Commit Graph
100 Commits
Author SHA1 Message Date
Liangsheng Yin a91ae6af9e nextn subclass owns post_load_weights is_nextn (#24333) 2026-05-03 22:04:44 -07:00
Liangsheng Yin 1dd8f6d5ae dedup state_kv_args setup into helper (#24340) 2026-05-03 20:45:26 -07:00
Liangsheng Yin 91fa2340ed extract adjust_hybrid_swa_layers_for_pp (#24334) 2026-05-03 18:52:54 -07:00
Liangsheng Yin 00d620b77d introduce arg_groups/ with nemotron_h hook (#24328) 2026-05-03 16:28:11 -07:00
Liangsheng Yin c3b6d20a80 Register deepseek_v32 alias instead of rewriting config.json (#24295) 2026-05-03 16:02:17 -07:00
Liangsheng Yin 53df43d0a3 rerun-test: route deepep h200 suite to deepep runner (#24325) 2026-05-03 15:57:53 -07:00
Liangsheng YinandDarkSharpness fcc8b7b126 Rename SGLANG_USE_JIT_ALL_REDUCE to SGLANG_OPT_USE_CUSTOM_ALL_REDUCE_V2 (#24297)
Co-authored-by: DarkSharpness <2040703891@qq.com>
2026-05-02 23:59:46 -07:00
3259a2c789 Encode routed_experts in the detokenizer, off the tokenizer hot path (#24263)
Co-authored-by: fzyzcjy <ch271828n@outlook.com>
Co-authored-by: Yueming Yuan <yym022502@gmail.com>
2026-05-02 02:44:32 -07:00
Liangsheng Yin cb8fbd53fc Reserve slot 0 as padding in all req pools (#24243) 2026-05-01 16:41:36 -07:00
Liangsheng Yin 8a530468fd [Bug] Size mamba mappings from req pool, not mamba pool (#24244) 2026-05-01 15:45:20 -07:00
Liangsheng Yin c54ada994b fix: rename mimo spec threshold attr to num_accepted_drafts_thres (#24118) 2026-04-29 21:00:55 -07:00
Liangsheng Yin bd448e51bd [Spec] Split accept_length into num_accepted_drafts and num_accepted_tokens (#23962) 2026-04-29 00:02:22 -07:00
Liangsheng Yin cf0061da43 [Spec] Fix spec_accept_rate and unify accept/draft naming (#23530) 2026-04-28 14:40:04 -07:00
Liangsheng Yin 0f21fe924a fix ngram greedy verify kwarg (#23521) 2026-04-22 20:49:54 -07:00
Liangsheng YinandSoluMilken f611dd24f1 fix retrive -> retrieve typo (#23503)
Co-authored-by: SoluMilken <19161836+solumilken@users.noreply.github.com>
2026-04-22 16:35:04 -07:00
Liangsheng Yin 6c2714f5ae guard adaptive speculative against unsupported configs (#23289) 2026-04-21 13:47:34 -07:00
Liangsheng Yin 2b2cad70d6 [Refactor] Move radix-cache utils onto RadixKey as methods (#23209) 2026-04-20 23:11:58 -07:00
a490632416 Opt-in strip of thinking tokens from radix cache (#23315)
Co-authored-by: ianliuy <ianl@alumni.usc.edu>
Co-authored-by: Wen-xuan-Xu <lilmeep727@gmail.com>
2026-04-20 22:59:50 -07:00
Liangsheng Yin 6cc2eee50d [misc] CI hygiene: enforce __main__ entry, drop silent-skipped tests, fix rerun-test protoc (#23305) 2026-04-20 21:16:24 -07:00
Liangsheng Yin c7a4ebf3c8 [Refactor] Replace page_align_keys helper with RadixKey.page_aligned method (#23107) 2026-04-20 18:10:42 -07:00
Liangsheng Yin 8cb957ccff [Perf] Make EAGLE bigram key an O(1) view on RadixKey (#23106) 2026-04-20 12:01:11 -07:00
Liangsheng Yin a2d30d27fe wait for reap in kill_process_tree (#23213) 2026-04-19 23:36:33 -07:00
Liangsheng Yin eb76aaba88 [core] Always-on StreamingSession in UnifiedRadixCache (#23202) 2026-04-19 21:19:43 -07:00
Liangsheng Yin a7276b623e integrate streaming session into UnifiedRadixCache (#23145) 2026-04-19 20:47:41 -07:00
