SGLang is a high-performance serving framework for large language models and multimodal models.
This report summarizes scanner evidence for sgl-project/sglang, a Python project with 36,969 GitHub stars. SynthScan examined 3,045,694 lines of code across 9529 source files, recording 44924 pattern matches distributed across 26 syntactic categories. The legacy adjusted pattern signal is 18.2 (Elevated pattern signal); it is not an authorship estimate.
Use the findings to ask whether maintainers can explain intent, verify dependencies, simplify unnecessary scaffolding, and focus review effort. Risk Payload v1 adds ownership, supply-chain, over-engineering, and review-effort vectors when available. These are prompts for human verification, not probabilities, security findings, or measured maintenance costs.
Use each vector to guide manual verification of the underlying findings.
Can maintainers explain intent, assumptions, and trade-offs? 33410 finding(s).
Are imports declared and package/API names verified? 3949 finding(s).
Are wrappers, guards, and repeated structures necessary? 35367 finding(s).
Which changes need focused human review? 44924 finding(s).
This chart tracks the legacy adjusted pattern score across scan runs. Changes can reflect source changes, scan coverage, or detector/scoring versions; they do not measure AI adoption or prove that a team refactored code. Compare runs only when their scoring versions are compatible. The dashed line tracks the raw pattern-hit count.
Groups findings by configured severity weight. Severity is not a confidence estimate, security rating, or proof that a package is missing. Verify imports against the project's manifests and inspect each finding in its local design context.
This horizontal bar chart shows where configured pattern scores accumulate across directories. Treat the bars as locations to inspect, not as evidence of authorship or code quality. Use the file-level findings to decide whether a directory needs a focused ownership, dependency, complexity, or review check.
The scanner identified 44924 distinct pattern matches across 26 syntactic categories. Each entry below represents a discrete location in the source code where the engine recorded a detector match to inspect in context. Expand any category row to inspect the individual file paths, line numbers, code snippets, and the lexical context (CODE, COMMENT, or STRING) in which each match was detected.
Reading the findings table: The Severity column indicates the configured finding weight (CRITICAL / HIGH / MEDIUM / LOW). The Context column identifies whether the match occurred inside executable code, an inline comment, or a string literal — legacy comment-context matches receive a ×1.5 score weight. The ⚡ bolt icon marks clustered matches: three or more patterns within a 10-line window, each receiving an additional ×1.5 density multiplier under the legacy scoring model.
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| LOW | benchmark/kernels/attention/sm90_config_search.py | 183 | def find_feasible_bwd_configs( | CODE |
| LOW | benchmark/kernels/attention/sm90_config_search.py | 327 | def find_feasible_fwd_configs( | CODE |
| LOW | …ing_window_attention_triton/bench_triton_swa_kernel.py | 11 | def extend_attention_fwd_torch( | CODE |
| LOW | benchmark/kernels/all_gather/benchmark_aiter.py | 175 | def expected_logical_all_gather( | CODE |
| LOW | benchmark/kernels/all_gather/benchmark_aiter.py | 230 | def install_aiter_aot_import_shims() -> None: | CODE |
| LOW | benchmark/kernels/all_gather/benchmark_mscclpp.py | 81 | def _validate_cuda_graph_updates( | CODE |
| LOW⚡ | …shinfer_allreduce_fusion/benchmark_fused_collective.py | 124 | def cleanup_flashinfer_workspace(ipc_handles): | CODE |
| LOW | …shinfer_allreduce_fusion/benchmark_fused_collective.py | 87 | def setup_flashinfer_workspace( | CODE |
| LOW | …shinfer_allreduce_fusion/benchmark_fused_collective.py | 155 | def get_trtllm_fused_allreduce_kwargs(self): | CODE |
| LOW | …shinfer_allreduce_fusion/benchmark_fused_collective.py | 165 | def flashinfer_fused_allreduce_rmsnorm( | CODE |
| LOW | …shinfer_allreduce_fusion/benchmark_fused_collective.py | 290 | def standard_allreduce_rmsnorm( | CODE |
| LOW | …shinfer_allreduce_fusion/benchmark_fused_collective.py | 400 | def standard_allreduce_rmsnorm_native( | CODE |
| LOW | …shinfer_allreduce_fusion/benchmark_fused_collective.py | 481 | def standard_allreduce_rmsnorm_native_compiled( | CODE |
| LOW | …shinfer_allreduce_fusion/benchmark_fused_collective.py | 897 | def prepare_results_with_speedups(results_dict): | CODE |
| LOW | …uler_batch/benchmark_write_req_to_token_pool_triton.py | 12 | def write_req_to_token_pool_triton( | CODE |
| LOW | …uler_batch/benchmark_write_req_to_token_pool_triton.py | 49 | def write_req_to_token_pool_triton_optimize( | CODE |
| LOW | …uler_batch/benchmark_write_req_to_token_pool_triton.py | 91 | def write_req_to_token_pool_reference( | CODE |
| LOW | …uler_batch/benchmark_write_req_to_token_pool_triton.py | 114 | def test_write_req_to_token_pool(): | CODE |
| LOW | benchmark/kernels/fused_moe_triton/common_utils.py | 24 | def calculate_shard_intermediate_size( | CODE |
| LOW | benchmark/kernels/fused_moe_triton/common_utils.py | 228 | def get_rocm_configs_compute_bound() -> List[Dict[str, int]]: | CODE |
| LOW | benchmark/kernels/fused_moe_triton/common_utils.py | 251 | def get_configs_compute_bound() -> List[Dict[str, int]]: | CODE |
| LOW | benchmark/kernels/deepseek/benchmark_q8kv8_kv_gather.py | 92 | def build_page_table_flattened( | CODE |
| LOW | …nels/deepseek/benchmark_deepgemm_fp8_gemm_blackwell.py | 182 | def get_benchmark_plot_friendly(tp_size): | CODE |
| LOW | …hmark/kernels/quantization/tuning_block_wise_kernel.py | 149 | def get_rocm_configs_compute_bound(): | CODE |
| LOW | …hmark/kernels/quantization/tuning_block_wise_kernel.py | 172 | def get_configs_compute_bound(): | CODE |
| LOW | …s/decoding_attention_triton/triton_flashinfer_cudnn.py | 98 | def decode_attention_flashinfer(dtype, head_num_q, head_num_kv): | CODE |
| LOW | benchmark/agentic-rollout/tests/test_simulate.py | 154 | async def test_all_modes_and_stream_formats_preserve_history(self): | CODE |
| LOW | benchmark/agentic-rollout/tests/test_simulate.py | 214 | async def test_omitted_rank_requires_verified_single_worker(self): | CODE |
| LOW | benchmark/agentic-rollout/tests/test_simulate.py | 238 | async def test_disable_dp_sticky_routing(self): | CODE |
| LOW | benchmark/agentic-rollout/tests/test_simulate.py | 247 | async def test_invalid_stream_fails_and_closes_owned_sessions(self): | CODE |
| LOW | benchmark/agentic-rollout/tests/test_simulate.py | 264 | async def test_server_rejection_is_preserved(self): | CODE |
| LOW⚡ | benchmark/agentic-rollout/tests/test_simulate.py | 289 | def test_inputs_are_repeatable_and_distinct(self): | CODE |
| LOW⚡ | benchmark/agentic-rollout/tests/test_simulate.py | 295 | def test_token_weighted_cache_and_missing_data(self): | CODE |
| LOW⚡ | benchmark/agentic-rollout/tests/test_simulate.py | 304 | def test_throughput_uses_arrivals_and_partial_window_duration(self): | CODE |
| LOW | benchmark/agentic-rollout/tests/test_simulate.py | 332 | def test_counter_labels_resets_and_missing_series(self): | CODE |
| LOW | benchmark/agentic-rollout/tests/test_explore.py | 27 | def test_occupancy_includes_evictable_and_matches_labels(self): | CODE |
| LOW | benchmark/agentic-rollout/tests/test_explore.py | 46 | def test_hits_include_storage_and_uncached_denominator(self): | STRING |
| LOW | benchmark/agentic-rollout/tests/test_explore.py | 58 | def test_per_label_reset_cannot_be_hidden_by_another_worker(self): | STRING |
| LOW⚡ | benchmark/agentic-rollout/tests/test_explore.py | 69 | def test_scrape_error_breaks_rates(self): | STRING |
| LOW⚡ | benchmark/agentic-rollout/tests/test_explore.py | 79 | def test_nonfinite_missing_and_zero_capacity(self): | STRING |
| LOW⚡ | benchmark/agentic-rollout/tests/test_explore.py | 88 | def test_old_failed_recording_and_offline_assets(self): | STRING |
| LOW | benchmark/agentic-rollout/tests/test_explore.py | 160 | def test_interrupted_empty_recording(self): | STRING |
| LOW | benchmark/mmmu/bench_sglang.py | 127 | async def process_sample_with_semaphore( | CODE |
| LOW | benchmark/mmmu/eval_utils.py | 292 | def _parse_explicit_multi_choice_answer(response, all_choices): | CODE |
| LOW | benchmark/mmmu/eval_utils.py | 303 | def parse_multi_choice_response(response, all_choices, index2ans): | CODE |
| LOW | …hmark/bench_pynccl_allocator/bench_segment_tracking.py | 56 | def bench_register_segments_with_comm( | CODE |
| LOW | …hmark/bench_pynccl_allocator/bench_segment_tracking.py | 114 | def bench_with_various_segment_counts( | CODE |
| LOW | benchmark/ocr/eval_utils.py | 440 | def eval_math_formula_accuracy(test: dict, ocr_output: str) -> bool: | CODE |
| LOW | benchmark/scheduler/bench_token_storage.py | 84 | def _batch_tensor_from_pyarrays(parts: list[array]) -> torch.Tensor: | CODE |
| LOW | benchmark/scheduler/bench_token_storage.py | 256 | def microbench_torch_tensor_paths( | CODE |
| LOW | benchmark/hicache/bench_mix.py | 299 | async def async_request_sglang_generate( | CODE |
| LOW | benchmark/hicache/bench_warm_cache.py | 90 | def _create_bench_client_session() -> aiohttp.ClientSession: | CODE |
| LOW | benchmark/hicache/bench_warm_cache.py | 97 | async def async_request_sglang_generate( | CODE |
| LOW | benchmark/hicache/bench_warm_cache.py | 413 | def maybe_write_summary_jsonl( | CODE |
| LOW | benchmark/hicache/bench_warm_cache.py | 472 | async def benchmark_shared_prefix_pct( | CODE |
| LOW | benchmark/hicache/bench_serving.py | 71 | async def async_request_openai_completions( | CODE |
| LOW | benchmark/hicache/data_processing.py | 150 | def sample_ultrachat_requests( | CODE |
| LOW | benchmark/hicache/data_processing.py | 434 | def sample_generated_shared_prefix_requests( | CODE |
| LOW | benchmark/hicache/bench_hicache_write_back.py | 69 | def _validate_mha_correctness( | CODE |
| LOW | benchmark/hicache/bench_hicache_write_back.py | 88 | def _validate_mla_correctness( | CODE |
| 26989 more matches not shown… | ||||
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| MEDIUM | benchmark/kernels/attention/sm90_config_search.py | 62 | # ============================================================================ | COMMENT |
| MEDIUM | benchmark/kernels/attention/sm90_config_search.py | 64 | # ============================================================================ | COMMENT |
| MEDIUM | benchmark/kernels/attention/sm90_config_search.py | 267 | # ============================================================================ | COMMENT |
| MEDIUM | benchmark/kernels/attention/sm90_config_search.py | 269 | # ============================================================================ | COMMENT |
| MEDIUM | benchmark/kernels/attention/sm90_config_search.py | 377 | # ============================================================================ | COMMENT |
| MEDIUM | benchmark/kernels/attention/sm90_config_search.py | 379 | # ============================================================================ | COMMENT |
| MEDIUM | benchmark/kernels/deepseek/benchmark_q8kv8_kv_gather.py | 170 | # ------------------------------------------------------------------ | COMMENT |
| MEDIUM | benchmark/kernels/deepseek/benchmark_q8kv8_kv_gather.py | 172 | # ------------------------------------------------------------------ | COMMENT |
| MEDIUM | benchmark/kernels/deepseek/benchmark_q8kv8_kv_gather.py | 201 | # ------------------------------------------------------------------ | COMMENT |
| MEDIUM | benchmark/kernels/deepseek/benchmark_q8kv8_kv_gather.py | 203 | # ------------------------------------------------------------------ | COMMENT |
| MEDIUM | …hmark/kernels/quantization/tuning_block_wise_kernel.py | 13 | # ============================================================================== | COMMENT |
| MEDIUM | benchmark/kernels/lora_csgmv/tune_lora_csgmv.py | 132 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM | benchmark/kernels/lora_csgmv/tune_lora_csgmv.py | 134 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM | benchmark/kernels/lora_csgmv/tune_lora_csgmv.py | 185 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM | benchmark/kernels/lora_csgmv/tune_lora_csgmv.py | 187 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM | benchmark/kernels/lora_csgmv/tune_lora_csgmv.py | 337 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM | benchmark/kernels/lora_csgmv/tune_lora_csgmv.py | 339 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM | benchmark/kernels/lora_csgmv/tune_lora_csgmv.py | 395 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM | benchmark/kernels/lora_csgmv/tune_lora_csgmv.py | 397 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | benchmark/ocr/bench_sglang.py | 367 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | benchmark/ocr/bench_sglang.py | 369 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | benchmark/ocr/bench_sglang.py | 372 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | benchmark/ocr/bench_sglang.py | 374 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM | benchmark/ocr/bench_sglang.py | 54 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM | benchmark/ocr/bench_sglang.py | 56 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM | benchmark/ocr/bench_sglang.py | 67 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM | benchmark/ocr/bench_sglang.py | 69 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM | benchmark/ocr/bench_sglang.py | 86 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM | benchmark/ocr/bench_sglang.py | 88 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM | benchmark/ocr/bench_sglang.py | 214 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM | benchmark/ocr/bench_sglang.py | 216 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM | benchmark/ocr/bench_sglang.py | 263 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM | benchmark/ocr/bench_sglang.py | 265 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM | benchmark/ocr/bench_sglang.py | 310 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM | benchmark/ocr/bench_sglang.py | 312 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM | benchmark/ocr/bench_sglang.py | 565 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM | benchmark/ocr/bench_sglang.py | 567 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM | benchmark/ocr/bench_sglang.py | 625 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM | benchmark/ocr/bench_sglang.py | 627 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM | benchmark/ocr/generate_report.py | 30 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM | benchmark/ocr/generate_report.py | 32 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM | benchmark/ocr/generate_report.py | 75 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM | benchmark/ocr/generate_report.py | 77 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM | benchmark/ocr/generate_report.py | 273 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM | benchmark/ocr/generate_report.py | 275 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM | benchmark/ocr/generate_report.py | 322 | # --------------------------------------------------------------------------- | STRING |
| MEDIUM | benchmark/ocr/generate_report.py | 324 | # --------------------------------------------------------------------------- | STRING |
| MEDIUM | benchmark/ocr/eval_utils.py | 22 | # ── Unicode normalization ───────────────────────────────────────────────────── | COMMENT |
| MEDIUM | benchmark/ocr/eval_utils.py | 50 | # ── Matching helpers ────────────────────────────────────────────────────────── | COMMENT |
| MEDIUM | benchmark/ocr/eval_utils.py | 96 | # ── olmOCR-bench test evaluators ───────────────────────────────────────────── | COMMENT |
| MEDIUM | benchmark/ocr/eval_utils.py | 190 | # ── Table parsing helpers ───────────────────────────────────────────────────── | COMMENT |
| MEDIUM | benchmark/ocr/eval_utils.py | 291 | # ── 1. Structured tables (HTML <tr>/<td> or Markdown) ──────────────────── | COMMENT |
| MEDIUM | benchmark/ocr/eval_utils.py | 342 | # ── 2. Flat <table>…</table> fallback (DeepSeek-OCR-2 format) ──────────── | COMMENT |
| MEDIUM | benchmark/ocr/eval_utils.py | 429 | # Math formula evaluation ───────────────────────────────────────────────────── | COMMENT |
| MEDIUM | benchmark/ocr/eval_utils.py | 467 | # ── Main dispatcher ─────────────────────────────────────────────────────────── | COMMENT |
| MEDIUM | benchmark/ocr/eval_utils.py | 510 | # ── Aggregation & reporting ─────────────────────────────────────────────────── | COMMENT |
| MEDIUM | benchmark/ocr/eval_utils.py | 602 | # ── Normalized Edit Distance (OmniDocBench-style text quality metric) ───────── | COMMENT |
| MEDIUM | benchmark/lora/lora_bench.py | 13 | # ============================================================================== | COMMENT |
| MEDIUM⚡ | …ark/bench_linear_attention/bench_kda_flashinfer_mtp.py | 50 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | …ark/bench_linear_attention/bench_kda_flashinfer_mtp.py | 52 | # --------------------------------------------------------------------------- | COMMENT |
| 2854 more matches not shown… | ||||
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| LOW | …nchmark/kernels/attention/bench_flash_attention_fp8.py | 13 | CODE | |
| LOW | …ernels/attention/bench_flash_attention_tmem_red_max.py | 10 | CODE | |
| LOW | …chmark/kernels/attention/bench_gdn_replayssm_decode.py | 30 | CODE | |
| LOW | benchmark/kernels/all_gather/benchmark_aiter.py | 16 | CODE | |
| LOW | …ls/deepseek/benchmark_cute_dsl_fp8_paged_mqa_logits.py | 3 | CODE | |
| LOW | benchmark/scheduler/bench_token_storage.py | 25 | CODE | |
| LOW | benchmark/hicache/perf.py | 1 | CODE | |
| LOW | benchmark/hicache/bench_hicache_write_back.py | 1 | CODE | |
| LOW | benchmark/unified_memory/bench_peer_aware_eviction.py | 22 | CODE | |
| LOW | benchmark/uno/math_data.py | 3 | CODE | |
| LOW | benchmark/uno/run_math_eval.py | 3 | CODE | |
| LOW | benchmark/uno/math_grader.py | 3 | CODE | |
| LOW | benchmark/bench_linear_attention/bench_gdn_qkv_split.py | 1 | CODE | |
| LOW | …router/tests/scripts/generate_kv_events_hash_parity.py | 48 | CODE | |
| LOW | …perimental/sgl-router/tests/e2e/test_tokenize_smoke.py | 5 | CODE | |
