A high-throughput and memory-efficient inference and serving engine for LLMs
This report summarizes scanner evidence for vllm-project/vllm, a Python project with 93,514 GitHub stars. SynthScan examined 2,323,866 lines of code across 7205 source files, recording 30615 pattern matches distributed across 28 syntactic categories. The legacy adjusted pattern signal is 16.9 (Elevated pattern signal); it is not an authorship estimate.
Use the findings to ask whether maintainers can explain intent, verify dependencies, simplify unnecessary scaffolding, and focus review effort. Risk Payload v1 adds ownership, supply-chain, over-engineering, and review-effort vectors when available. These are prompts for human verification, not probabilities, security findings, or measured maintenance costs.
Use each vector to guide manual verification of the underlying findings.
Can maintainers explain intent, assumptions, and trade-offs? 23910 finding(s).
Are imports declared and package/API names verified? 2229 finding(s).
Are wrappers, guards, and repeated structures necessary? 24361 finding(s).
Which changes need focused human review? 30615 finding(s).
This chart tracks the legacy adjusted pattern score across scan runs. Changes can reflect source changes, scan coverage, or detector/scoring versions; they do not measure AI adoption or prove that a team refactored code. Compare runs only when their scoring versions are compatible. The dashed line tracks the raw pattern-hit count.
Groups findings by configured severity weight. Severity is not a confidence estimate, security rating, or proof that a package is missing. Verify imports against the project's manifests and inspect each finding in its local design context.
This horizontal bar chart shows where configured pattern scores accumulate across directories. Treat the bars as locations to inspect, not as evidence of authorship or code quality. Use the file-level findings to decide whether a directory needs a focused ownership, dependency, complexity, or review check.
The scanner identified 30615 distinct pattern matches across 28 syntactic categories. Each entry below represents a discrete location in the source code where the engine recorded a detector match to inspect in context. Expand any category row to inspect the individual file paths, line numbers, code snippets, and the lexical context (CODE, COMMENT, or STRING) in which each match was detected.
Reading the findings table: The Severity column indicates the configured finding weight (CRITICAL / HIGH / MEDIUM / LOW). The Context column identifies whether the match occurred inside executable code, an inline comment, or a string literal — legacy comment-context matches receive a ×1.5 score weight. The ⚡ bolt icon marks clustered matches: three or more patterns within a 10-line window, each receiving an additional ×1.5 density multiplier under the legacy scoring model.
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| LOW⚡ | setup.py | 69 | def should_require_rust_frontend() -> bool: | CODE |
| LOW⚡ | setup.py | 74 | def get_precompiled_rust_extension_paths() -> list[Path]: | CODE |
| LOW⚡ | setup.py | 78 | def get_missing_precompiled_rust_extension_modules() -> list[str]: | CODE |
| LOW⚡ | setup.py | 652 | def detect_system_rocm_variant() -> str | None: | CODE |
| LOW⚡ | setup.py | 669 | def fetch_available_rocm_variants(commit: str) -> list[str]: | CODE |
| LOW⚡ | setup.py | 688 | def resolve_rocm_wheel_variant( | CODE |
| LOW | setup.py | 89 | def has_precompiled_rust_extensions() -> bool: | CODE |
| LOW | setup.py | 523 | def fetch_metadata_for_variant( | CODE |
| LOW | setup.py | 605 | def detect_system_cuda_variant() -> str: | CODE |
| LOW | setup.py | 737 | def warn_if_rocm_torch_version_mismatch( | CODE |
| LOW | setup.py | 793 | def fetch_wheel_from_pypi_index(index_url: str, package: str = "vllm") -> str: | CODE |
| LOW | setup.py | 914 | def find_published_wheel_commit( | CODE |
| LOW | setup.py | 1084 | def extract_precompiled_and_patch_package( | CODE |
| LOW | setup.py | 1211 | def get_base_commit_in_main_branch() -> str: | CODE |
| LOW | setup.py | 1401 | def _check_requirements_preinstalled(requirements: list[str]) -> None: | CODE |
| LOW | csrc/cpu/generate_cpu_attn_dispatch.py | 100 | def generate_cases_for_isa_group(isa_list: list[str], include_fp8: bool = False) -> str: | CODE |
| LOW | csrc/libtorch_stable/quantization/machete/generate.py | 334 | def generate_type_option_name(kernel_types: TypeConfig): | STRING |
| LOW | csrc/libtorch_stable/quantization/machete/generate.py | 370 | def unsigned_type_with_bitwidth(num_bits): | STRING |
| LOW | tools/generate_versions_json.py | 85 | def generate_bake_native_json(args: dict[str, str]) -> dict: | CODE |
| LOW | tools/benchmark_helion_kernels.py | 259 | def make_correctness_baseline( | CODE |
| LOW | tools/install_nixl_from_source_ubuntu.py | 65 | def install_system_dependencies(): | CODE |
| LOW | tools/install_nixl_from_source_ubuntu.py | 105 | def build_and_install_prerequisites(args): | CODE |
| LOW | tools/build_nixl_with_torch215_for_rubin.py | 161 | def get_default_cuda_arch_list(): | CODE |
| LOW | tools/build_nixl_with_torch215_for_rubin.py | 201 | def install_system_dependencies(build_nixl_ep=False): | CODE |
| LOW | tools/build_nixl_with_torch215_for_rubin.py | 257 | def install_python_build_dependencies(): | CODE |
| LOW | tools/build_nixl_with_torch215_for_rubin.py | 452 | def build_and_install_prerequisites(args): | CODE |
| LOW | tools/build_rust.py | 32 | def prepare_build_environment() -> str | None: | CODE |