Liangsheng Yin d3ce664612 move session to python/sglang/srt/session (#23144) 2026-04-19 17:34:19 -07:00
Liangsheng Yin 09b689b407 Apply HF transformers patches from sglang init (#23103) 2026-04-17 15:37:51 -07:00
Liangsheng Yin 573e12a7fc Merge /get_load into /v1/loads (#23010) 2026-04-17 13:36:51 -07:00
Liangsheng Yin 3df35ecc80 Lower TestPiecewiseCudaGraphQwen25VL gsm8k threshold to 0.80 (#23099) 2026-04-17 13:31:10 -07:00
Liangsheng Yin db7a751d48 refactor: extract FanOutCommunicator and use declarative spec table (#22967) 2026-04-16 15:37:19 -07:00
Liangsheng Yin c83ef4fdb6 use envs in server_args (#22994) 2026-04-16 15:01:33 -07:00
Liangsheng Yin bbd8f9ba09 migrate CPU-only unit tests from openai_server to unit/ (#22965) 2026-04-16 03:53:33 -07:00
Liangsheng Yin 62309f09db fix(loads): preserve include filtering after watching mode switch (#22959) 2026-04-16 03:04:53 -07:00
Liangsheng Yin a4cf2ea128 streaming session: spec v2 bonus accounting + comprehensive test matrix (#22651) 2026-04-15 17:12:41 -07:00
Liangsheng Yin f9792166c3 trim_overshoot: cap swa_evicted_seqlen + unit test (#22900) 2026-04-15 15:05:35 -07:00
Liangsheng Yin efc267ca29 streaming session: trim spec v2 overshoot in cache_finished_req (#22897) 2026-04-15 14:15:46 -07:00
Liangsheng Yin ce31934ca8 Streaming session: fix retract tail leak via _free_tail (#22862) 2026-04-15 01:44:27 -07:00
Liangsheng Yin aa78564e1a Refactor streaming session abort handling (#22790) 2026-04-15 00:13:05 -07:00
Liangsheng Yin 36891ab514 Rename _alive_streaming_session_count; use _is_streaming helper (#22755) 2026-04-14 13:26:03 -07:00
Liangsheng Yin 0cb7295698 Fix streaming session busy-check double-counting via active_pool_idxs (#22753) 2026-04-14 13:11:06 -07:00
Liangsheng Yin 33a3ba256f Delete dead rematch path in SessionAwareCache.release_session (#22735) 2026-04-13 17:02:40 -07:00
Liangsheng Yin 9e6d1c066e TestStreamingSessionAbortLeakRepro: inherit stdout/stderr instead of tempfile (#22668) 2026-04-13 00:37:29 -07:00
Liangsheng Yin c775c31cb3 Simplify test_chunked_prefill; remove redundant tests (#22652) 2026-04-12 20:26:54 -07:00
Liangsheng Yin da6b8e1448 Extract pause_resume_in_place kit; rename test_abort to test_scheduler_control (#22647) 2026-04-12 18:49:37 -07:00
Liangsheng Yin f1eb4ca90c Fix streaming session busy check double-counting; add compat CI tests (#22213) 2026-04-12 01:48:16 -07:00
Liangsheng Yin f2377a00cb Add SWA support for runtime busy memory check (#21499) 2026-04-12 00:39:51 -07:00
Liangsheng Yin 61a62c6503 [mem] Flatten memory checkers into composable per-pool invariant checks (#22562) 2026-04-11 02:56:22 -07:00
Liangsheng Yin 7ab94e438c [Test] Fix flaky test_function_call_required by adding strict=True (#22586) 2026-04-11 02:14:22 -07:00
Liangsheng Yin d11da2403c Add hisparse staging + decode offload guards to is_fully_idle() (#22577) 2026-04-11 00:12:10 -07:00
Liangsheng Yin c7f93a2ce7 [metrics] Add PoolStats.update_scheduler_stats to deduplicate metrics assignment (#22559) 2026-04-10 21:04:18 -07:00
Liangsheng Yin c2821dfbe9 [mem] Introduce PoolStats dataclass; unify pool metrics and token_usage (#22554) 2026-04-10 20:35:50 -07:00
Liangsheng Yin 6cd183ff6b Remove redundant test_page_size.py (#22571) 2026-04-10 20:35:04 -07:00