| LOW | experimental/sgl-router/tests/e2e/test_gpu_allotment.py | 13 | CODE | |
| LOW | experimental/sgl-router/tests/e2e/conftest.py | 22 | CODE | |
| LOW | experimental/sgl-router/tests/e2e/test_chat_smoke.py | 5 | CODE | |
| LOW | experimental/sgl-router/tests/e2e/test_chat_smoke.py | 8 | CODE | |
| LOW | experimental/sgl-router/tests/e2e/infra/gateway.py | 25 | CODE | |
| LOW | experimental/sgl-router/tests/e2e/infra/model_specs.py | 16 | CODE | |
| LOW | experimental/sgl-router/tests/e2e/infra/model_pool.py | 24 | CODE | |
| LOW | …al/sgl-router/tests/e2e/k8s_integration/fake_worker.py | 22 | CODE | |
| LOW | …sgl-router/tests/e2e/k8s_integration/test_lifecycle.py | 14 | CODE | |
| LOW | …ental/sgl-router/tests/e2e/k8s_integration/conftest.py | 13 | CODE | |
| LOW | …er/tests/e2e/k8s_integration/test_kv_peer_bootstrap.py | 40 | CODE | |
| LOW | …outer/tests/e2e/k8s_integration/test_h2c_forwarding.py | 20 | CODE | |
| LOW | …sgl-router/tests/e2e/k8s_integration/test_discovery.py | 8 | CODE | |
| LOW | …sgl-router/tests/e2e/k8s_integration/test_discovery.py | 11 | CODE | |
| LOW | …sgl-router/tests/e2e/k8s_integration/test_discovery.py | 12 | CODE | |
| LOW | …outer/tests/e2e/k8s_integration/test_shutdown_drain.py | 22 | CODE | |
| LOW | …sgl-router/tests/e2e/k8s_integration/fake_kv_worker.py | 21 | CODE | |
| LOW | …uter/tests/e2e/k8s_integration/test_cross_namespace.py | 15 | CODE | |
| LOW | …uter/tests/e2e/k8s_integration/test_cross_namespace.py | 18 | CODE | |
| LOW | …outer/tests/e2e/k8s_integration/test_reconciliation.py | 16 | CODE | |
| LOW | …outer/tests/e2e/k8s_integration/test_reconciliation.py | 18 | CODE | |
| LOW | …l-router/tests/e2e/chat_completions/test_validation.py | 14 | CODE | |
| LOW | …r/tests/e2e/chat_completions/test_load_based_policy.py | 6 | CODE | |
| LOW | …sts/e2e/chat_completions/test_hicache_storage_tiers.py | 58 | CODE | |
| LOW | …ang-simulator/src/sglang_simulator/dataset/__init__.py | 1 | CODE | |
| LOW | …ang-simulator/src/sglang_simulator/dataset/__init__.py | 1 | CODE | |
| LOW | …sglang-simulator/src/sglang_simulator/spec/__init__.py | 1 | CODE | |
| LOW | …sglang-simulator/src/sglang_simulator/spec/__init__.py | 2 | CODE | |
| LOW | …sglang-simulator/src/sglang_simulator/spec/__init__.py | 3 | CODE | |
| LOW | …ator/src/sglang_simulator/spec/accelerator/__init__.py | 1 | CODE | |
| LOW | …ator/src/sglang_simulator/spec/accelerator/__init__.py | 2 | CODE | |
| LOW | …-simulator/src/sglang_simulator/spec/model/__init__.py | 1 | CODE | |
| LOW | …/src/sglang_simulator/simulation/benchmark/__init__.py | 1 | CODE | |
| LOW | …/src/sglang_simulator/simulation/benchmark/__init__.py | 2 | CODE | |
| LOW | …or/src/sglang_simulator/simulation/manager/__init__.py | 1 | CODE | |
| LOW | …or/src/sglang_simulator/simulation/manager/__init__.py | 2 | CODE | |
| LOW | …or/src/sglang_simulator/simulation/manager/__init__.py | 3 | CODE | |
| LOW | …ulator/src/sglang_simulator/simulation/sglang/utils.py | 7 | CODE | |
| LOW | …ulator/src/sglang_simulator/simulation/sglang/utils.py | 8 | CODE | |
| LOW | …glang-simulator/src/sglang_simulator/utils/__init__.py | 1 | CODE | |
| LOW | test/manual/test_ray_engine.py | 25 | CODE | |
| LOW | test/manual/quant/test_block_fp8.py | 692 | CODE | |
| LOW | test/manual/kernels/test_hisparse_prefetch.py | 36 | CODE | |
| LOW | …ernels/benchmark/attention/bench_gdn_replayssm_fold.py | 9 | CODE | |
| LOW | test/manual/layers/moe/bench_mxfp4_sm90_kernels.py | 21 | CODE | |
| 3787 more matches not shown… | ||||
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| HIGH | benchmark/kernels/all_reduce/benchmark_aiter.py | 305 | except Exception: | CODE |
| HIGH | benchmark/kernels/all_reduce/benchmark_all_reduce.py | 326 | except Exception: | CODE |
| HIGH | benchmark/agentic-rollout/simulate.py | 144 | except asyncio.TimeoutError: | CODE |
| HIGH | benchmark/ocr/bench_sglang.py | 288 | except ImportError: | CODE |
| HIGH | benchmark/hicache/bench_buffer_mode.py | 242 | except Exception: | CODE |
| HIGH | benchmark/asr/bench_sglang.py | 114 | except json.JSONDecodeError: | CODE |
| HIGH | benchmark/uno/math_grader.py | 100 | except Exception: | CODE |
| HIGH | benchmark/uno/math_grader.py | 117 | except Exception: | CODE |
| HIGH | benchmark/uno/math_grader.py | 210 | except ValueError: | CODE |
| HIGH | …rk/bench_linear_attention/bench_gdn_prefill_cutedsl.py | 212 | except Exception: | CODE |
| HIGH | benchmark/bench_linear_attention/bench_gdn_prefill.py | 316 | except Exception: | CODE |
| HIGH | …/bench_linear_attention/bench_int8_checkpoint_reuse.py | 115 | except Exception: | CODE |
| HIGH | experimental/sgl-router/tests/e2e/conftest.py | 210 | except Exception: # noqa: BLE001 | CODE |
| HIGH | experimental/sgl-router/tests/e2e/infra/gateway.py | 247 | except ProcessLookupError: | CODE |
| HIGH | experimental/sgl-router/tests/e2e/infra/gateway.py | 261 | except ProcessLookupError: | CODE |
| HIGH | experimental/sgl-router/tests/e2e/infra/model_pool.py | 194 | except subprocess.TimeoutExpired: | CODE |
| HIGH | experimental/sgl-router/tests/e2e/infra/model_pool.py | 339 | except (httpx.RequestError, httpx.TimeoutException): | CODE |
| HIGH | …glang-simulator/test/test_simulation_sglang_serving.py | 66 | except requests.RequestException: | CODE |
| HIGH | …rc/sglang_simulator/simulation/sglang/mem_pool_host.py | 86 | except ImportError: | CODE |
| HIGH | …rc/sglang_simulator/simulation/sglang/mem_pool_host.py | 92 | except ImportError: | CODE |
| HIGH | …rc/sglang_simulator/simulation/sglang/mem_pool_host.py | 331 | except AssertionError: | CODE |
| HIGH | test/manual/test_weight_cache_e2e.py | 142 | except Exception: | CODE |
| HIGH | test/manual/test_dsa_alias_cli_registry_env.py | 113 | except Exception: | CODE |
| HIGH | test/manual/test_ray_engine.py | 138 | except Exception: | CODE |
| HIGH | test/manual/test_ray_engine.py | 142 | except Exception: | CODE |
| HIGH | test/manual/test_ray_engine.py | 464 | except Exception: | CODE |
| HIGH | test/manual/test_ray_engine.py | 468 | except Exception: | CODE |
| HIGH | test/manual/test_ray_engine.py | 452 | except req_lib.exceptions.RequestException: | CODE |
| HIGH | test/manual/test_cross_node_scheduler_info_sync.py | 145 | except Exception: | CODE |
| HIGH | test/manual/test_weight_loader_v2_equiv.py | 50 | except AssertionError: | CODE |
| HIGH | test/manual/ascend/disaggregation_utils.py | 80 | except Exception: | CODE |
| HIGH | test/manual/ep/test_elastic_scale.py | 290 | except Exception: | CODE |
| HIGH | test/manual/ep/test_elastic_scale.py | 297 | except Exception: | CODE |
| HIGH | test/manual/ep/test_elastic_scale.py | 302 | except Exception: | CODE |
| HIGH | test/manual/ep/test_elastic_scale.py | 385 | except Exception: | CODE |
| HIGH | test/manual/ep/test_flashinfer_dispatcher.py | 60 | except ImportError: | CODE |
| HIGH | test/manual/ep/test_mooncake_expert_backup.py | 105 | except requests.RequestException: | CODE |
| HIGH | test/manual/dsv4/test_dsv4_flash_mtp_dp4.py | 121 | except requests.exceptions.Timeout: | CODE |
| HIGH | test/manual/dsv4/test_dsv4_pd_disagg_fp8_nixl.py | 272 | except Exception: | CODE |
| HIGH | test/manual/dsv4/test_dsv4_pd_disagg_fp8_nixl.py | 294 | except Exception: | CODE |
| HIGH | test/manual/vlm/test_anthropic_vision.py | 83 | except json.JSONDecodeError: | CODE |
| HIGH | test/manual/lora/check_spec_baseline_divergence.py | 88 | except Exception: | CODE |
| HIGH | test/manual/lora/test_lora_cuda_graph.py | 103 | except RuntimeError: | CODE |
| HIGH | test/manual/lora/test_lora_spec_decoding.py | 49 | except RuntimeError: | CODE |
| HIGH | test/manual/lora/test_lora_backend.py | 71 | except RuntimeError: | CODE |
| HIGH | test/manual/lora/test_lora_xpu_basics.py | 41 | except RuntimeError: | CODE |
| HIGH | …red/attention/test_gemma4_swa_triton_oob_regression.py | 99 | except Exception: | CODE |
| HIGH | test/registered/attention/test_fused_fp8_kv_write.py | 268 | except _StopForward: | CODE |
| HIGH | test/registered/attention/test_fused_fp8_kv_write.py | 322 | except _StopForward: | CODE |
| HIGH | test/registered/attention/test_fused_fp8_kv_write.py | 378 | except _StopForward: | CODE |
| HIGH | test/registered/unit/test_runtime_context.py | 2375 | except Exception: | CODE |
| HIGH | test/registered/unit/test_runtime_context.py | 2413 | except Exception: | CODE |
| HIGH | …t/registered/unit/platforms/test_platform_interface.py | 274 | except Exception: | CODE |
| HIGH | …d/unit/disaggregation/test_mori_deferred_kv_release.py | 21 | except (ImportError, OSError): | CODE |
| HIGH | …hardware_backend/mlx/test_mlx_reference_correctness.py | 159 | except Exception: | CODE |
| HIGH | test/registered/unit/distributed/test_gated_launch.py | 45 | except requests.exceptions.RequestException: | CODE |
| HIGH | …registered/unit/models/test_deepseek_nextn_mm_embed.py | 117 | except Exception: | CODE |
| HIGH | …registered/unit/models/test_deepseek_nextn_mm_embed.py | 174 | except Exception: | CODE |
| HIGH | …tered/unit/multimodal/rust/shared/test_wrap_encoded.py | 181 | except FileNotFoundError: | CODE |
| HIGH | …registered/unit/entrypoints/test_ssl_cert_refresher.py | 34 | except OSError: | CODE |
| 482 more matches not shown… | ||||
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| LOW | benchmark/bench_adaptive_speculative.py | 70 | except Exception as e: | CODE |
| LOW | …nchmark/kernels/attention/bench_flash_attention_fp8.py | 363 | except Exception as e: | CODE |
| LOW | …nchmark/kernels/attention/bench_flash_attention_fp8.py | 445 | except Exception as e: | CODE |
| LOW | benchmark/kernels/all_gather/benchmark_aiter.py | 365 | except Exception: | CODE |
| LOW⚡ | …shinfer_allreduce_fusion/benchmark_fused_collective.py | 119 | except Exception as e: | CODE |
| LOW⚡ | …shinfer_allreduce_fusion/benchmark_fused_collective.py | 132 | except Exception as e: | CODE |
| LOW | …shinfer_allreduce_fusion/benchmark_fused_collective.py | 49 | except Exception: # pragma: no cover - fallback on non-supported platforms | CODE |
| LOW | …shinfer_allreduce_fusion/benchmark_fused_collective.py | 653 | except Exception as e: | CODE |
| LOW | …shinfer_allreduce_fusion/benchmark_fused_collective.py | 667 | except Exception as e: | CODE |
| LOW | …shinfer_allreduce_fusion/benchmark_fused_collective.py | 686 | except Exception as e: | CODE |
| LOW | …shinfer_allreduce_fusion/benchmark_fused_collective.py | 703 | except Exception as e: | CODE |
| LOW | …shinfer_allreduce_fusion/benchmark_fused_collective.py | 723 | except Exception as e: | CODE |
| LOW | …shinfer_allreduce_fusion/benchmark_fused_collective.py | 740 | except Exception as e: | CODE |
| LOW | …shinfer_allreduce_fusion/benchmark_fused_collective.py | 765 | except Exception as e: | CODE |
| LOW | …shinfer_allreduce_fusion/benchmark_fused_collective.py | 790 | except Exception as e: | CODE |
| LOW | …shinfer_allreduce_fusion/benchmark_fused_collective.py | 814 | except Exception as e: | CODE |
| LOW | …shinfer_allreduce_fusion/benchmark_fused_collective.py | 831 | except Exception as e: | CODE |
| LOW | …shinfer_allreduce_fusion/benchmark_fused_collective.py | 857 | except Exception as e: | CODE |
| LOW | …shinfer_allreduce_fusion/benchmark_fused_collective.py | 885 | except Exception as e: | CODE |
| LOW | …shinfer_allreduce_fusion/benchmark_fused_collective.py | 1086 | except Exception as e: | CODE |
| LOW | benchmark/kernels/all_reduce/benchmark_aiter.py | 227 | except Exception as e: | CODE |
| LOW | benchmark/kernels/all_reduce/benchmark_aiter.py | 237 | except Exception as e: | CODE |
| LOW | benchmark/kernels/all_reduce/benchmark_aiter.py | 253 | except Exception as e: | CODE |
| LOW | benchmark/kernels/all_reduce/benchmark_aiter.py | 265 | except Exception as e: | CODE |
| LOW | benchmark/kernels/all_reduce/benchmark_aiter.py | 305 | except Exception: | CODE |
| LOW | benchmark/kernels/all_reduce/benchmark_all_reduce.py | 247 | except Exception as e: | CODE |
| LOW | benchmark/kernels/all_reduce/benchmark_all_reduce.py | 257 | except Exception as e: | CODE |
| LOW | benchmark/kernels/all_reduce/benchmark_all_reduce.py | 276 | except Exception as e: | CODE |
| LOW | benchmark/kernels/all_reduce/benchmark_all_reduce.py | 286 | except Exception as e: | CODE |
| LOW | benchmark/kernels/all_reduce/benchmark_all_reduce.py | 326 | except Exception: | CODE |
| LOW | benchmark/kernels/deepseek/benchmark_q8kv8_q_prep.py | 335 | except Exception as e: | CODE |
| LOW | benchmark/kernels/deepseek/benchmark_q8kv8_q_prep.py | 383 | except Exception as e: | CODE |
| LOW | benchmark/kernels/lora_csgmv/tune_lora_csgmv.py | 232 | except Exception: | CODE |
| LOW | benchmark/kernels/lora_csgmv/tune_lora_csgmv.py | 309 | except Exception: | CODE |
| LOW | benchmark/mmmu/bench_sglang.py | 64 | except Exception: | CODE |
| LOW | benchmark/mmmu/bench_hf.py | 35 | except Exception as first_exception: | CODE |
| LOW | benchmark/mmmu/bench_hf.py | 60 | except Exception as second_exception: | CODE |
| LOW | benchmark/mmmu/eval_utils.py | 195 | except Exception as exc: | CODE |
| MEDIUM | benchmark/mmmu/eval_utils.py | 514 | print(f"Error to extract answer from: {pred_i}") | CODE |
| MEDIUM | …hmark/bench_pynccl_allocator/bench_segment_tracking.py | 195 | print("Error: CUDA is not available. This benchmark requires a GPU.") | CODE |
| LOW⚡ | benchmark/ocr/bench_sglang.py | 363 | except Exception: | CODE |
| LOW | benchmark/ocr/bench_sglang.py | 449 | except Exception as exc: | CODE |
| LOW | benchmark/ocr/bench_sglang.py | 460 | except Exception as exc: | CODE |
| LOW | benchmark/ocr/bench_sglang.py | 686 | except Exception as exc: | CODE |
| LOW | benchmark/ocr/eval_utils.py | 503 | except Exception as exc: | CODE |
| LOW | benchmark/hicache/bench_mix.py | 362 | except Exception as e: | CODE |
| LOW | benchmark/hicache/bench_mix.py | 423 | except Exception as e: | CODE |
| MEDIUM | benchmark/hicache/bench_mix.py | 479 | print(f"Error processing response for client {user_data}: {e}") | CODE |
| MEDIUM | benchmark/hicache/bench_mix.py | 417 | def handle_request(self, user_data): | CODE |
| LOW | benchmark/hicache/bench_warm_cache.py | 169 | except Exception as exc: | CODE |
| LOW | benchmark/hicache/bench_serving.py | 194 | except Exception: | CODE |
| LOW | benchmark/hicache/bench_serving.py | 214 | except Exception: | CODE |
| LOW | benchmark/hicache/bench_serving.py | 273 | except Exception as e: | CODE |
| MEDIUM | benchmark/hicache/bench_serving.py | 624 | print(f"Error running benchmark for request rate: {request_rate}") | CODE |
| LOW | benchmark/hicache/bench_serving.py | 731 | except Exception as e: | CODE |
| LOW | benchmark/hicache/bench_buffer_mode.py | 136 | except Exception: | CODE |
| LOW | benchmark/hicache/bench_buffer_mode.py | 242 | except Exception: | CODE |
| MEDIUM | benchmark/hicache/bench_buffer_mode.py | 133 | def scrape_metrics(base_url): | CODE |
| MEDIUM | benchmark/hicache/bench_multiturn.py | 181 | print(f"Error writing to JSONL file: {e}") | CODE |
| LOW | benchmark/hicache/bench_multiturn.py | 397 | except Exception as e: | CODE |
| 2133 more matches not shown… | ||||
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| LOW | benchmark/lean_kernel_sweep.py | 124 | CODE | |
| LOW | benchmark/kernels/bench_block_verification.py | 18 | CODE | |
| LOW | benchmark/kernels/attention/sm90_config_search.py | 183 | CODE | |
| LOW | benchmark/kernels/attention/sm90_config_search.py | 327 | CODE | |
| LOW | …nchmark/kernels/attention/bench_flash_attention_fp8.py | 244 | CODE | |
| LOW | benchmark/kernels/all_gather/benchmark_aiter.py | 290 | CODE | |
| LOW | …shinfer_allreduce_fusion/benchmark_fused_collective.py | 608 | CODE | |
| LOW | …shinfer_allreduce_fusion/benchmark_fused_collective.py | 897 | CODE | |
| LOW | …kernels/all_reduce/benchmark_fused_ar_rms_quant_amd.py | 224 | CODE | |
| LOW | …hmark/kernels/all_reduce/benchmark_fused_ar_rms_amd.py | 368 | CODE | |
| LOW | …/fused_moe_triton/benchmark_torch_compile_fused_moe.py | 16 | CODE | |
| LOW | …ernels/fused_moe_triton/tuning_fused_moe_triton_sep.py | 649 | CODE | |
| LOW | benchmark/kernels/fused_moe_triton/common_utils.py | 33 | CODE | |
| LOW | benchmark/kernels/fused_moe_triton/common_utils.py | 228 | CODE | |
| LOW | benchmark/kernels/fused_moe_triton/common_utils.py | 251 | CODE | |
| LOW | benchmark/kernels/deepseek/benchmark_q8kv8_q_prep.py | 272 | CODE | |
| LOW | …ls/deepseek/benchmark_cute_dsl_fp8_paged_mqa_logits.py | 89 | CODE | |
| LOW | …hmark/kernels/quantization/tuning_block_wise_kernel.py | 149 | CODE | |
| LOW | …hmark/kernels/quantization/tuning_block_wise_kernel.py | 172 | CODE | |
| LOW | benchmark/kernels/lora_csgmv/tune_lora_csgmv.py | 137 | CODE | |
| LOW | benchmark/kernels/lora_csgmv/tune_lora_csgmv.py | 155 | CODE | |
| LOW | benchmark/kernels/deepep/tuning_deepep.py | 54 | CODE | |
| LOW | …s/decoding_attention_triton/triton_flashinfer_cudnn.py | 159 | CODE | |
| LOW | benchmark/agentic-rollout/plot.py | 113 | CODE | |
| LOW | benchmark/agentic-rollout/simulate.py | 259 | CODE | |
| LOW | benchmark/agentic-rollout/tests/test_simulate.py | 154 | CODE | |
| LOW | benchmark/agentic-rollout/tests/test_simulate.py | 214 | CODE | |
| LOW | benchmark/dllm/bench_diffusion_gemma.py | 28 | CODE | |
| LOW | benchmark/mmmu/eval_utils.py | 154 | CODE | |
| LOW | benchmark/mmmu/eval_utils.py | 303 | CODE | |
| LOW | benchmark/mmmu/eval_utils.py | 423 | CODE | |
| LOW | benchmark/mmmu/eval_utils.py | 528 | CODE | |
| LOW | benchmark/mmmu/eval_utils.py | 430 | CODE | |
| LOW | benchmark/ocr/bench_sglang.py | 430 | CODE | |
| LOW | benchmark/ocr/generate_report.py | 148 | CODE | |
| LOW | benchmark/ocr/eval_utils.py | 263 | CODE | |
| LOW | benchmark/ocr/eval_utils.py | 390 | CODE | |
| LOW | benchmark/hicache/bench_mix.py | 299 | CODE | |
| LOW | benchmark/hicache/nextqa.py | 9 | CODE | |
| LOW | benchmark/hicache/bench_warm_cache.py | 97 | CODE | |
| LOW | benchmark/hicache/bench_serving.py | 71 | CODE | |
| LOW | benchmark/hicache/bench_serving.py | 286 | CODE | |
| LOW | benchmark/hicache/data_processing.py | 37 | CODE | |
| LOW | benchmark/hicache/data_processing.py | 198 | CODE | |
| LOW | benchmark/hicache/data_processing.py | 267 | CODE | |
| LOW | benchmark/hicache/data_processing.py | 520 | CODE | |
| LOW | benchmark/hicache/bench_buffer_mode.py | 369 | CODE | |
| LOW | benchmark/hicache/bench_multiturn.py | 437 | CODE | |
| LOW | benchmark/hicache/bench_multiturn.py | 559 | CODE | |
| LOW | benchmark/prefill_only/util.py | 106 | CODE | |