| LOW | tools/build_rust.py | 70 | def rust_py_extension_module_names() -> list[str]: | CODE |
| LOW | tools/pre_commit/check_test_tethering.py | 481 | def _looks_like_command_string(token: str) -> bool: | CODE |
| LOW | tools/pre_commit/check_test_tethering.py | 627 | def _exec_runs_matches_through_pytest(action: list[str]) -> bool: | CODE |
| LOW | tools/pre_commit/check_test_tethering.py | 795 | def allowlist_entries_missing_reason() -> list[str]: | CODE |
| LOW | tools/pre_commit/check_test_tethering.py | 884 | def _change_set_touches_ci_config(paths: list[str]) -> bool: | CODE |
| LOW | …s/gumbel_precision/prove_exponential_race_precision.py | 31 | def measure_exponential_lower_tail( | CODE |
| LOW | tools/recipes/hardware_detection.py | 67 | def build_numa_omp_threads_bind( | CODE |
| LOW | tools/recipes/model_constraints.py | 102 | def suggest_tensor_parallel_size( | CODE |
| LOW⚡ | tools/recipes/sweep/sweep_recommendation.py | 80 | def _scheduler_preference_key(value: object) -> tuple[int, int]: | CODE |
| LOW⚡ | tools/recipes/sweep/sweep_recommendation.py | 89 | def _has_fixed_parallel_layout(candidates: list[dict[str, Any]]) -> bool: | CODE |
| LOW⚡ | tools/recipes/sweep/sweep_recommendation.py | 98 | def _select_near_equivalent_scheduler_candidate( | CODE |
| LOW | tools/recipes/sweep/sweep_recommendation.py | 369 | def _build_recommended_config( | CODE |
| LOW | tools/recipes/sweep/runtime_tuning.py | 84 | def _get_active_sequence_count( | CODE |
| LOW | tools/recipes/sweep/runtime_tuning.py | 115 | def _estimate_prefills_per_step( | CODE |
| LOW | tools/recipes/sweep/runtime_tuning.py | 145 | def _resolve_tensor_parallel_size( | CODE |
| LOW | tools/recipes/sweep/runtime_tuning.py | 198 | def _resolve_gpu_memory_utilization( | CODE |
| LOW | tools/recipes/sweep/runtime_tuning.py | 263 | def _resolve_max_num_batched_tokens( | CODE |
| LOW | tools/recipes/sweep/runtime_tuning.py | 322 | def _resolve_data_parallel_size( | CODE |
| LOW | tools/recipes/sweep/runtime_tuning.py | 358 | def get_runtime_tuning_policies( | CODE |
| LOW⚡ | tools/recipes/sweep/scheduler_sweep_prepare.py | 32 | def _num_prompts_for_concurrency(concurrency: int) -> int: | CODE |
| LOW⚡ | tools/recipes/sweep/scheduler_sweep_prepare.py | 39 | def _strict_lower_power_of_two(value: int) -> int: | CODE |
| LOW⚡ | tools/recipes/sweep/scheduler_sweep_prepare.py | 45 | def _strict_upper_power_of_two(value: int) -> int: | CODE |
| LOW | tools/recipes/sweep/sweep_generation.py | 55 | def _num_prompts_for_concurrency(concurrency: int) -> int: | CODE |
| LOW | tools/recipes/sweep/sweep_generation.py | 69 | def _strict_lower_power_of_two(value: int) -> int: | CODE |
| LOW | tools/recipes/sweep/sweep_generation.py | 75 | def _strict_upper_power_of_two(value: int) -> int: | CODE |
| LOW | tools/recipes/sweep/sweep_generation.py | 160 | def _scheduler_baseline_for_dp( | CODE |
| LOW | tools/recipes/sweep/sweep_generation.py | 211 | def _first_positive_model_int(views: list[Any], keys: tuple[str, ...]) -> int | None: | CODE |
| LOW | tools/recipes/sweep/sweep_generation.py | 233 | def model_parallel_metadata_from_config(hf_config: Any) -> ModelParallelMetadata: | CODE |
| LOW | tools/recipes/sweep/sweep_generation.py | 253 | def inspect_model_parallel_metadata(config: dict[str, Any]) -> ModelParallelMetadata: | CODE |
| LOW | tools/recipes/sweep/sweep_generation.py | 271 | def cpu_supports_moe_dp_collectives() -> bool: | CODE |
| LOW | tools/recipes/sweep/sweep_generation.py | 299 | def _default_parallel_layout_sizes(numa_node_count: int) -> list[tuple[int, int]]: | CODE |
| LOW | tools/recipes/sweep/sweep_generation.py | 316 | def build_parallel_layout_plan( | CODE |
| LOW | tools/recipes/sweep/sweep_generation.py | 416 | def build_parallel_layout_params( | CODE |
| 19404 more matches not shown… | ||||
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| MEDIUM⚡ | docker/entrypoints/test_vllm_nonroot_entrypoint.sh | 107 | # ----------------------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | docker/entrypoints/test_vllm_nonroot_entrypoint.sh | 109 | # ----------------------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | docker/entrypoints/test_vllm_nonroot_entrypoint.sh | 118 | # ----------------------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | docker/entrypoints/test_vllm_nonroot_entrypoint.sh | 122 | # ----------------------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | docker/entrypoints/test_vllm_nonroot_entrypoint.sh | 136 | # ----------------------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | docker/entrypoints/test_vllm_nonroot_entrypoint.sh | 139 | # ----------------------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | docker/entrypoints/test_vllm_nonroot_entrypoint.sh | 151 | # ----------------------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | docker/entrypoints/test_vllm_nonroot_entrypoint.sh | 157 | # ----------------------------------------------------------------------------- | COMMENT |
| MEDIUM | docker/entrypoints/test_vllm_nonroot_entrypoint.sh | 66 | # ----------------------------------------------------------------------------- | COMMENT |
| MEDIUM | docker/entrypoints/test_vllm_nonroot_entrypoint.sh | 68 | # ----------------------------------------------------------------------------- | COMMENT |
| MEDIUM | docker/entrypoints/test_vllm_nonroot_entrypoint.sh | 79 | # ----------------------------------------------------------------------------- | COMMENT |
