A framework for efficient model inference with omni-modality models
This report presents the forensic synthetic code analysis of vllm-project/vllm-omni, a Python project with 6,436 GitHub stars. SynthScan v2.0 examined 944,682 lines of code across 3201 source files, recording 19064 pattern matches distributed across 24 syntactic categories. The overall adjusted score of 26.4 places this repository in the Moderate AI signal band.
The scanner applied 160+ deterministic lexical heuristics, multi-line block detectors, abstract syntax tree depth profilers, and a cross-file Jaccard similarity matrix to construct a statistically normalised synthetic code estimate. All matches are individually weighted by severity coefficient and contextual multiplier before summation, and the resulting headline score is temporally discounted to account for the repository's development history relative to the commercial emergence of large language model coding tooling (November 2022 onward).
This chart maps the temporal evolution of the adjusted synthetic code score across successive scan runs. An upward trajectory indicates ongoing incorporation of AI-generated code or expanding LLM-assisted scaffolding; a stable or declining trajectory may reflect active human refactoring, code removal, or the adoption of stricter authorship policies. The dashed secondary line (right axis) independently tracks total raw pattern hit count, which can diverge from the normalised score when codebase size changes significantly between scans.
Classifies detected patterns by their diagnostic confidence and structural impact. CRITICAL patterns (coefficient 10) represent definitive synthetic signatures — hallucinated imports, explicit LLM attribution metadata — virtually never produced by human authors. HIGH (5) indicates strong structural tells such as cross-file repetition or cross-linguistic idioms. MEDIUM (2) covers recognisable conversational padding and AI-specific vocabulary. LOW (1) captures subtle indicators like tautological comments and generic boilerplate that require density to carry independent signal.
This horizontal bar chart decomposes the repository's raw synthetic code score by top-level directory, allowing you to pinpoint precisely which modules or components carry the highest AI authorship density. Directories with disproportionately high scores relative to their size warrant targeted manual review: concentrated AI signatures often trace back to mass-generated configuration layers, auto-ported test suites, LLM-scaffolded boilerplate classes, or entire subsystems authored under heavy copilot assistance. Use this view to prioritise your human code-review effort.
The scanner identified 19064 distinct pattern matches across 24 syntactic categories. Each entry below represents a discrete location in the source code where the engine recorded a statistically significant AI authorship indicator. 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 diagnostic confidence level (CRITICAL / HIGH / MEDIUM / LOW). The Context column identifies whether the match occurred inside executable code, an inline comment, or a string literal — comment-context matches receive a ×1.5 weight because LLMs systematically over-annotate. The ⚡ bolt icon marks clustered matches: three or more patterns within a 10-line window, each receiving an additional ×1.5 density multiplier as dense clusters constitute far stronger evidence of synthetic authorship than isolated hits.
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| LOW | collect_env.py | 135 | def run_and_parse_first_match(run_lambda, command, regex): | CODE |
| LOW | collect_env.py | 146 | def run_and_return_first_line(run_lambda, command): | CODE |
| LOW | collect_env.py | 179 | def get_nvidia_driver_version(run_lambda): | CODE |
| LOW | collect_env.py | 318 | def summarize_vllm_build_flags(): | CODE |
| LOW | collect_env.py | 473 | def get_cachingallocator_config(): | CODE |
| LOW | collect_env.py | 478 | def get_cuda_module_loading_config(): | CODE |
| LOW | tools/run_jobs_common.sh | 526 | def timeout_kill_after_seconds() -> int: | CODE |
| LOW | tools/run_jobs_common.sh | 535 | def prepend_timeout_to_pytest(pytest_line: str, timeout_min: int | None) -> str: | CODE |
| LOW | tools/run_jobs_common.sh | 547 | def _write_job_timeouts_manifest(jobs_dir: Path, job_timeouts: dict[str, int]) -> None: | CODE |
| LOW | tools/nightly/buildkite_testcase_statistics.py | 86 | def extract_pytest_targets_from_line(line: str) -> list[str]: | CODE |
| LOW | tools/nightly/buildkite_testcase_statistics.py | 138 | def get_pytest_targets_from_step(step: dict) -> list[tuple[str, str]]: | CODE |
| LOW⚡ | tools/nightly/buildkite_testcase_statistics.py | 158 | def _strip_ignore_clauses_from_args_fragment(fragment: str) -> str: | CODE |
| LOW⚡ | tools/nightly/buildkite_testcase_statistics.py | 163 | def _strip_test_config_file_clauses_from_args_fragment(fragment: str) -> str: | CODE |
| LOW⚡ | tools/nightly/buildkite_testcase_statistics.py | 168 | def _pytest_collect_sidecar_args(raw_line: str) -> list[str]: | CODE |
| LOW | tools/nightly/buildkite_testcase_statistics.py | 208 | def _parse_extra_args_from_line(raw_line: str) -> list[str]: | CODE |
| LOW | tools/nightly/buildkite_testcase_statistics.py | 213 | def _parse_collect_only_stdout(stdout: str, *, raise_on_empty: bool = True, stderr: str = "") -> list[str]: | CODE |
| LOW | tools/nightly/buildkite_testcase_statistics.py | 268 | def _format_pytest_collect_failure( | CODE |
| LOW | tools/nightly/buildkite_testcase_statistics.py | 383 | def get_docstring_for_node_id(repo_root: Path, node_id: str) -> str: | CODE |
| LOW | tools/nightly/buildkite_testcase_statistics.py | 426 | def get_test_file_for_node_id(node_id: str) -> str: | CODE |
| LOW | tools/nightly/generate_nightly_perf_excel.py | 219 | def _ensure_omni_summary_columns(summary_columns: list[str]) -> list[str]: | CODE |
| LOW | tools/nightly/generate_nightly_perf_excel.py | 262 | def _default_diffusion_input_dir(input_dir: str) -> str: | CODE |
| LOW | tools/nightly/generate_nightly_perf_excel.py | 456 | def _parse_diffusion_result_from_filename(filename: str) -> dict[str, Any]: | CODE |
| LOW | tools/nightly/generate_nightly_perf_excel.py | 516 | def _collect_diffusion_records(diffusion_input_dir: str) -> list[dict[str, Any]]: | CODE |
| LOW | tools/nightly/generate_nightly_perf_excel.py | 545 | def _process_diffusion_record(record: dict[str, Any]) -> dict[str, Any]: | CODE |
| LOW | tools/nightly/generate_nightly_perf_excel.py | 561 | def _apply_build_metadata_to_latest_only( | CODE |
| LOW | tools/nightly/generate_nightly_perf_excel.py | 599 | def _sort_records_for_summary(records: list[dict[str, Any]]) -> list[dict[str, Any]]: | CODE |
| LOW | tools/nightly/generate_nightly_perf_excel.py | 608 | def _sort_diffusion_records_for_summary(records: list[dict[str, Any]]) -> list[dict[str, Any]]: | CODE |
| LOW | tools/nightly/generate_nightly_perf_html.py | 52 | def _default_diffusion_input_dir(input_dir: str) -> str: | CODE |
| LOW | tools/nightly/generate_nightly_perf_html.py | 173 | def _parse_diffusion_from_filename(filename: str) -> dict[str, Any]: | CODE |
| LOW | tools/nightly/generate_nightly_perf_html.py | 192 | def _iter_diffusion_json_records(input_dir: str) -> Iterable[dict[str, Any]]: | CODE |
| LOW | tools/nightly/generate_nightly_perf_html.py | 248 | def _collect_diffusion_records(input_dir: str) -> list[dict[str, Any]]: | CODE |
| LOW | tools/nightly/run_nightly_jobs.sh | 565 | def timeout_kill_after_seconds() -> int: | CODE |
| LOW | tools/nightly/run_nightly_jobs.sh | 574 | def prepend_timeout_to_pytest(pytest_line: str, timeout_min: int | None) -> str: | CODE |
| LOW | tools/nightly/run_nightly_jobs.sh | 586 | def _write_job_timeouts_manifest(jobs_dir: Path, job_timeouts: dict[str, int]) -> None: | CODE |
| LOW | tools/nightly/run_nightly_jobs.sh | 660 | def local_test_files_by_filename(substr: str) -> list[str]: | CODE |
| LOW | tools/nightly/run_nightly_jobs.sh | 694 | def perf_json_matches_model_type(json_basename: str, model_types: list[str]) -> bool: | CODE |
| LOW | tools/pre_commit/check_test_marks.py | 80 | def get_files_missing_markers( | CODE |
| LOW | tools/wan22/assemble_wan22_i2v_diffusers.py | 62 | def _load_shard_files_from_index(index_file: Path, role: str) -> tuple[Path, ...]: | CODE |
| LOW | tests/metrics/test_stats.py | 18 | def test_orchestrator_aggregator_builds_summary() -> None: | CODE |
| LOW | tests/metrics/test_stats.py | 94 | def test_build_and_log_summary_multiple_requests() -> None: | CODE |
| LOW | tests/metrics/test_prometheus.py | 53 | def test_all_metric_families_present(self, scrape_output: str) -> None: | CODE |
| LOW⚡ | tests/metrics/test_prometheus.py | 102 | def test_pipeline_metrics_carry_model_name(self, scrape_output: str) -> None: | CODE |
| LOW⚡ | tests/metrics/test_prometheus.py | 107 | def test_no_legacy_engine_label(self, scrape_output: str) -> None: | CODE |
| LOW⚡ | tests/metrics/test_prometheus.py | 110 | def test_no_legacy_seconds_or_ms_families(self, scrape_output: str) -> None: | CODE |
| LOW⚡ | tests/metrics/test_prometheus.py | 120 | def test_omni_metrics_in_default_registry(self, scrape_output: str) -> None: | CODE |
| LOW⚡ | tests/metrics/test_prometheus.py | 124 | def test_process_metrics_in_default_registry(self, scrape_output: str) -> None: | CODE |
| LOW | tests/metrics/test_prometheus.py | 147 | def test_running_and_waiting_zero_after_request_completes(self, registry: CollectorRegistry) -> None: | CODE |
| LOW | tests/metrics/test_prometheus.py | 176 | def test_gauges_reflect_remaining_requests_after_one_completes(self, registry: CollectorRegistry) -> None: | CODE |
| LOW | tests/metrics/test_metrics_utils.py | 25 | def test_count_audio_chunk_frames(audio_chunk: object, expected: int) -> None: | CODE |
| LOW | tests/metrics/test_metrics_utils.py | 29 | def test_count_audio_frames_sums_audio_chunks() -> None: | CODE |
| LOW | tests/metrics/test_metrics_utils.py | 40 | def test_count_audio_frames_supports_model_outputs() -> None: | CODE |
| LOW | tests/metrics/test_metrics_utils.py | 52 | def test_count_audio_frames_returns_zero_without_audio(mm_out: dict[str, object]) -> None: | CODE |
| LOW | tests/metrics/test_metrics_utils.py | 72 | def test_count_image_pixels_sums_nested_values() -> None: | CODE |
| LOW | tests/metrics/test_metrics_utils.py | 84 | def test_count_tokens_from_outputs() -> None: | CODE |
| LOW⚡ | tests/metrics/test_transfer.py | 42 | def test_all_four_families_present(self, tx: OmniTransferMetrics) -> None: | CODE |
| LOW⚡ | tests/metrics/test_transfer.py | 52 | def test_legacy_ms_family_names_dropped(self, tx: OmniTransferMetrics) -> None: | CODE |
| LOW⚡ | tests/metrics/test_transfer.py | 68 | def test_size_observed_with_correct_labels(self, tx: OmniTransferMetrics) -> None: | CODE |
| LOW⚡ | tests/metrics/test_transfer.py | 124 | def test_multiple_edges_produce_independent_series(self, tx: OmniTransferMetrics) -> None: | CODE |
| LOW⚡ | tests/metrics/test_transfer.py | 157 | def test_size_uses_bytes_buckets(self, tx: OmniTransferMetrics) -> None: | CODE |
| LOW⚡ | tests/metrics/test_transfer.py | 165 | def test_time_families_use_seconds_fast_buckets(self, tx: OmniTransferMetrics) -> None: | CODE |
| 10715 more matches not shown… | ||||
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| MEDIUM | tools/run_jobs_common.sh | 26 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM | tools/run_jobs_common.sh | 28 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM | tools/run_jobs_common.sh | 179 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM | tools/run_jobs_common.sh | 186 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | tests/metrics/test_transfer.py | 36 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | tests/metrics/test_transfer.py | 38 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | tests/metrics/test_transfer.py | 62 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | tests/metrics/test_transfer.py | 64 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | tests/metrics/test_transfer.py | 118 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | tests/metrics/test_transfer.py | 120 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | tests/metrics/test_transfer.py | 151 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | tests/metrics/test_transfer.py | 153 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | tests/metrics/test_transfer.py | 173 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | tests/metrics/test_transfer.py | 175 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | tests/metrics/test_modality.py | 183 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | tests/metrics/test_modality.py | 185 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | tests/metrics/test_modality.py | 227 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | tests/metrics/test_modality.py | 229 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM | tests/metrics/test_modality.py | 38 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM | tests/metrics/test_modality.py | 40 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM | tests/metrics/test_modality.py | 88 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM | tests/metrics/test_modality.py | 90 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | tests/metrics/test_stat_logger.py | 70 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | tests/metrics/test_stat_logger.py | 72 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | tests/metrics/test_stat_logger.py | 92 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | tests/metrics/test_stat_logger.py | 94 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | tests/metrics/test_stat_logger.py | 219 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | tests/metrics/test_stat_logger.py | 221 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | tests/metrics/test_stat_logger.py | 255 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | tests/metrics/test_stat_logger.py | 257 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | tests/metrics/test_stat_logger.py | 278 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | tests/metrics/test_stat_logger.py | 280 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | tests/metrics/test_stat_logger.py | 305 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | tests/metrics/test_stat_logger.py | 309 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | tests/metrics/test_stat_logger.py | 371 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | tests/metrics/test_stat_logger.py | 378 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | tests/metrics/test_stat_logger.py | 404 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | tests/metrics/test_stat_logger.py | 411 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM | tests/metrics/test_stat_logger.py | 35 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM | tests/metrics/test_stat_logger.py | 37 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | tests/metrics/test_emit_calls.py | 28 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | tests/metrics/test_emit_calls.py | 30 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | tests/metrics/test_emit_calls.py | 365 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | tests/metrics/test_emit_calls.py | 368 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM | tests/metrics/test_emit_calls.py | 417 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM | tests/metrics/test_emit_calls.py | 419 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | …/omni_connectors/test_mooncake_transfer_engine_rdma.py | 443 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | …/omni_connectors/test_mooncake_transfer_engine_rdma.py | 445 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM | …/omni_connectors/test_mooncake_transfer_engine_rdma.py | 35 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM | …/omni_connectors/test_mooncake_transfer_engine_rdma.py | 37 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM | …/omni_connectors/test_mooncake_transfer_engine_rdma.py | 190 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM | …/omni_connectors/test_mooncake_transfer_engine_rdma.py | 192 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM | …/omni_connectors/test_mooncake_transfer_engine_rdma.py | 250 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM | …/omni_connectors/test_mooncake_transfer_engine_rdma.py | 252 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM | …/omni_connectors/test_mooncake_transfer_engine_rdma.py | 471 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM | …/omni_connectors/test_mooncake_transfer_engine_rdma.py | 473 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM | …/omni_connectors/test_mooncake_transfer_engine_rdma.py | 519 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM | …/omni_connectors/test_mooncake_transfer_engine_rdma.py | 521 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | …ributed/omni_connectors/test_chunk_transfer_adapter.py | 1785 | # --------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | …ributed/omni_connectors/test_chunk_transfer_adapter.py | 1787 | # --------------------------------------------------------------- | COMMENT |