Liangsheng Yin b4a1d8fd71 [mem] Fix idle token_usage missing mamba_usage; add FIXME for naming (#22555) 2026-04-10 16:20:33 -07:00
Liangsheng Yin 6cf7f210bf Add page_size to admission token budget check (#22495) 2026-04-10 01:16:04 -07:00
Liangsheng Yin 722e25a621 Fix SWA eviction boundary and page-align chunked prefill (#22470) 2026-04-09 22:09:43 -07:00
Liangsheng Yin 9fed58805f [Doc] Clarify SWA HybridSWAPoolConfigurator comments on all-SWA vs hybrid semantics (#22443) 2026-04-09 03:02:16 -07:00
Liangsheng YinandKe Bao 8ff01d6841 [Test] Add CPU unit tests for MemoryPoolConfigurator (#22420)
Co-authored-by: Ke Bao <ispobaoke@gmail.com>
2026-04-09 00:39:19 -07:00
Liangsheng Yin de441ac6bb [core] Introduce MemoryPoolConfigurator class hierarchy (#22389) 2026-04-09 15:29:19 +08:00
Liangsheng Yin edfddda192 Move runai model loader test to nightly suite (#22418) 2026-04-08 21:39:32 -07:00
Liangsheng Yin 2c4e113dd7 [CI] Fast-fail on lint check failure in check-stage-health (#22400) 2026-04-08 17:17:07 -07:00
Liangsheng Yin 1e3f6ebea6 [core] Extract pool sizing logic to pool_configurator.py (#22384) 2026-04-08 16:13:21 -07:00
1c5c6dad5e [tiny] Fix TOCTOU race in pause-aware weight update locking (#22304)
Co-authored-by: maocheng23 <maocheng@berkeley.edu>
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
2026-04-07 18:54:28 -07:00
Liangsheng Yin 8c3d80eabe Only upload CUDA coredumps on test failure (#22301) 2026-04-07 18:07:28 -07:00
Liangsheng Yin 0e2a0260a1 Add fast-fail to multimodal-gen CI (#22284) 2026-04-07 15:56:12 -07:00
Liangsheng Yin cc35714b03 [tiny] migrate /get_server_info; print accept length in accuracy tests (#22282) 2026-04-07 13:08:35 -07:00
Liangsheng Yin be0277f9a0 [Spec][Ngram] Add output-as-corpus accept length benchmark for external SAM (#22199) 2026-04-06 19:09:52 -07:00
Liangsheng Yin e4b1366a46 [Spec][Ngram] Support multiple SAMs with dynamic HTTP API (#22203) 2026-04-06 18:49:22 -07:00
Liangsheng Yin 49cb7d546e Move hash utils out of hicache_storage to break CUDA import chain (#22214) 2026-04-06 18:16:40 -07:00
Liangsheng Yin 6de2ff2a80 [Spec][Ngram] Followup fixes for MatchState incremental advance (#22180) 2026-04-05 23:04:28 -07:00
Liangsheng YinandShangming Cai 3a4f4cbc52 DEBUG: reproduce flaky test_load_weights_from_remote_instance (#22150)
Co-authored-by: Shangming Cai <csmthu@gmail.com>
2026-04-05 22:46:36 +08:00
Liangsheng Yin df9c831ab8 Unify think_end_id to model_config as single source of truth (#22148) 2026-04-05 03:35:38 -07:00
Liangsheng Yin aeff9fb7c1 Add dump_metric to MMMU, lm-eval, and NeMo Skills eval paths (#22147) 2026-04-05 03:23:52 -07:00
Liangsheng Yin cd2d45e220 Isolate spec V1 path in decode post-processing (#22146) 2026-04-05 03:16:56 -07:00
Liangsheng Yin bd6a585605 Consolidate reasoning tests into test/registered/reasoning/ (#22139) 2026-04-05 01:09:11 -07:00
Liangsheng Yin 904bb476d8 Migrate reasoning_tokens tests to existing server fixtures (#22102) 2026-04-05 00:30:57 -07:00
Liangsheng Yin bb9e058f5b Add failfast flag to rerun-test workflow (#22141) 2026-04-05 00:24:50 -07:00
Liangsheng Yin 675100b53a Remove flaky TestToolChoiceLfm2Moe from test_tool_choice (#22137) 2026-04-04 22:53:11 -07:00