| LOW | benchmark/prefill_only/util.py | 284 | CODE | |
| LOW | benchmark/asr/bench_sglang.py | 75 | CODE | |
| LOW | benchmark/lora/lora_bench.py | 48 | CODE | |
| LOW | …ark/bench_linear_attention/bench_kda_flashinfer_mtp.py | 203 | CODE | |
| LOW | benchmark/bench_linear_attention/bench_gdn_decode.py | 359 | CODE | |
| LOW | benchmark/bench_linear_attention/bench_gdn_prefill.py | 520 | CODE | |
| LOW | …l/sgl-router/tests/scripts/generate_deepseek_parity.py | 49 | CODE | |
| LOW | experimental/sgl-router/tests/e2e/infra/gateway.py | 237 | CODE | |
| LOW | experimental/sgl-router/tests/e2e/infra/gateway.py | 354 | CODE | |
| LOW | …ental/sgl-router/tests/e2e/k8s_integration/conftest.py | 102 | CODE | |
| 2618 more matches not shown… | ||||
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| LOW | …nchmark/kernels/attention/bench_flash_attention_fp8.py | 1 | # Benchmark FP8 attention for FA4 (CuTe-DSL) on SM100. | COMMENT |
| LOW | …hmark/kernels/quantization/tuning_block_wise_kernel.py | 1 | # Copyright 2025 SGLang Team | COMMENT |
| LOW | benchmark/lora/lora_bench.py | 1 | # Copyright 2023-2024 SGLang Team | COMMENT |
| LOW | experimental/sgl-router/Cargo.toml | 41 | # axum's `http2` looks redundant and is not — please don't drop it. `axum::serve` | COMMENT |
| LOW | …rimental/sgl-router/tests/proxy/pd_protocol_binding.rs | 1 | // SPDX-FileCopyrightText: Copyright (c) 2026 The SGLang Authors | COMMENT |
| LOW | …ental/sgl-router/tests/proxy/pd_bootstrap_injection.rs | 1 | // SPDX-FileCopyrightText: Copyright (c) 2026 The SGLang Authors | COMMENT |
| LOW | experimental/sgl-router/tests/proxy/h2c_forward.rs | 1 | // SPDX-FileCopyrightText: Copyright (c) 2026 The SGLang Authors | COMMENT |
| LOW | …perimental/sgl-router/tests/proxy/graceful_shutdown.rs | 1 | // SPDX-FileCopyrightText: Copyright (c) 2026 The SGLang Authors | COMMENT |
| LOW | …mental/sgl-router/tests/proxy/cache_aware_input_ids.rs | 1 | // SPDX-FileCopyrightText: Copyright (c) 2026 The SGLang Authors | COMMENT |
| LOW | …perimental/sgl-router/tests/proxy/pd_pool_isolation.rs | 1 | // SPDX-FileCopyrightText: Copyright (c) 2026 The SGLang Authors | COMMENT |
| LOW | experimental/sgl-router/tests/proxy/chat_routing.rs | 1081 | "body drop must trip the breaker (threshold=1)" | COMMENT |
| LOW | experimental/sgl-router/tests/proxy/chat_routing.rs | 1641 | } | COMMENT |
| LOW | experimental/sgl-router/tests/proxy/sticky_input_ids.rs | 1 | // SPDX-FileCopyrightText: Copyright (c) 2026 The SGLang Authors | COMMENT |
| LOW | experimental/sgl-router/tests/proxy/inbound_h2c.rs | 1 | // SPDX-FileCopyrightText: Copyright (c) 2026 The SGLang Authors | COMMENT |
| LOW | experimental/sgl-router/tests/proxy/common/streaming.rs | 21 | use http_body_util::BodyExt; | COMMENT |
| LOW | …imental/sgl-router/tests/component/tokenizer/parity.rs | 1 | // SPDX-FileCopyrightText: Copyright (c) 2026 The SGLang Authors | COMMENT |
| LOW | …tal/sgl-router/tests/component/policies/fused_score.rs | 1 | // SPDX-FileCopyrightText: Copyright (c) 2026 The SGLang Authors | COMMENT |
| LOW | …r/tests/component/policies/kv_events_peer_bootstrap.rs | 1 | // SPDX-FileCopyrightText: Copyright (c) 2026 The SGLang Authors | COMMENT |
| LOW | …r/tests/component/policies/kv_events_peer_bootstrap.rs | 381 | for (w, seq) in &cursors { | COMMENT |
| LOW | …/tests/component/policies/kv_events_two_subscribers.rs | 1 | // SPDX-FileCopyrightText: Copyright (c) 2026 The SGLang Authors | COMMENT |
| LOW | …/tests/component/policies/kv_events_two_subscribers.rs | 141 | assert!(r.is_ok(), "router_b shutdown hung"); | COMMENT |
| LOW | …/tests/component/policies/kv_events_tree_concurrent.rs | 1 | // SPDX-FileCopyrightText: Copyright (c) 2026 The SGLang Authors | COMMENT |
| LOW | …/tests/component/policies/kv_events_tree_concurrent.rs | 161 | ); | COMMENT |
| LOW | …uter/tests/component/policies/kv_events_hash_parity.rs | 1 | // SPDX-FileCopyrightText: Copyright (c) 2026 The SGLang Authors | COMMENT |
| LOW | …sgl-router/tests/component/workers/concurrent_state.rs | 81 | let snapshot = r.workers_for(&model); | COMMENT |
| LOW | experimental/sgl-router/tests/e2e/requirements.txt | 1 | httpx==0.27.2 | COMMENT |
| LOW | …outer/tests/e2e/k8s_integration/test_shutdown_drain.py | 101 | # Budget for observing the /readyz flip, deliberately a fraction of the drain: | COMMENT |
| LOW | …outer/tests/e2e/k8s_integration/test_shutdown_drain.py | 201 | COMMENT | |
| LOW | …sts/e2e/chat_completions/test_hicache_storage_tiers.py | 281 | COMMENT | |
| LOW | experimental/sgl-router/sgl-kv-indexer/src/bridge.rs | 1 | // SPDX-FileCopyrightText: Copyright (c) 2026 The SGLang Authors | COMMENT |
| LOW | experimental/sgl-router/sgl-kv-indexer/src/bridge.rs | 741 | ))), | COMMENT |
| LOW | experimental/sgl-router/sgl-kv-indexer/src/service.rs | 21 | /// prefix query is exempt from the hash bound; see [`validate_hashes`]. | COMMENT |
| LOW | experimental/sgl-router/sgl-kv-indexer/src/service.rs | 481 | COMMENT | |
| LOW | experimental/sgl-router/benches/tree_lookup.rs | 1 | // SPDX-FileCopyrightText: Copyright (c) 2026 The SGLang Authors | COMMENT |
| LOW | experimental/sgl-router/benches/tree_lookup.rs | 141 | /// exactly one shard; a `Continuation` writer first resolves its parent | COMMENT |
| LOW | experimental/sgl-router/src/main.rs | 41 | const DRAIN_HEARTBEAT_INTERVAL: Duration = Duration::from_secs(5); | COMMENT |
| LOW | experimental/sgl-router/src/buckets_reorg.rs | 1 | // SPDX-FileCopyrightText: Copyright (c) 2026 The SGLang Authors | COMMENT |
| LOW | experimental/sgl-router/src/proxy/sse.rs | 121 | pub abort: Option<CancellationToken>, | COMMENT |
| LOW | experimental/sgl-router/src/proxy/mod.rs | 41 | enum BreakerOutcome { | COMMENT |
| LOW | experimental/sgl-router/src/proxy/mod.rs | 61 | /// A backpressure status — `503 Service Unavailable` or `429 Too Many | COMMENT |
| LOW | experimental/sgl-router/src/proxy/mod.rs | 101 | COMMENT | |
| LOW | experimental/sgl-router/src/proxy/mod.rs | 201 | /// Breaker-gated JSON POST: acquires a cancellation-safe permit first, classifies the | COMMENT |
| LOW | experimental/sgl-router/src/proxy/mod.rs | 281 | *out.status_mut() = status; | COMMENT |
| LOW | experimental/sgl-router/src/proxy/mod.rs | 361 | // SGLang's `text/event-stream; charset=utf-8`: clients that decode by the | COMMENT |
| LOW | experimental/sgl-router/src/config/types.rs | 121 | #[value(name = "random")] | COMMENT |
| LOW | experimental/sgl-router/src/config/types.rs | 361 | pub id: String, | COMMENT |
| LOW | experimental/sgl-router/src/config/types.rs | 461 | #[value(name = "indexer")] | COMMENT |
| LOW | experimental/sgl-router/src/config/types.rs | 501 | } | COMMENT |
| LOW | experimental/sgl-router/src/config/types.rs | 541 | /// Affinity preference; legacy routing uses Strict/Soft, reorg uses Prefer/Balanced. | COMMENT |
| LOW | experimental/sgl-router/src/config/types.rs | 621 | /// while depth proxies it badly (an engine can queue far below its running cap on | COMMENT |
| LOW | experimental/sgl-router/src/config/types.rs | 741 | pub struct StaticUrlsDiscoveryConfig { | COMMENT |
| LOW | experimental/sgl-router/src/config/types.rs | 761 | /// unambiguous choice. | COMMENT |
| LOW | experimental/sgl-router/src/config/sampling.rs | 1 | // SPDX-FileCopyrightText: Copyright (c) 2026 The SGLang Authors | COMMENT |
| LOW | experimental/sgl-router/src/config/sampling.rs | 41 | #[default] | COMMENT |
| LOW | experimental/sgl-router/src/config/cli.rs | 81 | #[arg(long)] | COMMENT |
| LOW | experimental/sgl-router/src/config/cli.rs | 121 | /// Address to bind the HTTP server to. | COMMENT |
| LOW | experimental/sgl-router/src/config/cli.rs | 141 | /// `/model_info`). Set it to the engines' `--api-key`. | COMMENT |
| LOW | experimental/sgl-router/src/config/cli.rs | 161 | /// jitter. 0 retries immediately. | COMMENT |
| LOW | experimental/sgl-router/src/config/cli.rs | 181 | } | COMMENT |
| LOW | experimental/sgl-router/src/config/cli.rs | 201 | pub selector: Vec<String>, | COMMENT |
| 1653 more matches not shown… | ||||
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| HIGH | …hmark/kernels/quantization/tuning_block_wise_kernel.py | 0 | this function performs matrix multiplication with block-wise quantization. it takes two input tensors `a` and `b` with s | STRING |
| HIGH | python/sglang/kernels/ops/gemm/int8_kernel.py | 0 | this function performs matrix multiplication with block-wise quantization. it takes two input tensors `a` and `b` with s | STRING |
| HIGH | python/sglang/kernels/ops/gemm/fp8_kernel.py | 0 | this function performs matrix multiplication with block-wise quantization. it takes two input tensors `a` and `b` with s | STRING |
| HIGH | test/manual/quant/test_nvfp4_gemm_archived.py | 0 | archived test classes split out of test/registered/4-gpu-models/test_qwen35_models.py. originally registered with `regis | STRING |
| HIGH | test/manual/distributed/test_dp_attention_archived.py | 0 | archived test classes split out of test/registered/4-gpu-models/test_qwen35_models.py. originally registered with `regis | STRING |
| HIGH | test/manual/mla/test_mla_flashinfer_archived.py | 0 | archived test classes split out of test/registered/4-gpu-models/test_qwen35_models.py. originally registered with `regis | STRING |
| HIGH | test/manual/mla/test_flashmla_archived.py | 0 | archived test classes split out of test/registered/4-gpu-models/test_qwen35_models.py. originally registered with `regis | STRING |
| HIGH | test/manual/mla/test_mla_int8_deepseek_v3_archived.py | 0 | archived test classes split out of test/registered/4-gpu-models/test_qwen35_models.py. originally registered with `regis | STRING |
| HIGH | …al/4-gpu-models/test_qwen3_next_models_mtp_archived.py | 0 | archived test classes split out of test/registered/4-gpu-models/test_qwen35_models.py. originally registered with `regis | STRING |
| HIGH | test/manual/4-gpu-models/test_qwen35_models_archived.py | 0 | archived test classes split out of test/registered/4-gpu-models/test_qwen35_models.py. originally registered with `regis | STRING |
| HIGH | test/manual/quant/test_block_fp8.py | 0 | this function performs matrix multiplication with block-wise quantization using native torch. it takes two input tensors | STRING |
| HIGH | test/manual/quant/test_block_fp8_deep_gemm_blackwell.py | 0 | this function performs matrix multiplication with block-wise quantization using native torch. it takes two input tensors | STRING |
| HIGH | test/registered/kernels/ops/moe/test_block_int8.py | 0 | this function performs matrix multiplication with block-wise quantization using native torch. it takes two input tensors | STRING |
| HIGH | test/manual/nightly/test_vlms_vit_cuda_graph.py | 0 | evaluate a vlm on the mmmu validation set with lmms-eval. only `model_version` (checkpoint) and `chat_template` vary; we | STRING |
| HIGH | test/manual/nightly/test_vlms_vit_flashinfer_cudnn.py | 0 | evaluate a vlm on the mmmu validation set with lmms-eval. only `model_version` (checkpoint) and `chat_template` vary; we | STRING |
| HIGH | test/manual/nightly/test_vlms_piecewise_cuda_graph.py | 0 | evaluate a vlm on the mmmu validation set with lmms-eval. only `model_version` (checkpoint) and `chat_template` vary; we | STRING |
| HIGH | python/sglang/test/kits/mmmu_vlm_kit.py | 0 | evaluate a vlm on the mmmu validation set with lmms-eval. only `model_version` (checkpoint) and `chat_template` vary; we | STRING |
| HIGH | test/manual/nightly/test_vlms_vit_cuda_graph.py | 0 | common method to run vlm mmmu benchmark test. args: model: model to test output_path: path for output logs test_name: op | STRING |
| HIGH | test/manual/nightly/test_vlms_vit_flashinfer_cudnn.py | 0 | common method to run vlm mmmu benchmark test. args: model: model to test output_path: path for output logs test_name: op | STRING |
| HIGH | test/manual/nightly/test_vlms_piecewise_cuda_graph.py | 0 | common method to run vlm mmmu benchmark test. args: model: model to test output_path: path for output logs test_name: op | STRING |
| HIGH | python/sglang/test/kits/mmmu_vlm_kit.py | 0 | common method to run vlm mmmu benchmark test. args: model: model to test output_path: path for output logs test_name: op | STRING |
| HIGH | …stered/kernels/ops/attention/test_flash_attention_4.py | 0 | arguments: hidden_states: (total_nnz, ...), where total_nnz = number of tokens in selected in attention_mask. indices: ( | STRING |
| HIGH | python/sglang/kernels/aot/tests/test_flash_attention.py | 0 | arguments: hidden_states: (total_nnz, ...), where total_nnz = number of tokens in selected in attention_mask. indices: ( | STRING |
| HIGH | …timodal_gen/runtime/layers/attention/backends/vmoba.py | 0 | arguments: hidden_states: (total_nnz, ...), where total_nnz = number of tokens in selected in attention_mask. indices: ( | STRING |
| HIGH | test/registered/kernels/ops/attention/test_pos_enc.py | 0 | args: x: [num_tokens, num_heads, head_size] or [num_tokens, head_size] cos: [num_tokens, head_size // 2] sin: [num_token | STRING |
| HIGH | …nels/aot/python/sgl_kernel/testing/rotary_embedding.py | 0 | args: x: [num_tokens, num_heads, head_size] or [num_tokens, head_size] cos: [num_tokens, head_size // 2] sin: [num_token | STRING |
| HIGH | python/sglang/srt/layers/rotary_embedding/utils.py | 0 | args: x: [num_tokens, num_heads, head_size] or [num_tokens, head_size] cos: [num_tokens, head_size // 2] sin: [num_token | STRING |
| HIGH | …ultimodal_gen/runtime/layers/rotary_embedding/utils.py | 0 | args: x: [num_tokens, num_heads, head_size] or [num_tokens, head_size] cos: [num_tokens, head_size // 2] sin: [num_token | STRING |
| HIGH | …t/registered/kernels/ops/layernorm/test_mhc_kernels.py | 0 | provide a tp-group placeholder for kernels with mocked symmetric memory. | STRING |
| HIGH | …gistered/kernels/ops/moe/test_minimax_quant_scatter.py | 0 | provide a tp-group placeholder for kernels with mocked symmetric memory. | STRING |
| HIGH | …ed/unit/layers/quantization/test_mxfp4_sm90_cutlass.py | 0 | provide a tp-group placeholder for kernels with mocked symmetric memory. | STRING |
| HIGH | …d/unit/layers/quantization/test_mxfp4_sm120_cutlass.py | 0 | provide a tp-group placeholder for kernels with mocked symmetric memory. | STRING |
| HIGH | …nit/layers/quantization/test_mxfp4_sm100_trtllm_gen.py | 0 | provide a tp-group placeholder for kernels with mocked symmetric memory. | STRING |
| HIGH | test/registered/kernels/ops/moe/test_int8_kernel.py | 0 | matrix multiplication function that supports per-token input quantization and per-column weight quantization | STRING |
| HIGH | …t/registered/moe/test_triton_moe_channel_fp8_kernel.py | 0 | matrix multiplication function that supports per-token input quantization and per-column weight quantization | STRING |
| HIGH | python/sglang/test/cpu_test_utils.py | 0 | matrix multiplication function that supports per-token input quantization and per-column weight quantization | STRING |
| HIGH | test/registered/kernels/ops/moe/test_int8_kernel.py | 0 | this function performs fused moe with per-column int8 quantization using native torch. | STRING |
| HIGH | …t/registered/moe/test_triton_moe_channel_fp8_kernel.py | 0 | this function performs fused moe with per-column int8 quantization using native torch. | STRING |
| HIGH | python/sglang/test/cpu_test_utils.py | 0 | this function performs fused moe with per-column int8 quantization using native torch. | STRING |
| HIGH | …/unit/disaggregation/test_specv2_kvcache_offloading.py | 0 | single-pool double. subclassing the base routes free_full / free_segment / free_segments into free(), so a new free api | STRING |
| HIGH | …st/registered/unit/layers/test_minicpm_sparse_cache.py | 0 | single-pool double. subclassing the base routes free_full / free_segment / free_segments into free(), so a new free api | STRING |
| HIGH | …gistered/unit/mem_cache/test_streaming_session_unit.py | 0 | single-pool double. subclassing the base routes free_full / free_segment / free_segments into free(), so a new free api | STRING |
| HIGH | …it/mem_cache/unified_tree_core_inspection_interface.py | 0 | delete childless tombstone ancestors until a live or locked node is reached. | STRING |
| HIGH | …gistered/unit/mem_cache/unified_tree_core_inspector.py | 0 | delete childless tombstone ancestors until a live or locked node is reached. | STRING |
| HIGH | …glang/srt/mem_cache/unified_cache/unified_tree_core.py | 0 | delete childless tombstone ancestors until a live or locked node is reached. | STRING |
| HIGH | test/registered/cpu/test_qwen3.py | 0 | derives `query`, `key` and `value` tensors from `mixed_qkvzba`. | STRING |
| HIGH | python/sglang/srt/models/qwen3_5.py | 0 | derives `query`, `key` and `value` tensors from `mixed_qkvzba`. | STRING |
| HIGH | python/sglang/srt/models/qwen3_next.py | 0 | derives `query`, `key` and `value` tensors from `mixed_qkvzba`. | STRING |
| HIGH | …t/registered/openai_server/basic/test_openai_server.py | 0 | extract the name, size, price, and color from this product description as a json object: <description> the smarthome min | STRING |
| HIGH | …ateway/e2e_test/chat_completions/test_openai_server.py | 0 | extract the name, size, price, and color from this product description as a json object: <description> the smarthome min | STRING |
| HIGH | examples/runtime/openai_chat_with_response_prefill.py | 0 | extract the name, size, price, and color from this product description as a json object: <description> the smarthome min | STRING |
| HIGH | …st/registered/npu/vlm_models/test_npu_gemma_3_4b_it.py | 0 | testcase: verify that the inference accuracy of the qwen/qwen2.5-vl-3b-instruct model on the mmmu dataset is no less tha | STRING |
| HIGH | …stered/npu/vlm_models/test_npu_qwen3_vl_8b_instruct.py | 0 | testcase: verify that the inference accuracy of the qwen/qwen2.5-vl-3b-instruct model on the mmmu dataset is no less tha | STRING |
| HIGH | …npu/vlm_models/test_npu_qwen3_vl_235b_a22b_instruct.py | 0 | testcase: verify that the inference accuracy of the qwen/qwen2.5-vl-3b-instruct model on the mmmu dataset is no less tha | STRING |
| HIGH | …stered/npu/vlm_models/test_npu_qwen3_vl_4b_instruct.py | 0 | testcase: verify that the inference accuracy of the qwen/qwen2.5-vl-3b-instruct model on the mmmu dataset is no less tha | STRING |
| HIGH | …st/registered/npu/vlm_models/test_npu_mimo_vl_7b_rl.py | 0 | testcase: verify that the inference accuracy of the qwen/qwen2.5-vl-3b-instruct model on the mmmu dataset is no less tha | STRING |