| 2037 more matches not shown… | ||||
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| LOW | tools/configure_stage_memory.py | 19 | CODE | |
| LOW | tools/nightly/send_nightly_email.py | 9 | CODE | |
| LOW | tools/nightly/buildkite_testcase_statistics.py | 18 | CODE | |
| LOW | tools/nightly/generate_nightly_perf_excel.py | 7 | CODE | |
| LOW | tools/nightly/generate_nightly_perf_html.py | 7 | CODE | |
| LOW | tools/pre_commit/mypy.py | 18 | CODE | |
| LOW | tools/pre_commit/check_tts_adapter.py | 20 | CODE | |
| LOW | tools/pre_commit/check_buildkite.py | 12 | CODE | |
| LOW | tools/wan22/assemble_wan22_i2v_diffusers.py | 22 | CODE | |
| LOW | tools/baselines/omni_hash_baseline.py | 44 | CODE | |
| LOW | tests/conftest.py | 13 | CODE | |
| LOW | tests/metrics/test_stats.py | 1 | CODE | |
| LOW | tests/metrics/test_prometheus.py | 1 | CODE | |
| LOW | tests/metrics/test_metrics_utils.py | 1 | CODE | |
| LOW | tests/metrics/test_transfer.py | 1 | CODE | |
| LOW | tests/metrics/test_modality.py | 4 | CODE | |
| LOW | tests/metrics/test_stat_logger.py | 1 | CODE | |
| LOW | tests/metrics/test_emit_calls.py | 9 | CODE | |
| LOW | tests/platforms/npu/test_310p_patches.py | 9 | CODE | |
| LOW | tests/platforms/npu/test_step_audio2_token2wav.py | 5 | CODE | |
| LOW | …/platforms/npu/test_diffusion_attn_backend_selector.py | 15 | CODE | |
| LOW | tests/platforms/npu/quant/test_kv_quant_npu.py | 10 | CODE | |
| LOW | tests/platforms/npu/quant/test_kv_quant_npu.py | 55 | CODE | |
| LOW | tests/tools/test_check_tts_adapter.py | 9 | CODE | |
| LOW | …e/sched/test_omni_ar_scheduler_free_request_cleanup.py | 8 | CODE | |
| LOW | …e/sched/test_omni_ar_scheduler_free_request_cleanup.py | 17 | CODE | |
| LOW | tests/core/sched/test_omni_ar_scheduler_stale_drain.py | 27 | CODE | |
| LOW | tests/core/sched/test_omni_ar_scheduler_stale_drain.py | 37 | CODE | |
| LOW | …sched/test_omni_scheduler_input_coordinator_cleanup.py | 1 | CODE | |
| LOW | …s/core/sched/test_omni_sched_prefill_stats_finalize.py | 13 | CODE | |
| LOW | …s/core/sched/test_omni_sched_prefill_stats_finalize.py | 22 | CODE | |
| LOW | …sched/test_omni_generation_scheduler_update_session.py | 9 | CODE | |
| LOW | …sched/test_omni_generation_scheduler_update_session.py | 20 | CODE | |
| LOW | tests/core/sched/test_omni_scheduling_coordinator.py | 12 | CODE | |
| LOW | tests/core/sched/test_omni_scheduler_mixin_timeouts.py | 17 | CODE | |
| LOW | tests/core/sched/test_omni_sched_deferred_free_fence.py | 24 | CODE | |
| LOW | tests/core/sched/test_omni_sched_deferred_free_fence.py | 33 | CODE | |
| LOW | tests/core/sched/test_sched_package_imports.py | 4 | CODE | |
| LOW | …ched/test_omni_ar_scheduler_unified_decode_deferral.py | 1 | CODE | |
| LOW | tests/core/sched/test_omni_ar_scheduler_streaming.py | 6 | CODE | |
| LOW | tests/core/sched/test_omni_ar_scheduler_streaming.py | 16 | CODE | |
| LOW | tests/core/sched/test_omni_sched_ec_request_finish.py | 15 | CODE | |
| LOW | tests/core/sched/test_omni_sched_ec_request_finish.py | 24 | CODE | |
| LOW | tests/core/sched/test_omni_ar_scheduler_logprobs.py | 3 | CODE | |
| LOW | …ore/sched/test_omni_scheduler_finish_requests_purge.py | 17 | CODE | |
| LOW | tests/host_weight_runtime/test_contracts.py | 5 | CODE | |
| LOW | tests/host_weight_runtime/test_runtime_resolution.py | 5 | CODE | |
| LOW | tests/host_weight_runtime/test_filesystem_store.py | 5 | CODE | |
| LOW | tests/config/test_omni_config.py | 5 | CODE | |
| LOW | tests/config/test_config_import_cycle.py | 5 | CODE | |
| LOW | tests/config/test_forced_aligner_injection.py | 3 | CODE | |
| LOW | tests/config/composable_parallel/test_translator.py | 5 | CODE | |
| LOW | …config/composable_parallel/test_factory_integration.py | 5 | CODE | |
| LOW | …sts/config/composable_parallel/test_strategy_loader.py | 5 | CODE | |
| LOW | tests/config/composable_parallel/test_spec_types.py | 5 | CODE | |
| LOW | tests/config/composable_parallel/test_apply.py | 5 | CODE | |
| LOW | tests/model_executor/test_cfg_pairing.py | 11 | CODE | |
| LOW | tests/model_executor/helpers.py | 27 | CODE | |
| LOW | tests/model_executor/models/registry.py | 10 | CODE | |
| LOW | …sts/model_executor/models/test_encoder_quant_config.py | 18 | CODE | |
| 1902 more matches not shown… | ||||
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| LOW | collect_env.py | 499 | except Exception: | CODE |
| MEDIUM | collect_env.py | 293 | def get_vllm_omni_version(run_lambda): | CODE |
| LOW | setup.py | 39 | except Exception as e: | CODE |
| LOW⚡ | setup.py | 94 | except Exception: | CODE |
| LOW⚡ | setup.py | 103 | except Exception: | CODE |
| LOW⚡ | setup.py | 112 | except Exception: | CODE |
| LOW | setup.py | 150 | except Exception as e: | CODE |
| LOW | tools/configure_stage_memory.py | 39 | except Exception: | CODE |
| LOW | tools/configure_stage_memory.py | 65 | except Exception as e: | CODE |
| MEDIUM | tools/configure_stage_memory.py | 317 | print(f"Error: {config_path} not found", file=sys.stderr) | STRING |
| LOW | tools/nightly/send_nightly_email.py | 277 | except Exception as e: | CODE |
| LOW | tools/nightly/buildkite_testcase_statistics.py | 402 | except Exception: | CODE |
| LOW | tools/pre_commit/check_buildkite.py | 109 | except Exception as exc: # noqa: BLE001 - surface expander errors as hook failures | CODE |
| LOW | tools/wan22/assemble_wan22_i2v_diffusers.py | 66 | except Exception as exc: | CODE |
| LOW | tools/baselines/omni_hash_baseline.py | 90 | except Exception: | CODE |
| LOW | tools/baselines/omni_hash_baseline.py | 99 | except Exception: | CODE |
| MEDIUM | tools/baselines/omni_hash_baseline.py | 87 | def _git_commit() -> str: | CODE |
| LOW⚡ | tests/core/sched/test_generation_scheduler_restore.py | 60 | except Exception: | CODE |
| LOW | tests/core/sched/test_generation_scheduler_restore.py | 83 | except Exception: | CODE |
| MEDIUM | …/omni_connectors/test_mooncake_transfer_engine_rdma.py | 402 | def worker(rid, data): | CODE |
| MEDIUM | …/omni_connectors/test_mooncake_transfer_engine_rdma.py | 613 | def producer_worker(rid, tensor): | CODE |
| LOW | …/omni_connectors/test_mooncake_transfer_engine_rdma.py | 75 | except Exception: | CODE |
| LOW | …/omni_connectors/test_mooncake_transfer_engine_rdma.py | 124 | except Exception as exc: | CODE |
| LOW | …/omni_connectors/test_mooncake_transfer_engine_rdma.py | 157 | except Exception as exc: | CODE |
| LOW | …/omni_connectors/test_mooncake_transfer_engine_rdma.py | 408 | except Exception as e: | CODE |
| LOW | …/omni_connectors/test_mooncake_transfer_engine_rdma.py | 562 | except Exception as e: | CODE |
| LOW | …/omni_connectors/test_mooncake_transfer_engine_rdma.py | 624 | except Exception as e: | CODE |
| LOW | …/omni_connectors/test_mooncake_transfer_engine_rdma.py | 657 | except Exception as e: | CODE |
| LOW⚡ | …ributed/omni_connectors/test_chunk_transfer_adapter.py | 1173 | except Exception as e: | CODE |
| LOW⚡ | …ributed/omni_connectors/test_chunk_transfer_adapter.py | 1182 | except Exception as e: | CODE |
| LOW | …mni_connectors/test_mooncake_transfer_engine_buffer.py | 78 | except Exception as e: | CODE |
| MEDIUM | …mni_connectors/test_mooncake_transfer_engine_buffer.py | 72 | def worker(worker_id): | CODE |
| LOW | …ributed/omni_connectors/test_omni_connector_configs.py | 89 | except Exception as e: | CODE |
| LOW | tests/model_executor/helpers.py | 54 | except Exception: | CODE |
| LOW | …s/model_executor/models/glm_image/test_glm_image_ar.py | 150 | except Exception: | CODE |
| LOW | …l_executor/models/qwen3_tts/test_cuda_graph_decoder.py | 33 | except Exception: | CODE |
| LOW | tests/utils/test_speaker_cache.py | 127 | except Exception as e: | CODE |
| MEDIUM | tests/utils/test_speaker_cache.py | 122 | def worker(tid): | CODE |
| LOW | tests/diffusion/test_diffusion_config_fields.py | 18 | except Exception: | CODE |
| LOW | tests/diffusion/test_diffusion_streaming_output.py | 86 | except Exception as exc: | CODE |
| LOW | tests/diffusion/test_diffusion_streaming_output.py | 204 | except Exception: | CODE |
| MEDIUM | tests/diffusion/test_diffusion_streaming_output.py | 83 | def step_streaming(self, request): | CODE |
| LOW | …iffusion/offloader/test_diffusion_layerwise_offload.py | 149 | except Exception: | CODE |
| LOW | tests/diffusion/offloader/test_diffusion_cpu_offload.py | 123 | except Exception: | CODE |
| LOW | tests/diffusion/attention/test_trtllm_attn.py | 28 | except Exception: | CODE |
| LOW | tests/diffusion/attention/test_flash_attn.py | 735 | except Exception as e: | CODE |
| LOW | tests/diffusion/attention/test_flash_attn.py | 744 | except Exception as e: | CODE |
| LOW | tests/diffusion/ar_diffusion/test_paged_attention.py | 81 | except Exception: | CODE |
| LOW | tests/diffusion/ar_diffusion/test_paged_attention.py | 92 | except Exception: | CODE |
| LOW | tests/diffusion/models/pi0/test_pi0_parity.py | 354 | except Exception as e: # noqa: BLE001 | CODE |
| LOW | …s/hunyuan_image3/test_hunyuan_image3_it2i_ar_format.py | 114 | except Exception: | CODE |
| LOW⚡ | tests/diffusion/batching/test_diffusion_batching.py | 514 | except Exception as e: | CODE |
| LOW | tests/diffusion/batching/test_diffusion_batching.py | 338 | except Exception as e: | CODE |
| LOW | tests/diffusion/batching/test_diffusion_batching.py | 378 | except Exception as e: | CODE |
| LOW | tests/diffusion/batching/test_diffusion_batching.py | 480 | except Exception as e: | CODE |
| LOW | tests/dfx/conftest.py | 172 | except Exception as e: | CODE |
| LOW⚡ | tests/dfx/reliability/test_reliability_qwen3_omni.py | 300 | except Exception: # noqa: BLE001 | CODE |
| LOW⚡ | tests/dfx/reliability/test_reliability_qwen3_omni.py | 314 | except Exception as exc: # noqa: BLE001 | CODE |
| LOW | tests/dfx/reliability/test_reliability_qwen3_omni.py | 335 | except Exception as exc: # noqa: BLE001 | CODE |
| LOW | tests/dfx/reliability/test_reliability_qwen3_omni.py | 451 | except Exception as exc: | CODE |
| 1262 more matches not shown… | ||||
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| LOW | collect_env.py | 187 | CODE | |
| LOW | collect_env.py | 356 | CODE | |
| LOW | setup.py | 43 | CODE | |
| LOW | setup.py | 123 | CODE | |
| LOW | tools/configure_stage_memory.py | 192 | CODE | |
| LOW | tools/nightly/send_nightly_email.py | 211 | CODE | |
| LOW | tools/nightly/buildkite_testcase_statistics.py | 324 | CODE | |
| LOW | tools/nightly/buildkite_testcase_statistics.py | 383 | CODE | |
| LOW | tools/nightly/buildkite_testcase_statistics.py | 437 | CODE | |
| LOW | tools/nightly/buildkite_testcase_statistics.py | 988 | CODE | |
| LOW | tools/nightly/generate_nightly_perf_excel.py | 727 | CODE | |
| LOW | tools/nightly/generate_nightly_perf_html.py | 192 | CODE | |
| LOW | tools/pre_commit/check_tts_adapter.py | 82 | CODE | |
| LOW | tools/pre_commit/check_spdx_header.py | 115 | CODE | |
| LOW | tools/pre_commit/check_spdx_header.py | 164 | CODE | |
| LOW | tools/baselines/omni_hash_baseline.py | 144 | CODE | |
| LOW | tests/host_weight_runtime/test_contracts.py | 208 | CODE | |
| LOW | tests/host_weight_runtime/test_filesystem_store.py | 1728 | CODE | |
| LOW | tests/host_weight_runtime/test_filesystem_store.py | 291 | CODE | |
| LOW | …/omni_connectors/test_mooncake_transfer_engine_rdma.py | 40 | CODE | |
| LOW | …/omni_connectors/test_mooncake_transfer_engine_rdma.py | 80 | CODE | |
| LOW | …/omni_connectors/test_mooncake_transfer_engine_rdma.py | 360 | CODE | |
| LOW | …/omni_connectors/test_mooncake_transfer_engine_rdma.py | 597 | CODE | |
| LOW | …/omni_connectors/test_mooncake_transfer_engine_rdma.py | 630 | CODE | |
| LOW | tests/config/test_environment_variables.py | 26 | CODE | |
| LOW | tests/config/test_environment_variables.py | 44 | CODE | |
| LOW | tests/config/test_environment_variables.py | 71 | CODE | |
| LOW | tests/config/test_environment_variables.py | 114 | CODE | |
| LOW | tests/config/test_environment_variables.py | 136 | CODE | |
| LOW | tests/model_executor/models/test_gepard_wiring.py | 168 | CODE | |
| LOW | tests/model_executor/models/test_omni_processing.py | 162 | CODE | |
| LOW | …executor/models/qwen3_tts/test_code_predictor_dtype.py | 821 | CODE | |
| LOW | …executor/models/qwen3_tts/test_code_predictor_dtype.py | 934 | CODE | |
| LOW | …executor/models/qwen3_tts/test_code_predictor_dtype.py | 1061 | CODE | |
| LOW | …el_executor/models/qwen2_5_omni/test_weight_loading.py | 7 | CODE | |
| LOW | tests/utils/test_mm_outputs.py | 32 | CODE | |
| LOW | tests/model_tests/conftest.py | 11 | CODE | |
| LOW | tests/model_tests/diffusion/test_common_online.py | 37 | CODE | |
| LOW | tests/model_tests/diffusion/test_common_offline.py | 30 | CODE | |
| LOW | tests/model_tests/diffusion/task_runners.py | 240 | CODE | |
| LOW | tests/model_tests/diffusion/task_runners.py | 266 | CODE | |
| LOW | tests/diffusion/test_multiproc_engine_concurrency.py | 100 | CODE | |
| LOW | tests/diffusion/test_multiproc_engine_concurrency.py | 105 | CODE | |
| LOW | tests/diffusion/test_diffusion_engine_metrics.py | 87 | CODE | |
| LOW | tests/diffusion/test_diffusion_step_pipeline.py | 437 | CODE | |
| LOW | tests/diffusion/attention/test_ulysses_uaa.py | 57 | CODE | |
| LOW | tests/diffusion/attention/test_flash_attn.py | 536 | CODE | |
| LOW | tests/diffusion/cache/test_cache_dit.py | 96 | CODE | |
| LOW | tests/diffusion/ar_diffusion/test_paged_attention.py | 63 | CODE | |
| LOW | …sion/models/lingbot_world/test_lingbot_world_typing.py | 21 | CODE | |
| LOW | …ion/models/hunyuan_image3/test_diffusion_kv_request.py | 240 | CODE | |