| MEDIUM | docker/entrypoints/test_vllm_nonroot_entrypoint.sh | 82 | # ----------------------------------------------------------------------------- | COMMENT |
| MEDIUM | docker/entrypoints/test_vllm_nonroot_entrypoint.sh | 94 | # ----------------------------------------------------------------------------- | COMMENT |
| MEDIUM | docker/entrypoints/test_vllm_nonroot_entrypoint.sh | 96 | # ----------------------------------------------------------------------------- | COMMENT |
| MEDIUM | docker/entrypoints/test_vllm_nonroot_entrypoint.sh | 175 | # ----------------------------------------------------------------------------- | COMMENT |
| MEDIUM | docker/entrypoints/test_vllm_nonroot_entrypoint.sh | 179 | # ----------------------------------------------------------------------------- | COMMENT |
| MEDIUM | docker/entrypoints/test_vllm_nonroot_entrypoint.sh | 192 | # ----------------------------------------------------------------------------- | COMMENT |
| MEDIUM | docker/entrypoints/test_vllm_nonroot_entrypoint.sh | 199 | # ----------------------------------------------------------------------------- | COMMENT |
| MEDIUM | docker/entrypoints/test_vllm_nonroot_entrypoint.sh | 214 | # ----------------------------------------------------------------------------- | COMMENT |
| MEDIUM | docker/entrypoints/test_vllm_nonroot_entrypoint.sh | 217 | # ----------------------------------------------------------------------------- | COMMENT |
| MEDIUM | docker/entrypoints/test_vllm_nonroot_entrypoint.sh | 240 | # ----------------------------------------------------------------------------- | COMMENT |
| MEDIUM | docker/entrypoints/test_vllm_nonroot_entrypoint.sh | 243 | # ----------------------------------------------------------------------------- | COMMENT |
| MEDIUM | tests/test_ray_env.py | 11 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM | tests/test_ray_env.py | 13 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | tests/test_ray_env.py | 46 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | tests/test_ray_env.py | 48 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | tests/test_ray_env.py | 61 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | tests/test_ray_env.py | 63 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM | tests/test_ray_env.py | 97 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM | tests/test_ray_env.py | 99 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | tests/test_ray_env.py | 133 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | tests/test_ray_env.py | 135 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | tests/test_ray_env.py | 147 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | tests/test_ray_env.py | 149 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | tests/v1/kv_offload/test_file_mapper.py | 17 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | tests/v1/kv_offload/test_file_mapper.py | 19 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | tests/v1/kv_offload/test_file_mapper.py | 73 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | tests/v1/kv_offload/test_file_mapper.py | 75 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | tests/v1/kv_offload/test_file_mapper.py | 154 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | tests/v1/kv_offload/test_file_mapper.py | 157 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | tests/v1/kv_offload/test_file_mapper.py | 227 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | tests/v1/kv_offload/test_file_mapper.py | 229 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | tests/v1/kv_offload/cpu/test_shared_offload_region.py | 247 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | tests/v1/kv_offload/cpu/test_shared_offload_region.py | 249 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | tests/v1/kv_offload/cpu/test_shared_offload_region.py | 287 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | tests/v1/kv_offload/cpu/test_shared_offload_region.py | 289 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | tests/v1/kv_offload/cpu/test_shared_offload_region.py | 312 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | tests/v1/kv_offload/cpu/test_shared_offload_region.py | 314 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | tests/v1/kv_offload/cpu/test_shared_offload_region.py | 344 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | tests/v1/kv_offload/cpu/test_shared_offload_region.py | 346 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | tests/v1/kv_offload/cpu/test_shared_offload_region.py | 460 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | tests/v1/kv_offload/cpu/test_shared_offload_region.py | 462 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | tests/v1/kv_offload/cpu/test_shared_offload_region.py | 623 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | tests/v1/kv_offload/cpu/test_shared_offload_region.py | 625 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | tests/v1/kv_offload/cpu/test_shared_offload_region.py | 726 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | tests/v1/kv_offload/cpu/test_shared_offload_region.py | 728 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | tests/v1/kv_offload/cpu/test_shared_offload_region.py | 877 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | tests/v1/kv_offload/cpu/test_shared_offload_region.py | 879 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | tests/v1/kv_offload/cpu/test_shared_offload_region.py | 933 | # --------------------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | tests/v1/kv_offload/cpu/test_shared_offload_region.py | 935 | # --------------------------------------------------------------------------- | COMMENT |