| HIGH | …red/npu/vlm_models/test_npu_qwen2_5_vl_72b_instruct.py | 0 | testcase: verify that the inference accuracy of the qwen/qwen2.5-vl-3b-instruct model on the mmmu dataset is no less tha | STRING |
| HIGH | …ed/npu/vlm_models/test_npu_phi4_multimodal_instruct.py | 0 | testcase: verify that the inference accuracy of the qwen/qwen2.5-vl-3b-instruct model on the mmmu dataset is no less tha | STRING |
| HIGH | …st/registered/npu/vlm_models/test_npu_minicpm_o_2_6.py | 0 | testcase: verify that the inference accuracy of the qwen/qwen2.5-vl-3b-instruct model on the mmmu dataset is no less tha | STRING |
| HIGH | …d/npu/vlm_models/test_npu_qwen3_vl_30b_a3b_instruct.py | 0 | testcase: verify that the inference accuracy of the qwen/qwen2.5-vl-3b-instruct model on the mmmu dataset is no less tha | STRING |
| 261 more matches not shown… | ||||
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| LOW | …shinfer_allreduce_fusion/benchmark_fused_collective.py | 56 | logger = logging.getLogger(__name__) | CODE |
| LOW | benchmark/hicache/bench_mix.py | 20 | logger = logging.getLogger(__name__) | CODE |
| LOW | benchmark/prefill_only/bench_embeddings.py | 35 | logger = logging.getLogger(__name__) | CODE |
| LOW | experimental/sgl-router/tests/e2e/conftest.py | 38 | logger = logging.getLogger(__name__) | CODE |
| LOW | experimental/sgl-router/tests/e2e/infra/gateway.py | 39 | logger = logging.getLogger(__name__) | CODE |
| LOW | experimental/sgl-router/tests/e2e/infra/model_pool.py | 41 | logger = logging.getLogger(__name__) | CODE |
| LOW | …ental/sgl-router/tests/e2e/k8s_integration/conftest.py | 24 | logger = logging.getLogger(__name__) | CODE |
| LOW | …er/tests/e2e/k8s_integration/test_kv_peer_bootstrap.py | 62 | logger = logging.getLogger(__name__) | CODE |
| LOW | …outer/tests/e2e/k8s_integration/test_shutdown_drain.py | 42 | logger = logging.getLogger(__name__) | CODE |
| LOW | …sglang-simulator/src/sglang_simulator/spec/__init__.py | 5 | __all__ = ["AcceleratorInfo", "ModelInfo", "DataType"] | CODE |
| LOW | …ator/src/sglang_simulator/spec/accelerator/__init__.py | 4 | __all__ = ["AcceleratorInfo", "NVIDIA"] | CODE |
| LOW | …-simulator/src/sglang_simulator/spec/model/__init__.py | 3 | __all__ = ["ModelInfo"] | CODE |
| LOW | …/src/sglang_simulator/simulation/benchmark/__init__.py | 4 | __all__ = ["BaseBenchmarkRunner", "BenchmarkConfig"] | CODE |
| LOW | …or/src/sglang_simulator/simulation/manager/__init__.py | 5 | __all__ = ["StateManager", "Envs", "ConfigManager"] | CODE |
| LOW⚡ | …lator/src/sglang_simulator/simulation/manager/state.py | 93 | def set_global_clock(cls, clock: float) -> None: | CODE |
| LOW⚡ | …lator/src/sglang_simulator/simulation/manager/state.py | 97 | def set_current_inference_dur(cls, dur: float) -> None: | CODE |
| LOW | …glang-simulator/src/sglang_simulator/utils/__init__.py | 3 | __all__ = ["get_logger"] | CODE |
| LOW | test/manual/ascend/disaggregation_utils.py | 18 | logger = logging.getLogger(__name__) | CODE |
| LOW⚡ | …d/unit/hardware_backend/mlx/test_attention_patching.py | 84 | def _set_runner_decode_context_defaults(runner) -> None: | CODE |
| LOW⚡ | …d/unit/hardware_backend/mlx/test_attention_patching.py | 91 | def _set_dummy_server_args_for_auxiliary_state_tests() -> None: | CODE |
| LOW | …st/registered/unit/spec/test_adaptive_runtime_state.py | 53 | def set_cuda_graph_bs(self, cuda_graph_bs: list[int] | None) -> None: | CODE |
| LOW⚡ | …it/mem_cache/unified_tree_core_inspection_interface.py | 214 | def update_duplicate_tracking(self, node_id: NodeId) -> None: | CODE |
| LOW⚡ | …gistered/unit/mem_cache/unified_tree_core_inspector.py | 205 | def update_duplicate_tracking(self, node_id: NodeId) -> None: | CODE |
| LOW⚡ | …red/unit/mem_cache/rust_unified_tree_core_inspector.py | 182 | def update_duplicate_tracking(self, node_id: NodeId) -> None: | CODE |
| LOW | …nit/lora/test_experimental_sgl_marlin_multi_prefill.py | 22 | def _set_mapping(mapping: torch.Tensor, num_slots: int, offset: int) -> None: | CODE |
| LOW | test/registered/xpu/test_xpu_flux2_dev.py | 33 | logger = logging.getLogger(__name__) | CODE |
| LOW | test/registered/xpu/test_xpu_zimage_turbo.py | 33 | logger = logging.getLogger(__name__) | CODE |
| LOW | test/registered/npu/interface/test_npu_api_encode.py | 21 | logger = logging.getLogger(__name__) | CODE |
| LOW | …gistered/npu/basic_function/quant/test_npu_gguf_moe.py | 21 | logger = logging.getLogger(__name__) | CODE |
| LOW | …gistered/npu/basic_function/quant/test_npu_gptq_moe.py | 23 | logger = logging.getLogger(__name__) | CODE |
| LOW | …red/npu/basic_function/quant/test_npu_autoround_moe.py | 22 | logger = logging.getLogger(__name__) | CODE |
| LOW | …t/registered/npu/basic_function/quant/test_npu_gguf.py | 19 | logger = logging.getLogger(__name__) | CODE |
| LOW | …d/npu/basic_function/quant/test_npu_autoround_dense.py | 21 | logger = logging.getLogger(__name__) | CODE |
| LOW | …istered/npu/reward_models/test_npu_gemma_2_27b_v0_2.py | 12 | logger = logging.getLogger(__name__) | CODE |
| LOW⚡ | test/registered/observability/test_tracing.py | 46 | logger = logging.getLogger(__name__) | CODE |
| LOW | …egistered/observability/test_tracing_disaggregation.py | 34 | logger = logging.getLogger(__name__) | CODE |
| LOW | test/registered/vlm/test_vision_chunked_prefill.py | 37 | logger = logging.getLogger(__name__) | CODE |
| LOW | test/registered/amd/test_zimage_turbo.py | 24 | logger = logging.getLogger(__name__) | CODE |
| LOW | test/registered/amd/test_wan22_fp8_mla.py | 26 | logger = logging.getLogger(__name__) | CODE |
| LOW | test/registered/rl/test_return_routed_experts.py | 36 | logger = logging.getLogger(__name__) | CODE |
| LOW | test/registered/rl/test_weight_version_spans.py | 121 | def _update_weights_from_disk(self, **fields) -> None: | CODE |
| LOW | test/registered/rl/test_return_indexer_topk.py | 32 | logger = logging.getLogger(__name__) | CODE |
| LOW | python/sglang/__init__.py | 70 | __all__ = [ | CODE |
| LOW | python/sglang/utils.py | 31 | logger = logging.getLogger(__name__) | CODE |
| LOW | python/sglang/_platform_stubs.py | 241 | def set_device(device: Any) -> None: # noqa: ARG001 | CODE |
| LOW | python/sglang/benchmark/dspark_sps_profiler.py | 23 | logger = logging.getLogger(__name__) | CODE |
| LOW | python/sglang/benchmark/dspark_sps_profiler.py | 835 | def set_forced_budget_frac(*, base_url: str, frac: Optional[float]) -> None: | CODE |
| LOW | python/sglang/benchmark/dspark_sts_fit.py | 16 | logger = logging.getLogger(__name__) | CODE |
| LOW | python/sglang/benchmark/datasets/__init__.py | 49 | __all__ = [ | CODE |
| LOW | python/sglang/kernels/kernel_api_logging.py | 66 | def _setup_logger() -> None: | CODE |
| LOW | python/sglang/kernels/__init__.py | 54 | __all__ = [ | CODE |
| LOW | python/sglang/kernels/fused_op.py | 98 | logger = logging.getLogger(__name__) | CODE |
| LOW | python/sglang/kernels/fused_op.py | 237 | def set_fused_op_backend(backend: Optional[KernelBackend]) -> None: | CODE |
| LOW | python/sglang/kernels/numerics.py | 66 | __all__ = [ | CODE |
| LOW | python/sglang/kernels/jit/cute_aot_cache.py | 24 | logger = logging.getLogger(__name__) | CODE |
| LOW | python/sglang/kernels/jit/benchmark/marker.py | 37 | __all__ = [ | CODE |
| LOW | python/sglang/kernels/jit/utils/occupancy.py | 13 | __all__ = ["NoSchedulableClustersError", "get_max_active_clusters"] | CODE |
| LOW | python/sglang/kernels/jit/utils/arch.py | 22 | logger = logging.getLogger(__name__) | CODE |
| LOW | python/sglang/kernels/jit/utils/__init__.py | 26 | __all__ = [ | CODE |
| LOW | python/sglang/kernels/jit/utils/compile/ninja.py | 33 | logger = logging.getLogger(__name__) | CODE |
| 1474 more matches not shown… | ||||
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| LOW | …chmark/kernels/deepseek/benchmark_deepgemm_fp8_gemm.py | 129 | CODE | |
| LOW | …nels/deepseek/benchmark_deepgemm_fp8_gemm_blackwell.py | 46 | CODE | |
| LOW | …nels/deepseek/benchmark_deepgemm_fp8_gemm_blackwell.py | 60 | CODE | |
| LOW | benchmark/hicache/bench_serving.py | 676 | CODE | |
| LOW | benchmark/fla/benchmark_layernorm_gated.py | 15 | CODE | |
| LOW | benchmark/lora/lora_bench.py | 350 | CODE | |
| LOW | test/manual/openai_server/features/test_cache_report.py | 68 | CODE | |
| LOW | test/manual/openai_server/features/test_cache_report.py | 80 | CODE | |
| LOW | test/manual/openai_server/features/test_cache_report.py | 91 | CODE | |
| LOW | …t/manual/entrypoints/http_server/test_abort_request.py | 78 | CODE | |
| LOW | …/kernels/ops/attention/test_flash_attention_4_sm103.py | 50 | CODE | |
| LOW | …stered/kernels/ops/attention/test_flash_attention_4.py | 462 | CODE | |
| LOW | …/kernels/ops/attention/test_flash_attention_4_sm120.py | 187 | CODE | |
| LOW | …tered/kernels/ops/attention/test_flash_mla_backends.py | 247 | CODE | |
| LOW | …st/registered/kernels/ops/kv_canary/test_write_hand.py | 73 | CODE | |
| LOW | …/registered/kernels/ops/kv_canary/test_pipeline_e2e.py | 183 | CODE | |
| LOW | test/registered/kernels/ops/diffusion/test_norm.py | 388 | CODE | |
| LOW | …istered/kernels/ops/moe/test_round_robin_ep_balance.py | 195 | CODE | |
| LOW | test/registered/attention/test_verify_shared_kv.py | 62 | CODE | |
| LOW | test/registered/attention/test_verify_splitkv.py | 85 | CODE | |
| LOW | test/registered/unit/test_model_overrides.py | 591 | CODE | |
| LOW | …ered/unit/layers/test_llama_inkling_parallel_groups.py | 21 | CODE | |
| LOW | …gistered/unit/model_executor/test_pool_configurator.py | 59 | CODE | |
| LOW | …t/registered/unit/sampling/test_sampling_batch_info.py | 685 | CODE | |
| LOW | test/registered/unit/mem_cache/test_registry.py | 35 | CODE | |
| LOW | …stered/unit/mem_cache/test_mooncake_group_semantics.py | 201 | CODE | |
| LOW | …stered/unit/mem_cache/test_mooncake_group_semantics.py | 238 | CODE | |
| LOW | …egistered/unit/mem_cache/test_hicache_file_lru_unit.py | 43 | CODE | |
| LOW | …egistered/unit/mem_cache/test_hicache_file_lru_unit.py | 76 | CODE | |
| LOW | …egistered/unit/mem_cache/test_fused_draft_placement.py | 128 | CODE | |
| LOW | …egistered/unit/mem_cache/test_page_interleave_shard.py | 89 | CODE | |
| LOW | …ed/unit/mem_cache/test_unified_radix_cache_unittest.py | 6294 | CODE | |
| LOW | …egistered/unit/layer_boundary/test_reduction_fusion.py | 402 | CODE | |
| LOW | test/registered/kv_canary/test_self_unit_violation.py | 25 | CODE | |
| LOW | test/registered/cpu/test_moe.py | 40 | CODE | |
| LOW | test/registered/cpu/test_extend.py | 184 | CODE | |
| LOW | …t/registered/openai_server/basic/test_openai_server.py | 619 | CODE | |
| LOW | …t/registered/openai_server/basic/test_openai_server.py | 655 | CODE | |
| LOW | …/openai_server/validation/test_large_max_new_tokens.py | 82 | CODE | |
| LOW | …t/registered/debug_utils/comparator/test_entrypoint.py | 2705 | CODE | |
| LOW | …g_utils/comparator/tensor_comparator/test_formatter.py | 301 | CODE | |
| LOW | test/registered/basic_perf/test_lora_latency.py | 130 | CODE | |
| LOW | test/registered/bench_fn/test_benchmark_datasets_api.py | 1036 | CODE | |
| LOW | test/registered/rl/test_update_weights_from_tensor.py | 234 | CODE | |
| LOW | test/registered/rl/test_update_weights_from_tensor.py | 242 | CODE | |
| LOW | test/registered/rl/test_update_weights_from_tensor.py | 257 | CODE | |
| LOW | python/sglang/benchmark/serving.py | 1377 | CODE | |
| LOW | python/sglang/benchmark/one_batch_server.py | 599 | CODE | |
| LOW | python/sglang/kernels/jit/benchmark/marker.py | 487 | CODE | |
| LOW | python/sglang/kernels/jit/utils/compile/loader.py | 48 | CODE | |
| LOW | …hon/sglang/kernels/ops/attention/flash_attention_v4.py | 48 | CODE | |
| LOW | …hon/sglang/kernels/ops/attention/flash_attention_v4.py | 196 | CODE | |
| LOW | …hon/sglang/kernels/ops/attention/flash_attention_v4.py | 339 | CODE | |
| LOW | python/sglang/kernels/ops/attention/decode_attention.py | 512 | CODE | |
| LOW | python/sglang/kernels/ops/attention/decode_attention.py | 770 | CODE | |
| LOW | python/sglang/kernels/ops/attention/decode_attention.py | 1098 | CODE | |
| LOW | python/sglang/kernels/ops/attention/decode_attention.py | 1158 | CODE | |
| LOW | …/kernels/ops/attention/cutedsl_fp8_paged_mqa_logits.py | 201 | CODE | |
| LOW | python/sglang/kernels/ops/attention/verify_mla.py | 725 | CODE | |
| LOW | python/sglang/kernels/ops/attention/deep_select.py | 190 | CODE | |
| 693 more matches not shown… | ||||
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| LOW⚡ | benchmark/bench_linear_attention/bench_gdn_decode.py | 104 | # Step 1: split (same as forward_decode) | COMMENT |
| LOW⚡ | benchmark/bench_linear_attention/bench_gdn_decode.py | 107 | # Step 2: view + reshape (same as forward_decode) | COMMENT |
| LOW⚡ | benchmark/bench_linear_attention/bench_gdn_decode.py | 112 | # Step 3: fused gating + recurrent update | COMMENT |
| LOW⚡ | …mark/bench_linear_attention/bench_fused_gate_cumsum.py | 79 | # Step 1: gate activation using torch ops | COMMENT |
| LOW⚡ | …mark/bench_linear_attention/bench_fused_gate_cumsum.py | 87 | # Step 2: chunk-local cumsum | COMMENT |
| LOW⚡ | …rimental/sgl-router/tests/e2e/k8s_integration/setup.sh | 38 | # Step 1: Create kind cluster (idempotent) | COMMENT |
| LOW⚡ | …rimental/sgl-router/tests/e2e/k8s_integration/setup.sh | 50 | # Step 2: Build Docker images (unless SKIP_DOCKER_BUILD=1) | COMMENT |
| LOW⚡ | …rimental/sgl-router/tests/e2e/k8s_integration/setup.sh | 84 | # Step 3: Load images into kind | COMMENT |
| LOW⚡ | …rimental/sgl-router/tests/e2e/k8s_integration/setup.sh | 92 | # Step 4: Apply namespace and RBAC | COMMENT |
| LOW⚡ | …rimental/sgl-router/tests/e2e/k8s_integration/setup.sh | 99 | # Step 5: Deploy 3 fake-worker replicas behind a Service | COMMENT |
| LOW⚡ | …rimental/sgl-router/tests/e2e/k8s_integration/setup.sh | 154 | # Step 6: Deploy sgl-router. It is configured entirely via CLI flags in | COMMENT |
| LOW⚡ | test/manual/test_tokenizer_manager.py | 341 | # Step 1: Detect format | COMMENT |
| LOW⚡ | test/manual/test_tokenizer_manager.py | 347 | # Step 2: Prepare input | COMMENT |
| LOW⚡ | test/manual/test_tokenizer_manager.py | 353 | # Step 3: Extract results (simulated tokenizer output) | COMMENT |
| LOW⚡ | test/manual/test_tokenizer_manager.py | 372 | # Step 1: Detect format | COMMENT |
| LOW⚡ | test/manual/test_tokenizer_manager.py | 378 | # Step 2: Prepare input | COMMENT |
| LOW⚡ | test/manual/test_tokenizer_manager.py | 384 | # Step 3: Extract results (simulated tokenizer output for cross-encoder) | COMMENT |
| LOW⚡ | test/manual/test_tokenizer_manager.py | 401 | # Step 1: Detect format | COMMENT |
| LOW⚡ | test/manual/test_tokenizer_manager.py | 407 | # Step 2: Prepare input | COMMENT |
| LOW⚡ | test/manual/test_tokenizer_manager.py | 413 | # Step 3: Extract results (simulated tokenizer output) | COMMENT |
| LOW | test/manual/dsv4/test_fused_compress_attn_hip.py | 311 | # Step 1: write current tokens to state (same for both paths) | COMMENT |
| LOW⚡ | …st/registered/kernels/ops/kv_canary/test_write_hand.py | 234 | # Step 1: write a chain from seed slot=7 → newly written slot=2. Then run verify with prev=7 and | COMMENT |
| LOW⚡ | …st/registered/kernels/ops/kv_canary/test_write_hand.py | 903 | # Step 1: build one RealKvSource with read_bytes=16 and a fixed byte pattern at slot 0. | COMMENT |
| LOW⚡ | …st/registered/kernels/ops/kv_canary/test_write_hand.py | 943 | # Step 3: run write kernel on slot 0 with the given mode. | COMMENT |
| LOW⚡ | …st/registered/kernels/ops/kv_canary/test_write_hand.py | 953 | # Step 4: assert stored real_kv_hash equals the hand-computed hex literal. | COMMENT |
| LOW | …st/registered/kernels/ops/kv_canary/test_write_hand.py | 256 | # Step 2: verify slot[2] with prev=7 — expects no violation. | COMMENT |
| LOW | …st/registered/kernels/ops/kv_canary/test_write_hand.py | 390 | # Step 1: compute the expected stored prev_hash sequence in pure Python via splitmix64. | COMMENT |
| LOW | …st/registered/kernels/ops/kv_canary/test_write_hand.py | 407 | # Step 2: verify every slot's stored 4 fields match the hardcoded expected sequence. | COMMENT |
| LOW | …st/registered/kernels/ops/kv_canary/test_write_hand.py | 925 | # Step 2: verify hand-computed fold matches the hex literal. | COMMENT |
| LOW⚡ | …t/registered/kernels/ops/kv_canary/test_verify_hand.py | 236 | # Step 1: stamp slot 5 such that stored.prev_hash already equals splitmix64(consts.CANARY_CHAIN_ANCHOR). | COMMENT |
| LOW⚡ | …t/registered/kernels/ops/kv_canary/test_verify_hand.py | 240 | # Step 2: a single-entry plan with prev_slot_idx = -1 should record no violation. | COMMENT |
| LOW⚡ | …t/registered/kernels/ops/kv_canary/test_verify_hand.py | 274 | # Step 1: compute the expected stored prev_hash sequence in pure Python via splitmix64. | COMMENT |
| LOW⚡ | …t/registered/kernels/ops/kv_canary/test_verify_hand.py | 284 | # Step 2: stamp each slot manually with the hardcoded expected prev_hash. | COMMENT |
| LOW⚡ | …t/registered/kernels/ops/kv_canary/test_verify_hand.py | 309 | # Step 4: also independently confirm the *stored* prev_hash at each slot matches the hardcoded sequence. | COMMENT |
| LOW⚡ | …t/registered/kernels/ops/kv_canary/test_verify_hand.py | 827 | # Step 1: build one RealKvSource with read_bytes=16 and a fixed byte pattern at slot 1. | COMMENT |
| LOW⚡ | …t/registered/kernels/ops/kv_canary/test_verify_hand.py | 865 | # Step 2: verify hand-computed fold matches the hex literal. | COMMENT |
| LOW⚡ | …t/registered/kernels/ops/kv_canary/test_verify_hand.py | 868 | # Step 3: stamp slot 1 with a chain-head entry whose real_kv_hash equals the expected value. | COMMENT |
| LOW⚡ | …t/registered/kernels/ops/kv_canary/test_verify_hand.py | 876 | # Step 4: 1-entry verify plan; no violation because stored matches recomputed. | COMMENT |
| LOW | …t/registered/kernels/ops/kv_canary/test_verify_hand.py | 298 | # Step 3: verify the 5-step chain — no violation expected and the ref vs CUDA state byte-equal. | COMMENT |
| LOW | …t/registered/kernels/ops/kv_canary/test_verify_hand.py | 888 | # Step 5: mutate one byte in the source so the recomputed hash diverges from stored. | COMMENT |
| LOW⚡ | test/registered/kernels/ops/kv_canary/test_plan_hand.py | 117 | # Step 1: build a one-req batch with no prefix and 5 extend tokens. | COMMENT |
| LOW⚡ | test/registered/kernels/ops/kv_canary/test_plan_hand.py | 957 | # Step 1: pin (prefix, extend) per req. | COMMENT |