| LOW | …ts/diffusion/models/minimax_h3/test_minimax_h3_lora.py | 46 | CODE | |
| LOW | tests/diffusion/quantization/test_fp8_config.py | 336 | CODE | |
| LOW | tests/diffusion/quantization/test_quantization_fp8.py | 191 | CODE | |
| LOW | tests/diffusion/batching/test_diffusion_batching.py | 284 | CODE | |
| LOW | tests/diffusion/lora/test_loader.py | 133 | CODE | |
| LOW | …/diffusion/diffusion_backend/test_diffusers_backend.py | 147 | CODE | |
| LOW | tests/dfx/conftest.py | 38 | CODE | |
| LOW | tests/dfx/stability/helpers.py | 32 | CODE | |
| LOW | tests/dfx/stability/helpers.py | 65 | CODE | |
| 841 more matches not shown… | ||||
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| HIGH⚡ | tools/nightly/buildkite_testcase_statistics.py | 877 | const fileName = (fileDetail.dataset.testFile || fileDetail.textContent || "").toLowerCase(); | CODE |
| HIGH⚡ | tools/nightly/buildkite_testcase_statistics.py | 878 | const pathOk = !query || fileName.includes(query); | CODE |
| HIGH⚡ | tools/nightly/buildkite_testcase_statistics.py | 883 | const matched = pathOk && anyRowVisible; | CODE |
| HIGH | tools/nightly/generate_nightly_perf_excel.py | 120 | # and max_concurrency (null); leave those as-is. If they become float in the future, they are | COMMENT |
| HIGH⚡ | tools/nightly/generate_nightly_perf_html.py | 648 | if (value === null || value === undefined) return ""; | CODE |
| HIGH⚡ | tools/nightly/generate_nightly_perf_html.py | 648 | if (value === null || value === undefined) return ""; | CODE |
| HIGH⚡ | tools/nightly/generate_nightly_perf_html.py | 658 | const items = new Set(arr.filter(v => v !== null && v !== undefined && String(v).trim() !== "")); | CODE |
| HIGH⚡ | tools/nightly/generate_nightly_perf_html.py | 663 | if (v === null || v === undefined) return null; | CODE |
| HIGH⚡ | tools/nightly/generate_nightly_perf_html.py | 663 | if (v === null || v === undefined) return null; | CODE |
| HIGH⚡ | tools/nightly/generate_nightly_perf_html.py | 664 | if (typeof v === "number") return Number.isFinite(v) ? v : null; | CODE |
| HIGH⚡ | tools/nightly/generate_nightly_perf_html.py | 666 | if (!s || s.toLowerCase() === "inf") return null; | CODE |
| HIGH⚡ | tools/nightly/generate_nightly_perf_html.py | 666 | if (!s || s.toLowerCase() === "inf") return null; | CODE |
| HIGH⚡ | tools/nightly/generate_nightly_perf_html.py | 668 | return Number.isFinite(n) ? n : null; | CODE |
| HIGH⚡ | tools/nightly/generate_nightly_perf_html.py | 1005 | if (x < padLeft || x > padLeft + innerW || y < padTop || y > padTop + innerH) {{ | CODE |
| HIGH⚡ | tools/nightly/generate_nightly_perf_html.py | 1012 | let best = null; | CODE |
| HIGH⚡ | tools/nightly/generate_nightly_perf_html.py | 1018 | if (!best || dist < best.dist) {{ | CODE |
| HIGH⚡ | tools/nightly/generate_nightly_perf_html.py | 1027 | .filter(([, value]) => value !== null && value !== undefined && String(value).trim() !== "") | CODE |
| HIGH⚡ | tools/nightly/generate_nightly_perf_html.py | 1179 | if (filters.model && String(row[filters.modelKey] || "") !== filters.model) return false; | CODE |
| HIGH⚡ | tools/nightly/generate_nightly_perf_html.py | 1180 | if (filters.testName && String(row.test_name || "") !== filters.testName) return false; | CODE |
| HIGH⚡ | tools/nightly/generate_nightly_perf_html.py | 1181 | if (filters.datasetName && String(row[filters.datasetKey] || "") !== filters.datasetName) return false; | CODE |
| HIGH | tools/nightly/generate_nightly_perf_html.py | 704 | row.max_concurrency !== undefined && row.max_concurrency !== null ? `c${{row.max_concurrency}}` : "", | CODE |
| HIGH | tools/nightly/generate_nightly_perf_html.py | 705 | row.num_prompts !== undefined && row.num_prompts !== null ? `p${{row.num_prompts}}` : "", | CODE |
| HIGH | tools/nightly/generate_nightly_perf_html.py | 720 | row.max_concurrency !== undefined && row.max_concurrency !== null ? `c${{row.max_concurrency}}` : "", | CODE |
| HIGH | tools/nightly/generate_nightly_perf_html.py | 721 | row.num_prompts !== undefined && row.num_prompts !== null ? `p${{row.num_prompts}}` : "", | CODE |
| HIGH | tools/nightly/generate_nightly_perf_html.py | 733 | const raw = value === null || value === undefined ? "" : String(value); | CODE |
| HIGH | tools/nightly/generate_nightly_perf_html.py | 786 | if (previous && values.map((v) => String(v)).includes(String(previous))) {{ | CODE |
| HIGH | tools/nightly/generate_nightly_perf_html.py | 871 | if (display.title && display.title !== display.text) {{ | CODE |
| HIGH | tools/nightly/generate_nightly_perf_html.py | 890 | const dpr = window.devicePixelRatio || 1; | CODE |
| HIGH | tools/nightly/generate_nightly_perf_html.py | 912 | xLabels.push(p.x); | CODE |
| HIGH | tools/nightly/generate_nightly_perf_html.py | 1057 | .filter((value) => value !== null && value !== undefined && String(value).trim() !== "") | CODE |
| HIGH | tools/nightly/generate_nightly_perf_html.py | 1062 | if (y === null) return; | CODE |
| HIGH | tools/nightly/generate_nightly_perf_html.py | 1074 | seriesByKey.get(key).points.push({{ x: dateStr, y, meta }}); | CODE |
| HIGH | tools/nightly/generate_nightly_perf_html.py | 1100 | if (latest && latest.date) {{ | CODE |
| HIGH | tools/nightly/generate_nightly_perf_html.py | 1101 | lines.push(`date: <code>${{escapeHtml(formatDate(latest.date))}}</code>`); | CODE |
| HIGH | tools/nightly/generate_nightly_perf_html.py | 1112 | lines.push(`${{escapeHtml(keyName)}}: <code>${{escapeHtml(String(latest[keyName]))}}</code>`); | CODE |
| HIGH | tools/nightly/generate_nightly_perf_html.py | 1116 | if (value) lines.push(`${{escapeHtml(metric)}}: <code>${{value}}</code>`); | CODE |
| HIGH | tools/nightly/generate_nightly_perf_html.py | 1134 | if (!r || r.width < 10 || r.height < 10) return; | CODE |
| HIGH | tools/nightly/generate_nightly_perf_html.py | 1194 | if (filters.backend && String(row[filters.backendKey] || "") !== filters.backend) return false; | CODE |
| HIGH | tools/nightly/generate_nightly_perf_html.py | 1211 | const values = rows.map((row) => toNumber(row[key])).filter((v) => v !== null); | CODE |
| HIGH | tools/nightly/generate_nightly_perf_html.py | 1400 | if (extraKey && extraInput && !extraInput.value.trim()) {{ | CODE |
| HIGH | tools/nightly/generate_nightly_perf_html.py | 1412 | const latest = filteredAsc[filteredAsc.length - 1] || null; | CODE |
| HIGH⚡ | tests/diffusion/ar_diffusion/test_slot_mapping.py | 61 | Every slot for the in-flight chunk must land inside a real (non-null) block | STRING |
| HIGH | tests/examples/test_minicpmo_realtime_web_static.py | 41 | assert "event.event || event" in source | CODE |
| HIGH | tests/examples/test_minicpmo_realtime_web_static.py | 147 | this.port = { onmessage: null, postMessage: () => {} }; | STRING |
| HIGH | tests/examples/test_minicpmo_realtime_web_static.py | 150 | let Processor = null; | STRING |
| HIGH⚡ | tests/examples/test_minicpmo_realtime_web_static.py | 227 | onmessage: null, | STRING |
| HIGH⚡ | tests/examples/test_minicpmo_realtime_web_static.py | 228 | postMessage: (message) => messages.push(message), | STRING |
| HIGH⚡ | tests/examples/test_minicpmo_realtime_web_static.py | 232 | let Processor = null; | STRING |
| HIGH | tests/examples/test_minicpmo_realtime_web_static.py | 260 | tail.every((sample, index) => index === 0 || sample <= tail[index - 1]), | STRING |
| HIGH | tests/dfx/stability/scripts/generate_report.py | 305 | charts.push(new Chart(canvas, {{ | CODE |
| HIGH | tests/entrypoints/test_stream_finish_reason.py | 123 | """text+audio: text finish sends finish_reason=null, audio sends stop.""" | STRING |
| HIGH | tests/entrypoints/openai_api/test_serving_speech.py | 3641 | '{"error":{"message":"engine dead","type":"InternalServerError","param":null,' | CODE |
| HIGH | …s/entrypoints/openai_api/test_serving_speech_stream.py | 454 | # Real-time audio chunk (timestamps null), then the timestamp frame. | COMMENT |
| HIGH⚡ | …ne_serving/text_to_speech/moss_tts_nano/gradio_demo.py | 55 | this.buf = null; | CODE |
| HIGH⚡ | …ne_serving/text_to_speech/moss_tts_nano/gradio_demo.py | 60 | if (e.data && e.data.type === 'clear') { | CODE |
| HIGH⚡ | …ne_serving/text_to_speech/moss_tts_nano/gradio_demo.py | 61 | this.queue = []; this.buf = null; this.pos = 0; this.played = 0; | CODE |
| HIGH⚡ | …ne_serving/text_to_speech/moss_tts_nano/gradio_demo.py | 182 | if (mTtfp && st.ttfp != null) mTtfp.textContent = st.ttfp.toFixed(0) + 'ms'; | CODE |
| HIGH⚡ | …ne_serving/text_to_speech/moss_tts_nano/gradio_demo.py | 182 | if (mTtfp && st.ttfp != null) mTtfp.textContent = st.ttfp.toFixed(0) + 'ms'; | CODE |
| HIGH⚡ | …ne_serving/text_to_speech/moss_tts_nano/gradio_demo.py | 185 | if (dur > 0 && elapsed > 0) {{ | CODE |
| HIGH | …ne_serving/text_to_speech/moss_tts_nano/gradio_demo.py | 73 | this.queue.push(mono); | CODE |
| 59 more matches not shown… | ||||
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| HIGH | …buted/omni_connectors/test_bagel_mooncake_connector.py | 0 | validate that image pixels match expected reference values. args: image: the pil image to validate. reference_pixels: li | STRING |
| HIGH | …/omni_connectors/test_bagel_shared_memory_connector.py | 0 | validate that image pixels match expected reference values. args: image: the pil image to validate. reference_pixels: li | STRING |
| HIGH | …line_inference/test_sensenova_u1_text2img_expansion.py | 0 | validate that image pixels match expected reference values. args: image: the pil image to validate. reference_pixels: li | STRING |
| HIGH | …fline_inference/test_sensenova_u1_img2img_expansion.py | 0 | validate that image pixels match expected reference values. args: image: the pil image to validate. reference_pixels: li | STRING |
| HIGH | …l_executor/models/voxcpm2/test_talker_output_marker.py | 0 | defer talker import (pulls vllm model_executor) until first use. | STRING |
| HIGH | …_executor/models/voxcpm2/test_talker_state_eviction.py | 0 | defer talker import (pulls vllm model_executor) until first use. | STRING |
| HIGH | tests/model_executor/models/dots_tts/test_talker.py | 0 | defer talker import (pulls vllm model_executor) until first use. | STRING |
| HIGH | …del_executor/models/moss_tts/test_per_request_codes.py | 0 | defer talker import (pulls vllm model_executor) until first use. | STRING |
| HIGH | …s/e2e/offline_inference/test_qwen2_5_omni_expansion.py | 0 | test audio and text input processing and text/audio output generation via openai api. deploy setting: default yaml input | STRING |
| HIGH | tests/e2e/online_serving/test_qwen3_tts_customvoice.py | 0 | test audio and text input processing and text/audio output generation via openai api. deploy setting: default yaml input | STRING |
| HIGH | tests/e2e/online_serving/test_qwen3_tts_base.py | 0 | test audio and text input processing and text/audio output generation via openai api. deploy setting: default yaml input | STRING |
| HIGH | tests/e2e/online_serving/test_qwen2_5_omni_expansion.py | 0 | test audio and text input processing and text/audio output generation via openai api. deploy setting: default yaml input | STRING |
| HIGH | tests/e2e/online_serving/test_qwen3_omni.py | 0 | test audio and text input processing and text/audio output generation via openai api. deploy setting: default yaml input | STRING |
| HIGH | docs/contributing/ci/test_writing_guide.md | 0 | test audio and text input processing and text/audio output generation via openai api. deploy setting: default yaml input | STRING |
| HIGH | tests/e2e/online_serving/test_mimo_audio_expansion.py | 0 | test audio and text input processing and text/audio output generation via openai api. deploy setting: default yaml input | STRING |
| HIGH | tests/e2e/online_serving/test_wan22_t2v.py | 0 | return a single default ``omniserverparams`` row (no extra ``server_args``). | STRING |
| HIGH | tests/e2e/online_serving/test_qwen_image.py | 0 | return a single default ``omniserverparams`` row (no extra ``server_args``). | STRING |
| HIGH | tests/e2e/online_serving/test_qwen_image_layered.py | 0 | return a single default ``omniserverparams`` row (no extra ``server_args``). | STRING |
| HIGH | tests/e2e/online_serving/test_boogu_image.py | 0 | return a single default ``omniserverparams`` row (no extra ``server_args``). | STRING |
| HIGH | …ts/e2e/online_serving/test_qwen3_tts_base_expansion.py | 0 | e2e online tests for qwen3-tts model with text input and audio output. these tests verify the /v1/audio/speech endpoint | STRING |
| HIGH | …online_serving/test_qwen3_tts_customvoice_expansion.py | 0 | e2e online tests for qwen3-tts model with text input and audio output. these tests verify the /v1/audio/speech endpoint | STRING |
| HIGH | tests/e2e/online_serving/test_qwen3_tts_customvoice.py | 0 | e2e online tests for qwen3-tts model with text input and audio output. these tests verify the /v1/audio/speech endpoint | STRING |
| HIGH | tests/e2e/online_serving/test_qwen3_tts_base.py | 0 | e2e online tests for qwen3-tts model with text input and audio output. these tests verify the /v1/audio/speech endpoint | STRING |
| HIGH | …ts/e2e/online_serving/test_qwen3_tts_base_expansion.py | 0 | text prompt for text-to-audio tests (same as test_qwen3_omni - beijing test case). | STRING |
| HIGH | …online_serving/test_qwen3_tts_customvoice_expansion.py | 0 | text prompt for text-to-audio tests (same as test_qwen3_omni - beijing test case). | STRING |
| HIGH | tests/e2e/online_serving/test_qwen3_tts_customvoice.py | 0 | text prompt for text-to-audio tests (same as test_qwen3_omni - beijing test case). | STRING |
| HIGH | tests/e2e/online_serving/test_qwen3_tts_base.py | 0 | text prompt for text-to-audio tests (same as test_qwen3_omni - beijing test case). | STRING |
| HIGH | …ts/e2e/online_serving/test_qwen3_tts_base_expansion.py | 0 | batch size for concurrent requests (same as test_qwen3_omni). | STRING |
| HIGH | …online_serving/test_qwen3_tts_customvoice_expansion.py | 0 | batch size for concurrent requests (same as test_qwen3_omni). | STRING |
| HIGH | tests/e2e/online_serving/test_qwen3_tts_customvoice.py | 0 | batch size for concurrent requests (same as test_qwen3_omni). | STRING |
| HIGH | tests/e2e/online_serving/test_qwen3_tts_base.py | 0 | batch size for concurrent requests (same as test_qwen3_omni). | STRING |
| HIGH | …ts/e2e/online_serving/test_qwen3_tts_base_expansion.py | 0 | test text input processing and audio output via openai api. deploy setting: default yaml input modal: text output modal: | STRING |
| HIGH | …online_serving/test_qwen3_tts_customvoice_expansion.py | 0 | test text input processing and audio output via openai api. deploy setting: default yaml input modal: text output modal: | STRING |
| HIGH | tests/e2e/online_serving/test_qwen3_tts_customvoice.py | 0 | test text input processing and audio output via openai api. deploy setting: default yaml input modal: text output modal: | STRING |
| HIGH | tests/e2e/online_serving/test_qwen3_tts_base.py | 0 | test text input processing and audio output via openai api. deploy setting: default yaml input modal: text output modal: | STRING |
| HIGH | tests/e2e/online_serving/test_qwen2_5_omni_expansion.py | 0 | test text input processing and text-only output generation via openai api. deploy setting: default yaml input modal: tex | STRING |
| HIGH | tests/e2e/online_serving/test_qwen3_omni.py | 0 | test text input processing and text-only output generation via openai api. deploy setting: default yaml input modal: tex | STRING |