| 2390 more matches not shown… | ||||
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| HIGH | tests/v1/logits_processors/utils.py | 0 | the request-level logits processor masks out all logits except the token id identified by `target_token` | STRING |
| HIGH | examples/features/logits_processor/custom_req.py | 0 | the request-level logits processor masks out all logits except the token id identified by `target_token` | STRING |
| HIGH | examples/features/logits_processor/custom_req_init.py | 0 | the request-level logits processor masks out all logits except the token id identified by `target_token` | STRING |
| HIGH | tests/v1/logits_processors/utils.py | 0 | this method returns a new request-level logits processor, customized to the `target_token` value associated with a parti | STRING |
| HIGH | examples/features/logits_processor/custom_req.py | 0 | this method returns a new request-level logits processor, customized to the `target_token` value associated with a parti | STRING |
| HIGH | examples/features/logits_processor/custom_req_init.py | 0 | this method returns a new request-level logits processor, customized to the `target_token` value associated with a parti | STRING |
| HIGH | …s/v1/kv_connector/nixl_integration/toy_proxy_server.py | 0 | lifespan context manager to handle startup and shutdown events. | STRING |
| HIGH | …cache/disagg_prefill_lmcache_v1/disagg_proxy_server.py | 0 | lifespan context manager to handle startup and shutdown events. | STRING |
| HIGH | …regated/mooncake_connector/mooncake_connector_proxy.py | 0 | lifespan context manager to handle startup and shutdown events. | STRING |
| HIGH | tests/tool_parsers/test_granite_tool_parser.py | 0 | <longcat_tool_call>{ "name": "test_function", "arguments": { "string_field": "hello", "int_field": 42, "float_field": 3. | STRING |
| HIGH | tests/tool_parsers/test_granite_20b_fc_tool_parser.py | 0 | <longcat_tool_call>{ "name": "test_function", "arguments": { "string_field": "hello", "int_field": 42, "float_field": 3. | STRING |
| HIGH | tests/tool_parsers/test_phi4mini_tool_parser.py | 0 | <longcat_tool_call>{ "name": "test_function", "arguments": { "string_field": "hello", "int_field": 42, "float_field": 3. | STRING |
| HIGH | tests/tool_parsers/test_longcat_tool_parser.py | 0 | <longcat_tool_call>{ "name": "test_function", "arguments": { "string_field": "hello", "int_field": 42, "float_field": 3. | STRING |
| HIGH | tests/models/language/pooling/conftest.py | 0 | reclaim cached gpu memory between tests in this shard. these tests start many vllm/hf engines sequentially in one pytest | STRING |
| HIGH | tests/models/language/pooling_mteb_test/conftest.py | 0 | reclaim cached gpu memory between tests in this shard. these tests start many vllm/hf engines sequentially in one pytest | STRING |
| HIGH | tests/models/multimodal/pooling/conftest.py | 0 | reclaim cached gpu memory between tests in this shard. these tests start many vllm/hf engines sequentially in one pytest | STRING |
| HIGH | tests/models/multimodal/pooling/test_colpali.py | 0 | create a small solid-color png image and return its base64 data uri. | STRING |
| HIGH | tests/models/multimodal/pooling/test_colqwen3.py | 0 | create a small solid-color png image and return its base64 data uri. | STRING |
| HIGH | tests/entrypoints/pooling/scoring/util.py | 0 | create a small solid-color png image and return its base64 data uri. | STRING |
| HIGH | tests/models/multimodal/pooling/test_colpali.py | 0 | build a scoremultimodalparam containing an image (and optional text). | STRING |
| HIGH | tests/models/multimodal/pooling/test_colqwen3.py | 0 | build a scoremultimodalparam containing an image (and optional text). | STRING |
| HIGH | tests/entrypoints/pooling/scoring/util.py | 0 | build a scoremultimodalparam containing an image (and optional text). | STRING |
| HIGH | tests/models/multimodal/pooling/test_colpali.py | 0 | verify per-token embedding shape and l2 normalization. | STRING |
| HIGH | tests/models/multimodal/pooling/test_colqwen3.py | 0 | verify per-token embedding shape and l2 normalization. | STRING |
| HIGH | tests/models/multimodal/pooling/test_colqwen3_5.py | 0 | verify per-token embedding shape and l2 normalization. | STRING |
| HIGH | tests/models/multimodal/pooling/test_colpali.py | 0 | verify that relevant documents score higher than irrelevant ones. | STRING |
| HIGH | tests/models/multimodal/pooling/test_colqwen3.py | 0 | verify that relevant documents score higher than irrelevant ones. | STRING |
| HIGH | tests/models/multimodal/pooling/test_colqwen3_5.py | 0 | verify that relevant documents score higher than irrelevant ones. | STRING |
| HIGH | …ts/models/multimodal/generation/test_granite_speech.py | 0 | sanitize vllm output [phi3v] to be comparable with hf output. | STRING |
| HIGH | tests/models/multimodal/generation/test_phi4mm.py | 0 | sanitize vllm output [phi3v] to be comparable with hf output. | STRING |
| HIGH | …/models/multimodal/generation/vlm_utils/model_utils.py | 0 | sanitize vllm output [phi3v] to be comparable with hf output. | STRING |
| HIGH | tests/entrypoints/openai/completion/test_completion.py | 0 | root ::= select_statement select_statement ::= "select " column " from " table " where " condition column ::= "col_1 " | | STRING |
| HIGH | …ypoints/openai/chat_completion/test_chat_completion.py | 0 | root ::= select_statement select_statement ::= "select " column " from " table " where " condition column ::= "col_1 " | | STRING |
