / GPU post-CG headroom for mooncake transfer + activation
         // peaks while giving ~1M-token KV pool.
         if (isGB300 && modelSize === "big") {

python/sglang/multimodal_gen/.claude/skills/sglang-diffusion-modelopt-quant/SKILL.md

@@ -69,7 +69,7 @@ Validated documentation and CI coverage currently center on six ModelOpt diffusi
 Treat a new family, a new precision, or a new checkpoint layout as unsupported until it has a documented matrix row and a matching validation story.
 Before writing CLI examples, re-read the active branch's `docs/diffusion/quantization.md`: FLUX.2 NVFP4 is an official `black-forest-labs/*` repo rather than a `BBuf/*` converted repo, and its preferred flag depends on the current documented loader flow. Use `--transformer-path` for a component override directory with `config.json`; use `--transformer-weights-path` when the repo or path should be probed as raw weights.
 
-B200 CI coverage can include loose BF16-vs-quantized quality smoke checks. Inspect the active branch's `run_suite.py` before assuming they are part of the suite; mainline and feature branches may differ. Those checks are intended to catch blank, corrupted, or obviously divergent images, not exact image parity.
+B200 CI coverage can include loose BF16-vs-quantized quality checks. Inspect the active branch's `run_suite.py` before assuming they are part of the suite; mainline and feature branches may differ. Those checks are intended to catch blank, corrupted, or obviously divergent images, not exact image parity.
 
 ## Documentation Maintenance
 

python/sglang/multimodal_gen/tools/compare_diffusion_trajectory_similarity.py

@@ -4,7 +4,7 @@ This tool runs two SGLang diffusion variants with the same prompt and seed,
 captures intermediate denoising latents via `return_trajectory_latents`, and
 reports cosine / error metrics for each timestep plus final frame metrics.
 
-The intended use is quant validation on reduced deterministic smoke settings:
+The intended use is quant validation with reduced deterministic settings:
 - same prompt / seed / resolution / step count for both variants
 - BF16 reference on the base model
 - FP8 candidate via `--candidate-transformer-path` and/or component overrides

python/sglang/test/attention/test_trtllm_mla_backend.py

@@ -91,7 +91,7 @@ TEST_CASES = {
             "batch_size": 1,
             "max_seq_len": 32,
             "page_size": 32,
-            "description": "Minimal smoke test",
+            "description": "Minimal sanity check",
         },
         {
             "name": "batch",
@@ -731,7 +731,7 @@ class TestTRTLLMMLA(CustomTestCase):
                 self.assertFalse(torch.isinf(output).any(), "Output contains Inf")
 
     def test_shape_sanity(self):
-        """Smoke test decode across several configurations."""
+        """Check decode shapes across several configurations."""
         print(f"\nRunning shape sanity tests...")
 
         for test_case in TEST_CASES["shape_sanity_tests"]:

test/manual/test_mori_transfer_engine_e2e.py

@@ -130,7 +130,7 @@ class TestMoriTransferEngineE2E(PDDisaggregationServerBase):
             other_args=decode_args,
         )
 
-    def test_generate_smoke(self):
+    def test_generate_basic(self):
         resp = requests.post(
             self.lb_url + "/generate",
             json={
@@ -273,7 +273,7 @@ class TestMoriTransferEngineTPMismatchE2E(PDDisaggregationServerBase):
             other_args=decode_args,
         )
 
-    def test_generate_smoke_tp_mismatch(self):
+    def test_generate_with_tp_mismatch(self):
         resp = requests.post(
             self.lb_url + "/generate",
             json={

test/registered/debug_utils/comparator/test_entrypoint.py

@@ -285,7 +285,7 @@ class TestEntrypointGroupingRaw:
         assert summary.total == 2
         assert summary.passed == 2
 
-    def test_text_output_smoke(self, tmp_path, capsys):
+    def test_text_output_format(self, tmp_path, capsys):
         """Text output format renders without errors and contains Config/Summary sections."""
         baseline_path, target_path = _create_dumps(tmp_path, ["tensor_a"])
         argv = _make_argv(

test/registered/debug_utils/test_cuda_coredump.py

@@ -1,7 +1,7 @@
-"""Smoke test: intentionally trigger a CUDA illegal memory access
+"""Intentionally trigger a CUDA illegal memory access
 to verify the coredump collection pipeline works end-to-end.
 
-Manual use:  python3 test/registered/debug_utils/test_cuda_coredump_smoke.py
+Manual use:  python3 test/registered/debug_utils/test_cuda_coredump.py
 """
 
 import unittest
@@ -17,7 +17,7 @@ register_cuda_ci(
 )
 
 
-class TestCudaCoredumpSmoke(unittest.TestCase):
+class TestCudaCoredump(unittest.TestCase):
     def test_trigger_illegal_memory_access(self):
         x = torch.zeros(10, device="cuda")
         y = torch.arange(10, device="cuda")

test/registered/hicache/test_hicache_storage_runtime_attach_detach.py

@@ -1,5 +1,5 @@
 """
-E2E smoke test for HiCache storage runtime attach/detach.
+E2E check for HiCache storage runtime attach/detach.
 
 This test launches an SGLang server with hierarchical cache enabled but WITHOUT
 any storage backend at startup, then attaches/detaches a storage backend via the

test/registered/moe/test_cutedsl_moe.py

@@ -581,7 +581,7 @@ class TestCuteDslV2(unittest.TestCase):
 
         Also checks both match the pure-PyTorch reference, and that a second
         cuda_graph pass reuses buffers deterministically (subsumes the former
-        cuda_graph_smoke test).
+        cuda_graph check).
         """
         test_cases = [
             # (num_tokens, hidden_size, intermediate_size, num_experts, top_k)