| HIGH | tests/e2e/online_serving/test_mimo_audio_expansion.py | 0 | test text input processing and text-only output generation via openai api. deploy setting: default yaml input modal: tex | STRING |
| HIGH | …ne_serving/text_to_speech/moss_tts_nano/gradio_demo.py | 0 | <div id="tts-player"> <div style="display:flex; align-items:center; gap:10px;"> <div id="tts-status-dot" style="width:10 | STRING |
| HIGH | …online_serving/text_to_speech/qwen3_tts/gradio_demo.py | 0 | <div id="tts-player"> <div style="display:flex; align-items:center; gap:10px;"> <div id="tts-status-dot" style="width:10 | STRING |
| HIGH | …s/online_serving/text_to_speech/voxcpm2/gradio_demo.py | 0 | <div id="tts-player"> <div style="display:flex; align-items:center; gap:10px;"> <div id="tts-status-dot" style="width:10 | STRING |
| HIGH | …ne_serving/text_to_speech/moss_tts_nano/gradio_demo.py | 0 | <script> const sr = {sample_rate}; const wc = {json.dumps(worklet_js)}; let ctx = null, node = null, abort = null, gen = | STRING |
| HIGH | …s/online_serving/text_to_speech/voxcpm2/gradio_demo.py | 0 | <script> const sr = {sample_rate}; const wc = {json.dumps(worklet_js)}; let ctx = null, node = null, abort = null, gen = | STRING |
| HIGH | …online_serving/text_to_speech/qwen3_tts/gradio_demo.py | 0 | <script> const sr = {sample_rate}; const wc = {json.dumps(worklet_js)}; let ctx = null, node = null, abort = null, gen = | STRING |
| HIGH | …ne_serving/text_to_speech/moss_tts_nano/gradio_demo.py | 0 | encode gradio audio input (sample_rate, numpy_array) to base64 data url. | STRING |
| HIGH | …s/online_serving/text_to_speech/glm_tts/gradio_demo.py | 0 | encode gradio audio input (sample_rate, numpy_array) to base64 data url. | STRING |
| HIGH | …line_serving/text_to_speech/fish_speech/gradio_demo.py | 0 | encode gradio audio input (sample_rate, numpy_array) to base64 data url. | STRING |
| HIGH | …/online_serving/text_to_speech/qwen3_tts/tts_common.py | 0 | encode gradio audio input (sample_rate, numpy_array) to base64 data url. | STRING |
| HIGH | …ne_serving/text_to_speech/moss_tts_nano/gradio_demo.py | 0 | #generate-btn button { width: 100%; } #streaming-player { border: 1px solid var(--border-color-primary) !important; bord | STRING |
| HIGH | …s/online_serving/text_to_speech/voxcpm2/gradio_demo.py | 0 | #generate-btn button { width: 100%; } #streaming-player { border: 1px solid var(--border-color-primary) !important; bord | STRING |
| HIGH | …online_serving/text_to_speech/qwen3_tts/gradio_demo.py | 0 | #generate-btn button { width: 100%; } #streaming-player { border: 1px solid var(--border-color-primary) !important; bord | STRING |
| HIGH | …ne_serving/text_to_speech/moss_tts_nano/gradio_demo.py | 0 | <div style="text-align:center; padding:8px 0; margin-top:4px;"> <a href="https://github.com/vllm-project/vllm-omni" targ | STRING |
| HIGH | …s/online_serving/text_to_speech/voxcpm2/gradio_demo.py | 0 | <div style="text-align:center; padding:8px 0; margin-top:4px;"> <a href="https://github.com/vllm-project/vllm-omni" targ | STRING |
| HIGH | …online_serving/text_to_speech/qwen3_tts/gradio_demo.py | 0 | <div style="text-align:center; padding:8px 0; margin-top:4px;"> <a href="https://github.com/vllm-project/vllm-omni" targ | STRING |
| HIGH | benchmarks/glm_image/benchmark_glm_image.py | 0 | return path to prompt.json, downloading from huggingface if needed. | STRING |
| HIGH | benchmarks/glm_image/vllm-omni/inference.py | 0 | return path to prompt.json, downloading from huggingface if needed. | STRING |
| HIGH | benchmarks/glm_image/huggingface/inference.py | 0 | return path to prompt.json, downloading from huggingface if needed. | STRING |
| HIGH | …model_executor/models/hunyuan_image3/hunyuan_image3.py | 0 | embeds scalar timesteps into vector representations. | STRING |
| HIGH | …odels/indextts2/s2mel/modules/diffusion_transformer.py | 0 | embeds scalar timesteps into vector representations. | STRING |
| HIGH | …on/models/hunyuan_image3/hunyuan_image3_transformer.py | 0 | embeds scalar timesteps into vector representations. | STRING |
| 61 more matches not shown… | ||||
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| LOW | tools/nightly/generate_nightly_perf_excel.py | 753 | def _set_column_width(ws, columns: Sequence[str], col_name: str, width: int) -> None: | CODE |
| LOW⚡ | tests/config/test_config_factory.py | 3036 | def set_files(self, files: list[tuple[str, str]]) -> None: | CODE |
| LOW | …utor/stage_input_processors/test_nemotron_voicechat.py | 341 | def set_init_inputs(self, speaker_name: str) -> None: | CODE |
| LOW | tests/model_tests/diffusion/utils.py | 10 | logger = logging.getLogger(__name__) | CODE |
| LOW⚡ | tests/diffusion/test_diffusion_step_pipeline.py | 67 | def _update_environment_variables(envs_dict: dict[str, str]) -> None: | CODE |
| LOW | tests/diffusion/test_diffusion_step_pipeline.py | 897 | def set_active_adapter(self, adapter, scale: float = 1.0) -> None: | CODE |
| LOW | tests/diffusion/distributed/test_comm.py | 33 | def _update_environment_variables(envs_dict: dict[str, str]) -> None: | CODE |
| LOW⚡ | tests/diffusion/distributed/test_pipeline_parallel.py | 320 | def _set_rank(monkeypatch, world_size: int, first_stage: bool) -> None: | CODE |
| LOW | tests/diffusion/distributed/test_cfg_parallel.py | 34 | def _set_random_seeds(seed: int) -> None: | CODE |
| LOW | tests/diffusion/distributed/test_cfg_parallel.py | 65 | def _update_environment_variables(envs_dict: dict[str, str]) -> None: | CODE |
| LOW | …n/models/lingbot_world/test_lingbot_world_attention.py | 61 | def set_weight_attrs(weight: torch.Tensor, attrs: dict) -> None: | CODE |
| LOW | …n/models/lingbot_world/test_lingbot_world_attention.py | 267 | def _set_identity_attention(attention: nn.Module) -> None: | CODE |
| LOW | tests/diffusion/models/wan2_2/conftest.py | 36 | def set_timesteps(self, num_steps: int, device: torch.device) -> None: | CODE |
| LOW | …diffusion/models/wan2_2/test_wan22_pipeline_diffuse.py | 39 | def set_timesteps(self, num_steps: int, device: torch.device) -> None: | CODE |
| LOW | tests/diffusion/diffusion_kv/test_initialization.py | 35 | def set_kv_cache_configs(self, configs, resolved_max_model_len) -> None: | CODE |
| LOW | tests/dfx/reliability/helpers.py | 34 | logger = logging.getLogger(__name__) | CODE |
| LOW | …/entrypoints/openai_api/test_serving_audio_generate.py | 26 | logger = logging.getLogger(__name__) | CODE |
| LOW | tests/entrypoints/openai_api/test_serving_speech.py | 68 | logger = logging.getLogger(__name__) | CODE |
| LOW⚡ | tests/e2e/features/helpers/custom_pipeline.py | 37 | logger = logging.getLogger(__name__) | CODE |
| LOW⚡ | tests/e2e/features/helpers/custom_pipeline.py | 262 | def set_death_signal() -> None: | CODE |
| LOW | tests/e2e/features/helpers/verl_omni_server.py | 34 | logger = logging.getLogger(__name__) | CODE |
| LOW | tests/e2e/accuracy/test_qwen_image.py | 99 | def _set_reference_attention_backend(pipe: DiffusionPipeline) -> None: | CODE |
| LOW | …sts/e2e/accuracy/test_hunyuan_image3_pixel_accuracy.py | 23 | logger = logging.getLogger(__name__) | CODE |
| LOW | tests/e2e/accuracy/test_diffusers_backend_similarity.py | 52 | def _set_matched_attention_backend(pipe: DiffusionPipeline) -> None: | CODE |
| LOW | …ts/e2e/online_serving/test_cosyvoice3_tts_expansion.py | 53 | def _set_rocm_request_timeout(request_config: dict) -> None: | CODE |
| LOW | tests/engine/test_orchestrator_stage_input_bridge.py | 73 | def set_engine_outputs(self, _outputs) -> None: | CODE |
| LOW | tests/engine/test_orchestrator.py | 166 | def set_engine_outputs(self, outputs) -> None: | CODE |
| LOW | tests/engine/test_orchestrator.py | 262 | def update_scheduler_stats(self, _scheduler_stats) -> None: | CODE |
| LOW | tests/helpers/media.py | 32 | logger = logging.getLogger(__name__) | CODE |
| LOW | tests/helpers/media.py | 866 | __all__ = [ | CODE |
| LOW | tests/helpers/clean.py | 20 | logger = logging.getLogger(__name__) | CODE |
| LOW | tests/helpers/clean.py | 255 | __all__ = [ | CODE |
| LOW | tests/helpers/monitor.py | 55 | __all__ = ["DeviceMemoryMonitor"] | CODE |
| LOW | tests/helpers/client.py | 2129 | __all__ = [ | CODE |
| LOW | tests/helpers/assertions.py | 1032 | __all__ = [ | CODE |
| LOW | tests/helpers/runtime.py | 1208 | __all__ = [ | CODE |
| LOW | tests/helpers/stage_config.py | 746 | __all__ = [ | CODE |
| LOW | …ence/custom_pipeline/image_to_image/custom_pipeline.py | 12 | logger = logging.getLogger(__name__) | CODE |
| LOW⚡ | …es/offline_inference/qwen3_omni/end2end_async_chunk.py | 46 | logger = logging.getLogger(__name__) | CODE |
| LOW | …es/offline_inference/text_to_speech/glm_tts/end2end.py | 48 | logger = logging.getLogger(__name__) | CODE |
| LOW | …ffline_inference/text_to_speech/fish_speech/end2end.py | 38 | logger = logging.getLogger(__name__) | CODE |
| LOW | …/offline_inference/text_to_speech/qwen3_tts/end2end.py | 21 | logger = logging.getLogger(__name__) | CODE |
| LOW | …ffline_inference/text_to_speech/voxtral_tts/end2end.py | 35 | logger = logging.getLogger(__name__) | CODE |
| LOW | examples/online_serving/minicpmo/realtime_web/server.py | 21 | logger = logging.getLogger(__name__) | CODE |
| LOW | …mples/online_serving/minicpmo/realtime_web/__init__.py | 5 | __all__ = ["build_app"] | CODE |
| LOW | …e_serving/step_audio2/openai_chat_completion_client.py | 34 | logger = logging.getLogger(__name__) | CODE |
| LOW⚡ | …ne_serving/text_to_speech/moss_tts_nano/gradio_demo.py | 38 | logger = logging.getLogger(__name__) | CODE |
| LOW⚡ | …s/online_serving/text_to_speech/voxcpm2/gradio_demo.py | 32 | logger = logging.getLogger(__name__) | CODE |
| LOW | …rving/text_to_speech/qwen3_tts/word_timestamps_demo.py | 47 | logger = logging.getLogger(__name__) | CODE |
| LOW⚡ | …online_serving/text_to_speech/qwen3_tts/gradio_demo.py | 45 | logger = logging.getLogger(__name__) | CODE |
| LOW⚡ | benchmarks/kernels/mot_linear_benchmarks.py | 56 | logger = logging.getLogger(__name__) | CODE |
| LOW | benchmarks/diffusion/diffusion_benchmark_serving.py | 112 | logger = logging.getLogger(__name__) | CODE |
| LOW | benchmarks/accuracy/image_to_image/gedit_bench.py | 41 | logger = logging.getLogger(__name__) | CODE |
| LOW | vllm_omni/version.py | 51 | __all__ = ["__version__", "__version_tuple__"] | CODE |
| LOW | vllm_omni/__init__.py | 59 | __all__ = [ | CODE |
| LOW⚡ | vllm_omni/metrics/definitions.py | 26 | logger = logging.getLogger(__name__) | CODE |
| LOW | vllm_omni/metrics/__init__.py | 10 | __all__ = [ | CODE |
| LOW | vllm_omni/metrics/prometheus.py | 122 | def set_running(self, n: int) -> None: | CODE |
| LOW | vllm_omni/metrics/prometheus.py | 127 | def set_waiting(self, n: int) -> None: | CODE |
| LOW⚡ | vllm_omni/metrics/prometheus.py | 183 | def set_stage_waiting_requests(self, stage: int, n_waiting: int) -> None: | CODE |
| 492 more matches not shown… | ||||
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| LOW | tools/run_merge_jobs.sh | 1 | #!/usr/bin/env bash | COMMENT |
| LOW | tools/run_merge_jobs.sh | 21 | # Usage: | COMMENT |
| LOW | tools/run_ready_jobs.sh | 1 | #!/usr/bin/env bash | COMMENT |
| LOW | tools/run_ready_jobs.sh | 21 | # Usage: | COMMENT |
| LOW | tools/run_jobs_common.sh | 1 | #!/usr/bin/env bash | COMMENT |
| LOW | tools/nightly/run_nightly_jobs.sh | 1 | #!/usr/bin/env bash | COMMENT |
| LOW | tools/nightly/run_nightly_jobs.sh | 21 | # function — label has neither "Perf Test" nor "Accuracy Test" (incl. Doc, Multi-Replica, etc.) | COMMENT |
| LOW | tools/nightly/run_nightly_jobs.sh | 41 | # From repo root: pytest -sv -m "<markers> and local_model" (markers from MODEL_TYPE: omni, tts, | COMMENT |
| LOW | tools/nightly/run_nightly_jobs.sh | 61 | # • Set REPO_ROOT (or pass --repo-root) — default YAML is $REPO_ROOT/.buildkite/cuda/test-nightly.yml | COMMENT |
| LOW | tests/attention/test_fish_kvcache_attn.py | 261 | assert captured["key"] == expected | COMMENT |
| LOW | tests/model_executor/models/voxcpm2/__init__.py | 1 | # SPDX-License-Identifier: Apache-2.0 | COMMENT |
| LOW | tests/model_executor/models/dots_tts/__init__.py | 1 | # SPDX-License-Identifier: Apache-2.0 | COMMENT |
| LOW | …l_executor/models/common/test_fused_group_norm_silu.py | 41 | H, W = spatial_size | COMMENT |
| LOW | tests/model_executor/models/minicpmo_4_5/__init__.py | 1 | # SPDX-License-Identifier: Apache-2.0 | COMMENT |
| LOW | tests/model_tests/diffusion/config_types.py | 61 | COMMENT | |
| LOW | tests/model_tests/diffusion/model_settings.py | 1 | from tests.model_tests.diffusion import diff_model_builders | COMMENT |
| LOW | tests/diffusion/attention/test_flash_attn.py | 461 | query = torch.randn(1, 8, 2, 4) | COMMENT |
| LOW | tests/diffusion/models/pi0/__init__.py | 1 | # SPDX-License-Identifier: Apache-2.0 | COMMENT |
| LOW | tests/diffusion/models/pi0/test_pi0_units.py | 481 | masks = [bool(m.item()) for m in image_masks] | COMMENT |
| LOW | …s/hunyuan_image3/test_hunyuan_image3_it2i_ar_format.py | 61 | # it kept skipping because instantiating | COMMENT |
| LOW | tests/diffusion/models/lingbot_video/__init__.py | 1 | # SPDX-License-Identifier: Apache-2.0 | COMMENT |
| LOW | tests/diffusion/models/hunyuan_video/__init__.py | 1 | # SPDX-License-Identifier: Apache-2.0 | COMMENT |
| LOW | tests/diffusion/models/minimax_h3/__init__.py | 1 | # SPDX-License-Identifier: Apache-2.0 | COMMENT |
| LOW | tests/diffusion/models/cosmos3/__init__.py | 1 | # SPDX-License-Identifier: Apache-2.0 | COMMENT |
| LOW | tests/diffusion/quantization/__init__.py | 1 | # SPDX-License-Identifier: Apache-2.0 | COMMENT |
| LOW | tests/diffusion/quantization/test_mxfp4_config.py | 361 | # Row-parallel (to_out, ffn.net_2): input is sharded (K/TP), output is full (N). | COMMENT |
| LOW | tests/dfx/stability/scripts/resource_monitor.sh | 1 | #!/bin/bash | COMMENT |
| LOW | tests/entrypoints/test_async_omni.py | 361 | # | COMMENT |
| LOW | tests/entrypoints/test_pd_disaggregation.py | 1161 | # Tests: Failure mode & memory leak prevention | COMMENT |
| LOW | tests/e2e/offline_inference/run_quantization_e2e.sh | 1 | #!/bin/bash | COMMENT |
| LOW | …eatures/fullduplex/engine/test_duplex_deploy_config.py | 81 | # Guards the admission probe. A capacity edit, a config that declares none, | COMMENT |
| LOW | tests/e2e/accuracy/sana_video/test_sana_video_golden.py | 321 | # The native prompt cross-attention uses vLLM-Omni Attention rather | COMMENT |
| LOW | tests/helpers/runtime.py | 1041 | print("OmniRunner started successfully") | COMMENT |
| LOW | requirements/common.txt | 1 | # Common dependencies for all platforms | COMMENT |
| LOW | examples/offline_inference/lance/gradio_demo.py | 41 | # canonical (W, H) via: | COMMENT |
| LOW | examples/offline_inference/lance/gradio_demo.py | 1021 | ) | COMMENT |
| LOW | examples/offline_inference/audex/speech_to_speech.py | 41 | from vllm_omni.model_executor.models.audex.prompt import build_cond_prompt | COMMENT |
| LOW | …_inference/qwen3_omni/run_single_prompt_async_chunk.sh | 1 | #!/bin/bash | COMMENT |
| LOW | …ine_inference/qwen2_5_omni/strategy_stage_replica.yaml | 1 | # Composable-parallel strategy for Qwen2.5-Omni. | COMMENT |
| LOW | …mples/offline_inference/qwen2_5_omni/strategy_tp2.yaml | 1 | # Composable-parallel strategy for Qwen2.5-Omni: tensor-parallel the thinker. | COMMENT |
| LOW | …s/offline_inference/text_to_speech/moss_tts/end2end.py | 161 | gc.collect() | COMMENT |