| HIGH | docs/features/structured_outputs.md | 0 | root ::= select_statement select_statement ::= "select " column " from " table " where " condition column ::= "col_1 " | | STRING |
| HIGH | …atures/structured_outputs/structured_outputs_client.py | 0 | root ::= select_statement select_statement ::= "select " column " from " table " where " condition column ::= "col_1 " | | STRING |
| HIGH | …tures/structured_outputs/structured_outputs_offline.py | 0 | root ::= select_statement select_statement ::= "select " column " from " table " where " condition column ::= "col_1 " | | STRING |
| HIGH | benchmarks/benchmark_serving_structured_output.py | 0 | root ::= select_statement select_statement ::= "select " column " from " table " where " condition column ::= "col_1 " | | STRING |
| HIGH | tests/lora/test_qwen3moe_tp.py | 0 | i want you to act as a sql terminal in front of an example database, you need only to return the sql command to me. do n | STRING |
| HIGH | tests/lora/test_llama_tp.py | 0 | i want you to act as a sql terminal in front of an example database, you need only to return the sql command to me. do n | STRING |
| HIGH | tests/lora/test_olmoe_tp.py | 0 | i want you to act as a sql terminal in front of an example database, you need only to return the sql command to me. do n | STRING |
| HIGH | vllm/sampling_params.py | 0 | whether to include the stop strings in output text. | STRING |
| HIGH | …m/entrypoints/speech_to_text/transcription/protocol.py | 0 | whether to include the stop strings in output text. | STRING |
| HIGH | vllm/entrypoints/speech_to_text/translation/protocol.py | 0 | whether to include the stop strings in output text. | STRING |
| HIGH | vllm/v1/attention/backends/flash_attn.py | 0 | forward pass with aiterflashattention. args: layer: the attention layer, providing the q/k/v quantization scales. query: | STRING |
| HIGH | vllm/v1/attention/backends/flash_attn_diffkv.py | 0 | forward pass with aiterflashattention. args: layer: the attention layer, providing the q/k/v quantization scales. query: | STRING |
| HIGH | vllm/v1/attention/backends/rocm_aiter_fa.py | 0 | forward pass with aiterflashattention. args: layer: the attention layer, providing the q/k/v quantization scales. query: | STRING |
| HIGH | vllm/v1/attention/backends/flash_attn.py | 0 | forward pass for encoder attention without kv cache. args: query: shape = [num_encoder_tokens, num_heads, head_size] key | STRING |
| HIGH | vllm/v1/attention/backends/triton_attn.py | 0 | forward pass for encoder attention without kv cache. args: query: shape = [num_encoder_tokens, num_heads, head_size] key | STRING |
| HIGH | vllm/v1/attention/backends/rocm_attn.py | 0 | forward pass for encoder attention without kv cache. args: query: shape = [num_encoder_tokens, num_heads, head_size] key | STRING |
| HIGH | vllm/v1/attention/backends/triton_attn.py | 0 | forward pass with flexattention. args: layer: the attention layer, providing the q/k/v quantization scales. query: shape | STRING |
| HIGH | vllm/v1/attention/backends/rocm_attn.py | 0 | forward pass with flexattention. args: layer: the attention layer, providing the q/k/v quantization scales. query: shape | STRING |
| HIGH | vllm/v1/attention/backends/rocm_aiter_unified_attn.py | 0 | forward pass with flexattention. args: layer: the attention layer, providing the q/k/v quantization scales. query: shape | STRING |
| HIGH | vllm/v1/attention/backends/cpu_attn.py | 0 | forward pass with flexattention. args: layer: the attention layer, providing the q/k/v quantization scales. query: shape | STRING |
| HIGH | vllm/v1/attention/backends/flex_attention.py | 0 | forward pass with flexattention. args: layer: the attention layer, providing the q/k/v quantization scales. query: shape | STRING |
| HIGH | vllm/v1/attention/backends/mamba_attn.py | 0 | this method builds the metadata for full cudagraph capture. currently, only decode is supported for full cudagraphs with | STRING |
| HIGH | vllm/v1/attention/backends/gdn_attn.py | 0 | this method builds the metadata for full cudagraph capture. currently, only decode is supported for full cudagraphs with | STRING |
| HIGH | vllm/model_executor/layers/attention/mla_attention.py | 0 | this method builds the metadata for full cudagraph capture. currently, only decode is supported for full cudagraphs with | STRING |
| HIGH | vllm/v1/core/sched/interface.py | 0 | reset the encoder cache to invalidate all cached encoder outputs. this should be called when model weights are updated t | STRING |
| HIGH | vllm/v1/core/sched/scheduler.py | 0 | reset the encoder cache to invalidate all cached encoder outputs. this should be called when model weights are updated t | STRING |
| HIGH | vllm/v1/engine/llm_engine.py | 0 | reset the encoder cache to invalidate all cached encoder outputs. this should be called when model weights are updated t | STRING |
| 355 more matches not shown… | ||||
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| LOW | …c/cutlass_extensions/vllm_cutlass_library_extension.py | 6 | CODE | |
| LOW | tools/autotune_helion_kernels.py | 32 | CODE | |
| LOW | tools/build_rust.py | 6 | CODE | |
| LOW | tools/ci/kernel_symbol_map.py | 62 | CODE | |
| LOW | …s/gumbel_precision/prove_exponential_race_precision.py | 18 | CODE | |
| LOW | tools/recipes/hardware_detection.py | 10 | CODE | |
| LOW | tools/recipes/model_constraints.py | 6 | CODE | |
| LOW | tools/recipes/recipe_json_to_vllm_config.py | 43 | CODE | |
| LOW | tools/recipes/sweep/sweep_recommendation.py | 7 | CODE | |