Liangsheng Yin abc297521f Fix killall_sglang missing the main sglang serve process (#22103) 2026-04-04 03:43:08 -07:00
Liangsheng Yin e9d92b0e33 Relax spec decoding accuracy threshold to fix flaky test (#22100) 2026-04-04 02:38:35 -07:00
Liangsheng Yin 95cdbce34f [Test] Extract common PD server setup into base fixture (#22080) 2026-04-03 16:37:12 -07:00
Liangsheng Yin 5118295f7b [CI] Support CPU stage and auto-batch same-stage files in /rerun-test (#22081) 2026-04-03 15:56:54 -07:00
Liangsheng Yin 4cc970290d [CI] Fix duplicate job names that bypass branch protection (#22001) 2026-04-02 23:59:35 -07:00
Liangsheng YinandDarkSharpness 9d9537fbd3 Migrate ngram corpus from torch cpp_extension to TVM FFI jit_kernel (#21920)
Co-authored-by: DarkSharpness <2040703891@qq.com>
2026-04-02 02:18:11 -07:00
Liangsheng Yin f25bf86065 Fix ngram doc for speculative_num_draft_tokens default (#21910) 2026-04-01 22:18:24 -07:00
Liangsheng Yin 269589ad71 Return HTTP 400 for streaming validation errors (#21900) 2026-04-01 21:58:12 -07:00
Liangsheng YinandClaude Opus 4.6 875a615993 fix(ci): update est_time for 57 tests based on runtime analysis (#21896)
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
2026-04-01 20:16:13 -07:00
Liangsheng Yin ae3b207dfd Allow /rerun-test to checkout fork PR branch for trusted users (#21890) 2026-04-01 18:20:29 -07:00
Liangsheng Yin d7256eb69a Unify GSM8K eval path to Chat API for regression CI readiness (#21667) 2026-04-01 17:12:19 -07:00
Liangsheng Yin 70fc4ce3ef Add merge prohibition policy during CI maintenance mode (#21882) 2026-04-01 15:08:09 -07:00
Liangsheng YinandClaude Opus 4.6 0138708576 [Misc] Add network timeout to eval dataset downloads (#21873)
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
2026-04-01 13:16:14 -07:00
Liangsheng Yin ac039bd04e Use CustomTestCase for TestSessionControl to enable CI retry (#21830) 2026-04-01 04:26:11 -07:00
Liangsheng Yin 09907795e1 Add latency and throughput metrics to run_eval (#21793) 2026-03-31 18:36:14 -07:00
Liangsheng Yin e4c565f2f2 [Misc] Tiny: Add test network timeouts and dynamic max-parallel for 5090/2-gpu runners (#21800) 2026-03-31 18:27:39 -07:00
Liangsheng Yin a8759dd9af Fix killall.py crash when sglang is not yet installed (#21797) 2026-03-31 17:40:58 -07:00
Liangsheng Yin b6fe0cca99 Switch MooncakeSpec to EAGLE3 + Llama-3.1 (#21794) 2026-03-31 17:12:20 -07:00
Liangsheng Yin d047d41bad Increase hicache eval to 200 examples (#21791) 2026-03-31 16:58:44 -07:00
Liangsheng Yin 7932e4c3e6 Remove redundant test_moe_eval_accuracy_large (#21787) 2026-03-31 16:45:03 -07:00
Liangsheng Yin 7581d814ae Add CompletionSampler for non-chat eval in run_eval (#21785) 2026-03-31 16:33:07 -07:00
Liangsheng Yin 53d8aa23ae Cache nvidia wheels locally to skip repeated 830 MB downloads in CI (#21778) 2026-03-31 16:06:09 -07:00
Liangsheng Yin f4505e2ee3 Fix ineffective is_base_mistral CI patch for HF API rate limiting (#21729) 2026-03-31 12:54:34 -07:00
Liangsheng Yinandwan4ch c06ca1526c Fix circular reference in CustomTestCase.__init_subclass__ (#21650)
Co-authored-by: wan4ch <wan4ch@gmail.com>
2026-03-29 20:38:12 -07:00