| LOW | …line_serving/replica_data_parallel/wan2_2_ti2v_dp.yaml | 1 | # Replica data-parallel recipe for Wan2.2 TI2V-5B (video DiT). | COMMENT |
| LOW | …ples/online_serving/dreamzero/droid_sim_eval_client.py | 61 | except ImportError as exc: # pragma: no cover - runtime dependency guard | COMMENT |
| LOW | …/online_serving/joyvl_interaction/scripts/start_all.sh | 1 | #!/bin/bash | COMMENT |
| LOW | examples/online_serving/audex/client.py | 261 | # The chat-pass recipe is size-conditional (see the TEXT_INSTRUCTION | COMMENT |
| LOW | examples/online_serving/audex/run_server.sh | 1 | #!/bin/bash | COMMENT |
| LOW | …amples/online_serving/text_to_video/run_server_ltx2.sh | 1 | #!/bin/bash | COMMENT |
| LOW | …ne_serving/text_to_speech/higgs_audio_v2/run_server.sh | 1 | #!/bin/bash | COMMENT |
| LOW | …ne_serving/text_to_speech/qwen3_tts/run_gradio_demo.sh | 1 | #!/bin/bash | COMMENT |
| LOW | benchmarks/kernels/mot_linear_benchmarks.py | 321 | # Mask predicate calculation, TMA state machine, etc. | COMMENT |
| LOW | .buildkite/npu/pipeline-npu-a3.yml | 1 | # steps: | COMMENT |
| LOW | .buildkite/npu/pipeline-npu-a3.yml | 21 | # vault.hashicorp.com/agent-inject-template-ca.pem: "{{- with secret \"internal/data/ascend/buildkitd\" -} | COMMENT |
| LOW | .buildkite/npu/pipeline-npu-a3.yml | 41 | # path: token | COMMENT |
| LOW | .buildkite/npu/pipeline-npu-a3.yml | 61 | # tar -xzf /tmp/buildkit.tar.gz -C /tmp/buildkit | COMMENT |
| LOW | .buildkite/npu/pipeline-npu-a3.yml | 81 | # echo "$${IMAGE_REGISTRY}/$${IMAGE_NAME}:$${VLLM_IMAGE_TAG}" | COMMENT |
| LOW | .buildkite/npu/pipeline-npu-a3.yml | 101 | # build.pull_request.labels includes "omni-test" || | COMMENT |
| LOW | .buildkite/release/scripts/publish-release-images.sh | 1 | #!/bin/bash | COMMENT |
| LOW | .buildkite/amd/test-amd-ready.yml | 201 | # mirror_hardwares: [amdproduction] | COMMENT |
| LOW | .buildkite/amd/test-amd-ready.yml | 221 | COMMENT | |
| LOW | .buildkite/amd/test-amd-merge.yml | 241 | # grade: Blocking | COMMENT |
| 323 more matches not shown… | ||||
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| HIGH | …buted/omni_connectors/test_bagel_mooncake_connector.py | 122 | Generate an image using Bagel model with configured parameters. Args: omni: The Omni instance to use for ge | STRING |
| HIGH | …/omni_connectors/test_bagel_shared_memory_connector.py | 163 | Generate an image using Bagel model with img2img pipeline. Args: omni: The Omni instance to use for generat | STRING |
| HIGH | …line_inference/test_sensenova_u1_text2img_expansion.py | 120 | Generate an image using SenseNova-U1 model with configured parameters. Args: omni: The Omni instance to use | STRING |
| HIGH | …fline_inference/test_sensenova_u1_img2img_expansion.py | 127 | Generate an edited image using SenseNova-U1 model with img2img pipeline. Args: omni: The Omni instance to u | STRING |
| HIGH | vllm_omni/distributed/omni_coordinator/load_balancer.py | 47 | Route a task to one of the available replicas. Args: task: The task to route. Not used by the rando | STRING |
| HIGH | vllm_omni/config/composable_parallel/apply.py | 220 | Overlay per-role strategy specs onto a merged stage list. Args: stages: the ``list[StageConfig]`` returned | STRING |
| HIGH | vllm_omni/model_executor/models/cosyvoice3/utils.py | 253 | Make mask tensor containing indices of padded part. See description of make_non_pad_mask. Args: length | STRING |
| HIGH | …/nemotron_voicechat/nemo_vendored/ear_tts_vae_codec.py | 175 | Converts a spectrogram back into a waveform using the overlap-add method. This function is an approximate inver | STRING |
| HIGH | …els/nemotron_voicechat/nemo_vendored/duplex_ear_tts.py | 444 | Retrieve a cached audio prompt latent and adapt it to the requested batch size. This fetches a latent | STRING |
| HIGH | …els/nemotron_voicechat/nemo_vendored/duplex_ear_tts.py | 651 | Returns a dictionary of initial inputs for inference, using registered buffers. Args: B (i | STRING |
| HIGH | …r/models/nemotron_voicechat/nemo_vendored/tokenizer.py | 303 | Adds a dictionary of special tokens (eos, pad, cls...). If special tokens are NOT in the vocabulary, they are | STRING |
| HIGH | …utor/models/nemotron_voicechat/nemo_vendored/fusion.py | 346 | Factory function to create the appropriate fusion module. Args: fuse_method: One of None, "add", "conc | STRING |
| HIGH | …ls/nemotron_voicechat/nemo_vendored/ear_tts_commons.py | 50 | Loads a configuration from a JSON or Python file. - For JSON files (`*.json`), it parses the file directly. | STRING |
| HIGH | …ls/nemotron_voicechat/nemo_vendored/ear_tts_commons.py | 230 | Finds the path of the latest checkpoint file or directory in a directory. The latest checkpoint is determined | STRING |
| HIGH | …m_omni/model_executor/models/minimax_music3/weights.py | 89 | Load the audio embedding table and the RVQ depth decoder in place. Args: vllm_config: The stage's config, u | STRING |
| HIGH | …m_omni/model_executor/models/minimax_music3/staging.py | 46 | Stage ``count`` entries of each row and return the device view. Args: values: One sequence of Pytho | STRING |
| HIGH | vllm_omni/model_executor/models/minimax_music3/dit.py | 412 | Solve an aligned condition into a vocoder latent ``[1, 128, T_mel]``. The previous window's latent is re-impose | STRING |
| HIGH | vllm_omni/model_executor/models/minimax_music3/dav.py | 173 | Decode a latent to interleaved stereo. Args: latent: ``[B, 128, T]`` vocoder latent. Retur | STRING |
| HIGH | …m_omni/model_executor/models/indextts2/utils/common.py | 66 | Make mask tensor containing indices of padded part. See description of make_non_pad_mask. Args: length | STRING |
| HIGH | …model_executor/models/qwen3_tts/qwen3_tts_tokenizer.py | 68 | Initialize tokenizer with HuggingFace `from_pretrained` style. Args: pretrained_model_name | STRING |
| HIGH | …odel_executor/stage_input_processors/minimax_music3.py | 166 | Emit the unconditioned companion for a guided MiniMax Music 3 request. Classifier-free guidance is not optional for | STRING |
| HIGH | vllm_omni/diffusion/registry.py | 393 | Initialize a diffusion model from the registry. This function: 1. Loads the model class from the registry 2 | STRING |
| HIGH | vllm_omni/diffusion/data.py | 1002 | Find an available port with retry logic. Args: port: Initial port to check por | STRING |
| HIGH | vllm_omni/diffusion/attention/backends/registry.py | 124 | Register or override a diffusion backend implementation. Args: backend: The DiffusionAttentionBackendEnum m | STRING |
| HIGH | vllm_omni/diffusion/cache/selector.py | 12 | Get cache backend instance based on cache_backend string. This is a selector function that routes to the appropriat | STRING |
| HIGH | vllm_omni/diffusion/cache/teacache/extractors.py | 1458 | Get extractor function for given transformer class. This function looks up the extractor based on the exact tr | STRING |
| HIGH | …omni/diffusion/cache/teacache/coefficient_estimator.py | 221 | Estimate polynomial coefficients from collected data. Args: poly_order: Order of polynomial fit (de | STRING |
| HIGH | vllm_omni/diffusion/cache/magcache/strategy.py | 473 | Get strategy function for given transformer class. This function looks up the strategy based on the exact transform | STRING |
| HIGH | vllm_omni/diffusion/distributed/sp_sharding.py | 32 | Shard a tensor along the specified dimension for sequence parallelism. The tensor is split into world_size chunks a | STRING |
| HIGH | vllm_omni/diffusion/distributed/parallel_state.py | 505 | Initialize Ulysses, Ring, and AllGather-KV process groups. AllGather-KV is mutually exclusive with Ulysses and | STRING |
| HIGH | …on/models/hunyuan_image3/hunyuan_image3_transformer.py | 2956 | The call function to the pipeline for generation. Args: prompt (`str` or `List[str]`): | STRING |
| HIGH | …ni/diffusion/models/glm_image/glm_image_transformer.py | 64 | Validate GLM-Image TP constraints without requiring a distributed context. Args: dim: Model hidden dimensio | STRING |
| HIGH | …ni/diffusion/models/glm_image/glm_image_transformer.py | 428 | Get cache for a specific layer. Args: layer_idx: Index of the layer (0-indexed). Returns: | STRING |
| HIGH | …mni/diffusion/models/sana_video/pipeline_sana_video.py | 813 | Function invoked when calling the pipeline for generation. Args: prompt (`str` or `list[st | STRING |
| HIGH | …lm_omni/diffusion/models/omnigen2/pipeline_omnigen2.py | 817 | Get prompt embeddings from the Qwen2 text encoder. Args: prompt: The prompt or list of pro | STRING |
| HIGH | …on/models/gr00t/dataio/state_action/action_chunking.py | 131 | Convert a relative action chunking to an absolute action chunking by applying the relative poses on top | STRING |
| HIGH | …on/models/gr00t/dataio/state_action/action_chunking.py | 168 | Convert action chunking to the specified action format. Must be implemented by subclasses. Arg | STRING |
| HIGH | …on/models/gr00t/dataio/state_action/action_chunking.py | 259 | Interpolate the joint action chunking to generate intermediate configurations. Uses linear interpolati | STRING |
| HIGH | …on/models/gr00t/dataio/state_action/action_chunking.py | 338 | Convert a relative joint action chunking to an absolute action chunking by adding the relative joint po | STRING |
| HIGH | …on/models/gr00t/dataio/state_action/action_chunking.py | 373 | Convert trajectory to the desired format. Args: action_format: The desired output format | STRING |
| HIGH | …on/models/gr00t/dataio/state_action/action_chunking.py | 479 | Interpolate the action chunking to generate intermediate poses. Uses linear interpolation for translat | STRING |
| HIGH | …on/models/gr00t/dataio/state_action/action_chunking.py | 635 | Convert trajectory to the desired format. Args: action_format: The desired output format | STRING |
| HIGH | …mni/diffusion/models/gr00t/dataio/state_action/pose.py | 101 | Compute relative transformation between two poses. For EndEffectorPose: Computes the relative transfor | STRING |
| HIGH | …mni/diffusion/models/gr00t/dataio/state_action/pose.py | 243 | Compute relative joint displacement. Args: other: Reference joint pose Returns: | STRING |
| HIGH | …ls/gr00t/dataio/state_action/state_action_processor.py | 227 | Reverse state processing (denormalization). Args: state: Dict mapping joint_group -> proce | STRING |
| HIGH | …ls/gr00t/dataio/state_action/state_action_processor.py | 286 | Apply action processing (absolute->relative conversion, normalization). Processing order: 1. C | STRING |
| HIGH | …ls/gr00t/dataio/state_action/state_action_processor.py | 373 | Reverse action processing (denormalization, relative->absolute conversion). Processing order: | STRING |
| HIGH | …iffusion/models/ernie_image/ernie_image_transformer.py | 60 | Validate ErnieImage TP constraints without requiring a distributed context. Args: heads: Number of attentio | STRING |
| HIGH | vllm_omni/diffusion/hooks/sequence_parallel.py | 80 | Get a parameter value from args or kwargs by name. Args: identifier: The parameter name to look up. | STRING |
| HIGH | vllm_omni/diffusion/hooks/sequence_parallel.py | 538 | Get a submodule by dotted name, supporting wildcards. Args: model: The root module. name: Dotted pa | STRING |
| HIGH | vllm_omni/diffusion/worker/diffusion_worker.py | 1616 | Execute a method on the worker. Args: method: Method name (str) or serialized callable (by | STRING |
| HIGH | vllm_omni/entrypoints/async_omni.py | 480 | Generate outputs for the given prompt(s) asynchronously. Coordinates multi-stage pipeline execution. Processes | STRING |
| HIGH | vllm_omni/entrypoints/utils.py | 93 | Filter dict-like object by removing callables and recursively converting values. Converts dict-like objects to regu | STRING |
| HIGH | vllm_omni/entrypoints/utils.py | 261 | Resolve the stage/deploy config file path from the model name. Resolves configuration path based on the model type | STRING |
| HIGH | vllm_omni/entrypoints/openai/image_api_utils.py | 21 | Parse size string to width and height tuple. Args: size_str: Size in format "WIDTHxHEIGHT" (e.g., "1024x102 | STRING |
| HIGH | vllm_omni/entrypoints/openai/api_server.py | 1336 | Generate speech audio from text using the loaded TTS model. Args: request: Speech synthesis request in Open | STRING |
| HIGH | vllm_omni/entrypoints/openai/api_server.py | 1864 | Generate images from text prompts using diffusion models. OpenAI DALL-E compatible endpoint for text-to-image gener | STRING |
| HIGH | vllm_omni/entrypoints/openai/api_server.py | 3629 | Retrieve metadata for a previously created video job. Args: video_id: Identifier returned by ``POST /v1/vid | STRING |
| HIGH | vllm_omni/entrypoints/openai/api_server.py | 3657 | Delete a stored video job and any generated output. If the job is still queued or running, this endpoint first atte | STRING |
| HIGH | vllm_omni/entrypoints/openai/api_server.py | 3709 | Download the generated file for a completed video job. Args: video_id: Identifier of the video job whose ou | STRING |
| 3 more matches not shown… | ||||
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| LOW | tests/distributed/omni_connectors/test_tp_rank_aware.py | 41 | CODE | |
| LOW | …stage_input_processors/test_fish_speech_async_chunk.py | 14 | CODE | |
| LOW | …r/stage_input_processors/test_qwen3_tts_async_chunk.py | 47 | CODE | |
| LOW | tests/diffusion/attention/test_attention_sp.py | 122 | CODE | |
| LOW | …on/models/lingbot_world/test_pipeline_lingbot_world.py | 220 | CODE | |
| LOW | …ffusion/models/hunyuan_image3/test_multi_resolution.py | 564 | CODE | |
| LOW | tests/diffusion/quantization/test_quantization_fp8.py | 112 | CODE | |
| LOW | …/diffusion/diffusion_backend/test_diffusers_backend.py | 379 | CODE | |
| LOW | tests/profile/test_omni_torch_profiler.py | 42 | CODE | |
| LOW | tests/entrypoints/openai_api/test_duplex_handler.py | 81 | CODE | |
| LOW | tests/e2e/features/helpers/custom_pipeline.py | 74 | CODE | |
| LOW | tests/e2e/features/helpers/custom_pipeline.py | 484 | CODE | |
| LOW | tests/engine/test_orchestrator.py | 118 | CODE | |
| LOW | …les/offline_inference/mimo_audio/process_speechdata.py | 131 | CODE | |
| LOW | …line_inference/text_to_speech/moss_tts_nano/end2end.py | 52 | CODE | |
| LOW | …/offline_inference/text_to_speech/indextts2/end2end.py | 43 | CODE | |
| LOW | examples/online_serving/lance/openai_chat_client.py | 17 | CODE | |
| LOW | …es/online_serving/image_to_image/openai_chat_client.py | 31 | CODE | |
| LOW | …les/online_serving/text_to_image/openai_chat_client.py | 17 | CODE | |
| LOW | benchmarks/accuracy/common.py | 141 | CODE | |
| LOW | benchmarks/accuracy/image_to_image/gedit_bench.py | 419 | CODE | |
| LOW | vllm_omni/tokenizers/mammoth_moda2_tokenizer.py | 103 | CODE | |
| LOW | vllm_omni/experimental/fullduplex/client.py | 502 | CODE | |
| LOW | vllm_omni/experimental/fullduplex/joyvl/memory/brain.py | 21 | CODE | |
| LOW | …_omni/experimental/fullduplex/joyvl/decision/policy.py | 26 | CODE | |
| LOW | vllm_omni/platforms/npu/worker/npu_model_runner.py | 112 | CODE | |
| LOW | …ni/platforms/npu/worker/npu_generation_model_runner.py | 623 | CODE | |
| LOW | vllm_omni/core/sched/omni_scheduler_mixin.py | 462 | CODE | |
| LOW | …m_omni/model_executor/models/voxcpm2/minicpm4_paged.py | 123 | CODE | |
| LOW | …model_executor/models/hunyuan_image3/hunyuan_image3.py | 2189 | CODE | |
| LOW | vllm_omni/model_executor/models/cosyvoice3/utils.py | 23 | CODE | |
| LOW | …/model_executor/models/cosyvoice3/code2wav_core/cfm.py | 233 | CODE | |