| LOW | tools/recipes/sweep/sweep_visualization.py | 7 | CODE | |
| LOW | tools/recipes/sweep/concurrency_recommendation.py | 7 | CODE | |
| LOW | tools/recipes/sweep/runtime_tuning.py | 14 | CODE | |
| LOW | tools/recipes/sweep/scheduler_sweep_prepare.py | 7 | CODE | |
| LOW | tools/recipes/sweep/adaptive_concurrency.py | 7 | CODE | |
| LOW | tools/recipes/sweep/sweep_generation.py | 6 | CODE | |
| LOW | tools/recipes/sweep/sweep_html_report.py | 7 | CODE | |
| LOW | tests/conftest.py | 26 | CODE | |
| LOW | tests/conftest.py | 78 | CODE | |
| LOW | tests/conftest.py | 78 | CODE | |
| LOW | tests/conftest.py | 79 | CODE | |
| LOW | tests/utils.py | 53 | CODE | |
| LOW | tests/utils.py | 59 | CODE | |
| LOW | tests/utils.py | 2140 | CODE | |
| LOW | tests/v1/kv_offload/tiering/p2p/test_sessions.py | 13 | CODE | |
| LOW | tests/v1/kv_offload/tiering/p2p/test_zmq_transport.py | 5 | CODE | |
| LOW | tests/v1/kv_offload/tiering/p2p/test_data_transport.py | 5 | CODE | |
| LOW | tests/v1/kv_offload/tiering/p2p/test_manager.py | 9 | CODE | |
| LOW | tests/v1/attention/test_rocm_aiter_mla_dcp_cprr.py | 12 | CODE | |
| LOW | tests/v1/attention/test_mla_backends.py | 301 | CODE | |
| LOW | …ts/v1/attention/test_sparse_indexer_decode_seq_lens.py | 28 | CODE | |
| LOW | …/v1/attention/test_rocm_aiter_mla_dcp_cprr_numerics.py | 27 | CODE | |
| LOW | tests/v1/attention/test_attention_backends.py | 77 | CODE | |
| LOW | tests/v1/logits_processors/utils.py | 13 | CODE | |
| LOW | tests/v1/cudagraph/test_cudagraph_mode.py | 38 | CODE | |
| LOW | tests/v1/cudagraph/test_breakable_cudagraph.py | 5 | CODE | |
| LOW | tests/v1/kv_connector/unit/test_tp_mapping.py | 10 | CODE | |
| LOW | tests/v1/kv_connector/unit/test_hf3fs_client.py | 16 | CODE | |
| LOW | tests/v1/kv_connector/unit/test_hf3fs_client.py | 16 | CODE | |
| LOW | tests/v1/kv_connector/unit/test_hf3fs_client.py | 16 | CODE | |
| LOW | tests/v1/kv_connector/unit/test_hf3fs_client.py | 16 | CODE | |
| LOW | tests/v1/kv_connector/unit/test_hf3fs_client.py | 16 | CODE | |
| LOW | tests/v1/kv_connector/unit/test_nixl_push_connector.py | 21 | CODE | |
| LOW | …kv_connector/unit/test_simple_cpu_offload_connector.py | 5 | CODE | |
| LOW | …1/kv_connector/unit/test_simple_cpu_offload_metrics.py | 5 | CODE | |
| LOW | …/v1/kv_connector/unit/offloading_connector/conftest.py | 3 | CODE | |
| LOW | …onnector/mooncake_integration/test_mooncake_imports.py | 22 | CODE | |
| LOW | tests/v1/spec_decode/test_ngram_gpu.py | 14 | CODE | |
| LOW | tests/v1/ec_connector/unit/test_scheduler_nixl_ctor.py | 66 | CODE | |
| LOW | tests/v1/ec_connector/unit/cpu/test_connector.py | 61 | CODE | |
| LOW | tests/v1/engine/conftest.py | 19 | CODE | |
| LOW | tests/v1/engine/conftest.py | 19 | CODE | |
| LOW | tests/v1/simple_kv_offload/test_scheduler.py | 5 | CODE | |
| LOW | tests/v1/simple_kv_offload/test_kv_events.py | 11 | CODE | |
| LOW | tests/v1/simple_kv_offload/test_worker.py | 10 | CODE | |
| LOW | tests/tool_use/test_gemma4_responses_adjust_request.py | 32 | CODE | |
| LOW | tests/tool_parsers/test_poolside_v1_tool_parser.py | 21 | CODE | |
| LOW | tests/renderers/test_chat_utils_prompt_embeds.py | 6 | CODE | |
| LOW | tests/kernels/test_cpu_fused_sampling.py | 14 | CODE | |
| LOW | …rnels/attention/test_rocm_aiter_mla_op_registration.py | 46 | CODE | |
| LOW | …rnels/attention/test_rocm_aiter_mla_op_registration.py | 84 | CODE | |
| 1993 more matches not shown… | ||||
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| LOW | setup.py | 1347 | CODE | |
| LOW | setup.py | 1469 | CODE | |
| LOW | setup.py | 215 | CODE | |
| LOW | setup.py | 830 | CODE | |
| LOW | setup.py | 987 | CODE | |
| LOW | setup.py | 1084 | CODE | |
| LOW | csrc/cpu/generate_cpu_attn_dispatch.py | 100 | CODE | |
| LOW | …ibtorch_stable/quantization/marlin/generate_kernels.py | 173 | CODE | |
| LOW | …btorch_stable/moe/marlin_moe_wna16/generate_kernels.py | 173 | CODE | |
| LOW | tools/benchmark_helion_kernels.py | 532 | CODE | |
| LOW | tools/autotune_helion_kernels.py | 88 | CODE | |
| LOW | tools/autotune_helion_kernels.py | 238 | CODE | |
| LOW | tools/use_existing_torch.py | 23 | CODE | |
| LOW | tools/build_nixl_with_torch215_for_rubin.py | 80 | CODE | |
| LOW | tools/report_build_time_ninja.py | 151 | CODE | |
| LOW | tools/ci/kernel_symbol_map.py | 203 | CODE | |
| LOW | tools/ci/kernel_symbol_map.py | 232 | CODE | |
| LOW | tools/pre_commit/check_label_rules.py | 43 | CODE | |
| LOW | tools/pre_commit/check_agent_files.py | 76 | CODE | |
| LOW | tools/pre_commit/check_agent_files.py | 116 | CODE | |
| LOW | tools/pre_commit/check_agent_files.py | 143 | CODE | |
| LOW | tools/pre_commit/check_test_tethering.py | 260 | CODE | |
| LOW | tools/pre_commit/check_test_tethering.py | 301 | CODE | |
| LOW | tools/pre_commit/check_test_tethering.py | 419 | CODE | |
| LOW | tools/pre_commit/check_test_tethering.py | 448 | CODE | |
| LOW | tools/pre_commit/check_test_tethering.py | 708 | CODE | |
| LOW | tools/pre_commit/validate_config.py | 70 | CODE | |
| LOW | tools/pre_commit/check_boolean_context_manager.py | 21 | CODE | |
| LOW | tools/pre_commit/check_spdx_header.py | 60 | CODE | |
| LOW | tools/pre_commit/check_spdx_header.py | 94 | CODE | |
| LOW | tools/pre_commit/check_spdx_header.py | 139 | CODE | |
| LOW | tools/recipes/recipe_json_to_vllm_config.py | 301 | CODE | |
| LOW | tools/recipes/recipe_json_to_vllm_config.py | 614 | CODE | |
| LOW | tools/recipes/recipe_json_to_vllm_config.py | 857 | CODE | |
| LOW | tools/recipes/sweep/adaptive_concurrency.py | 119 | CODE | |