| LOW | test/registered/kernels/ops/kv_canary/test_plan_hand.py | 132 | # Step 2: prefix=0 → no verify entries; seed = -1 because prefix==0. | COMMENT |
| LOW | test/registered/kernels/ops/kv_canary/test_plan_hand.py | 976 | # Step 2: hand-compute expected write_offsets (exclusive cumsum of extends) and verify_num_valid. | COMMENT |
| LOW | …/registered/kernels/ops/kv_canary/test_pipeline_e2e.py | 598 | # Step 1: pre-pollute canary_buf slots [0..n_slots) with wrong prev_hash so verify fires n_slots violations. | COMMENT |
| LOW | …/registered/kernels/ops/kv_canary/test_pipeline_e2e.py | 613 | # Step 2: run real pipeline (plan + no write + verify); overflow ring capacity=4 with all n_slots violations. | COMMENT |
| LOW⚡ | …/registered/kernels/ops/kv_canary/test_pipeline_e2e.py | 684 | # Step 3: write_index byte-equal; ring contents relaxed (atomic order not guaranteed under overflow). | COMMENT |
| LOW⚡ | test/registered/kernels/ops/kvcache/test_hisparse.py | 453 | # Step 1: touch tokens [4, 2], so LRU becomes [0, 3, 1, 2]. | COMMENT |
| LOW⚡ | test/registered/kernels/ops/kvcache/test_hisparse.py | 454 | # Step 2: query token 6, which is a miss. | COMMENT |
| LOW⚡ | …gistered/kernels/ops/quantization/test_hadamard_jit.py | 210 | # Step 1: apply H_N (standard power-of-2 Hadamard) along the N dimension | COMMENT |
| LOW⚡ | …gistered/kernels/ops/quantization/test_hadamard_jit.py | 214 | # Step 2: apply H_M (special ±1 matrix) along the M dimension | COMMENT |
| LOW⚡ | test/registered/unit/managers/test_hisparse_unit.py | 473 | # Step 1: load the first TOP_K positions from host (no newest token — | COMMENT |
| LOW⚡ | test/registered/unit/managers/test_hisparse_unit.py | 482 | # Step 2: half overlap (hit) + half new (miss). | COMMENT |
| LOW⚡ | test/registered/unit/parser/test_reasoning_parser.py | 1029 | # Step 1: Send partial end tag when not in reasoning mode | COMMENT |
| LOW⚡ | test/registered/unit/parser/test_reasoning_parser.py | 1035 | # Step 2: Send normal text that doesn't complete the end tag | COMMENT |
| LOW⚡ | …egistered/unit/mem_cache/test_decode_radix_lock_ref.py | 265 | # Step 2: checkpoint (dec old lock, inc new lock) | COMMENT |
| LOW⚡ | …egistered/unit/mem_cache/test_decode_radix_lock_ref.py | 268 | # Step 3: release_kv_cache (insert, free the rest, dec lock) | COMMENT |
| LOW⚡ | …egistered/unit/mem_cache/test_decode_radix_lock_ref.py | 316 | # Step 2: checkpoint (dec root=no-op, inc new leaf) | COMMENT |
| LOW⚡ | …egistered/unit/mem_cache/test_decode_radix_lock_ref.py | 319 | # Step 3: cache_finished_req (dec leaf) | COMMENT |
| LOW | …egistered/unit/mem_cache/test_decode_radix_lock_ref.py | 248 | # Step 1: lock (pop_preallocated locks the matched node) | COMMENT |
| 288 more matches not shown… | ||||
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| LOW⚡ | benchmark/bench_linear_attention/bench_gdn_decode.py | 104 | # Step 1: split (same as forward_decode) | COMMENT |
| LOW⚡ | benchmark/bench_linear_attention/bench_gdn_decode.py | 107 | # Step 2: view + reshape (same as forward_decode) | COMMENT |
| LOW⚡ | benchmark/bench_linear_attention/bench_gdn_decode.py | 112 | # Step 3: fused gating + recurrent update | COMMENT |
| LOW⚡ | …mark/bench_linear_attention/bench_fused_gate_cumsum.py | 79 | # Step 1: gate activation using torch ops | COMMENT |
| LOW⚡ | …mark/bench_linear_attention/bench_fused_gate_cumsum.py | 87 | # Step 2: chunk-local cumsum | COMMENT |
| LOW⚡ | …rimental/sgl-router/tests/e2e/k8s_integration/setup.sh | 38 | # Step 1: Create kind cluster (idempotent) | COMMENT |
| LOW⚡ | …rimental/sgl-router/tests/e2e/k8s_integration/setup.sh | 50 | # Step 2: Build Docker images (unless SKIP_DOCKER_BUILD=1) | COMMENT |
| LOW⚡ | …rimental/sgl-router/tests/e2e/k8s_integration/setup.sh | 84 | # Step 3: Load images into kind | COMMENT |
| LOW⚡ | …rimental/sgl-router/tests/e2e/k8s_integration/setup.sh | 92 | # Step 4: Apply namespace and RBAC | COMMENT |
| LOW⚡ | …rimental/sgl-router/tests/e2e/k8s_integration/setup.sh | 99 | # Step 5: Deploy 3 fake-worker replicas behind a Service | COMMENT |
| LOW⚡ | …rimental/sgl-router/tests/e2e/k8s_integration/setup.sh | 154 | # Step 6: Deploy sgl-router. It is configured entirely via CLI flags in | COMMENT |
| LOW⚡ | test/manual/test_tokenizer_manager.py | 341 | # Step 1: Detect format | COMMENT |
| LOW⚡ | test/manual/test_tokenizer_manager.py | 347 | # Step 2: Prepare input | COMMENT |
| LOW⚡ | test/manual/test_tokenizer_manager.py | 353 | # Step 3: Extract results (simulated tokenizer output) | COMMENT |
| LOW⚡ | test/manual/test_tokenizer_manager.py | 372 | # Step 1: Detect format | COMMENT |
| LOW⚡ | test/manual/test_tokenizer_manager.py | 378 | # Step 2: Prepare input | COMMENT |
| LOW⚡ | test/manual/test_tokenizer_manager.py | 384 | # Step 3: Extract results (simulated tokenizer output for cross-encoder) | COMMENT |
| LOW⚡ | test/manual/test_tokenizer_manager.py | 401 | # Step 1: Detect format | COMMENT |
| LOW⚡ | test/manual/test_tokenizer_manager.py | 407 | # Step 2: Prepare input | COMMENT |
| LOW⚡ | test/manual/test_tokenizer_manager.py | 413 | # Step 3: Extract results (simulated tokenizer output) | COMMENT |
| LOW | test/manual/dsv4/test_fused_compress_attn_hip.py | 311 | # Step 1: write current tokens to state (same for both paths) | COMMENT |
| LOW⚡ | …st/registered/kernels/ops/kv_canary/test_write_hand.py | 234 | # Step 1: write a chain from seed slot=7 → newly written slot=2. Then run verify with prev=7 and | COMMENT |
| LOW⚡ | …st/registered/kernels/ops/kv_canary/test_write_hand.py | 903 | # Step 1: build one RealKvSource with read_bytes=16 and a fixed byte pattern at slot 0. | COMMENT |
| LOW⚡ | …st/registered/kernels/ops/kv_canary/test_write_hand.py | 943 | # Step 3: run write kernel on slot 0 with the given mode. | COMMENT |
| LOW⚡ | …st/registered/kernels/ops/kv_canary/test_write_hand.py | 953 | # Step 4: assert stored real_kv_hash equals the hand-computed hex literal. | COMMENT |
| LOW | …st/registered/kernels/ops/kv_canary/test_write_hand.py | 256 | # Step 2: verify slot[2] with prev=7 — expects no violation. | COMMENT |
| LOW | …st/registered/kernels/ops/kv_canary/test_write_hand.py | 390 | # Step 1: compute the expected stored prev_hash sequence in pure Python via splitmix64. | COMMENT |
| LOW | …st/registered/kernels/ops/kv_canary/test_write_hand.py | 407 | # Step 2: verify every slot's stored 4 fields match the hardcoded expected sequence. | COMMENT |
| LOW | …st/registered/kernels/ops/kv_canary/test_write_hand.py | 925 | # Step 2: verify hand-computed fold matches the hex literal. | COMMENT |
| LOW⚡ | …t/registered/kernels/ops/kv_canary/test_verify_hand.py | 236 | # Step 1: stamp slot 5 such that stored.prev_hash already equals splitmix64(consts.CANARY_CHAIN_ANCHOR). | COMMENT |
| LOW⚡ | …t/registered/kernels/ops/kv_canary/test_verify_hand.py | 240 | # Step 2: a single-entry plan with prev_slot_idx = -1 should record no violation. | COMMENT |
| LOW⚡ | …t/registered/kernels/ops/kv_canary/test_verify_hand.py | 274 | # Step 1: compute the expected stored prev_hash sequence in pure Python via splitmix64. | COMMENT |
| LOW⚡ | …t/registered/kernels/ops/kv_canary/test_verify_hand.py | 284 | # Step 2: stamp each slot manually with the hardcoded expected prev_hash. | COMMENT |
| LOW⚡ | …t/registered/kernels/ops/kv_canary/test_verify_hand.py | 309 | # Step 4: also independently confirm the *stored* prev_hash at each slot matches the hardcoded sequence. | COMMENT |
| LOW⚡ | …t/registered/kernels/ops/kv_canary/test_verify_hand.py | 827 | # Step 1: build one RealKvSource with read_bytes=16 and a fixed byte pattern at slot 1. | COMMENT |
| LOW⚡ | …t/registered/kernels/ops/kv_canary/test_verify_hand.py | 865 | # Step 2: verify hand-computed fold matches the hex literal. | COMMENT |
| LOW⚡ | …t/registered/kernels/ops/kv_canary/test_verify_hand.py | 868 | # Step 3: stamp slot 1 with a chain-head entry whose real_kv_hash equals the expected value. | COMMENT |
| LOW⚡ | …t/registered/kernels/ops/kv_canary/test_verify_hand.py | 876 | # Step 4: 1-entry verify plan; no violation because stored matches recomputed. | COMMENT |
| LOW | …t/registered/kernels/ops/kv_canary/test_verify_hand.py | 298 | # Step 3: verify the 5-step chain — no violation expected and the ref vs CUDA state byte-equal. | COMMENT |
| LOW | …t/registered/kernels/ops/kv_canary/test_verify_hand.py | 888 | # Step 5: mutate one byte in the source so the recomputed hash diverges from stored. | COMMENT |
| LOW⚡ | test/registered/kernels/ops/kv_canary/test_plan_hand.py | 117 | # Step 1: build a one-req batch with no prefix and 5 extend tokens. | COMMENT |
| LOW⚡ | test/registered/kernels/ops/kv_canary/test_plan_hand.py | 957 | # Step 1: pin (prefix, extend) per req. | COMMENT |
| LOW | test/registered/kernels/ops/kv_canary/test_plan_hand.py | 132 | # Step 2: prefix=0 → no verify entries; seed = -1 because prefix==0. | COMMENT |
| LOW | test/registered/kernels/ops/kv_canary/test_plan_hand.py | 976 | # Step 2: hand-compute expected write_offsets (exclusive cumsum of extends) and verify_num_valid. | COMMENT |
| LOW | …/registered/kernels/ops/kv_canary/test_pipeline_e2e.py | 598 | # Step 1: pre-pollute canary_buf slots [0..n_slots) with wrong prev_hash so verify fires n_slots violations. | COMMENT |
| LOW | …/registered/kernels/ops/kv_canary/test_pipeline_e2e.py | 613 | # Step 2: run real pipeline (plan + no write + verify); overflow ring capacity=4 with all n_slots violations. | COMMENT |
| LOW⚡ | …/registered/kernels/ops/kv_canary/test_pipeline_e2e.py | 684 | # Step 3: write_index byte-equal; ring contents relaxed (atomic order not guaranteed under overflow). | COMMENT |
| LOW⚡ | test/registered/kernels/ops/kvcache/test_hisparse.py | 453 | # Step 1: touch tokens [4, 2], so LRU becomes [0, 3, 1, 2]. | COMMENT |
| LOW⚡ | test/registered/kernels/ops/kvcache/test_hisparse.py | 454 | # Step 2: query token 6, which is a miss. | COMMENT |
| LOW⚡ | …gistered/kernels/ops/quantization/test_hadamard_jit.py | 210 | # Step 1: apply H_N (standard power-of-2 Hadamard) along the N dimension | COMMENT |
| LOW⚡ | …gistered/kernels/ops/quantization/test_hadamard_jit.py | 214 | # Step 2: apply H_M (special ±1 matrix) along the M dimension | COMMENT |
| LOW⚡ | test/registered/unit/managers/test_hisparse_unit.py | 473 | # Step 1: load the first TOP_K positions from host (no newest token — | COMMENT |
| LOW⚡ | test/registered/unit/managers/test_hisparse_unit.py | 482 | # Step 2: half overlap (hit) + half new (miss). | COMMENT |
| LOW⚡ | test/registered/unit/parser/test_reasoning_parser.py | 1029 | # Step 1: Send partial end tag when not in reasoning mode | COMMENT |
| LOW⚡ | test/registered/unit/parser/test_reasoning_parser.py | 1035 | # Step 2: Send normal text that doesn't complete the end tag | COMMENT |
| LOW⚡ | …egistered/unit/mem_cache/test_decode_radix_lock_ref.py | 265 | # Step 2: checkpoint (dec old lock, inc new lock) | COMMENT |
| LOW⚡ | …egistered/unit/mem_cache/test_decode_radix_lock_ref.py | 268 | # Step 3: release_kv_cache (insert, free the rest, dec lock) | COMMENT |
| LOW⚡ | …egistered/unit/mem_cache/test_decode_radix_lock_ref.py | 316 | # Step 2: checkpoint (dec root=no-op, inc new leaf) | COMMENT |
| LOW⚡ | …egistered/unit/mem_cache/test_decode_radix_lock_ref.py | 319 | # Step 3: cache_finished_req (dec leaf) | COMMENT |
| LOW | …egistered/unit/mem_cache/test_decode_radix_lock_ref.py | 248 | # Step 1: lock (pop_preallocated locks the matched node) | COMMENT |
| 230 more matches not shown… | ||||
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| MEDIUM | benchmark/hicache/bench_mix.py | 24 | # Create a lock for thread-safe debug log writing | COMMENT |
| MEDIUM | benchmark/prefill_only/bench_score.py | 157 | # Create the request builder function with shared tokenizer | COMMENT |
| MEDIUM⚡ | benchmark/prefill_only/util.py | 173 | # Create the full request list by cycling through unique requests | COMMENT |
| MEDIUM | experimental/sgl-router/tests/e2e/infra/gateway.py | 43 | # This file is at `experimental/sgl-router/tests/e2e/infra/gateway.py`, | COMMENT |
| MEDIUM | test/manual/test_weight_validation.py | 75 | # Create a minimal valid safetensors file | COMMENT |
| MEDIUM | test/manual/test_weight_validation.py | 186 | # Create a broken symlink for the index file | COMMENT |
| MEDIUM | test/manual/test_modelopt_fp8kvcache.py | 15 | # Create a ModelOptFp8Config object | COMMENT |
| MEDIUM | test/manual/test_config_integration.py | 27 | # Create a temporary config file | COMMENT |
| MEDIUM | test/manual/test_config_integration.py | 82 | # Create a temporary config file | COMMENT |
| MEDIUM | test/manual/test_config_integration.py | 113 | # Create a temporary config file | COMMENT |
| MEDIUM | test/manual/test_async_dynamic_batch_tokenizer.py | 54 | # Create a proper BatchEncoding-like object that supports dict operations | COMMENT |
| MEDIUM | test/manual/test_async_dynamic_batch_tokenizer.py | 193 | # Create a new async tokenizer with a failing tokenizer | COMMENT |
| MEDIUM | test/manual/attention/test_flashattn_backend.py | 65 | # Create a large enough req_to_token_pool to fit the test usage. | COMMENT |
| MEDIUM | test/manual/attention/test_flashattn_mla_backend.py | 55 | # Create a proper req_to_token_pool with the req_to_token attribute | COMMENT |
| MEDIUM⚡ | …ual/lang_frontend/test_separate_reasoning_execution.py | 72 | # Create a mock backend to avoid AttributeError in __del__ | COMMENT |
| MEDIUM⚡ | …ual/lang_frontend/test_separate_reasoning_execution.py | 75 | # Create a StreamExecutor with necessary setup | COMMENT |
| MEDIUM | …ual/lang_frontend/test_separate_reasoning_execution.py | 106 | # Create a gen expression and a separate_reasoning expression | COMMENT |
| MEDIUM | test/manual/debug_utils/run_with_retry.py | 37 | # Create a TestFile with a reasonable estimated time | COMMENT |
| MEDIUM | …t/registered/kernels/ops/attention/test_dsa_indexer.py | 80 | # Create a simple page table for testing | COMMENT |
| MEDIUM | …t/registered/kernels/ops/attention/test_dsa_indexer.py | 96 | # Create a simple page table for testing with page size 1 | COMMENT |
| MEDIUM | …st/registered/attention/test_wave_attention_kernels.py | 189 | # Define the varying parameter values | COMMENT |
| MEDIUM | …/registered/attention/test_triton_attention_kernels.py | 316 | # Define the varying parameter values | COMMENT |
| MEDIUM | …egistered/attention/test_normal_decode_set_metadata.py | 149 | # Create a simple SWA KV pool for testing | COMMENT |
| MEDIUM | …egistered/attention/test_normal_decode_set_metadata.py | 170 | # Create a minimal mock that inherits from SWAKVPool to pass isinstance check | COMMENT |
| MEDIUM | test/registered/unit/distributed/test_parallel_state.py | 326 | # Create a mock group object | COMMENT |
| MEDIUM | …stered/unit/function_call/test_function_call_parser.py | 946 | # Create a concrete implementation of BaseFormatDetector for testing | COMMENT |
| MEDIUM | …tered/unit/mem_cache/test_radix_cache_slru_accuracy.py | 49 | # Create a cache with the memory pools | COMMENT |
| MEDIUM | …ed/unit/mem_cache/test_unified_radix_cache_unittest.py | 3288 | # Create a chain: root -> A -> B -> C (3 levels) | COMMENT |
| MEDIUM | …t/registered/unit/configs/test_quant_config_parsing.py | 23 | # Create a raw instance | COMMENT |
| MEDIUM | …egistered/unit/entrypoints/openai/test_serving_chat.py | 1764 | # Create a mock conversation object that will be returned by generate_chat_conv | COMMENT |
| MEDIUM | …egistered/unit/entrypoints/openai/test_serving_chat.py | 2943 | # Create a mock conversation object | COMMENT |
| MEDIUM | test/registered/cpu/test_mamba.py | 24 | """This function is intended to align with the l2norm implementation in the FLA library.""" | STRING |
| MEDIUM | …t/registered/openai_server/basic/test_openai_server.py | 595 | # This class is a Chat gate; Completion already has its own Rust matrix. | COMMENT |
| MEDIUM | …registered/hicache/test_hicache_storage_3fs_backend.py | 26 | # Create a temporary JSON config file for HF3FS | COMMENT |
| MEDIUM | test/registered/rl/test_update_weights_from_tensor.py | 140 | # Create a small set of parameters for testing | COMMENT |
| MEDIUM⚡ | test/registered/rl/test_update_weights_from_tensor.py | 154 | # Create a flattened bucket | COMMENT |
| MEDIUM⚡ | test/registered/rl/test_update_weights_from_tensor.py | 161 | # Create the dict format expected by _update_weights_from_flattened_bucket | COMMENT |
| MEDIUM⚡ | test/registered/rl/test_update_weights_from_tensor.py | 171 | # Create a list where each rank contains the same serialized data | COMMENT |
| MEDIUM | python/sglang/benchmark/serving.py | 1453 | # Create a temporary DatasetRow object for warmup | COMMENT |
| MEDIUM | python/sglang/benchmark/serving.py | 1469 | # Create the test input once | COMMENT |
| MEDIUM | python/sglang/benchmark/serving.py | 2619 | # Create a mutually exclusive group for profiling URLs | COMMENT |
| MEDIUM | python/sglang/kernels/ops/attention/deepseek_v4_rope.py | 12 | # This module is imported during model-registry discovery. Keep it free of | COMMENT |
| MEDIUM | …sglang/kernels/ops/attention/dsa/index_buf_accessor.py | 664 | # Define the token range within the block and the K dimension range handled by the thread. | COMMENT |
| MEDIUM | python/sglang/kernels/ops/attention/fla/kda.py | 4 | # This file contains code copied from the flash-linear-attention project. | COMMENT |
| MEDIUM⚡ | python/sglang/kernels/ops/layernorm/mhc.py | 120 | # This module is imported during model-registry discovery. Do not import the real | COMMENT |
| MEDIUM | …nels/ops/quantization/nvfp4_gemm_swiglu_nvfp4_quant.py | 115 | """This class implements batched matrix multiplication (C = A x SFA x B x SFB) with support for various data types | STRING |
| MEDIUM | python/sglang/kernels/ops/quantization/awq_triton.py | 152 | # Create the necessary shifts to use to unpack. | COMMENT |
| MEDIUM | …thon/sglang/kernels/ops/speculative/ngram_embedding.py | 135 | # Create an empty tensor for ignore_tokens | COMMENT |
| MEDIUM | …/sglang/kernels/kda_kernels/qwen3x_nvfp4_gemm_sm120.py | 34 | # This file is ported from the CUTLASS dense block-scaled GEMM example and | COMMENT |
| MEDIUM | …els/aot/benchmark/bench_amd_deterministic_allreduce.py | 127 | # Create a gloo group for custom AR (it requires non-NCCL backend) | COMMENT |
| MEDIUM | python/sglang/test/cpu_test_utils.py | 478 | # Create a new tensor with alternating values | COMMENT |
| MEDIUM | python/sglang/test/cpu_test_utils.py | 501 | # Create a tensor with the E2M1 values | COMMENT |