| LOW | …el_executor/models/cosyvoice3/code2wav_core/hifigan.py | 383 | CODE | |
| LOW | …el_executor/models/cosyvoice3/code2wav_core/hifigan.py | 612 | CODE | |
| LOW | …el_executor/models/cosyvoice3/code2wav_core/hifigan.py | 859 | CODE | |
| LOW | vllm_omni/model_executor/models/glm_tts/glm_tts_dit.py | 436 | CODE | |
| LOW | vllm_omni/model_executor/models/glm_tts/glm_tts_dit.py | 659 | CODE | |
| LOW | …i/model_executor/models/glm_tts/glm_tts_dit_wrapper.py | 565 | CODE | |
| LOW | …m_omni/model_executor/models/glm_image/glm_image_ar.py | 1800 | CODE | |
| LOW | …i/model_executor/models/gepard/configuration_gepard.py | 40 | CODE | |
| LOW | …ecutor/models/fish_speech/configuration_fish_speech.py | 23 | CODE | |
| LOW | …ecutor/models/fish_speech/configuration_fish_speech.py | 77 | CODE | |
| LOW | …odel_executor/models/mimo_audio/mimo_audio_code2wav.py | 997 | CODE | |
| LOW | …executor/models/mimo_audio/modeling_audio_tokenizer.py | 1135 | CODE | |
| LOW | …executor/models/mimo_audio/modeling_audio_tokenizer.py | 1176 | CODE | |
| LOW | …lm_omni/model_executor/models/mimo_audio/mimo_audio.py | 766 | CODE | |
| LOW | …/model_executor/models/mimo_audio/config_mimo_audio.py | 17 | CODE | |
| LOW | …/model_executor/models/mimo_audio/config_mimo_audio.py | 122 | CODE | |
| LOW | …lm_omni/model_executor/models/dots_tts/dots_tts_dit.py | 124 | CODE | |
| LOW | …mni/model_executor/models/dots_tts/dots_tts_vocoder.py | 90 | CODE | |
| LOW | …mni/model_executor/models/dots_tts/dots_tts_vocoder.py | 133 | CODE | |
| LOW | …mni/model_executor/models/dots_tts/dots_tts_vocoder.py | 235 | CODE | |
| LOW | …mni/model_executor/models/dots_tts/dots_tts_vocoder.py | 727 | CODE | |
| LOW | …omni/model_executor/models/ming_tts/config_ming_tts.py | 113 | CODE | |
| LOW | …omni/model_executor/models/ming_tts/prompt_assembly.py | 257 | CODE | |
| LOW | …/nemotron_voicechat/nemo_vendored/ear_tts_vae_codec.py | 519 | CODE | |
| LOW | …/nemotron_voicechat/nemo_vendored/ear_tts_vae_codec.py | 598 | CODE | |
| LOW | …r/models/nemotron_voicechat/nemo_vendored/tokenizer.py | 131 | CODE | |
| LOW | …dels/nemotron_voicechat/nemo_vendored/ear_tts_model.py | 786 | CODE | |
| LOW | …dels/nemotron_voicechat/nemo_vendored/ear_tts_model.py | 1228 | CODE | |
| 265 more matches not shown… | ||||
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| LOW⚡ | tests/core/sched/test_generation_scheduler_restore.py | 52 | # Step 1: process_pending_chunks moves req-B out | COMMENT |
| LOW⚡ | tests/core/sched/test_generation_scheduler_restore.py | 57 | # Step 2: simulate the try/except/finally pattern | COMMENT |
| LOW⚡ | tests/core/sched/test_generation_scheduler_restore.py | 66 | # Step 3: verify request is restored | COMMENT |
| LOW | tests/diffusion/attention/test_attention_sp.py | 473 | # Step 1: Run without SP (baseline with ulysses_degree=1, ring_degree=1) | COMMENT |
| LOW | tests/diffusion/attention/test_attention_sp.py | 502 | # Step 2: Run with SP enabled | COMMENT |
| LOW | tests/diffusion/attention/test_attention_sp.py | 534 | # Step 3: Verify input consistency and compare outputs | COMMENT |
| LOW⚡ | tests/e2e/offline_inference/test_moss_tts_realtime.py | 166 | # Step 1: locate the realtime processor module in the snapshot. | COMMENT |
| LOW⚡ | tests/e2e/offline_inference/test_moss_tts_realtime.py | 191 | # Step 2: encode the reference audio clip via MOSS-Audio-Tokenizer. | COMMENT |
| LOW | tests/e2e/offline_inference/test_moss_tts_realtime.py | 209 | # Step 3: build the (L, 17) prefill grid. | COMMENT |
| LOW⚡ | …ures/custom_pipeline/test_async_omni_collective_rpc.py | 126 | # Step 1: list_loras (the call that was crashing) | COMMENT |
| LOW⚡ | …ures/custom_pipeline/test_async_omni_collective_rpc.py | 130 | # Step 2: generate (should still work after list_loras) | COMMENT |
| LOW⚡ | .claude/skills/vllm-omni-test/SKILL.md | 31 | ### Step 1: Classify Test Goal | COMMENT |
| LOW⚡ | .claude/skills/vllm-omni-test/SKILL.md | 37 | ### Step 2: Select Test Level | COMMENT |
| LOW⚡ | .claude/skills/vllm-omni-test/SKILL.md | 45 | ### Step 3: Pick Markers | COMMENT |
| LOW | .claude/skills/vllm-omni-test/SKILL.md | 91 | ### Step 4: Generate Test Case Skeleton | COMMENT |
| LOW | .claude/skills/vllm-omni-test/SKILL.md | 627 | ### Step 5: Wire Buildkite (when CI must run the new test) | COMMENT |
| LOW | .claude/skills/vllm-omni-test/SKILL.md | 829 | ### Step 6: Run Tests | COMMENT |
| LOW | .claude/skills/vllm-omni-test/SKILL.md | 856 | ### Step 7: Validate Result Quality | COMMENT |
| LOW | .claude/skills/precheck-pr/SKILL.md | 21 | ### Step 1: Detect Base Branch | COMMENT |
| LOW | .claude/skills/precheck-pr/SKILL.md | 31 | ### Step 2: Validate PR Title | COMMENT |
| LOW | .claude/skills/precheck-pr/SKILL.md | 50 | ### Step 3: Categorize the PR | COMMENT |
| LOW | .claude/skills/precheck-pr/SKILL.md | 62 | ### Step 4: Run Checklist | COMMENT |
| LOW | .claude/skills/precheck-pr/SKILL.md | 74 | ### Step 5: Print Report | COMMENT |
| LOW | .claude/skills/add-diffusion-model/SKILL.md | 21 | ## Step 0: Classify the Migration Path | COMMENT |
| LOW | .claude/skills/add-diffusion-model/SKILL.md | 263 | ### Step 4: Register Model in registry.py | COMMENT |
| LOW | .claude/skills/add-diffusion-model/SKILL.md | 283 | ### Step 5: Run, Test, Debug | COMMENT |
| LOW⚡ | .claude/skills/add-diffusion-model/SKILL.md | 301 | ### Step 6: Add Example Scripts | COMMENT |
| LOW⚡ | .claude/skills/add-diffusion-model/SKILL.md | 308 | ### Step 7: Update Documentation | COMMENT |
| LOW⚡ | .claude/skills/add-diffusion-model/SKILL.md | 318 | ### Step 8: Add E2E Tests | COMMENT |
| LOW | .claude/skills/add-diffusion-model/SKILL.md | 349 | ### Step 9: Add Cache-DiT Acceleration | COMMENT |
| LOW | .claude/skills/add-diffusion-model/SKILL.md | 453 | ### Step 10: Add Parallelism Support | COMMENT |
| LOW | .claude/skills/add-diffusion-model/SKILL.md | 571 | ### Step 11: Add CPU Offload Support | COMMENT |
| LOW | .claude/skills/add-diffusion-model/SKILL.md | 602 | ### Step 12: Performance Profiling | COMMENT |
| LOW | …add-diffusion-model/references/parallelism-patterns.md | 186 | #### Step 1: Identify module boundaries | COMMENT |
| LOW | …add-diffusion-model/references/parallelism-patterns.md | 208 | #### Step 2: Handle inline operations | COMMENT |
| LOW | …add-diffusion-model/references/parallelism-patterns.md | 227 | #### Step 3: Write `_sp_plan` | COMMENT |
| LOW | .claude/skills/vllm-omni-npu-upgrade/SKILL.md | 120 | ### Step 1: Preparation | COMMENT |
| LOW | .claude/skills/vllm-omni-npu-upgrade/SKILL.md | 137 | ### Step 2: Analyze Omni-Specific Logic | COMMENT |
| LOW | .claude/skills/vllm-omni-npu-upgrade/SKILL.md | 151 | ### Step 3: Update Base Class (OmniNPUModelRunner) | COMMENT |
| LOW | .claude/skills/vllm-omni-npu-upgrade/SKILL.md | 173 | ### Step 4: Update AR Model Runner | COMMENT |
| LOW | .claude/skills/vllm-omni-npu-upgrade/SKILL.md | 188 | ### Step 5: Update Generation Model Runner | COMMENT |
| LOW | .claude/skills/vllm-omni-npu-upgrade/SKILL.md | 205 | ### Step 6: Update Imports | COMMENT |
| LOW | .claude/skills/vllm-omni-npu-upgrade/SKILL.md | 226 | ### Step 7: Sync GPU-Side Omni Changes | COMMENT |
| LOW | .claude/skills/vllm-omni-npu-upgrade/SKILL.md | 238 | ### Step 8: Validation | COMMENT |
| LOW | docs/design/feature/cfg_parallel.md | 84 | ### Step 1: Inherit `CFGParallelMixin` | COMMENT |
| LOW | docs/design/feature/cfg_parallel.md | 123 | ### Step 2: Call `diffuse` | COMMENT |
| LOW | docs/design/feature/tensor_parallel.md | 46 | ### Step 1: Identify Linear Layers | COMMENT |
| LOW | docs/design/feature/tensor_parallel.md | 54 | ### Step 2: Replace Linear Layers with Parallel Equivalents | COMMENT |
| LOW | docs/design/feature/tensor_parallel.md | 148 | ### Step 3: Validate TP Constraints | COMMENT |
| LOW | docs/design/feature/pipeline_parallel.md | 97 | ### Step 1: Inherit `PipelineParallelMixin` and `CFGParallelMixin` | COMMENT |
| LOW | docs/design/feature/pipeline_parallel.md | 123 | ### Step 2: Make model forward and `predict_noise()` PP-aware | COMMENT |
| LOW | docs/design/feature/pipeline_parallel.md | 159 | ### Step 3: Partition the transformer layers | COMMENT |
| LOW | docs/design/feature/pipeline_parallel.md | 235 | ### Step 4: Use the standard denoising contract | COMMENT |
| LOW | docs/design/feature/hsdp.md | 44 | ### Step 1: Identify Modules to Shard | COMMENT |
| LOW | docs/design/feature/hsdp.md | 55 | ### Step 2: Define Shard Conditions | COMMENT |
| LOW | docs/design/feature/sequence_parallel.md | 122 | ### Step 1: Understand Module Boundaries | COMMENT |
| LOW | docs/design/feature/sequence_parallel.md | 145 | ### Step 2: Handle Inline Operations | COMMENT |
| LOW | docs/design/feature/sequence_parallel.md | 191 | ### Step 3: Write `_sp_plan` for Your Model | COMMENT |
| LOW | docs/design/feature/expert_parallel.md | 59 | ### Step 1: Configure Expert Parallelism Settings | COMMENT |
| LOW | docs/design/feature/expert_parallel.md | 72 | ### Step 2: Use Sparse MoE Block to enable EP routing. | COMMENT |
| 72 more matches not shown… | ||||
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| MEDIUM | …ributed/omni_connectors/test_omni_connector_configs.py | 23 | # This file is 4 levels deep from root if we count from tests? | COMMENT |
| MEDIUM | tests/config/test_config_factory.py | 908 | # Create the model | COMMENT |
| MEDIUM | tests/config/test_config_factory.py | 949 | # Create the model | COMMENT |
| MEDIUM | tests/config/test_config_factory.py | 1084 | # Create a config with the autodetected type, and write the | COMMENT |
| MEDIUM | …s/model_executor/models/glm_image/test_glm_image_ar.py | 137 | # Create a mock for the parent parse_mm_data | COMMENT |
| MEDIUM | tests/diffusion/test_worker_wrapper_base.py | 465 | # Create a mock worker class with re_init_pipeline | COMMENT |
| MEDIUM | tests/diffusion/cache/test_cache_backends.py | 428 | # Create a cache config with an scm policy & enable it | COMMENT |
| MEDIUM | …iffusion/distributed/test_autoencoder_kl_wan_encode.py | 220 | # Create a simple task with 2 temporal chunks | COMMENT |
| MEDIUM⚡ | tests/diffusion/distributed/test_sp_plan_hooks.py | 517 | # Create a test tensor | COMMENT |
| MEDIUM | tests/diffusion/distributed/test_sp_plan_hooks.py | 534 | # Create a test tensor with [text, image] concatenated | COMMENT |
| MEDIUM⚡ | …n/models/wan2_2/test_wan22_quant_config_propagation.py | 202 | # Create a fake module that has WanRMS_norm | COMMENT |
| MEDIUM | tests/diffusion/models/ltx2/test_ltx2_pipeline.py | 2128 | # Create a minimal od_config mock | COMMENT |
| MEDIUM | …fusion/models/glm_image/test_glm_image_quantization.py | 544 | # Create a minimal mock od_config | COMMENT |
| MEDIUM | …els/sensenova_u1/test_sensenova_u1_logits_processor.py | 52 | # Create a vLLM config with a wrapped diffusion config | COMMENT |
| MEDIUM | tests/diffusion/quantization/test_hsdp_fp8.py | 53 | # Create the quant-method instance without calling __init__ (avoids | COMMENT |
| MEDIUM | tests/entrypoints/openai_api/test_serving_speech.py | 548 | # Create a file larger than 10MB | COMMENT |
| MEDIUM | tests/entrypoints/openai_api/test_serving_speech.py | 878 | # Create a TTS stage | COMMENT |
| MEDIUM | tests/entrypoints/openai_api/test_image_server.py | 97 | # Create a simple test image | COMMENT |
| MEDIUM | tests/worker/test_omni_gpu_model_runner.py | 216 | # Create an instance without calling OmniGPUModelRunner.__init__ | COMMENT |
| MEDIUM | tests/engine/test_arg_utils.py | 216 | # Create a vLLM ModelConfig | COMMENT |
| MEDIUM | tests/engine/test_arg_utils.py | 219 | # Create a mock talking config with a dummy value for position_id_per_seconds | COMMENT |
| MEDIUM | docs/mkdocs/hooks/generate_examples.py | 398 | # Create the EXAMPLE_DOC_DIR if it doesn't exist | COMMENT |
| MEDIUM⚡ | …ctors/connectors/mooncake_transfer_engine_connector.py | 313 | # Create a socket to determine the local IP used for external communication | COMMENT |
| MEDIUM | …del_executor/models/fish_speech/fish_speech_fast_ar.py | 532 | # Create a seeded generator for deterministic residual codebook sampling. | COMMENT |
| MEDIUM | …odel_executor/models/mimo_audio/mimo_audio_code2wav.py | 1019 | # Create an all-zero logits tensor (sampler will choose the first token, usually pad token) | COMMENT |
| MEDIUM⚡ | …/nemotron_voicechat/nemo_vendored/ear_tts_vae_codec.py | 96 | # Create a range tensor from 0 to max_length - 1 | COMMENT |
| MEDIUM | …/nemotron_voicechat/nemo_vendored/ear_tts_vae_codec.py | 148 | # Create the window tensor on the same device as the waveform. | COMMENT |
| MEDIUM | …els/nemotron_voicechat/nemo_vendored/duplex_ear_tts.py | 375 | # Create a random dropout decision for each BOS instance | COMMENT |
| MEDIUM | …els/nemotron_voicechat/nemo_vendored/duplex_ear_tts.py | 386 | # Create a mask of the same shape as target_text_tokens | COMMENT |
| MEDIUM | …dels/nemotron_voicechat/nemo_vendored/ear_tts_model.py | 258 | # Create a range tensor from 0 to max_length - 1 | COMMENT |
| MEDIUM | …dels/nemotron_voicechat/nemo_vendored/ear_tts_model.py | 431 | # Create a new, dense character vocabulary sorted by the original token ID | COMMENT |
| MEDIUM | …dels/nemotron_voicechat/nemo_vendored/ear_tts_model.py | 860 | # Create a padded tensor for the character IDs | COMMENT |
| MEDIUM | …odels/nemotron_voicechat/nemo_vendored/asr/features.py | 33 | # This file contains code artifacts adapted from https://github.com/ryanleary/patter | COMMENT |
| MEDIUM | …otron_voicechat/nemo_vendored/asr/conformer_encoder.py | 490 | # Create the self-attention and padding masks | COMMENT |
| MEDIUM | …on_voicechat/nemo_vendored/asr/multi_head_attention.py | 596 | # Create a helper tensor to find the local indices of global attention | COMMENT |
| MEDIUM | …_tts/tokenizer_25hz/modeling_qwen3_tts_tokenizer_v1.py | 210 | """This class implements an attentive statistic pooling layer for each channel. | STRING |
| MEDIUM | …lm_omni/model_executor/models/qwen3_omni/qwen3_omni.py | 185 | # Create a new vllm_config with thinker_config as the hf_config | COMMENT |
| MEDIUM | …lm_omni/model_executor/models/qwen3_omni/qwen3_omni.py | 221 | # Create a new vllm_config with talker_config as the hf_config | COMMENT |
| MEDIUM | …lm_omni/model_executor/models/qwen3_omni/qwen3_omni.py | 256 | # Create a new vllm_config with code2wav_config as the hf_config | COMMENT |
| MEDIUM | …el_executor/models/qwen3_omni/qwen3_omni_moe_talker.py | 365 | # Create a vllm_config for the talker's text model | COMMENT |
| MEDIUM | …executor/models/qwen2_5_omni/qwen2_5_omni_token2wav.py | 168 | """This class implements an attentive statistic pooling layer for each channel. | STRING |
| MEDIUM | …_executor/models/minicpmo_4_5/minicpmo_4_5_omni_llm.py | 4381 | # Create a sequence tensor of shape (batch_size, max_seq_len) | COMMENT |
| MEDIUM | vllm_omni/diffusion/layers/mot/ops/mot_gemm.py | 228 | # Initialize VAE variables first and overwrite for Text | COMMENT |
| MEDIUM | vllm_omni/diffusion/distributed/comm.py | 251 | # Create a contiguous buffer for receiving | COMMENT |
| MEDIUM | vllm_omni/diffusion/utils/kv_utils.py | 45 | # Create the boolean mask for the input sequences | COMMENT |
| MEDIUM | …on/models/hunyuan_image3/hunyuan_image3_transformer.py | 686 | # This class is just for type hinting purposes. Attribute `i` should be defined | COMMENT |
