Commit Graph
97 Commits
Author SHA1 Message Date
Liangsheng Yin d507accadc [Test] Drop dead and strictly-subsumed CI test registrations (#40264) 2026-09-18 17:49:13 -07:00
Yanbin Jiang ec30f19e4a [LoRA] Support MoE in full and breakable prefill CUDA graphs (#38578) 2026-09-11 14:40:02 -07:00
Yikai Zhang b3dc0388ed test(lora): enable ROCm logprob accuracy coverage (#38791) 2026-09-10 20:29:06 -07:00
Anupa Sajikumar 91a45ea37e [Intel][XPU][LoRA] Enable LoRA on Intel XPU (#30345) 2026-09-08 12:46:41 +08:00
Cheng Wan b99175dc7d [Config] Round 6.4: the runtime reads the bags, not the record (#38049)
Last of four; stacked on #38048.

The record is the operator's input; the bags are what is in effect. A reader
that takes the record and reads a field off it gets the input, which is the
wrong one of the two whenever resolution decided something -- and the mistake is
silent, because for most fields and most launches the two agree. Several of
these files already read both ways, sometimes in the same expression:

```python
get_tokenizer(
    get_serving().tokenizer_path,
    tokenizer_mode=server_args.tokenizer_mode,   # the input, not the decision
    ...
)
```

Sixty-odd files convert. Record field reads in runtime code go from 199 to 11.
Nine parameters that the conversion emptied are dropped along with the argument
at every call site -- the dead-parameter ratchet is what names them.

### "Runs after its process publishes" is a per-entry-point claim

Most converted reads sit in the serving and model-executor layers, which only
exist after publication, or in the two subprocess entry points, which publish
first thing. Three places are not like that, and they keep reading the record
they were handed:

- **`HttpServerEngineAdapter`** launches the server as a *child*. The parent
  resolves the record and never publishes, so the adapter's own reads -- the
  launch banner, the API key in its readiness loop, the TP width in
  `update_weights_from_tensor` -- are of `self.server_args`. A bag read here
  fails closed in a bare process, or answers for an unrelated engine in one that
  happens to have published.
- **`serve_grpc`** reads its sidecar port before the integrated servicer builds
  the `Engine` that publishes. The comment above that line already said so and
  already bound `cfg = resolving_view(server_args)` for it; the sidecar port and
  the port it derives from read `cfg`.
- **`initialize_dp_attention`** runs from callers whose publish is not
  guaranteed, so its one predicate stays on the resolution view.

`ROLE_NAMESPACE_SETS["dp_controller"]` gains `observability` and `serving`,
because the controller's metrics gate, tracing setup and worker-port broadcast
now read those namespaces. Under `SGLANG_ROLE_NAMESPACES=enforce` that set is
what the process may read, so a conversion that reaches a new namespace has to
widen it in the same change.

## Three things worth a reviewer's attention

**Eleven reads were `getattr(record, "field", default)`.** An AST scan for
attribute access does not see those, so the census that said "43 readers" was
counting the shape it could match rather than the thing it was after.
`incremental_streaming_output` was read that way twice, and the transcription
tests were the only reason it surfaced.

**Not every record read is a bag read waiting to happen.** A multimodal
processor's `base_gpu_id` is the instance's, not the process's: two engines in
one process keep different ones, and
`test_publishing_another_config_does_not_move_the_device` exists to say so. It
stays on the record while `rl_on_policy_target` beside it moves.
`RequestMetricsExporter` is the same shape -- it is handed the directory it
writes to, and a test builds several with different ones. `configure_logger` is
a third: 17 call sites, one of which passes an `argparse.Namespace`, so it is
not a global-context reader at all. Those eleven remaining reads are the ones
with a reason.

**The fixtures move with the code.** Tests that hung config off a mock manager
now publish a record, which is what the serving layer reads; where a test states
a value it says so with `override_server_args` instead of assigning through the
mock. `test_hisparse_unit` is the last of them: it stubbed a `server_args` onto
a fake scheduler to say the decode radix cache was off, and the value it was
standing in for is the published default, so the stub goes and the class
publishes.