| MEDIUM | python/sglang/srt/layers/rocm_linear_utils.py | 18 | # This module is imported wherever AITER is on, gfx942 included, but the fused | COMMENT |
| MEDIUM | python/sglang/srt/layers/attention/local_attention.py | 331 | # Create a block_table for the local attention blocks | COMMENT |
| MEDIUM | python/sglang/srt/layers/quantization/fp8.py | 642 | # Create the weight | COMMENT |
| MEDIUM | python/sglang/srt/layers/quantization/mxfp4_tensor.py | 90 | # Create a new tensor with alternating values | COMMENT |
| MEDIUM | python/sglang/srt/layers/quantization/mxfp4_tensor.py | 113 | # Create a tensor with the E2M1 values | COMMENT |
| MEDIUM | python/sglang/srt/layers/quantization/mxfp4.py | 804 | # Create a new shape with pairs swapped along specified axis | COMMENT |
| MEDIUM | python/sglang/srt/eplb/eplb_algorithms/deepseek.py | 1 | # This file is copied from https://github.com/deepseek-ai/EPLB/blob/main/eplb.py since that one is not a pypi package | COMMENT |
| MEDIUM | python/sglang/srt/eplb/eplb_algorithms/deepseek_vec.py | 1 | # This file is copied from https://github.com/deepseek-ai/EPLB/blob/main/eplb.py since that one is not a pypi package | COMMENT |
| 62 more matches not shown… | ||||
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| LOW | …shinfer_allreduce_fusion/benchmark_fused_collective.py | 1162 | # Check if running with torchrun (required for collective operations) | STRING |
| LOW⚡ | benchmark/prefill_only/util.py | 163 | # Check if we have any valid requests | COMMENT |
| LOW | benchmark/boolq/bench_sglang.py | 89 | # Print results | COMMENT |
| LOW | test/manual/test_schedule_policy.py | 93 | # Check if FCFS keeps the original order | COMMENT |
| LOW | test/manual/test_schedule_policy.py | 118 | # Check if priority enabled fcfs ordering is applied. | COMMENT |
| LOW | test/manual/test_schedule_policy.py | 144 | # Check if priority enabled fcfs ordering is applied. | COMMENT |
| LOW | test/manual/test_schedule_policy.py | 166 | # Check if priority enabled fcfs ordering is applied. | COMMENT |
| LOW | test/manual/test_schedule_policy.py | 190 | # Check if priority enabled fcfs ordering is applied. | COMMENT |
| LOW | test/manual/test_schedule_policy.py | 216 | # Check if priority enabled fcfs ordering is applied. | COMMENT |
| LOW | …nual/kv_transfer/test_mooncake_transfer_engine_init.py | 204 | # Check if mooncake should be used | COMMENT |
| LOW | test/registered/kernels/ops/attention/test_pos_enc.py | 479 | # Print results | COMMENT |
| LOW | …mmunication/test_amd_deterministic_custom_allreduce.py | 93 | # Check if inputs fit in buffer | COMMENT |
| LOW | …tered/unit/mem_cache/test_radix_cache_slru_accuracy.py | 101 | # Check if the frequently accessed key-value is still present | COMMENT |
| LOW | …tered/unit/mem_cache/test_radix_cache_slru_accuracy.py | 107 | # Check if the first low-frequency key-value has been evicted | COMMENT |
| LOW | …tered/unit/mem_cache/test_radix_cache_slru_accuracy.py | 119 | # Check if the tensor is empty, which indicates the key was not found (evicted) | COMMENT |
| LOW | …registered/unit/model_loader/test_presharded_loader.py | 283 | # Write the file with stored_keys mapped to tensor content. | COMMENT |
| LOW | …d/unit/batch_invariant_ops/test_batch_invariant_ops.py | 53 | # Check if results are identical | COMMENT |
| LOW | …d/unit/batch_invariant_ops/test_batch_invariant_ops.py | 208 | # Check if results are identical | COMMENT |
| LOW⚡ | …registered/disaggregation/test_disaggregation_basic.py | 213 | # Check if servers are still healthy | COMMENT |
| LOW | …t/registered/openai_server/basic/test_openai_server.py | 92 | # Check if ignore_eos resulted in more tokens or exactly max_tokens | COMMENT |
| LOW | …red/npu/interface/test_npu_openai_server_ignore_eos.py | 84 | # Check if ignore_eos resulted in more tokens or exactly max_tokens | COMMENT |
| LOW | …pu/basic_function/offloading/test_npu_offload_modes.py | 69 | # Check if server is running (basic functionality test) | COMMENT |
| LOW | test/registered/observability/test_tracing.py | 80 | # Check if it's a RequestStageConfig (has stage_name and level attributes) | COMMENT |
| LOW⚡ | …stered/amd/disaggregation/test_disaggregation_basic.py | 324 | # Check if servers are still healthy | COMMENT |
| LOW | …t/registered/amd/accuracy/mi30x/test_gsm8k_eval_amd.py | 281 | # Print results | COMMENT |
| LOW | …gistered/amd/accuracy/mi30x/test_vlms_mmmu_eval_amd.py | 258 | # Print results | COMMENT |
| LOW | …/registered/rl/test_update_weights_from_distributed.py | 651 | # Check if the weights of lm_head are tied with embed_tokens. | COMMENT |
| LOW | python/sglang/utils.py | 392 | # Check if the cache file already exists | COMMENT |
| LOW | python/sglang/benchmark/serving.py | 1499 | # Check if at least one warmup request succeeded | COMMENT |
| LOW | python/sglang/benchmark/one_batch.py | 739 | # Print output texts | COMMENT |
| LOW | python/sglang/benchmark/utils.py | 100 | # Check if the cache file already exists | COMMENT |
| LOW | python/sglang/benchmark/one_batch_server.py | 967 | # Print results | COMMENT |
| LOW | python/sglang/kernels/ops/attention/metadata.py | 651 | # Check if we should use the specialized fast path for page_size=1, no SWA | COMMENT |
| LOW | python/sglang/kernels/ops/attention/extend_attention.py | 1382 | # Check if we can skip this tile | COMMENT |
| LOW | python/sglang/kernels/ops/attention/pad.py | 295 | # Check if we're in CUDA graph mode (buffers are pre-allocated) | COMMENT |
| LOW | …n/nsa_triton_decode/triton_mla_kernels_decode_fused.py | 2888 | # Check if buffer ops should be disabled | COMMENT |
| LOW | …ang/kernels/ops/attention/flash_attn/cute/flash_fwd.py | 171 | # Check if block size setting is out of shared memory capacity | COMMENT |
| LOW | …ang/kernels/ops/attention/flash_attn/cute/flash_fwd.py | 186 | # Check if twice the block size is divisible by the number of threads | COMMENT |
| LOW | …nels/ops/quantization/nvfp4_gemm_swiglu_nvfp4_quant.py | 490 | # Check if input data types are compatible with MMA instruction | COMMENT |
| LOW | …thon/sglang/kernels/ops/lora/dense/embedding_lora_a.py | 56 | # Check if this token is within the segment | COMMENT |
| LOW | …thon/sglang/kernels/ops/lora/dense/embedding_lora_a.py | 70 | # Check if this is an extra token | COMMENT |
| LOW | python/sglang/kernels/ops/speculative/spec_tree.py | 223 | # Check if we've reached end of sibling list | COMMENT |
| LOW | python/sglang/kernels/ops/moe/inkling_moe.py | 988 | # Set grid_m to the max number of M blocks and skip padding-only blocks | COMMENT |
| LOW | …els/aot/benchmark/bench_amd_deterministic_allreduce.py | 243 | # Check if input size fits in buffer | COMMENT |
| LOW | …hon/sglang/kernels/aot/python/sgl_kernel/load_utils.py | 36 | # Check if it's a compiled extension (including complex names like .abi3.so, .cpython-312.so) | COMMENT |
| LOW | python/sglang/test/send_one.py | 238 | # Print results | COMMENT |
| LOW | python/sglang/test/runners.py | 513 | # Check if the text is empty or only whitespace. | COMMENT |
| LOW | python/sglang/test/runners.py | 813 | # Check if the text is empty or only whitespace. | COMMENT |
| LOW | python/sglang/test/few_shot_gsm8k.py | 138 | # Print results | COMMENT |
| LOW | python/sglang/test/ascend/e2e/run_npu_e2e_test.py | 389 | # Check if current line matches expected pattern | COMMENT |
| LOW | python/sglang/test/ascend/e2e/run_npu_e2e_test.py | 476 | # Check if pattern was successfully matched | COMMENT |
| LOW | python/sglang/test/ci/ci_utils.py | 363 | # Check if we should retry | COMMENT |
| LOW | …/sglang/test/server_fixtures/disaggregation_fixture.py | 336 | # Check if port is active | COMMENT |
| LOW⚡ | python/sglang/srt/environ.py | 1573 | # Set True to dequantize the FP4 experts to FP8 at runtime | COMMENT |
| LOW | python/sglang/srt/dllm/mixin/scheduler.py | 377 | # Check if batch is full | COMMENT |
| LOW | python/sglang/srt/disaggregation/decode.py | 2837 | # Check if request was aborted due to corruption | COMMENT |
| LOW | python/sglang/srt/layers/logprob_processor.py | 967 | # Check if any requests actually need logprobs computation | COMMENT |
| LOW | python/sglang/srt/layers/attention/xpu_backend.py | 569 | # Check if we should use local attention | COMMENT |
| LOW | python/sglang/srt/layers/attention/dsa_backend.py | 1162 | # Check if MHA FP8 dequantization is needed | COMMENT |
| LOW | python/sglang/srt/layers/attention/dsa_backend.py | 3772 | # Check if sequence meets criteria for MHA_ONE_SHOT | COMMENT |
| 182 more matches not shown… | ||||
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| HIGH | benchmark/hicache/bench_mix.py | 465 | self.user_generator.push( | CODE |
| HIGH | experimental/sgl-router/tests/e2e/infra/gateway.py | 335 | "Build it first: `cd experimental/sgl-router && cargo build --release` " | CODE |
| HIGH | test/manual/models/test_qwen3_asr.py | 630 | f"first commit's previous_item_id must be JSON null, got {committed_1!r}", | CODE |
| HIGH | test/manual/debug_utils/test_log_parser.py | 16 | """[{"line":"(SGLangEngine pid=35555) [2025-10-31 03:45:20 TP0] Decode batch [51341], #running-req: 317, #to | STRING |
| HIGH | …stered/unit/function_call/test_deepseekv41_detector.py | 328 | "[true, null, 2]", "invalid" | CODE |
| HIGH | …stered/unit/function_call/test_function_call_parser.py | 2739 | [true, null, {"enabled": false}] | CODE |
| HIGH | …stered/unit/function_call/test_function_call_parser.py | 3763 | ({"enum": [{"x": 1}, [True, None]]}, ['{"x": 1}', "[true, null]"], ["{}"]), | CODE |
| HIGH | …gistered/unit/function_call/test_iquest_q1_detector.py | 43 | "<arg_key>items</arg_key><arg_value>[1, null]</arg_value>" | CODE |
| HIGH | …gistered/unit/function_call/test_iquest_q1_detector.py | 167 | wire = '{ "code": "你好", "items": [1, null] }' | CODE |
| HIGH | …registered/unit/models/test_inkling_per_expert_sync.py | 10 | - w13 row layout: Inkling-interleaved vs contiguous [gate || up] | STRING |
| HIGH | test/registered/unit/parser/test_chat_parsing.py | 477 | 'null_value:null,number_value:1,string_value:<|"|>foo<|"|>,' | CODE |
| HIGH⚡ | test/registered/unit/api/v1/test_json_contract.py | 74 | p = sp('{"temperature": null, "top_k": null, "ignore_eos": null}') | CODE |
| HIGH⚡ | test/registered/unit/api/v1/test_json_contract.py | 82 | self.assertIsNone(sp('{"max_new_tokens": null}').max_new_tokens) | CODE |
| HIGH⚡ | test/registered/unit/api/v1/test_json_contract.py | 220 | self.assertEqual(compact(encode_OptionalStringOrList(v)), '[null,"h"]') | CODE |
| HIGH⚡ | test/registered/unit/api/v1/test_json_contract.py | 269 | self.assertEqual(compact(e.to_json_value()), "[-0.25,7,null]") | CODE |
| HIGH⚡ | test/registered/unit/api/v1/test_json_contract.py | 275 | self.assertEqual(compact(first_prefill.to_json_value()), "[null,10,null]") | CODE |
| HIGH | test/registered/unit/api/v1/test_json_contract.py | 298 | '{"text":"ok","meta_info":{"id":"client-rid","prompt_tokens":5,"completion_tokens":2,"finish_reason":null}," | CODE |
| HIGH⚡ | test/registered/unit/api/v1/test_json_contract.py | 217 | """PD bootstrap columns: [null, "h"] keeps per-item presence.""" | STRING |
| HIGH⚡ | test/registered/unit/api/v1/test_json_contract.py | 267 | """[logprob, token_id, text|null]; the logprob slot is nullable.""" | STRING |
| HIGH | …xer_replay/test_npu_return_indexer_topk_deepseek_v4.py | 101 | "source /usr/local/Ascend/ascend-toolkit/latest/opp/vendors/customize/bin/set_env.bash || true; " | CODE |
| HIGH | …xer_replay/test_npu_return_indexer_topk_deepseek_v4.py | 102 | "source /usr/local/Ascend/ascend-toolkit/latest/opp/vendors/custom_transformer/bin/set_env.bash || true; " | CODE |
| HIGH | …nels/ops/quantization/nvfp4_gemm_swiglu_nvfp4_quant.py | 2552 | :param c_sf_ptr: Pointer to scale factor tensor for C (can be null) | STRING |
| HIGH | …nels/ops/quantization/nvfp4_gemm_swiglu_nvfp4_quant.py | 2556 | :param norm_const_ptr: Pointer to normalization constant for SFC generation (can be null) | STRING |
| HIGH | …tion_unittest/runner_modes/cuda_graph_decode_runner.py | 401 | # capture, which makes the capture-time output undefined; only | COMMENT |
| HIGH | python/sglang/srt/layers/attention/dsa/dsa_indexer.py | 428 | # builds wk_weights_proj and leaves wk/weights_proj undefined, and a | COMMENT |
| HIGH | python/sglang/srt/layers/quantization/humming.py | 460 | # explicit null) instead of silently quantizing with the defaults. | COMMENT |
| HIGH⚡ | …ng/srt/managers/scheduler_components/internal_state.py | 219 | raise UpdateRejected(f"{key} must be in (0, 1] or null, got {value}.") | CODE |
| HIGH | python/sglang/srt/function_call/utils.py | 357 | return "string" # If only null, default to string | CODE |
| HIGH | python/sglang/srt/function_call/mimo_detector.py | 147 | <parameter=command>pwd && ls</parameter> | STRING |
| HIGH | python/sglang/srt/utils/field_validators.py | 66 | raise ValueError(f"must be list or null; got {type(v).__name__}") | CODE |
| HIGH | python/sglang/srt/utils/hf_transformers/tokenizer.py | 376 | # When the config omits the key or has null, use the v4 default so that | COMMENT |
| HIGH | python/sglang/srt/models/inkling.py | 1370 | when the config stores w13 interleaved, contiguous [gate || up] under | STRING |
| HIGH⚡ | python/sglang/srt/models/inkling_common/util.py | 16 | """Use the contiguous ``[gate || up]`` layout required by LoRA slicing.""" | STRING |
| HIGH | python/sglang/srt/models/inkling_common/util.py | 59 | """Return whether BF16 weights require TRT-LLM's ``[up || gate]`` layout.""" | STRING |
| HIGH | python/sglang/srt/entrypoints/http_server.py | 564 | {"error": {"message": "...", "type": "...", "param": null, "code": <status>}} | STRING |
| HIGH | …timodal_gen/test/unit/kandinsky6_sr_release_configs.py | 26 | r"""{"_class_name":"FlowMatchEulerDiscreteScheduler","_diffusers_version":"0.39.0","base_image_seq_len":256,"base_sh | CODE |
| HIGH | …timodal_gen/test/unit/kandinsky6_sr_release_configs.py | 45 | r"""{"models":[{"model":{"architecture":"multi_scale","bare_stem":true,"bottleneck_channels":null,"depthwise":false, | CODE |
| HIGH⚡ | …en/test/unit/realtime/test_realtime_control_signals.py | 16 | queue.push("camera_actions", [["w"], ["d"]]) | CODE |
| HIGH⚡ | …en/test/unit/realtime/test_realtime_control_signals.py | 28 | queue.push( | CODE |
| HIGH⚡ | …en/test/unit/realtime/test_realtime_control_signals.py | 205 | queue.push("camera_actions", []) | CODE |
| HIGH⚡ | …en/test/unit/realtime/test_realtime_control_signals.py | 265 | queue.push(ControlStateTransition(payload=["w"], seq_id=7)) | CODE |
| HIGH⚡ | …en/test/unit/realtime/test_realtime_control_signals.py | 266 | queue.push(ControlStateTransition(payload=[], seq_id=8)) | CODE |
| HIGH⚡ | …en/test/unit/realtime/test_realtime_control_signals.py | 277 | queue.push(ControlStateTransition(payload=["w"], seq_id=7)) | CODE |
| HIGH⚡ | …en/test/unit/realtime/test_realtime_control_signals.py | 287 | queue.push(ControlStateTransition(payload=["w"], seq_id=7)) | CODE |
| HIGH⚡ | …en/test/unit/realtime/test_realtime_control_signals.py | 288 | queue.push(ControlStateTransition(payload=["w", "d"], seq_id=8)) | CODE |
| HIGH⚡ | …en/test/unit/realtime/test_realtime_control_signals.py | 289 | queue.push(ControlStateTransition(payload=["d"], seq_id=9)) | CODE |
| HIGH | …en/test/unit/realtime/test_realtime_control_signals.py | 46 | queue.push( | CODE |
| HIGH | …en/test/unit/realtime/test_realtime_control_signals.py | 70 | queue.push("camera_actions", [["w"], ["a"], ["s"]]) | CODE |
| HIGH | …en/test/unit/realtime/test_realtime_control_signals.py | 87 | queue.push("camera_actions", [["w"]]) | CODE |
| HIGH | …en/test/unit/realtime/test_realtime_control_signals.py | 104 | queue.push("camera_actions", [["w"]]) | CODE |
| HIGH | …en/test/unit/realtime/test_realtime_control_signals.py | 122 | queue.push("camera_actions", [[]]) | CODE |
| HIGH | …en/test/unit/realtime/test_realtime_control_signals.py | 139 | queue.push( | CODE |
| HIGH | …en/test/unit/realtime/test_realtime_control_signals.py | 174 | queue.push("camera_actions", [["w"], ["w"], ["w"], ["w"]]) | CODE |
| HIGH | …ultimodal_gen/runtime/pipelines/minimax_h3_pipeline.py | 49 | # scheduler intentionally absent: model_index carries scheduler=null; | COMMENT |
| HIGH | …ang/multimodal_gen/runtime/pipelines/ltx_2_pipeline.py | 416 | # model_index.json records absent optional components as [null, null]. | COMMENT |
| HIGH | …ang/multimodal_gen/runtime/realtime/control_signals.py | 190 | self.push( | CODE |
| HIGH | …ang/multimodal_gen/runtime/realtime/control_signals.py | 351 | self._signals.push(self.kind, script, event_id=event_id) | CODE |
| HIGH | …ang/multimodal_gen/runtime/realtime/control_signals.py | 422 | self.push(transition) | CODE |
| HIGH | …ai/realtime/adapters/lingbot_world_realtime_adapter.py | 57 | self.prompt_queue.push("prompt", prompt, event_id=event_id) | CODE |
| HIGH | …rify/scripts/mechanical_refactor_reproduction_utils.py | 48 | exec_command(f"git add -A && git commit -m {shlex.quote(message)}", cwd=cwd) | CODE |
| 3 more matches not shown… | ||||
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| MEDIUM | benchmark/prefill_only/bench_setwise_score.py | 153 | # this harness sends exactly one item, so the per-candidate row count is | COMMENT |
| MEDIUM | docker/k8s-sglang-distributed-sts.yaml | 58 | privileged: true # to leverage RDMA/InfiniBand device, co-work with HostNetwork=true | CODE |
| MEDIUM | …t/manual/layers/attention/dsa/test_act_quant_triton.py | 269 | # Run comprehensive benchmark | COMMENT |
| MEDIUM⚡ | test/manual/layers/moe/bench_mxfp4_sm90_kernels.py | 328 | # Timing harness | COMMENT |
| MEDIUM | test/manual/4-gpu-models/test_qwen35_models_archived.py | 17 | # This eval harness applies the chat_template, which is critical for qwen3.5 | COMMENT |
| MEDIUM | test/manual/4-gpu-models/test_qwen35_fp4_triton.py | 5 | # This eval harness applies the chat_template, which is critical for qwen3.5 | COMMENT |
| MEDIUM | …ered/kernels/ops/attention/test_minimax_decode_topk.py | 41 | """Per-row sorted-desc multiset of the scores the kernel selected (tie-robust).""" | STRING |
| MEDIUM | …ered/kernels/ops/attention/test_minimax_decode_topk.py | 114 | # tie-robust: the multiset of selected scores must match torch.topk's | COMMENT |
| MEDIUM | …ered/kernels/ops/attention/test_minimax_decode_topk.py | 151 | # Tie-robust: compare the selected scores, not the indices. | COMMENT |