| MEDIUM | …iffusion/models/qwen_image/autoencoder_kl_qwenimage.py | 347 | # Create the components | STRING |
| MEDIUM | …/scheduling_flow_match_euler_discrete_time_shifting.py | 12 | # This file is adapted by Boogu Team from prior open-source scheduler work. | COMMENT |
| MEDIUM | …mni/diffusion/models/nextstep_1_1/modeling_flux_vae.py | 295 | # Create a config-like object for compatibility | COMMENT |
| MEDIUM | …on/models/gr00t/dataio/state_action/action_chunking.py | 203 | # Create a joint action chunking | STRING |
| MEDIUM | …on/models/gr00t/dataio/state_action/action_chunking.py | 399 | # Create an end-effector action chunking | STRING |
| MEDIUM | …mni/diffusion/models/gr00t/dataio/state_action/pose.py | 172 | # Create a 6-DOF joint configuration | STRING |
| MEDIUM | vllm_omni/diffusion/hooks/sequence_parallel.py | 5 | # This module is adapted from HuggingFace diffusers library: | COMMENT |
| MEDIUM | vllm_omni/diffusion/worker/diffusion_worker.py | 1480 | # Create the actual worker instance | COMMENT |
| MEDIUM | vllm_omni/benchmarks/data_modules/daily_omni_dataset.py | 672 | # Create an iterator-like wrapper for compatibility | COMMENT |
| MEDIUM | vllm_omni/entrypoints/utils.py | 716 | # The following code detects if the process is running in a container and if | COMMENT |
| MEDIUM | vllm_omni/engine/arg_utils.py | 291 | # Create a temp dir with a patched config.json | COMMENT |
| MEDIUM | vllm_omni/inputs/data.py | 352 | # This class is changed to only represent a single prompt request | COMMENT |
| MEDIUM | …ps/ComfyUI-vLLM-Omni/comfyui_vllm_omni/utils/logger.py | 57 | # Create a copy with redacted values | COMMENT |
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| MEDIUM | tests/e2e/offline_inference/test_audex_s2s.py | 43 | # the official temp-1.0 seed-0 draw hallucinates off-language text (yet closes | COMMENT |
| MEDIUM | …uracy/sana_video/test_sana_video_pipeline_alignment.py | 396 | # Isolate scheduler orchestration from native-attention error by | COMMENT |
| MEDIUM | …ts/e2e/online_serving/test_cosyvoice3_tts_expansion.py | 33 | # Official CosyVoice zero-shot prompt audio and its transcript. Vendored under | COMMENT |
| MEDIUM⚡ | tests/engine/test_orchestrator_event_driven.py | 71 | # Parity: the legacy scenario matrix, re-run through the event-driven loop | COMMENT |
| MEDIUM | tests/engine/test_orchestrator_event_driven.py | 89 | # KV/queue gauges under the event-driven loop. | COMMENT |
| MEDIUM | tests/helpers/assertions.py | 755 | # may hallucinate extra context around the short utterance. Use | COMMENT |
| MEDIUM | requirements/common.txt | 39 | # Required for running the Cosmos3 model with guardrails enabled (default behavior). | COMMENT |
| MEDIUM | docs/design/module/entrypoints.md | 91 | ### ENTRY-INV-001: Entrypoints adapt but do not orchestrate | COMMENT |
| MEDIUM | docs/design/module/engine_orchestration.md | 57 | # Engine orchestration | COMMENT |
| MEDIUM | docs/contributing/profiling.md | 329 | [Speech API](../serving/speech_api.md#orchestration-loop-experimental)) an idle | CODE |
| MEDIUM | docs/contributing/model/adding_tts_model.md | 600 | # Build context window: left_context + chunk | COMMENT |
| MEDIUM | examples/offline_inference/audex/speech_to_speech.py | 54 | # the official temp-1.0 seed-0 draw hallucinates off-language text (yet | COMMENT |
| MEDIUM | …mples/offline_inference/text_to_image/text_to_image.py | 477 | # Cosmos3 loads its (gated) guardrail models at build time, so the guardrails | COMMENT |
| MEDIUM | …mples/offline_inference/text_to_image/text_to_image.py | 478 | # gate is an engine-level config (offline analog of the server's --no-guardrails). | COMMENT |
| MEDIUM | …mples/offline_inference/text_to_video/text_to_video.py | 525 | # Cosmos3 loads its (gated) guardrail models at build time, so the guardrails | COMMENT |
| MEDIUM | …mples/offline_inference/text_to_video/text_to_video.py | 526 | # gate is an engine-level config (offline analog of the server's --no-guardrails). | COMMENT |
| MEDIUM | examples/offline_inference/text_to_speech/README.md | 432 | ### Quick start (zero-shot, default voice) | COMMENT |
| MEDIUM | …offline_inference/text_to_speech/cosyvoice3/end2end.py | 18 | # Upstream zero-shot reference clip | COMMENT |
| MEDIUM | …les/offline_inference/image_to_video/image_to_video.py | 566 | # Cosmos3 loads its (gated) guardrail models at build time, so the guardrails | COMMENT |
| MEDIUM | …les/offline_inference/image_to_video/image_to_video.py | 567 | # gate is an engine-level config (offline analog of the server's --no-guardrails). | COMMENT |
| MEDIUM | examples/online_serving/audex/client.py | 46 | # official temp-1.0 seed-0 draw measurably hallucinates off-language text on | COMMENT |
| MEDIUM | examples/online_serving/audex/client.py | 299 | # hallucinate off-language answers (see the module-level comment). | COMMENT |
| MEDIUM | …ine_serving/text_to_speech/cosyvoice3/speech_client.py | 30 | # Official CosyVoice zero-shot prompt and its transcript. | COMMENT |
| MEDIUM | …ctors/connectors/mooncake_transfer_engine_connector.py | 172 | # The orchestration layer (get_connectors_config_for_stage / | COMMENT |
| MEDIUM | vllm_omni/deploy/lance.yaml | 6 | # stage orchestration. | COMMENT |
| MEDIUM | …model_executor/models/hunyuan_image3/hunyuan_image3.py | 1972 | # The <timestep> slot at the head of each per-image scaffold is NOT | COMMENT |
| MEDIUM | …i/model_executor/models/gepard/configuration_gepard.py | 102 | # Carried for the cloning follow-up; zero-shot uses null_prefix only. | COMMENT |
| MEDIUM | …omni/model_executor/models/dots_tts/dots_tts_prompt.py | 68 | # V1 zero-shot leaves this empty; preprocess() ignores it. | COMMENT |
| MEDIUM | …omni/model_executor/models/dots_tts/dots_tts_talker.py | 704 | # upstream zero-shot emits every DiT patch including the first | COMMENT |
| MEDIUM | …omni/model_executor/models/dots_tts/dots_tts_talker.py | 1158 | # runner.py:1114). Otherwise prefill scaffold (24 hidden × 1536 | COMMENT |
| MEDIUM | …el_executor/models/personaplex/personaplex_temporal.py | 200 | # Capacity = context window (the mask truncates at `context` anyway). | COMMENT |
| MEDIUM | …_omni/diffusion/models/hunyuan_image3/system_prompt.py | 195 | # Think task: use chain-of-thought recaption prompt | STRING |
| MEDIUM | vllm_omni/diffusion/models/lance/pipeline_lance.py | 1854 | # the chat scaffolding. | COMMENT |
| MEDIUM | …iffusion/models/qwen_image/pipeline_qwen_image_edit.py | 443 | # The edit template contains fixed multimodal scaffolding around the | COMMENT |
| MEDIUM | …/models/longcat_video/pipeline_longcat_video_avatar.py | 53 | # Whisper uses a 30-second context window: 30s * 16kHz raw samples, | COMMENT |
| MEDIUM | vllm_omni/diffusion/models/cosmos3/guardrails.py | 65 | # ``model_config["guardrails"]`` is False. | COMMENT |
| MEDIUM⚡ | vllm_omni/diffusion/models/cosmos3/guardrails.py | 133 | # caller has opted in to guardrails. | COMMENT |
| MEDIUM | …quantization/tools/merge_mxfp4_dualscale_checkpoint.py | 331 | # BF16 base: provides the scaffold for non-MXFP4 tensors (norms, embeddings, | COMMENT |
| MEDIUM | …quantization/tools/merge_mxfp4_dualscale_checkpoint.py | 378 | # Merge: base_state provides BF16 scaffold; MXFP4 tensors override their BF16 counterparts | COMMENT |
| MEDIUM | vllm_omni/entrypoints/async_omni.py | 56 | # Blocking-wait interval for the event-driven final-output drain | COMMENT |
| MEDIUM | vllm_omni/entrypoints/async_omni.py | 891 | # event-driven loop (vllm_omni/engine/orchestrator.py). | COMMENT |
| MEDIUM | vllm_omni/entrypoints/cli/serve.py | 874 | # Disable safety guardrails for this server (currently only applicable for Cosmos3) | COMMENT |
| MEDIUM | vllm_omni/entrypoints/openai/serving_chat.py | 668 | # scaffold. | COMMENT |
| MEDIUM | vllm_omni/entrypoints/openai/tts_adapters/voxcpm2.py | 91 | # VoxCPM2 has no predefined speaker presets — "default" means zero-shot | COMMENT |
| MEDIUM⚡ | vllm_omni/engine/orchestrator.py | 76 | # VLLM_OMNI_EVENT_DRIVEN_ORCH=1 switches the orchestration loop (and the | COMMENT |
| MEDIUM⚡ | vllm_omni/engine/orchestrator.py | 78 | # 1 ms poll cadence to event-driven wakeups: one reader task per live LLM stage | COMMENT |
| MEDIUM⚡ | vllm_omni/engine/orchestrator.py | 84 | # How often the event-driven loop reconciles its reader-task set against | COMMENT |
| MEDIUM | vllm_omni/engine/orchestrator.py | 166 | # min_tokens wider than the leftover context window never reaches | COMMENT |
| MEDIUM | vllm_omni/engine/orchestrator.py | 593 | # The orchestration loop isolates per-replica EngineDeadError | COMMENT |
| MEDIUM | vllm_omni/engine/orchestrator.py | 675 | # orchestration loop observes _shutdown_event and exits. | COMMENT |
| MEDIUM | vllm_omni/inputs/preprocess.py | 70 | # defaults and scaffold). | COMMENT |
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| LOW | tools/pre_commit/check_test_marks.py | 51 | # Check if a file is located under tests/ and matches test_<something>.py | COMMENT |
| LOW | tools/pre_commit/check_forbidden_imports.py | 249 | # Check if it's allowed | COMMENT |
| LOW | tests/distributed/omni_connectors/test_kv_flow.py | 126 | # Check if data was put into connector | COMMENT |
| LOW | …_executor/models/higgs_audio_v3/test_higgs_audio_v3.py | 118 | # Set weights to identity-like pattern so we can verify offsets | COMMENT |
| LOW | …ts/diffusion/models/flux2/test_flux2_transformer_tp.py | 243 | # Check if to_out related weights are loaded | COMMENT |
| LOW⚡ | tests/benchmarks/patch/test_patch.py | 470 | # Check if attribute exists and has a value (should be 0.0 or similar default) | COMMENT |
| LOW | tests/e2e/offline_inference/compute_lpips.py | 95 | # Print results table | COMMENT |
| LOW | tests/e2e/features/helpers/custom_pipeline.py | 312 | # Check if model is a local path | COMMENT |
| LOW | tests/helpers/stage_config.py | 90 | # Check if next key is a digit (list index) or string (dict key) | COMMENT |
| LOW | tests/helpers/stage_config.py | 224 | # Check if this is a simple key (not dot-separated) | COMMENT |
| LOW | examples/offline_inference/step_audio2/end2end.py | 203 | # Check if it's a local path | COMMENT |
| LOW | examples/offline_inference/image_to_image/image_edit.py | 639 | # Check if this is a layered output (list of images) | COMMENT |
| LOW | examples/offline_inference/image_to_image/image_edit.py | 650 | # Check if this is a layered output (list of images) | COMMENT |
| LOW | …ai_chat_completion_client_for_multimodal_generation.py | 57 | # Check if it's a URL (starts with http:// or https://) | COMMENT |
| LOW | …ai_chat_completion_client_for_multimodal_generation.py | 96 | # Check if it's a URL (starts with http:// or https://) | COMMENT |
| LOW | …ai_chat_completion_client_for_multimodal_generation.py | 133 | # Check if it's a URL (starts with http:// or https://) | COMMENT |
| LOW | …ai_chat_completion_client_for_multimodal_generation.py | 64 | # Check if it's a URL (starts with http:// or https://) | COMMENT |
| LOW | …ai_chat_completion_client_for_multimodal_generation.py | 103 | # Check if it's a URL (starts with http:// or https://) | COMMENT |
| LOW | …ai_chat_completion_client_for_multimodal_generation.py | 141 | # Check if it's already a base64 data URL | COMMENT |
| LOW | …ai_chat_completion_client_for_multimodal_generation.py | 145 | # Check if it's a URL (starts with http:// or https://) | COMMENT |
| LOW | examples/online_serving/lance/openai_chat_client.py | 49 | # Check if local file | COMMENT |
| LOW | examples/online_serving/qwen3_omni/run_gradio_demo.sh | 154 | # Check if startup flag file exists (startup complete) | COMMENT |
| LOW | examples/online_serving/qwen3_omni/run_gradio_demo.sh | 163 | # Check if server process is still running | COMMENT |
| LOW | examples/online_serving/qwen2_5_omni/run_gradio_demo.sh | 154 | # Check if startup flag file exists (startup complete) | COMMENT |
| LOW | examples/online_serving/qwen2_5_omni/run_gradio_demo.sh | 163 | # Check if server process is still running | COMMENT |
| LOW | …online_serving/text_to_speech/qwen3_tts/gradio_demo.py | 365 | # Check if this is a request ID referencing a stored payload | COMMENT |
| LOW | benchmarks/build_dataset/extract_tts_prompts.py | 45 | # Print result stats | COMMENT |
| LOW | benchmarks/build_dataset/extract_tts_prompts.py | 64 | # Check if input file exists | COMMENT |
| LOW | vllm_omni/platforms/cuda/platform.py | 91 | # Check if FA packages are available | COMMENT |
| LOW | vllm_omni/platforms/musa/platform.py | 72 | # Check if FA packages are available | COMMENT |
| LOW | vllm_omni/platforms/rocm/platform.py | 95 | # Check if aiter is available for Flash Attention support | COMMENT |
| LOW⚡ | …ctors/connectors/mooncake_transfer_engine_connector.py | 1104 | # Check if this is a query request | COMMENT |
| LOW⚡ | …mni/distributed/omni_connectors/utils/serialization.py | 210 | # Check if this looks like an OmniRequestOutput (check before RequestOutput | COMMENT |
| LOW⚡ | …mni/distributed/omni_connectors/utils/serialization.py | 215 | # Check if this looks like a RequestOutput | COMMENT |
| LOW⚡ | …mni/distributed/omni_connectors/utils/serialization.py | 219 | # Check if this looks like a CompletionOutput | COMMENT |
| LOW | vllm_omni/plugins/__init__.py | 34 | # Check if the only discovered plugin is the default one. | COMMENT |
| LOW | …model_executor/models/hunyuan_image3/hunyuan_image3.py | 1813 | # Check if T dimension is 1, then squeeze it | COMMENT |
| LOW | …m_omni/model_executor/models/glm_image/glm_image_ar.py | 178 | # Check if we're in a subdirectory (vision_language_encoder) | COMMENT |
| LOW | …m_omni/model_executor/models/glm_image/glm_image_ar.py | 891 | # Check if we have images (i2i mode) | COMMENT |
| LOW | …m_omni/model_executor/models/glm_image/glm_image_ar.py | 945 | # Check if there are any images to process | COMMENT |
| LOW | …m_omni/model_executor/models/glm_image/glm_image_ar.py | 2808 | # Check if this is a text-to-image request: | COMMENT |
| LOW | …m_omni/model_executor/models/glm_image/glm_image_ar.py | 2840 | # Check if there's a token after this start and it's an image token | COMMENT |
| LOW | …odel_executor/models/mimo_audio/mimo_audio_code2wav.py | 955 | # Check if in CUDA graph capture phase | COMMENT |
| LOW | …lm_omni/model_executor/models/mimo_audio/mimo_audio.py | 879 | # Check if in CUDA graph capture phase | COMMENT |
| LOW | …_executor/models/minicpmo_4_5/minicpmo_4_5_omni_llm.py | 461 | # Check if the args are a device or a dtype | STRING |
| LOW | …mni/model_executor/stage_input_processors/glm_image.py | 338 | # Check if elements are tensors or Python lists | COMMENT |
| LOW | …_omni/model_executor/stage_input_processors/glm_tts.py | 340 | # Check if enough tokens accumulated for next chunk | COMMENT |
| LOW | …i/diffusion/offloader/distributed_layerwise_backend.py | 1196 | # Check if this buffer is non-persistent on its OWNING module | COMMENT |
| LOW | vllm_omni/diffusion/attention/parallel/ulysses.py | 366 | # Check if Ring Attention is also active (Hybrid mode) | COMMENT |
| LOW | vllm_omni/diffusion/cache/cachedit/model_specific.py | 57 | # Set timesteps to calculate the split | COMMENT |
| LOW | vllm_omni/diffusion/distributed/sp_plan.py | 403 | # Check if it's an output specification (SequenceParallelOutput or list/tuple thereof) | COMMENT |
| LOW | vllm_omni/diffusion/distributed/cfg_parallel.py | 305 | # Assign branches to ranks via round-robin | COMMENT |