## Two things CI caught that a local sweep could not

**`unittest.TestCase.enterContext` is Python 3.11+.** The converted fixtures used
it at 18 sites; `requires-python` is `>=3.10` and CI runs 3.10, so every one of
them raised `AttributeError` there while passing on a newer local interpreter.
They call `enter_override(self, ...)` now -- a four-line helper in
`sglang/test/test_utils.py` over the override's own `install()` / `restore()`.

**A batched sweep cannot see a missing publish.** Three fixtures needed a
published config and did not have one; each *passed* inside a shard where some
other file had published, and failed when run alone. The affected cases are
`test_serving_completions` (which set `incremental_streaming_output` on the mock
manager's record, where nothing reads it now), `test_qwen3_vl_feature_materialization`
(same shape for `mm_enable_dp_encoder`), and the two Qwen Rust tests -- whose
fixture already carried the comment `# Non-auto: get_resolved_model_impl would
choke on a SimpleNamespace` next to the `model_impl` it sets, which is exactly
what happened once `get_mm_processor_cls` started reading that value from the
bag. Its `publish` mirrors `model_impl` now, like the four fields it already
mirrored.

## Verification

A full registered-unit sweep (648 files) against this stack's merge-base:
19 failures on both sides, the same 19, none of them config. That sweep is what
caught 23 failures the file-scoped runs missed -- and, later, that the narrower
139-file list did not even contain the files this change reaches. It is also
what caught the `test_hisparse_unit` fixture above: the file passes inside a
shard where something else published, and fails when it is run on its own,
which is why every failing file is re-run alone before it is counted.
2026-09-06 21:41:46 -07:00
sglang-botandsglang-bot 6252993afe chore: update CI test est_time values (#38238)
Co-authored-by: sglang-bot <sglang-bot@users.noreply.github.com>
2026-09-06 17:49:41 -07:00
junduandMa Mingfei 8770c1db1f [CPU][CI]: rename Xeon CPU CI suites to stage-*-intel (#37395)
Co-authored-by: Ma Mingfei <mingfei.ma@intel.com>
2026-09-04 10:08:41 +08:00
Alex NailsandAlison Shao 28262c20df [CI][RFC] Replace black-jupyter with ruff-format (#37210)
Co-authored-by: Alison Shao <a.shao@wustl.edu>
2026-09-02 19:46:08 -07:00
Liangsheng YinandBaizhou Zhang 95f0f41021 [CI] Move tests onto the right CI stages (#34074)
Co-authored-by: Baizhou Zhang <sobereddiezhang@gmail.com>
2026-08-31 12:39:59 -07:00
514b997e6c Register CPU CI for 17 e2e tests and partition xeon base-c suite (#35227)
Co-authored-by: Zhang, Mingxu <mingxu.zhang@intel.com>
Co-authored-by: MingxuZh <109504044+MingxuZh@users.noreply.github.com>
Co-authored-by: Copilot Autofix powered by AI <175728472+Copilot@users.noreply.github.com>
Co-authored-by: Claude <noreply@anthropic.com>
2026-08-24 10:49:38 +08:00
Yanbin Jiang 7d893255c3 [Spec][LoRA] Support multi-adapter LoRA with EAGLE/NEXTN/DFLASH/DSPARK speculative decoding (#34337) 2026-08-21 14:21:53 -07:00