| MEDIUM | …t/registered/kernels/ops/kv_canary/test_verify_fuzz.py | 110 | # harness exercises the chain-check-skip branch (added for SWA-evicted ancestor handling). | COMMENT |
| MEDIUM | …st/registered/kernels/ops/moe/test_moe_topk_softmax.py | 124 | # Tie-robust index check: the probability at each returned index must equal | COMMENT |
| MEDIUM | …tered/attention/unittests/dense/test_flex_attention.py | 60 | # from logical positions, so it's robust to all non-tidy layouts. | COMMENT |
| MEDIUM | test/registered/attention/unittests/dense/test_fa3.py | 42 | # FlashAttention kernels are most stable in this harness with FA-friendly dims. | COMMENT |
| MEDIUM | test/registered/attention/unittests/dense/test_fa4.py | 38 | # FlashAttention kernels are most stable in this harness with FA-friendly dims. | COMMENT |
| MEDIUM⚡ | …egistered/attention/unittests/swa/test_torch_native.py | 100 | # arithmetic, so it's robust to all non-tidy layouts. | COMMENT |
| MEDIUM⚡ | …unit/lora/test_experimental_sgl_marlin_runtime_unit.py | 135 | # Runtime-flow harness | COMMENT |
| MEDIUM | …/registered/chunked_prefill/test_scripted_core_1gpu.py | 82 | # output_ids are no longer observable through the harness. That case's | COMMENT |
| MEDIUM | test/registered/hicache/test_qwen35_hicache.py | 14 | # This eval harness applies the chat_template, which is critical for qwen3.5 | COMMENT |
| MEDIUM | test/registered/observability/test_tracing.py | 585 | # Only run trace_level_3 — the most comprehensive check. | COMMENT |
| MEDIUM | test/registered/vlm/test_vision_openai_server_a.py | 42 | # coordinate, so the check is robust to normalized-vs-pixel conventions. | COMMENT |
| MEDIUM | …ered/amd/accuracy/mi35x/test_glm53_flash_eval_mi35x.py | 34 | defaults are not costing accuracy relative to #36607's harness. | STRING |
| LOW | test/registered/debug_utils/test_dumper.py | 3339 | # code can simply call `from sglang.srt.debug_utils.dumper import dumper` | COMMENT |
| MEDIUM | test/registered/e2e/models/test_inkling.py | 212 | # Sized to the 20480-token pool the harness pins: large enough to | COMMENT |
| MEDIUM⚡ | test/registered/e2e/models/test_inkling_small_nvfp4.py | 236 | # The multi-turn branching harness, unlike the single-turn helpers above, replays | COMMENT |
| MEDIUM | …istered/e2e/models_large/test_inkling_nvfp4_nightly.py | 101 | # The harness pins a tight pool; this checkpoint needs room for | COMMENT |
| MEDIUM | test/registered/e2e/scoring/test_setwise_scoring.py | 208 | """Ranking (argsort over label 0) agrees with HF — robust to fp noise.""" | STRING |
| MEDIUM | …on/sglang/kernels/ops/attention/dsa/tilelang_kernel.py | 1101 | # to better utilize FP8 dynamic range, then apply the inverse scale after GEMM. | COMMENT |
| MEDIUM | …on/sglang/kernels/ops/attention/fa4_sm120/flash_fwd.py | 342 | # N128 is robust before the compact-SM steady state. N256's | COMMENT |
| MEDIUM | …rnels/ops/attention/flash_attn/cute/flash_fwd_sm100.py | 1452 | # CLC buffers placed here to utilize padding before sO's 1024-byte alignment. | COMMENT |
| LOW | …ernels/ops/attention/flash_attn/cute/flash_fwd_sm90.py | 1975 | # 2 elements. So we just call ptx directly. | COMMENT |
| LOW | …ernels/ops/attention/flash_attn/cute/flash_fwd_sm90.py | 2082 | # 2 elements. So we just call ptx directly. | COMMENT |
| LOW | python/sglang/kernels/ops/mamba/causal_conv1d_triton.py | 177 | # first chunk and does not have prior-token, so just set to 0 | COMMENT |
| MEDIUM | python/sglang/kernels/aot/csrc/cpu/decode.cpp | 1657 | // use smaller BLOCK_H when batches is small to utilize all cores | COMMENT |
| MEDIUM | python/sglang/test/simple_eval_common.py | 278 | # The harness passes the OpenAI base (`.../v1`); `/generate` lives at root. | COMMENT |
| MEDIUM | python/sglang/test/kernels/utils.py | 47 | # single test exceeds half the harness budget, so a hung collective | COMMENT |
| MEDIUM | …tion_unittest/runner_modes/speculative_draft_runner.py | 181 | # in EagleDraftWorker.__init__, which the harness skips), so build them | COMMENT |
| MEDIUM | …tion_unittest/runner_modes/speculative_draft_runner.py | 199 | # canary_manager); for the harness that's the fixture runner. | COMMENT |
| MEDIUM | …attention_unittest/attention_methods/dsv4_attention.py | 1307 | # In runner-harness flows the reference is called BEFORE | COMMENT |
| LOW | python/sglang/srt/disaggregation/nixl/conn.py | 194 | ) # hacky just add it into the message that will be sent | CODE |
| MEDIUM | python/sglang/srt/disaggregation/common/conn.py | 2175 | # Enable higher PP ranks to be bootstrapped earlier to make PP PD requests bootstrap more robust | COMMENT |
| LOW | python/sglang/srt/layers/attention/local_attention.py | 266 | # then we can simply use a cdiv for the rest. | COMMENT |
| MEDIUM | python/sglang/srt/layers/quantization/fp8.py | 352 | # Keep both "model." and non-"model." variants for robust prefix matching. | COMMENT |
| MEDIUM | python/sglang/srt/layers/quantization/fp8.py | 506 | # For GPUs that lack FP8 hardware support, we can leverage the Marlin | COMMENT |
| MEDIUM | python/sglang/srt/layers/quantization/fp8_utils.py | 1237 | # TODO: add more robust shape check here | COMMENT |
| MEDIUM | …hon/sglang/srt/layers/quantization/marlin_utils_fp8.py | 68 | # For GPUs that lack FP8 hardware support, we can leverage the | COMMENT |
| LOW | …srt/layers/quantization/mxfp4_flashinfer_trtllm_moe.py | 490 | # in-place. Otherwise `routed` is already scale-final and we just add | COMMENT |
| MEDIUM | python/sglang/srt/layers/moe/moe_runner/humming.py | 227 | # Neighboring nodes are required to utilize distinct workspaces. | COMMENT |
| LOW | python/sglang/srt/managers/io_struct.py | 1083 | lora_id: Optional[str] = None # None means just use the base model | CODE |
| LOW | python/sglang/srt/managers/io_struct.py | 1498 | lora_id: Optional[str] = None # None means just use the base model | CODE |
| LOW | python/sglang/srt/managers/scheduler_pp_mixin.py | 462 | # otherwise, just pass along previous consensus | COMMENT |
| LOW | python/sglang/srt/managers/scheduler.py | 5512 | # In-place pause: just set the flag and return immediately. | COMMENT |
| MEDIUM | …thon/sglang/srt/model_executor/runner_backend/utils.py | 89 | # with a disabled config (e.g. the draft-runner test harness) historically | COMMENT |
| LOW | python/sglang/srt/function_call/qwen3_coder_detector.py | 45 | # Base class already initializes _buffer, we just use it directly | COMMENT |
| MEDIUM | python/sglang/srt/function_call/inkling_detector.py | 310 | # the harness sees a structured tool_call, returns a tool error, and | COMMENT |
| MEDIUM | python/sglang/srt/models/inkling_common/kernels/comm.py | 800 | # JIT module / bench harness as alternatives. | COMMENT |
| LOW | python/sglang/srt/parser/harmony_parser.py | 492 | # If no emit, just return the held content | COMMENT |
| MEDIUM | python/sglang/multimodal_gen/test/test_utils.py | 1518 | """Probe whether a remote GT file exists, robust to transient failures.""" | STRING |
| MEDIUM | …dal_gen/test/unit/test_diffusion_nightly_comparison.py | 30 | # An outside benchmark launches the published recipe as-is, so the harness | COMMENT |
| MEDIUM | …l_gen/test/unit/test_kandinsky6_sr_reference_parity.py | 85 | # Reference harness (test-only): stubs, tiny resolutions, tiny random components | COMMENT |
| MEDIUM | …ultimodal_gen/test/unit/sana_wm/test_realtime_chain.py | 167 | # Tick 2: horizon exhausted -> seamless continuation, uniform chunk from | COMMENT |
| 31 more matches not shown… | ||||
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| CRITICAL | …/attention/unittests/dsv4/test_dsv41_fused_compress.py | 232 | latent[live].float(), t.layer.indexer.wk.weight.float() | CODE |
| CRITICAL | …it/disaggregation/test_prefill_abort_result_cleanup.py | 226 | scheduler.tree_cache.req_to_token_pool.mamba_allocator.free.assert_called_once() | CODE |
| CRITICAL | test/registered/unit/mem_cache/test_registry.py | 338 | ctx.params.token_to_kv_pool_allocator.get_kvcache.return_value.register_layer_transfer_counter.assert_called_onc | CODE |
| CRITICAL⚡ | …/unit/layer_boundary/test_declared_decoder_boundary.py | 411 | max_len, communicator.ffn.plan.output.plan.path_for(max_len) | CODE |
| CRITICAL⚡ | …/unit/layer_boundary/test_declared_decoder_boundary.py | 415 | communicator.ffn.plan.output.plan.path_for(max_len), step | CODE |
| CRITICAL | …/unit/layer_boundary/test_declared_decoder_boundary.py | 392 | communicator.ffn.plan.output.plan.path_for( | CODE |
| CRITICAL | …/unit/layer_boundary/test_declared_decoder_boundary.py | 1170 | communicator.ffn.plan.output.plan.path_for( | CODE |
| CRITICAL | …red/unit/layer_boundary/test_dp_attention_cp_gather.py | 298 | r.communicator.ffn.plan.paths.get(BatchVariant.CONTEXT_PARALLEL), | CODE |
| CRITICAL | …/sglang/kernels/ops/attention/dsa/triton_sparse_mla.py | 145 | properties = triton.runtime.driver.active.utils.get_device_properties(device) | CODE |
| CRITICAL | python/sglang/kernels/ops/attention/fla/utils.py | 231 | return triton.runtime.driver.active.utils.get_device_properties(tensor_idx)[ | CODE |
| CRITICAL | python/sglang/kernels/ops/attention/fla/utils.py | 286 | triton.runtime.driver.active.utils.get_device_properties(i)[ | CODE |
| CRITICAL | …tention_unittest/attention_methods/mamba2_attention.py | 553 | self.mixer.norm.weight.data = self.mixer.norm.weight.data.float() | CODE |
| CRITICAL | python/sglang/srt/models/glm4_moe.py | 494 | and self.shared_experts.gate_up_proj.quant_method.quant_config.get_name() | CODE |
| CRITICAL | python/sglang/srt/models/deepseek_v2.py | 866 | and self.shared_experts.gate_up_proj.quant_method.quant_config.get_name() | CODE |
| CRITICAL | python/sglang/srt/mem_cache/buffer_mode/pipeline.py | 341 | entry = self._cache.cache_controller.mem_pool_host.entry_map.get(transfer.name) | CODE |
| CRITICAL | python/sglang/srt/mem_cache/buffer_mode/pipeline.py | 500 | entry = self._cache.cache_controller.mem_pool_host.entry_map.get(transfer.name) | CODE |
| CRITICAL | …ng/multimodal_gen/test/unit/test_prompt_enhancement.py | 317 | assert server.args.pipeline_config.task_type.name.lower() == task | CODE |
| CRITICAL | …l_gen/runtime/pipelines_core/composed_pipeline_base.py | 146 | task_name = self.server_args.pipeline_config.task_type.name.lower() | CODE |
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| HIGH | python/sglang/kernels/ops/attention/fla/chunk.py | 150 | Args: q (torch.Tensor): queries of shape `[B, T, H, K]` if `head_first=False` else `[B, H, T, K | STRING |
| HIGH | …on/sglang/kernels/ops/attention/fla/fused_recurrent.py | 775 | Args: q (torch.Tensor): queries of shape `[B, T, H, K]`. k (torch.Tensor): | STRING |
| HIGH | python/sglang/kernels/ops/quantization/fp8_kernel.py | 1167 | Quantize input tensor to FP8 (8-bit floating point) format. Args: input (torch.Tensor): Input tensor to be quantiz | STRING |
| HIGH | python/sglang/kernels/ops/quantization/fp8_cvt.py | 27 | Cast fp32 to fp8 e4m3 through the same ``pack_fp8`` the fp8 stores use. Nothing in the serving path calls this -- i | STRING |
| HIGH | python/sglang/srt/layers/model_parallel.py | 32 | Locally shards a full tensor based on indicated sharding arrangement, and returns a DTensor containing the loca | STRING |
| HIGH | …thon/sglang/srt/layers/attention/flashinfer_backend.py | 586 | Process multi-item scoring tensors for FlashInfer attention. This method handles sequences containing multiple | STRING |
| HIGH | python/sglang/srt/layers/moe/cutlass_moe.py | 55 | Performs Fused MoE computation using CUTLASS-like kernels with FP8 weights and activations. This function implement | STRING |
| HIGH | python/sglang/srt/function_call/utils.py | 264 | Get consolidated $defs from all tools, validating for conflicts. Args: tools: List of tools to process | STRING |
| HIGH | python/sglang/srt/utils/network.py | 140 | Bind a TCP socket on the first available address family (IPv4/IPv6). Iterates over address families returned by _ge | STRING |
| HIGH | python/sglang/srt/utils/network.py | 358 | Automatically detect the local IP address using multiple fallback strategies. This function attempts to obtain | STRING |
| HIGH | python/sglang/srt/utils/server_args_config_parser.py | 53 | Merge configuration file arguments with command-line arguments. Configuration arguments are inserted a | STRING |
| HIGH | python/sglang/srt/utils/server_args_config_parser.py | 102 | Parse YAML configuration file and convert to argument list. Args: file_path: Path to the Y | STRING |
| HIGH | python/sglang/srt/utils/nvtx_pytorch_hooks.py | 46 | Descends iterators that contains Tensors and prints the Tensor Recursive function that descends iterator type a | STRING |
| HIGH | python/sglang/srt/utils/nvtx_pytorch_hooks.py | 74 | Extract the static parameters from LLM and VLM relevant layer types Args: module_obj(class): Module | STRING |
| HIGH | python/sglang/srt/utils/nvtx_pytorch_hooks.py | 198 | Callback function that ends the NVTX marker Records the module name and tensor information Called after | STRING |
| HIGH | python/sglang/srt/utils/nvtx_pytorch_hooks.py | 218 | Creates an NVTX marker with the module name in it. This function is called before the module executes | STRING |
| HIGH | python/sglang/srt/utils/nvtx_pytorch_hooks.py | 256 | User level function that activates all the hooks The user needs to call this method from the network source cod | STRING |
| HIGH | python/sglang/srt/models/minicpmo.py | 138 | In streaming audio generation, determine which `text` positions the TTS model can attend to when generating each ch | STRING |
| HIGH | python/sglang/srt/models/minicpmo.py | 617 | Merge `input_ids` and `lm_spk_emb_last_hidden_states` to `inputs_embeds`. Args: input_ids (torch.Te | STRING |
| HIGH | python/sglang/srt/multimodal/processors/qwen_vl.py | 204 | calculate the number of frames for video used for model inputs. Args: ele (dict): a dict contains the confi | STRING |
| HIGH | python/sglang/srt/multimodal/processors/ernie45_vl.py | 130 | calculate the number of frames for video used for model inputs. Args: ele (dict): a dict contains the confi | STRING |
| HIGH | python/sglang/srt/mem_cache/storage/backend_factory.py | 73 | Create a storage backend instance. Args: backend_name: Name of the backend to create sto | STRING |
| HIGH | python/sglang/srt/speculative/dflash_utils.py | 367 | Select target layer indices used to build DFlash context features. Args: num_target_layers: Number of trans | STRING |
| HIGH | python/sglang/srt/speculative/dflash_utils.py | 897 | Compute DFlash accept lengths and bonus tokens (greedy verify rule). Args: candidates: Token ids proposed b | STRING |
| HIGH | …/runtime/managers/memory_managers/layerwise_offload.py | 2488 | Update consolidated CPU buffers with new weights. When layerwise offload (--dit-layerwise-offload) is enabled, | STRING |
| HIGH | …ang/multimodal_gen/runtime/utils/nvtx_pytorch_hooks.py | 97 | Walk ``model`` and attach forward pre/post hooks to every module. Args: model: Root module to instr | STRING |
| HIGH | …ng/multimodal_gen/runtime/models/encoders/qwen2_5vl.py | 653 | Calculate the 3D rope index based on image and video's temporal, height and width in LLM. Explanation: | STRING |
| HIGH | …imodal_gen/runtime/pipelines_core/stages/validators.py | 345 | Add a validation check for a field. Args: field_name: Name of the field being checked | STRING |
| HIGH | sgl-model-gateway/e2e_test/infra/model_pool.py | 727 | Get a model instance by model_id, mode, and worker_type. If the model is not running, it will be launched on-de | STRING |
| HIGH | sgl-model-gateway/e2e_test/responses/test_basic_crud.py | 24 | Wait for background task to complete. Args: client: OpenAI client response_id: Response ID to poll | STRING |
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| MEDIUM | …t/manual/prefill_only/test_encoder_embedding_models.py | 12 | # Encoder embedding model tests (CUDA only) | COMMENT |
| MEDIUM | …rnels/ops/attention/test_inkling_checkpoint_indices.py | 727 | # Only model/distributed capture scaffolding is replaced. __init__ and | COMMENT |
| MEDIUM⚡ | …t/registered/kernels/ops/kv_canary/test_verify_hand.py | 65 | # Shared per-test scaffolding helpers. | COMMENT |
| MEDIUM⚡ | …t/registered/unit/managers/test_setwise_score_mixin.py | 230 | # A base embedding model likewise does not route through score_and_pool. | COMMENT |
| MEDIUM⚡ | …tered/unit/layer_boundary/test_moe_output_reduction.py | 169 | # Execute the real orchestration methods without constructing weights. | COMMENT |
| MEDIUM | test/registered/pp/test_pp_gemma4.py | 57 | # Gemma4 is instruction-tuned and doesn't follow few-shot completion | COMMENT |
| MEDIUM | …red/disaggregation/test_disaggregation_different_tp.py | 450 | # Small enough that the shared gsm8k few-shot prefix (~900 tokens) spans | COMMENT |
| MEDIUM | …istered/accuracy/models/test_text_models_gsm8k_eval.py | 28 | # sgl-eval (zero-shot chat, \boxed{}, math_verify grading). Thresholds are | COMMENT |
| MEDIUM | …registered/e2e/models/test_deepseek_v41_flash_gb300.py | 28 | # Use the zero-shot chat evaluator, not the legacy five-shot completion path. | COMMENT |
| MEDIUM | python/sglang/benchmark/datasets/agentic_trace.py | 13 | # ~220-token average assistant reply of OpenHands-style agentic traces. | COMMENT |
| MEDIUM | python/sglang/kernels/ops/attention/flash_attention.py | 31 | window_size=(-1, -1), # -1 means infinite context window | CODE |
| MEDIUM | …hon/sglang/kernels/ops/attention/flash_attention_v3.py | 124 | window_size=(-1, -1), # -1 means infinite context window | CODE |
| MEDIUM | …hon/sglang/kernels/aot/python/sgl_kernel/flash_attn.py | 58 | window_size=(-1, -1), # -1 means infinite context window | CODE |
| MEDIUM | python/sglang/test/simple_eval_mixed_prefix_gsm8k.py | 64 | # The evaluation data should not include the few-shot examples to prevent data leakage. | COMMENT |
| MEDIUM | python/sglang/test/accuracy_test_runner.py | 35 | num_shots: Optional[int] = None # few-shot count; None = run_eval's default | CODE |
| MEDIUM | …/srt/layers/attention/deepseek_v4_backend_hip_radix.py | 852 | # the ratio-1/2 compressor and indexer orchestration is the CUDA backend's, taken | COMMENT |
| MEDIUM | …ang/srt/layers/attention/deepseek_v4_trtllm_backend.py | 344 | # DSpark attends noncausally to its context window plus the whole | COMMENT |
| MEDIUM | python/sglang/srt/managers/multi_tokenizer_mixin.py | 573 | # read SHM only. Drain it event-driven via the socket's fd instead of | COMMENT |