| LOW | tools/vllm-rocm/pin_rocm_dependencies.py | 39 | CODE | |
| LOW | tools/vllm-rocm/pin_rocm_dependencies.py | 93 | CODE | |
| LOW | tests/conftest.py | 412 | CODE | |
| LOW | tests/conftest.py | 577 | CODE | |
| LOW | tests/conftest.py | 1067 | CODE | |
| LOW | tests/utils.py | 111 | CODE | |
| LOW | tests/utils.py | 1197 | CODE | |
| LOW | tests/utils.py | 1589 | CODE | |
| LOW | tests/utils.py | 1753 | CODE | |
| LOW | tests/utils.py | 555 | CODE | |
| LOW | tests/utils.py | 1759 | CODE | |
| LOW | tests/v1/utils.py | 12 | CODE | |
| LOW | tests/v1/kv_offload/cpu/test_gpu_worker.py | 433 | CODE | |
| LOW | tests/v1/kv_offload/cpu/test_gpu_worker.py | 639 | CODE | |
| LOW | tests/v1/tracing/test_tracing.py | 22 | CODE | |
| LOW | tests/v1/attention/test_mm_prefix.py | 683 | CODE | |
| LOW | tests/v1/attention/test_mla_backends.py | 368 | CODE | |
| LOW | tests/v1/attention/test_mla_backends.py | 1587 | CODE | |
| LOW | tests/v1/attention/test_sparse_mla_backends.py | 892 | CODE | |
| LOW | tests/v1/attention/test_deepseek_v4_swa_visible.py | 39 | CODE | |
| LOW | tests/v1/attention/test_kpool_tail_slot_mapping.py | 356 | CODE | |
| LOW | tests/v1/logits_processors/test_correctness.py | 305 | CODE | |
| LOW | tests/v1/logits_processors/test_correctness.py | 562 | CODE | |
| LOW | tests/v1/core/test_scheduler.py | 3482 | CODE | |
| LOW | tests/v1/core/utils.py | 249 | CODE | |
| 1415 more matches not shown… | ||||
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| LOW | CMakeLists.txt | 1 | cmake_minimum_required(VERSION 3.26) | COMMENT |
| LOW | CMakeLists.txt | 61 | # | COMMENT |
| LOW | CMakeLists.txt | 281 | # setup.py will override FETCHCONTENT_BASE_DIR to play nicely with sccache. | COMMENT |
| LOW | CMakeLists.txt | 1461 | target_compile_definitions(_moe_C_stable_libtorch PRIVATE USE_CUDA) | COMMENT |
| LOW | csrc/torch_utils.h | 1 | #pragma once | COMMENT |
| LOW | csrc/torch_bindings.cpp | 1 | // Provides torch::Tensor for ops.h (previously included transitively via | COMMENT |
| LOW | csrc/cumem_allocator_compat.h | 101 | } // extern "C" | COMMENT |
| LOW | csrc/cumem_allocator.cpp | 1 | // A CUDAPluggableAllocator based on cumem* APIs. | COMMENT |
| LOW | csrc/fs_io.cpp | 1 | // SPDX-License-Identifier: Apache-2.0 | COMMENT |
| LOW | csrc/cuda_compat.h | 41 | #define VLLM_LDG(arg) __ldg(arg) | COMMENT |
| LOW | csrc/cuda_compat.h | 61 | #endif | COMMENT |
| LOW | csrc/cuda_utils.h | 1 | #pragma once | COMMENT |
| LOW | csrc/spinloop.cpp | 1 | #include <Python.h> | COMMENT |
| LOW | csrc/attention/attention_dtypes.h | 1 | #pragma once | COMMENT |
| LOW | csrc/core/scalar_type.hpp | 1 | #pragma once | COMMENT |
| LOW | csrc/core/registration.h | 1 | #pragma once | COMMENT |
| LOW | csrc/cpu/mamba_kernels.hpp | 1 | // SPDX-License-Identifier: Apache-2.0 | COMMENT |
| LOW | csrc/cpu/mamba_kernels.hpp | 21 | // state_ptr may point to a NON-CONTIGUOUS paged KV cache tensor. | COMMENT |
| LOW | csrc/cpu/mamba_kernels.hpp | 81 | } | COMMENT |
| LOW | csrc/cpu/mamba_kernels.hpp | 261 | // | COMMENT |
| LOW | csrc/cpu/cpu_attn_neon.hpp | 1 | #ifndef CPU_ATTN_NEON_HPP | COMMENT |
| LOW | csrc/cpu/utils.cpp | 1 | #ifndef VLLM_NUMA_DISABLED | COMMENT |
| LOW | csrc/cpu/cpu_wna16.cpp | 1 | #include "cpu/cpu_types.hpp" | COMMENT |
| LOW | csrc/cpu/cpu_fused_moe.cpp | 1 | #include "cpu/cpu_types.hpp" | COMMENT |
| LOW | csrc/cpu/cpu_types.hpp | 1 | #ifndef CPU_TYPES_HPP | COMMENT |
| LOW | csrc/cpu/cpu_types.hpp | 21 | #include "cpu_types_scalar.hpp" | COMMENT |
| LOW | csrc/cpu/cpu_types_riscv.hpp | 1 | #ifndef CPU_TYPES_RISCV_HPP | COMMENT |
| LOW | csrc/cpu/cpu_attn_impl.hpp | 1021 | // - max_buffer: [MaxQHeadNumPerIteration, 1], store max logits | COMMENT |
| LOW | csrc/cpu/cpu_attn_impl.hpp | 1881 | blocksize_alignment); | COMMENT |
| LOW | csrc/cpu/cpu_attn_impl.hpp | 1901 | float* curr_partial_q_buffer = | COMMENT |
| LOW | csrc/cpu/mamba_cpu.cpp | 1 | // SPDX-License-Identifier: Apache-2.0 | COMMENT |
| LOW | csrc/cpu/mamba_cpu.cpp | 121 | if (z.has_value() && z.value().defined()) z_in = ensure_input(z.value()); | COMMENT |
| LOW | csrc/cpu/cpu_attn_vxe.hpp | 381 | } // namespace cpu_attention | COMMENT |
| LOW | csrc/cpu/generate_cpu_attn_dispatch.py | 141 | """Generate the complete header file content.""" | COMMENT |
| LOW | csrc/cpu/generate_cpu_attn_dispatch.py | 161 | #endif | COMMENT |
| LOW | csrc/cpu/cpu_arch_macros.h | 61 | #endif | COMMENT |
| LOW | csrc/cpu/cpu_arch_macros.h | 161 | #include <riscv_vector.h> | COMMENT |
| LOW | csrc/cpu/cpu_attn_amx_fp8.hpp | 1 | // SPDX-License-Identifier: Apache-2.0 | COMMENT |
| LOW | csrc/cpu/cpu_attn_amx_fp8.hpp | 21 | // Scale bookkeeping for PV: | COMMENT |
| LOW | csrc/cpu/cpu_attn_amx_fp8.hpp | 41 | // --------------------------------------------------------------------------- | COMMENT |
| LOW | csrc/cpu/cpu_attn_rvv.hpp | 1 | // SPDX-License-Identifier: Apache-2.0 | COMMENT |
| LOW | csrc/cpu/cpu_attn_fp8.hpp | 1 | // SPDX-License-Identifier: Apache-2.0 | COMMENT |
| LOW | csrc/cpu/cpu_types_vxe.hpp | 1 | COMMENT | |
| LOW | csrc/cpu/cpu_types_vxe.hpp | 21 | #define vec_mul(a, b) ((a) * (b)) | COMMENT |
| LOW | csrc/cpu/cpu_types_x86.hpp | 1 | COMMENT | |
| LOW | csrc/cpu/shm.cpp | 1 | #include "cpu/cpu_types.hpp" | COMMENT |
| LOW | csrc/cpu/cpu_types_riscv_impl.hpp | 1 | #ifndef CPU_TYPES_RISCV_IMPL_HPP | COMMENT |
| LOW | csrc/cpu/cpu_fused_moe_activations.hpp | 1 | // SPDX-License-Identifier: Apache-2.0 | COMMENT |
| LOW | csrc/cpu/cpu_types_riscv_defs.hpp | 1 | #ifndef CPU_TYPES_RISCV_DEFS_HPP | COMMENT |