Yanbin Jiang 1ce515a53d Refocus LoRA tests on regression coverage (#34464) 2026-08-11 17:37:52 -07:00
Liangsheng Yin 7c90840bad [CI] Key scheduled CUDA suites by runner_config instead of hand-written jobs (#34186) 2026-08-09 16:44:53 -07:00
Liangsheng YinandBaizhou Zhang f6a6f5bf1e [CI] Trim redundant nightly test registrations (#34070)
Co-authored-by: Baizhou Zhang <sobereddiezhang@gmail.com>
2026-08-08 01:42:46 -07:00
Liangsheng Yin 7395ee833e [CI] Share VLM engines and prune launch matrices on the per-commit H100/H200 suites (#33944) 2026-08-07 01:22:22 -07:00
Liangsheng Yin 0753663b8e [CI] Trim redundant B200 test registrations (#33586) 2026-08-04 13:22:00 -07:00
Xiaoyu ZhangandClaude Opus 4.8 74338e94f1 [Kernel] Phase 4 batch-3: migrate tangled JIT subsystems + new groups into kernels.ops (RFC #29630) (#32045)
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
2026-07-22 21:15:03 +08:00
Liangsheng Yin ff6c755952 [Refactor] Unify logprob results into a single LogprobResult and rename chunk env vars (#31733) 2026-07-20 12:44:27 -07:00
02236fa38c Add Inkling model support (#31681)
Co-authored-by: Chunan Zeng <zcnrex@gmail.com>
Co-authored-by: Ke Bao <ispobaoke@gmail.com>
Co-authored-by: Yanbin Jiang <jybsuper@gmail.com>
Co-authored-by: Yuhao Yang <47235274+yhyang201@users.noreply.github.com>
Co-authored-by: Qiaolin Yu <qiaolin.yu@radixark.ai>
Co-authored-by: Zhichen Zeng <zczeng@uw.edu>
Co-authored-by: Aurick Qiao <aurick@thinkingmachines.ai>
Co-authored-by: Joseph <jk@thinkingmachines.ai>
2026-07-19 22:57:37 -07:00
Liangsheng Yin 2c856abbe3 [Fix] Enable chunked input-logprob processing by default to cap peak memory (#31498) 2026-07-17 13:19:03 -07:00
Liangsheng Yin c0f1f7e062 [Spec] Rename num_tokens_per_bs to num_tokens_per_req (#30977) 2026-07-13 13:47:53 -05:00
Xiaoyu ZhangandClaude Opus 4.8 6ed9843b57 [Kernel] Introduce sglang.kernels namespace and migrate scattered triton_ops kernels (RFC #29630, Phase 2) (#30044)
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
2026-07-10 21:41:08 +08:00
Michael 60f502a4fd [AMD] Register 2 hardware-agnostic 1-GPU PR tests for AMD CI (#30207) 2026-07-07 14:46:22 -07:00
Cheng Wan 88db9e033a Adjust KL_THRESHOLD for log probability calculations (#30101) 2026-07-03 23:53:51 -07:00
jundu e04ed05193 [CI] reduce CPU CI scope with base-c suite (#28623) 2026-06-24 16:18:06 +08:00
Michael 420004827c [AMD] register 3 tests to stage-b-test-1-gpu-large-amd (batch-6) (#28736) 2026-06-19 15:45:16 -07:00
Yanbin Jiang 093908d4c0 [LoRA] Fix chunked SGMV (csgmv) CUDA graph segment replay (#28371) 2026-06-16 19:19:07 -07:00
Shu Wang 32685874f3 Reenable MNNVL backend for FlashInfer allreduce fusion (#23402) 2026-06-15 20:19:15 -07:00
2495c02c2c [Refactor] Cuda Graph Runner/Backend Refactor (#23906)
Co-authored-by: BBuf <1182563586@qq.com>
Co-authored-by: Cheng Wan <54331508+ch-wan@users.noreply.github.com>
Co-authored-by: Lianmin Zheng <lianminzheng@gmail.com>
2026-06-09 21:36:57 -07:00
5c8a04ac4e [XPU CI] Expand stage-a and consolidate stage-b tests into stage-a (#27156)
Co-authored-by: vshekhawat-hlab <vshekhawat@habana.ai>