| MEDIUM | python/sglang/srt/managers/detokenizer_manager.py | 223 | # If it is embedding model, no detokenization is needed. | COMMENT |
| MEDIUM | python/sglang/srt/models/locate_anything.py | 365 | # <box>: a closed <box>...</box> in a few-shot / multi-turn prompt is | COMMENT |
| MEDIUM | python/sglang/srt/parser/reasoning_parser.py | 1406 | # detect_and_parse; without it the trailing reasoning trace is dropped. | COMMENT |
| MEDIUM | python/sglang/srt/parser/conversation.py | 92 | # The number of few shot examples | COMMENT |
| MEDIUM | …l_gen/test/unit/test_kandinsky6_sr_reference_parity.py | 539 | # End to end: pure orchestration vs kandinsky_sr.pipeline.stages.run_tiled_sr* | COMMENT |
| MEDIUM | …/runtime/models/sensenova_u1/neo_unify/conversation.py | 54 | # The number of few shot examples | COMMENT |
| MEDIUM | .agents/skills/large-class-style/SKILL.md | 29 | ### 1.3 Allowed: orchestration | COMMENT |
| MEDIUM | .agents/skills/large-class-style/SKILL.md | 41 | # model_runner.py — orchestration only. | COMMENT |
| MEDIUM | …-verify/scripts/mechanical_refactor_proof_generator.py | 803 | # are exactly the relocated symbols (the prep commit inlined them, scaffolding included, as | COMMENT |
| MEDIUM | …-verify/scripts/mechanical_refactor_proof_generator.py | 1157 | # A move source the commit deletes (its defs all relocated, leaving only scaffolding) is | COMMENT |
| MEDIUM | docs/src/snippets/_deployment.jsx | 104 | // agenticLink optional — heading id of the page's agentic long-context | COMMENT |
| MEDIUM | docs/src/snippets/autoregressive/qwen35-deployment.jsx | 18 | // 397B-A17B: H100 tp=8, H200 tp=8 ep=8, B200 tp=4, B300 tp=4, GB200 tp=4, GB300 tp=4, MI300X tp=4, MI325X tp=2, MI3 | COMMENT |
| MEDIUM | docs/src/snippets/autoregressive/qwen35-deployment.jsx | 122 | // separately for the 397B FP8 long-context agentic recipe. | COMMENT |
| MEDIUM | docs/src/snippets/autoregressive/qwen35-deployment.jsx | 370 | // 397B MI355X FP8 with HiCache: the long-context agentic recipe runs tp=4. | COMMENT |
| MEDIUM | docs/src/snippets/autoregressive/qwen35-deployment.jsx | 372 | // to feed a host-DRAM tier at agentic context lengths. | COMMENT |
| MEDIUM | docs/src/snippets/autoregressive/qwen35-deployment.jsx | 574 | // MI355X FP8 long-context agentic recipe: a host-DRAM KV tier plus an FP8 KV | COMMENT |
| MEDIUM | docs/src/snippets/autoregressive/qwen35-deployment.jsx | 575 | // cache, which together keep enough prefix resident to reach agentic | COMMENT |
| MEDIUM | docs/src/snippets/configs/poolside/laguna-m1.jsx | 157 | // agentic coding, repeated long contexts. "Enable" emits --enable-hierarchical-cache (+host L2); | COMMENT |
| MEDIUM | docs/src/snippets/configs/deepseek-ai/deepseek-v4.jsx | 553 | // The AMD role flags are the MI355X 1P x 1D agentic recipe. Both roles | COMMENT |
| MEDIUM | docs/src/snippets/configs/deepseek-ai/deepseek-v4.jsx | 771 | // High-throughput on Pro Official is the DP-attention arm of the agentic | COMMENT |
| MEDIUM | docs/src/snippets/configs/deepseek-ai/deepseek-v4.jsx | 876 | // MI355X fronts the 1P x 1D agentic pair with a cache-aware router rather | COMMENT |
| MEDIUM | docs/src/snippets/configs/deepseek-ai/deepseek-v4.jsx | 883 | // agentic run should notice a wedged worker, but it is slack enough that | COMMENT |
| MEDIUM | docs/src/snippets/configs/deepseek-ai/deepseek-v4.jsx | 2368 | // The DP + DSpark agentic path is documented in cookbook §3.7. | COMMENT |
| MEDIUM | docs/src/snippets/configs/Qwen/qwen3.8-flash-next.jsx | 815 | warn: "Long-context recipe: reserve 224 GiB of host RAM for HiCache, approximately 48 GiB for pinned PLE offload, | CODE |
| MEDIUM | …les/frontend_language/usage/openai_chat_speculative.py | 150 | # expect answer with same format as in the few shot | COMMENT |
| MEDIUM⚡ | examples/runtime/chain_of_verification.py | 98 | # `content` and the chain-of-thought in `reasoning_content`. In some | COMMENT |
| MEDIUM⚡ | scripts/ci_monitor/ci_auto_bisect.py | 1060 | # Main orchestration | COMMENT |
| MEDIUM | scripts/ci/utils/ci_coverage_report.py | 360 | # adding scaffolding for an unused backend would widen every row with a | COMMENT |
| MEDIUM | …-model-gateway/e2e_test/embeddings/test_correctness.py | 151 | # Get model path from MODEL_SPECS for the embedding model | COMMENT |
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| MEDIUM | experimental/sgl-router/tests/e2e/test_gpu_allotment.py | 138 | CODE | |
| MEDIUM | …ed/unit/managers/test_tokenizer_manager_rid_cleanup.py | 940 | CODE | |
| MEDIUM | …ed/unit/managers/test_tokenizer_manager_rid_cleanup.py | 861 | CODE | |
| MEDIUM | …ed/unit/managers/test_tokenizer_manager_rid_cleanup.py | 889 | CODE | |
| MEDIUM | …/registered/unit/entrypoints/anthropic/test_serving.py | 1274 | CODE | |
| MEDIUM | …egistered/unit/entrypoints/openai/test_serving_chat.py | 3287 | CODE | |
| MEDIUM | python/sglang/srt/entrypoints/http_server.py | 2165 | CODE | |
| MEDIUM | python/sglang/srt/entrypoints/openai/serving_base.py | 145 | CODE | |
| MEDIUM | python/sglang/srt/entrypoints/openai/serving_base.py | 148 | CODE | |
| MEDIUM | …ime/models/sensenova_u1/neo_unify/modeling_neo_chat.py | 370 | CODE | |
| MEDIUM | …ime/models/sensenova_u1/neo_unify/modeling_neo_chat.py | 374 | CODE | |
| MEDIUM | …ime/models/sensenova_u1/neo_unify/modeling_neo_chat.py | 375 | CODE | |
| MEDIUM | …ime/models/sensenova_u1/neo_unify/modeling_neo_chat.py | 377 | CODE | |
| MEDIUM | …ime/models/sensenova_u1/neo_unify/modeling_neo_chat.py | 378 | CODE | |
| MEDIUM | …ime/models/sensenova_u1/neo_unify/modeling_neo_chat.py | 380 | CODE | |
| MEDIUM | …ime/models/sensenova_u1/neo_unify/modeling_neo_chat.py | 381 | CODE | |
| MEDIUM | …ime/models/sensenova_u1/neo_unify/modeling_neo_chat.py | 386 | CODE | |
| MEDIUM | …ime/models/sensenova_u1/neo_unify/modeling_neo_chat.py | 387 | CODE | |
| MEDIUM | …ime/models/sensenova_u1/neo_unify/modeling_neo_chat.py | 388 | CODE | |
| MEDIUM | …ime/models/sensenova_u1/neo_unify/modeling_neo_chat.py | 403 | CODE | |
| MEDIUM | …ime/models/sensenova_u1/neo_unify/modeling_neo_chat.py | 405 | CODE | |
| MEDIUM | …ime/models/sensenova_u1/neo_unify/modeling_neo_chat.py | 415 | CODE | |
| MEDIUM | …ime/models/sensenova_u1/neo_unify/modeling_neo_chat.py | 417 | CODE | |
| MEDIUM | …ime/models/sensenova_u1/neo_unify/modeling_neo_chat.py | 418 | CODE | |
| MEDIUM | …ime/models/sensenova_u1/neo_unify/modeling_neo_chat.py | 430 | CODE | |
| MEDIUM | …ime/models/sensenova_u1/neo_unify/modeling_neo_chat.py | 434 | CODE | |
| MEDIUM | …ime/models/sensenova_u1/neo_unify/modeling_neo_chat.py | 478 | CODE | |
| MEDIUM | …ime/models/sensenova_u1/neo_unify/modeling_neo_chat.py | 482 | CODE | |
| MEDIUM | …ime/models/sensenova_u1/neo_unify/modeling_neo_chat.py | 488 | CODE | |
| MEDIUM | …ime/models/sensenova_u1/neo_unify/modeling_neo_chat.py | 489 | CODE | |
| MEDIUM | …ime/models/sensenova_u1/neo_unify/modeling_neo_chat.py | 491 | CODE | |
| MEDIUM | …ime/models/sensenova_u1/neo_unify/modeling_neo_chat.py | 495 | CODE | |
| MEDIUM | …ime/models/sensenova_u1/neo_unify/modeling_neo_chat.py | 496 | CODE | |
| MEDIUM | …ime/models/sensenova_u1/neo_unify/modeling_neo_chat.py | 510 | CODE | |
| MEDIUM | …ime/models/sensenova_u1/neo_unify/modeling_neo_chat.py | 511 | CODE | |
| MEDIUM | …ime/models/sensenova_u1/neo_unify/modeling_neo_chat.py | 512 | CODE | |
| MEDIUM | …ime/models/sensenova_u1/neo_unify/modeling_neo_chat.py | 513 | CODE | |
| MEDIUM | …ime/models/sensenova_u1/neo_unify/modeling_neo_chat.py | 514 | CODE | |
| MEDIUM | …ime/models/sensenova_u1/neo_unify/modeling_neo_chat.py | 515 | CODE | |
| MEDIUM | …ime/models/sensenova_u1/neo_unify/modeling_neo_chat.py | 516 | CODE | |
| MEDIUM | …ime/models/sensenova_u1/neo_unify/modeling_neo_chat.py | 522 | CODE | |
| MEDIUM | …ime/models/sensenova_u1/neo_unify/modeling_neo_chat.py | 523 | CODE | |
| MEDIUM | …ime/models/sensenova_u1/neo_unify/modeling_neo_chat.py | 526 | CODE | |
| MEDIUM | …ang/multimodal_gen/runtime/models/dits/hunyuanvideo.py | 960 | CODE | |
| MEDIUM | …ang/multimodal_gen/runtime/models/dits/hunyuanvideo.py | 961 | CODE | |
| MEDIUM | …ang/multimodal_gen/runtime/models/dits/hunyuanvideo.py | 962 | CODE | |
| MEDIUM | …ang/multimodal_gen/runtime/models/dits/hunyuanvideo.py | 965 | CODE | |
| MEDIUM | …ang/multimodal_gen/runtime/models/dits/hunyuanvideo.py | 966 | CODE | |
| MEDIUM | …ang/multimodal_gen/runtime/models/dits/hunyuanvideo.py | 968 | CODE | |
| MEDIUM | …ang/multimodal_gen/runtime/models/dits/hunyuanvideo.py | 970 | CODE | |
| MEDIUM | …ang/multimodal_gen/runtime/models/dits/hunyuanvideo.py | 973 | CODE | |
| MEDIUM | …ang/multimodal_gen/runtime/models/dits/hunyuanvideo.py | 974 | CODE | |
| MEDIUM | …ang/multimodal_gen/runtime/models/dits/hunyuanvideo.py | 976 | CODE | |
| MEDIUM | …ang/multimodal_gen/runtime/models/dits/hunyuanvideo.py | 986 | CODE | |
| MEDIUM | …ang/multimodal_gen/runtime/models/dits/hunyuanvideo.py | 1001 | CODE | |
| MEDIUM | …ang/multimodal_gen/runtime/models/dits/hunyuanvideo.py | 1009 | CODE | |
| MEDIUM | …ang/multimodal_gen/runtime/models/dits/hunyuanvideo.py | 1010 | CODE | |
| MEDIUM | …ang/multimodal_gen/runtime/models/dits/hunyuanvideo.py | 1011 | CODE | |
| MEDIUM | …ang/multimodal_gen/runtime/models/dits/hunyuanvideo.py | 1039 | CODE | |
| MEDIUM | …ang/multimodal_gen/runtime/models/dits/hunyuanvideo.py | 1040 | CODE | |
| 1 more matches not shown… | ||||
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| MEDIUM | …manual/openai_server/function_call/test_tool_choice.py | 844 | # # Start the local OpenAI Server. If necessary, you can add other parameters such as --enable-tools. | COMMENT |
| MEDIUM | …i_server/function_call/test_openai_function_calling.py | 69 | # Start the local OpenAI Server. If necessary, you can add other parameters such as --enable-tools. | COMMENT |
| MEDIUM | …i_server/function_call/test_openai_function_calling.py | 1002 | # # Start the local OpenAI Server. If necessary, you can add other parameters such as --enable-tools. | COMMENT |
| MEDIUM | …hon/sglang/kernels/ops/attention/flash_attention_v3.py | 92 | # That means if you use A100/A*0/L20/L40/L40s/4090 you can use fa3. | COMMENT |
| MEDIUM | …hon/sglang/kernels/aot/python/sgl_kernel/flash_attn.py | 24 | # That means if you use A100/A*0/L20/L40/L40s/4090 you can use fa3. | COMMENT |
| MEDIUM | python/sglang/kernels/aot/tests/test_flash_attention.py | 28 | # That means if you use A100/A*0/L20/L40/L40s/4090 you can use fa3. | COMMENT |
| LOW | python/sglang/test/runners.py | 504 | # make sure to disable compile | COMMENT |
| LOW | python/sglang/srt/models/hunyuan.py | 772 | # make sure to leave KV cache scale factors in a known good (dummy) state | COMMENT |
| LOW | python/sglang/srt/models/apertus.py | 389 | # make sure to leave KV cache scale factors in a known good (dummy) state | COMMENT |
| LOW | python/sglang/srt/models/mimo_v2.py | 1097 | # make sure to leave KV cache scale factors in a known good (dummy) state | COMMENT |
| MEDIUM | python/sglang/srt/models/inkling.py | 1053 | # (BCG) setup treats Inkling as a language model: model_runner's | COMMENT |
| LOW | python/sglang/srt/models/llama.py | 502 | # make sure to leave KV cache scale factors in a known good (dummy) state | COMMENT |
| LOW | python/sglang/srt/models/glm4.py | 417 | # make sure to leave KV cache scale factors in a known good (dummy) state | COMMENT |
| LOW | python/sglang/srt/models/qwen2.py | 472 | # make sure to leave KV cache scale factors in a known good (dummy) state | COMMENT |
| LOW | …glang/multimodal_gen/runtime/models/vaes/hunyuanvae.py | 295 | # If you are encountering an error here, make sure to try running encoding/decoding with | COMMENT |
| MEDIUM | …ntend_language/quick_start/local_example_llava_next.py | 55 | # Or you can use the 72B model | COMMENT |
| MEDIUM | …ntend_language/quick_start/local_example_llava_next.py | 62 | # Or you can use API models | COMMENT |
| MEDIUM | …-gateway/examples/wasm/wasm-guest-ratelimit/src/lib.rs | 37 | // This is a simplified example for demonstration purposes | COMMENT |
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| LOW | …l/sgl-router/tests/scripts/generate_parity_fixtures.py | 41 | "Lorem ipsum dolor sit amet, consectetur adipiscing elit, sed do eiusmod " | CODE |
| LOW | …l/sgl-router/tests/scripts/generate_parity_fixtures.py | 41 | "Lorem ipsum dolor sit amet, consectetur adipiscing elit, sed do eiusmod " | CODE |
| LOW⚡ | test/registered/unit/models/test_kimi_k25.py | 201 | with pytest.raises(ValueError, match="placeholder"): | CODE |
| LOW⚡ | …st/registered/unit/multimodal/rust/qwen/test_driver.py | 80 | self.assert_rejected(ids, images, "placeholder") | CODE |
| LOW⚡ | …registered/unit/model_loader/test_presharded_loader.py | 748 | fake_key = ("_test_meta_rope_cleanup_sentinel",) | CODE |
| LOW⚡ | …registered/unit/model_loader/test_presharded_loader.py | 751 | _ROPE_DICT[fake_key] = fake_module | CODE |
| LOW | …registered/unit/model_loader/test_presharded_loader.py | 777 | self.assertNotIn(fake_key, _ROPE_DICT) | CODE |
| LOW | python/sglang/test/kits/ebnf_constrained_kit.py | 61 | self.__class__.ebnf_grammar = 'root ::= "user@example.com"' | CODE |
| LOW⚡ | python/sglang/srt/batch_overlap/two_batch_overlap.py | 880 | # (MAX_LEN) count and would undo the idle-rank dummy-token mask. | COMMENT |
| LOW | python/sglang/srt/compilation/backend.py | 238 | if node.op in ("output", "placeholder"): | CODE |
| LOW | rust/sglang-mm/src/driver.rs | 254 | assert!(err.contains("placeholder")); | CODE |
| LOW | …-model-gateway/tests/security/auth_integration_test.rs | 174 | let claims = create_claims("admin@example.com", vec!["admin"]); | CODE |
| LOW | …-model-gateway/tests/security/auth_integration_test.rs | 187 | assert_eq!(validated.subject, "admin@example.com"); | CODE |
| LOW | …-model-gateway/tests/security/auth_integration_test.rs | 213 | let claims = create_claims("user@example.com", vec!["user", "viewer"]); | CODE |
| LOW | …-model-gateway/tests/security/auth_integration_test.rs | 221 | assert_eq!(validated.subject, "user@example.com"); | CODE |
| LOW | …-model-gateway/tests/security/auth_integration_test.rs | 252 | sub: "user@example.com".to_string(), | CODE |
| LOW | …-model-gateway/tests/security/auth_integration_test.rs | 301 | sub: "user@example.com".to_string(), | CODE |
| LOW | …-model-gateway/tests/security/auth_integration_test.rs | 338 | let claims = create_claims("user@example.com", vec!["admin"]); | CODE |
| LOW | …-model-gateway/tests/security/auth_integration_test.rs | 590 | let claims = create_claims("user@example.com", vec!["admin"]); | CODE |
| LOW | …-model-gateway/tests/security/auth_integration_test.rs | 713 | let claims = create_claims("user@example.com", vec!["admin"]); | CODE |
| LOW | sgl-model-gateway/e2e_test/benchmarks/conftest.py | 38 | "dummy-token", | CODE |
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| LOW | benchmark/hicache/bench_mix.py | 417 | async def handle_request(self, user_data): | CODE |
| LOW | benchmark/hicache/bench_multiturn.py | 390 | async def handle_request(self, item): | CODE |
| LOW | python/sglang/srt/debug_utils/dumper.py | 1376 | def handle_request(self, *, method: str, body: dict[str, Any]) -> list[dict]: | CODE |
| LOW | python/sglang/srt/entrypoints/openai/serving_base.py | 72 | async def handle_request( | CODE |
| LOW | …/scripts/tests/proof_generator/test_script_and_diff.py | 137 | "def keep():\n return 0\n\n\ndef helper(x):\n return x + 1\n" | CODE |
| LOW | …/scripts/tests/proof_generator/test_script_and_diff.py | 160 | "def helper(x):\n" | CODE |
| LOW | …rify/scripts/tests/proof_generator/test_infer_moves.py | 49 | "def keep():\n return 0\n\n\ndef helper(x):\n return x + 1\n" | CODE |
| LOW | …rify/scripts/tests/proof_generator/test_infer_moves.py | 72 | "def helper(x):\n" | CODE |
| LOW | …ipts/tests/reproduction_utils/test_extract_function.py | 180 | signature=" def helper(self):", | CODE |
| LOW | …ipts/tests/reproduction_utils/test_extract_function.py | 190 | assert " def helper(self):\n val = compute_thing()\n" in out | CODE |
| LOW⚡ | …y/scripts/tests/reproduction_utils/test_move_symbol.py | 84 | "class Old:\n @staticmethod\n def helper(x):\n return x * 2\n" | CODE |
| LOW⚡ | …y/scripts/tests/reproduction_utils/test_move_symbol.py | 91 | assert "def helper(x):\n return x * 2\n" in (tmp_path / "dst.py").read_text() | CODE |
| LOW⚡ | …y/scripts/tests/reproduction_utils/test_move_symbol.py | 286 | " def helper(y):\n" | CODE |
| LOW⚡ | …y/scripts/tests/reproduction_utils/test_move_symbol.py | 296 | assert " @staticmethod\n def helper(y):\n" in dst_out | CODE |
| LOW⚡ | …y/scripts/tests/reproduction_utils/test_move_symbol.py | 415 | (tmp_path / "src.py").write_text("def helper(x):\n return x\n") | CODE |
| LOW⚡ | …y/scripts/tests/reproduction_utils/test_move_symbol.py | 421 | assert " def helper(x):\n return x\n" in (tmp_path / "dst.py").read_text() | CODE |
| LOW | …y/scripts/tests/reproduction_utils/test_move_symbol.py | 341 | (tmp_path / "src.py").write_text("def helper(x):\n return x + 1\n") | CODE |
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| LOW | …rimental/sgl-router/tests/e2e/k8s_integration/setup.sh | 6 | # Usage: | COMMENT |
| LOW | scripts/build_sgl_deepep.sh | 4 | # Usage: | COMMENT |
| LOW | scripts/killall_sglang.sh | 8 | # Usage: | COMMENT |
| LOW | scripts/playground/launch_pr_babysitters.sh | 4 | # Usage: | COMMENT |
| LOW | scripts/ci/musa/rename_wheels_musa.sh | 10 | # Usage: | COMMENT |
| LOW | sgl-model-gateway/e2e_test/k8s_integration/setup.sh | 8 | # Usage: | COMMENT |
| LOW | …ngs/golang/examples/oai_server/scripts/analyze_tpot.sh | 6 | # Usage: | COMMENT |
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| HIGH | test/manual/attention/test_trtllm_mla_backend.py | 1235 | # Reshape as requested | COMMENT |
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| LOW | test/manual/test_expert_distribution.py | 20 | # TODO: Add tests for DeepEP gatherer (currently our CI cannot run that) | COMMENT |
| LOW | python/sglang/kernels/aot/csrc/cpu/common.h | 369 | // TODO: implement reverse order of [MB / cache_blocks_mb, NB, cache_blocks_mb] | COMMENT |
| LOW | python/sglang/kernels/aot/python/sgl_kernel/top_k.py | 12 | # TODO: implement faster cuda kernels for large vocab sizes | COMMENT |
| LOW | python/sglang/srt/layers/moe/fused_moe_triton/layer.py | 1631 | # TODO: consider using symmetric memory | COMMENT |