| LOW | …lm_omni/diffusion/models/wan2_2/pipeline_wan2_2_i2v.py | 228 | # Check if this is a two-stage model (MoE with transformer_2) | COMMENT |
| LOW | vllm_omni/diffusion/models/wan2_2/pipeline_wan2_2.py | 352 | # Check if this is a two-stage model (MoE with transformer_2) | COMMENT |
| LOW | vllm_omni/diffusion/models/flux/pipeline_flux.py | 106 | # Check if model is a local path | COMMENT |
| LOW | …_omni/diffusion/models/glm_image/pipeline_glm_image.py | 753 | # Check if prior_token_ids are provided externally (from AR stage in multistage mode) | COMMENT |
| LOW | …ni/diffusion/models/glm_image/glm_image_transformer.py | 582 | # Check if SP is enabled | COMMENT |
| LOW | vllm_omni/diffusion/models/flux2/pipeline_flux2.py | 235 | # Check if it's a list of lists or a list of images | COMMENT |
| LOW | vllm_omni/diffusion/models/flux2/pipeline_flux2.py | 381 | # Check if model is a local path | COMMENT |
| LOW | …ion/models/qwen_image/pipeline_qwen_image_edit_plus.py | 214 | # Check if model is a local path | COMMENT |
| 25 more matches not shown… | ||||
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| MEDIUM | …executor/models/qwen3_tts/test_code_predictor_dtype.py | 769 | # The test harness mocks ``default_weight_loader`` as a no-op; install a | COMMENT |
| MEDIUM | …executor/models/qwen3_tts/test_code_predictor_dtype.py | 830 | # The test harness mocks ``default_weight_loader`` as a no-op; install a | COMMENT |
| MEDIUM | …executor/models/qwen3_tts/test_code_predictor_dtype.py | 1136 | # Find any actual self_attn layer prefix, to make the test robust | COMMENT |
| MEDIUM | tests/model_tests/diffusion/test_alignment.py | 64 | # The common harness does not yet define an image-to-video task or supply | COMMENT |
| MEDIUM | tests/diffusion/kernels/mot/test_mot_linear.py | 320 | # Cosine similarity is the primary correctness gate: robust to scale | COMMENT |
| LOW | …iffusion/distributed/test_autoencoder_kl_wan_encode.py | 47 | # Simple mock: just return the input | COMMENT |
| MEDIUM⚡ | tests/diffusion/models/pi0/test_pi0_units.py | 494 | # vision_tower / embed_scale version-robust paths against regressions. | COMMENT |
| MEDIUM⚡ | …s/e2e/offline_inference/test_sensenova_u1_expansion.py | 48 | # regardless of GQA, so this is the robust LoRA target. Verify against config.json. | COMMENT |
| LOW | …e/features/fullduplex/test_personaplex_server_lease.py | 63 | # The drain must genuinely wait for the worker, not just return. | COMMENT |
| MEDIUM | …line_serving/text_to_speech/voxtral_tts/gradio_demo.py | 43 | # Default fallback voices - comprehensive list | COMMENT |
| LOW | vllm_omni/metrics/stat_logger.py | 114 | # rewritten label set already, so just pass through. | COMMENT |
| LOW | vllm_omni/platforms/cuda/platform.py | 319 | # users can just use IR op priority directly | COMMENT |
| MEDIUM | …ni/platforms/npu/worker/npu_generation_model_runner.py | 253 | # We simply utilize the implementation in vLLM. | COMMENT |
| MEDIUM | vllm_omni/platforms/npu/worker/npu_ar_model_runner.py | 610 | # We simply utilize the implementation in vLLM. | COMMENT |
| LOW⚡ | …ctors/connectors/mooncake_transfer_engine_connector.py | 314 | # We don't actually connect, just use the socket to get routing info | COMMENT |
| LOW | vllm_omni/config/model.py | 66 | # would just use itself as prefix — safe but worth | COMMENT |
| MEDIUM | vllm_omni/model_executor/models/utils.py | 10 | """Make ``trust_remote_code`` weight loading robust to the transformers 5.9 | STRING |
| LOW | …m_omni/model_executor/models/glm_image/glm_image_ar.py | 3140 | # For intermediate tensors (PP), just return hidden states | COMMENT |
| LOW | …lm_omni/model_executor/models/mimo_audio/mimo_audio.py | 475 | # If no bos/eos tokens, just use audio tokens | COMMENT |
| MEDIUM | …els/nemotron_voicechat/nemo_vendored/duplex_ear_tts.py | 325 | # EOS dropout to make the model more robust | COMMENT |
| MEDIUM | …els/nemotron_voicechat/nemo_vendored/duplex_ear_tts.py | 342 | # BOS dropout to make the model more robust | COMMENT |
| MEDIUM | …els/nemotron_voicechat/nemo_vendored/duplex_ear_tts.py | 361 | # BOS dropout to make the model more robust | COMMENT |
| LOW | …del_executor/models/indextts2/gpt/conformer_encoder.py | 141 | # It's better we just return None if no cache is required, | COMMENT |
| MEDIUM | …m_omni/diffusion/attention/backends/ring/ring_utils.py | 107 | # Initialize LSE with robust logic (same as _update) | COMMENT |
| MEDIUM | vllm_omni/diffusion/models/pi0/modeling_pi0.py | 635 | # embedding (version-robust: self-scaled at transformers ≥ 5.4, manually | COMMENT |
| LOW | vllm_omni/diffusion/hooks/base.py | 313 | # Otherwise just call the original forward. | COMMENT |
| MEDIUM | vllm_omni/diffusion/lora/layers/base_linear.py | 51 | # `__dict__` for robust lookups in `__getattr__`. | COMMENT |
| MEDIUM | vllm_omni/benchmarks/data_modules/daily_omni_eval.py | 133 | # ``\b`` after the letter avoids "Because"/"Definitely" false positives | COMMENT |
| MEDIUM | vllm_omni/entrypoints/openai/serving_chat.py | 1438 | # Always track previous_texts for comprehensive output logging | COMMENT |
| MEDIUM | vllm_omni/entrypoints/openai/serving_chat.py | 1905 | # Update for comprehensive logging even in simple case | COMMENT |
| LOW | vllm_omni/worker/gpu_ar_model_runner.py | 2100 | # just use zeros for the draft tokens. | COMMENT |
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| LOW⚡ | tests/core/sched/test_generation_scheduler_restore.py | 52 | # Step 1: process_pending_chunks moves req-B out | COMMENT |
| LOW⚡ | tests/core/sched/test_generation_scheduler_restore.py | 57 | # Step 2: simulate the try/except/finally pattern | COMMENT |
| LOW⚡ | tests/core/sched/test_generation_scheduler_restore.py | 66 | # Step 3: verify request is restored | COMMENT |
| LOW | tests/diffusion/attention/test_attention_sp.py | 473 | # Step 1: Run without SP (baseline with ulysses_degree=1, ring_degree=1) | COMMENT |
| LOW | tests/diffusion/attention/test_attention_sp.py | 502 | # Step 2: Run with SP enabled | COMMENT |
| LOW | tests/diffusion/attention/test_attention_sp.py | 534 | # Step 3: Verify input consistency and compare outputs | COMMENT |
| LOW⚡ | tests/e2e/offline_inference/test_moss_tts_realtime.py | 166 | # Step 1: locate the realtime processor module in the snapshot. | COMMENT |
| LOW⚡ | tests/e2e/offline_inference/test_moss_tts_realtime.py | 191 | # Step 2: encode the reference audio clip via MOSS-Audio-Tokenizer. | COMMENT |
| LOW | tests/e2e/offline_inference/test_moss_tts_realtime.py | 209 | # Step 3: build the (L, 17) prefill grid. | COMMENT |
| LOW⚡ | …ures/custom_pipeline/test_async_omni_collective_rpc.py | 126 | # Step 1: list_loras (the call that was crashing) | COMMENT |
| LOW⚡ | …ures/custom_pipeline/test_async_omni_collective_rpc.py | 130 | # Step 2: generate (should still work after list_loras) | COMMENT |
| LOW⚡ | vllm_omni/core/prefix_cache.py | 211 | # Step 1: consume the previous step's write FIRST. The previous | COMMENT |
| LOW⚡ | vllm_omni/core/prefix_cache.py | 220 | # Step 2: schedule the D2H on the copy stream. PyTorch implicitly | COMMENT |
| LOW | vllm_omni/core/prefix_cache.py | 243 | # Step 3: stash for next step's consume. | COMMENT |
| LOW | …utor/models/ming_flash_omni/ming_flash_omni_thinker.py | 934 | # Step 1: if vision/audio features were extracted, merge them into | COMMENT |
| LOW | …utor/models/ming_flash_omni/ming_flash_omni_thinker.py | 945 | # Step 2: image-gen query tokens are appended at the tail of the prompt | COMMENT |
| LOW⚡ | …odel_executor/stage_input_processors/higgs_audio_v3.py | 139 | # Step 1: Revert delay pattern | COMMENT |
| LOW⚡ | …odel_executor/stage_input_processors/higgs_audio_v3.py | 147 | # Step 2: Replace out-of-range codes (BOC=1024, EOC=1025, -1) with 0. | COMMENT |
| LOW⚡ | …odel_executor/stage_input_processors/higgs_audio_v3.py | 157 | # Step 3: Trim the last frame. After de-delay, the final frame | COMMENT |
| LOW | vllm_omni/diffusion/layers/mot/ops/mot_rms_norm.py | 25 | # Step 0: MoT Routing | COMMENT |
| LOW | vllm_omni/diffusion/layers/mot/ops/mot_rms_norm.py | 45 | # Step 1: Compute sum of squares in float32 to avoid overflow | COMMENT |
| LOW⚡ | vllm_omni/diffusion/layers/mot/ops/mot_rms_norm.py | 57 | # Step 2: Compute RMS (root mean square) in float32 | COMMENT |
| LOW⚡ | vllm_omni/diffusion/layers/mot/ops/mot_rms_norm.py | 62 | # Step 3: Normalize and apply weight | COMMENT |
| LOW⚡ | vllm_omni/diffusion/models/gr00t/policy.py | 336 | # Step 1: Split batched observation into individual observations | COMMENT |
| LOW⚡ | vllm_omni/diffusion/models/gr00t/policy.py | 340 | # Step 2: Process each observation through the VLA processor | COMMENT |
| LOW⚡ | vllm_omni/diffusion/models/gr00t/policy.py | 348 | # Step 3: Collate processed inputs into a single batch for model | COMMENT |
| LOW⚡ | vllm_omni/diffusion/models/gr00t/policy.py | 352 | # Step 4: Run model inference to predict actions | COMMENT |
| LOW⚡ | vllm_omni/diffusion/models/gr00t/policy.py | 357 | # Step 5: Decode actions from normalized space back to physical units | COMMENT |
| LOW | …ls/gr00t/dataio/state_action/state_action_processor.py | 310 | # Step 1: Convert absolute actions to relative (if needed) | COMMENT |
| LOW | …ls/gr00t/dataio/state_action/state_action_processor.py | 343 | # Step 2: Normalize actions | COMMENT |
| LOW | …ls/gr00t/dataio/state_action/state_action_processor.py | 395 | # Step 1: Unnormalize actions | COMMENT |
| LOW | …ls/gr00t/dataio/state_action/state_action_processor.py | 418 | # Step 2: Convert relative actions to absolute (if needed) | COMMENT |
| LOW | vllm_omni/quantization/tools/merge_mxfp8_checkpoint.py | 231 | # Step 1: Copy original model, skipping transformer dirs (will be replaced). | COMMENT |
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| CRITICAL | …executor/models/minicpmo_4_5/test_code2wav_batching.py | 364 | token2wav.flow.decoder.estimator.cfg_batches.clear() | CODE |
| CRITICAL | tests/helpers/client.py | 1479 | with self.client.audio.speech.with_streaming_response.create( | CODE |
| CRITICAL | tests/helpers/client.py | 1519 | with self.client.audio.speech.with_streaming_response.create( | CODE |
| CRITICAL | examples/offline_inference/hunyuan_image3/README.md | 153 | The shared `vllm_omni.diffusion.models.hunyuan_image3.prompt_utils.build_prompt_tokens()` | CODE |
| CRITICAL⚡ | …executor/models/qwen2_5_omni/qwen2_5_omni_token2wav.py | 1231 | self.codec_embed_size = self.code2wav_dit_model.text_embed.codec_embed.weight.size(0) | STRING |
| CRITICAL⚡ | vllm_omni/diffusion/models/gr00t/modeling/gr00t_n1d7.py | 326 | self.model.model.language_model.layers.pop(-1) | CODE |
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| LOW | tools/run_merge_jobs.sh | 21 | # Usage: | COMMENT |
| LOW | tools/run_ready_jobs.sh | 21 | # Usage: | COMMENT |
| LOW | tools/nightly/run_nightly_jobs.sh | 55 | # Usage: | COMMENT |
| LOW | tests/e2e/offline_inference/run_quantization_e2e.sh | 10 | # Usage: | COMMENT |
| LOW | …_inference/qwen3_omni/run_single_prompt_async_chunk.sh | 12 | # Usage: | COMMENT |
| LOW | …ference/qwen3_omni/run_multiple_prompts_async_chunk.sh | 8 | # Usage: | COMMENT |
| LOW | …ine_inference/qwen2_5_omni/strategy_stage_replica.yaml | 14 | # Usage: | COMMENT |
| LOW | …mples/offline_inference/qwen2_5_omni/strategy_tp2.yaml | 12 | # Usage: | COMMENT |
| LOW | …les/online_serving/replica_data_parallel/run_server.sh | 4 | # Usage: | COMMENT |
| LOW | examples/online_serving/audex/run_server.sh | 13 | # Usage: | COMMENT |
| LOW | …amples/online_serving/text_to_video/run_server_ltx2.sh | 7 | # Usage: | COMMENT |
| LOW | examples/online_serving/qwen3_omni/run_gradio_demo.sh | 4 | # Usage: | COMMENT |
| LOW | examples/online_serving/qwen2_5_omni/run_gradio_demo.sh | 4 | # Usage: | COMMENT |
| LOW | …erving/text_to_speech/moss_tts_nano/run_gradio_demo.sh | 4 | # Usage: | COMMENT |
| LOW | …ine_serving/text_to_speech/moss_tts_nano/run_server.sh | 4 | # Usage: | COMMENT |
| LOW | …rving/text_to_speech/ming_flash_omni_tts/run_server.sh | 4 | # Usage: | COMMENT |
| LOW | …online_serving/text_to_speech/cosyvoice3/run_server.sh | 4 | # Usage: | COMMENT |
| LOW | …line_serving/text_to_speech/glm_tts/run_gradio_demo.sh | 4 | # Usage: | COMMENT |
| LOW | …es/online_serving/text_to_speech/glm_tts/run_server.sh | 4 | # Usage: | COMMENT |
| LOW | …/online_serving/text_to_speech/omnivoice/run_server.sh | 4 | # Usage: | COMMENT |
| LOW | …_serving/text_to_speech/fish_speech/run_gradio_demo.sh | 4 | # Usage: | COMMENT |
| LOW | …nline_serving/text_to_speech/fish_speech/run_server.sh | 4 | # Usage: | COMMENT |
| LOW | …s/online_serving/text_to_speech/ming_tts/run_server.sh | 4 | # Usage: | COMMENT |
| LOW | …ne_serving/text_to_speech/higgs_audio_v2/run_server.sh | 8 | # Usage: | COMMENT |
| LOW | …ne_serving/text_to_speech/higgs_audio_v3/run_server.sh | 6 | # Usage: | COMMENT |
| LOW | …ne_serving/text_to_speech/qwen3_tts/run_gradio_demo.sh | 4 | # Usage: | COMMENT |
| LOW | …/online_serving/text_to_speech/qwen3_tts/run_server.sh | 4 | # Usage: | COMMENT |
| LOW | .buildkite/release/scripts/publish-release-images.sh | 6 | # Usage: | COMMENT |
| LOW | vllm_omni/deploy/qwen3_omni_moe_mori_intranode.yaml | 26 | # Usage: | COMMENT |
| LOW | vllm_omni/deploy/indextts2_low_latency.yaml | 8 | # Usage: | COMMENT |
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| MEDIUM | tests/diffusion/test_inline_stage_diffusion_client.py | 142 | CODE | |
| MEDIUM | …/entrypoints/openai_api/test_serving_audio_generate.py | 418 | CODE | |
| MEDIUM | …/entrypoints/openai_api/test_serving_audio_generate.py | 477 | CODE | |
| MEDIUM | …/entrypoints/openai_api/test_serving_audio_generate.py | 492 | CODE | |
| MEDIUM | tests/entrypoints/openai_api/test_serving_speech.py | 3005 | CODE | |
| MEDIUM | tests/entrypoints/openai_api/test_serving_speech.py | 4815 | CODE | |
| MEDIUM | tests/entrypoints/openai_api/test_image_server.py | 615 | CODE | |
| MEDIUM | tests/entrypoints/openai_api/test_image_server.py | 937 | CODE |
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| HIGH | .github/ISSUE_TEMPLATE/400-bug-report.yml | 20 | Consider redacting or replacing sensitive values with placeholders like `<YOUR_TOKEN_HERE>` when sharing configura | CODE |
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| LOW | tests/diffusion/test_stage_diffusion_proc.py | 100 | async def run_task(req_data): | CODE |
| LOW | tests/diffusion/test_diffusion_engine.py | 811 | async def run_task(rid): | CODE |
| LOW | examples/offline_inference/lance/gradio_demo.py | 794 | def run_task( | CODE |
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| LOW | …lm_omni/experimental/fullduplex/personaplex/runtime.py | 301 | dummy_user = self._initial[9:].view(1, 8).expand(B, 8) | CODE |
| LOW | …lm_omni/experimental/fullduplex/personaplex/runtime.py | 309 | ran, _ = self._tick(dummy_user, dummy_moshi, zero_text, all_rows, embed_override=(ov, all_rows)) | CODE |
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| LOW | …_executor/models/minicpmo_4_5/minicpmo_4_5_omni_llm.py | 1904 | # When lifting this restriction, don't forget to either | STRING |
| LOW | …_executor/models/minicpmo_4_5/minicpmo_4_5_omni_llm.py | 1904 | # When lifting this restriction, don't forget to either | STRING |
| MEDIUM | vllm_omni/diffusion/models/bagel/bagel_transformer.py | 369 | although it's worth noting that this is currently only used | STRING |