Co-authored-by: Ma Mingfei <mingfei.ma@intel.com>
2026-06-04 12:00:25 +08:00
Erik Wijmans 54b06f199c [lora] Share MoE LoRA Info (#24160) 2026-05-29 11:01:47 +09:00
MingxuZh 21ba329dac [Xeon] CPU CI enhancement for Intel Xeon platforms (#24649) 2026-05-28 10:49:04 +08:00
Erik Wijmans 87e69d57c4 [lora] Fix overlap loading for cancelled requests (#25413) 2026-05-25 15:18:36 +09:00
Liangsheng Yin b7d62bd724 [CI] Rename basic CI stage-a/b/c -> base-a/b/c for symmetry with extra CI (#25420) 2026-05-15 18:26:55 -07:00
ba214ef3d3 ci: tag-gated nightly migration — foundation + 40 whole-file moves (#24725)
Co-authored-by: hnyls2002 <lsyincs@gmail.com>
Co-authored-by: Liangsheng Yin <hnyls2002@gmail.com>
2026-05-14 16:11:15 -07:00
Alison Shao 36c9495aaa ci: H200 conditional split + dsv4 est_time recalibration (h200 partition 6→2) (#25236) 2026-05-13 22:28:37 -07:00
Alison Shao 1e308aec66 ci: B200 conditional split + LPT_SLOP removal (stage-c partition 8→3) (#25203) 2026-05-13 18:40:45 -07:00
Liangsheng Yin 22d3f3996c ci: decouple stage and runner for cuda registry (#25197) 2026-05-13 17:28:21 -07:00
Mick ff70aeac30 [diffusion] feat: add performance mode server args (#24491) 2026-05-14 00:57:46 +08:00
e9dea79755 (3/n - prefill optimize)[LoRA][MoE] Optimize virtual experts: remove CPU-GPU sync & multi-block CUDA JIT histogram (#24262)
Co-authored-by: Cursor <cursoragent@cursor.com>
Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-05-11 16:36:57 -07:00
Alison Shao 5fbec0e445 ci: prune per-commit CUDA tests — move 25 files + 13 testcases to test/manual/ (#24721) 2026-05-08 15:53:23 -07:00
Vladimir Serovandronnie_zheng 263cb3b222 [LoRA] Torch Native enhancement: embedding and graph optimization (#21885)
Co-authored-by: ronnie_zheng <zl19940307@163.com>
2026-05-07 17:28:38 +03:00
Liangsheng Yin eaf074d50e propagate pytest exit code from test __main__ entries (#24487) 2026-05-06 18:46:52 -07:00
Glen Liu 76b9c8de6f [Feature] add LoRADrainer to address high P99 TTFT (#17913) 2026-05-02 16:13:43 -07:00
Yanbin Jiang 8975479f87 [LoRA][MOE] Fix EP correctness in MoE LoRA slicing and virtual-experts kernels (#24171) 2026-04-30 22:42:10 -07:00
Alison ShaoandAlison Shao 694ef516cb Revert "[ci] split stage-c-test-4-gpu-b200 to enable a low-disk runner pool" (#24163)
Co-authored-by: Alison Shao <alisonshao@radixark.ai>
2026-04-30 15:57:19 -07:00
Opher LieberandYanbin Jiang c8c1c9261d LoRA support for qwen3.5 and nemotron3 (#23594)
Co-authored-by: Yanbin Jiang <jybsuper@gmail.com>
2026-04-29 21:51:53 -07:00
sglang-botandsglang-bot da175b964d chore: update CI test est_time values (#23785)
Co-authored-by: sglang-bot <sglang-bot@users.noreply.github.com>
2026-04-26 20:17:50 -07:00
Kangyan-ZhouandClaude Opus 4.7 77fd86f89e [ci] split stage-c-test-4-gpu-b200 to enable a low-disk runner pool (#23417)
Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-04-21 18:33:33 -07:00
Cheng WanandClaude Opus 4.7 5f7aee726a refactor(moe): de-duplicate triton MoE runner path into shared helpers (#23019)
Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-04-17 17:05:13 -07:00