| LOW | csrc/cpu/cpu_types_riscv_defs.hpp | 21 | #define BOOL_256 b16 | COMMENT |
| LOW | csrc/cpu/cpu_fused_moe_int8.cpp | 1 | // SPDX-License-Identifier: Apache-2.0 | COMMENT |
| LOW | csrc/cpu/cpu_attn_vsx.hpp | 1 | // SPDX-License-Identifier: Apache-2.0 | COMMENT |
| LOW | csrc/cpu/cpu_attn_vsx.hpp | 361 | COMMENT | |
| LOW | csrc/cpu/sgl-kernels/paged_mqa_logits.cpp | 1 | // Adapted from | COMMENT |
| LOW | csrc/cpu/sgl-kernels/paged_mqa_logits.cpp | 21 | // server.log`), so the upstream `kvcache_fp8.size(1) == 64` shape | COMMENT |
| LOW | csrc/cpu/sgl-kernels/gemm.cpp | 81 | constexpr int BLOCK_N = block_size_n(); | COMMENT |
| LOW | csrc/cpu/sgl-kernels/indexer.cpp | 1 | // vLLM-native: DeepSeek-V4 sparse indexer Q-side RoPE + FP8 quant kernel. | COMMENT |
| LOW | csrc/cpu/sgl-kernels/indexer.cpp | 21 | // (x_even, x_odd) pairs via in-register permutes (`_mm512_permute_ps` / | COMMENT |
| LOW | csrc/cpu/sgl-kernels/indexer.cpp | 101 | amax = _mm512_reduce_max_ps(vamax); | COMMENT |
| LOW | csrc/cpu/sgl-kernels/fla.cpp | 1 | // Adapted from | COMMENT |
| 1279 more matches not shown… | ||||
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| MEDIUM | tools/install_deepgemm.sh | 86 | # Create a temporary directory for the build | COMMENT |
| MEDIUM | tools/report_build_time_ninja.py | 200 | # Create a list that is in order by time stamp and has entries for the | COMMENT |
| MEDIUM | tools/pre_commit/update-dockerfile-graph.sh | 26 | # Define the target file path | COMMENT |
| MEDIUM | tests/conftest.py | 603 | # Create a copy to avoid modifying the original dict | COMMENT |
| MEDIUM⚡ | tests/test_uvicorn_logging.py | 257 | # Create a logger with our filter (simulating uvicorn.access) | COMMENT |
| MEDIUM⚡ | tests/test_uvicorn_logging.py | 264 | # Create a custom handler that tracks messages | COMMENT |
| MEDIUM | tests/utils.py | 770 | # Create a dedicated process group so we can kill | COMMENT |
| MEDIUM | tests/utils.py | 1762 | # Create a unique temporary file to store exception info from child | COMMENT |
| MEDIUM | tests/test_config.py | 2776 | # Create a new mock and run the method with the same S3 URL | COMMENT |
| MEDIUM | tests/v1/test_tensor_ipc_queue.py | 262 | # Create a CPU tensor | COMMENT |
| MEDIUM | tests/v1/test_tensor_ipc_queue.py | 589 | # Create a CPU tensor | COMMENT |
| MEDIUM | tests/v1/test_tensor_ipc_queue.py | 738 | # Create a CPU tensor | COMMENT |
| MEDIUM | tests/v1/test_tensor_ipc_queue.py | 1000 | # Create a tensor queue | COMMENT |
| MEDIUM | tests/v1/test_serial_utils.py | 188 | # Create a sample Python object | COMMENT |
| MEDIUM | tests/v1/test_serial_utils.py | 206 | # Create a sample tensor | COMMENT |
| MEDIUM | tests/v1/test_serial_utils.py | 226 | # Create a sample numpy array | COMMENT |
| MEDIUM | tests/v1/test_serial_utils.py | 311 | # Create a request with a non-multimodal tensor | COMMENT |
| MEDIUM | tests/v1/test_serial_utils.py | 352 | # Create a request with None for the tensor field | COMMENT |
| MEDIUM | tests/v1/metrics/test_ray_metrics.py | 74 | # Create the actor and call the async method | COMMENT |
| MEDIUM | tests/v1/attention/test_mla_backends.py | 626 | # Create a realistic slot mapping that corresponds to the block table | COMMENT |
| MEDIUM | tests/v1/attention/test_mla_backends.py | 2041 | # Create a summary for the single-line failure message | COMMENT |
| MEDIUM | tests/v1/attention/test_attention_backends.py | 236 | # Create a realistic slot mapping that corresponds to the block table | COMMENT |
| MEDIUM | tests/v1/logits_processors/test_correctness.py | 914 | # Define a shuffled batch of requests which individually use a different | COMMENT |
| MEDIUM | tests/v1/logits_processors/test_custom_offline.py | 30 | # Create a mixture of requests which do and don't utilize the dummy logitproc | COMMENT |
| MEDIUM | tests/v1/logits_processors/test_custom_offline.py | 91 | # Create a vLLM instance and load custom logitproc | COMMENT |
| MEDIUM | tests/v1/logits_processors/test_custom_offline.py | 98 | # Create a reference vLLM instance without custom logitproc | COMMENT |
| MEDIUM⚡ | tests/v1/core/test_kv_cache_utils.py | 763 | # Create a list of KVCacheBlock objects | COMMENT |
| MEDIUM⚡ | tests/v1/core/test_kv_cache_utils.py | 766 | # Create a FreeKVCacheBlockQueue with these blocks | COMMENT |
| MEDIUM⚡ | tests/v1/core/test_kv_cache_utils.py | 986 | # Create a list of KVCacheBlock objects | COMMENT |
| MEDIUM⚡ | tests/v1/core/test_kv_cache_utils.py | 989 | # Create a FreeKVCacheBlockQueue with these blocks | COMMENT |
| MEDIUM | tests/v1/core/test_kv_cache_utils.py | 809 | # Create an empty FreeKVCacheBlockQueue with these blocks | COMMENT |
| MEDIUM | tests/v1/core/test_kv_cache_utils.py | 857 | # Create an empty FreeKVCacheBlockQueue | COMMENT |
| MEDIUM | tests/v1/core/test_kv_cache_utils.py | 911 | # Create an empty FreeKVCacheBlockQueue with these blocks | COMMENT |
| MEDIUM | tests/v1/core/test_kv_cache_utils.py | 2180 | # Create a VllmConfig | STRING |
| MEDIUM | tests/v1/core/test_kv_cache_utils.py | 2216 | # Create a VllmConfig | STRING |
| MEDIUM⚡ | tests/v1/core/test_scheduler.py | 4255 | # Create a request and schedule it | COMMENT |
| MEDIUM | tests/v1/core/test_scheduler.py | 4282 | # Create a high priority request and schedule it | COMMENT |
| MEDIUM | tests/v1/core/test_scheduler.py | 5133 | # Create a request and schedule it (and to be preempted) | COMMENT |
| MEDIUM | tests/v1/core/test_scheduler.py | 5183 | # Create a high priority request and schedule it | COMMENT |
| MEDIUM | tests/v1/core/test_scheduler.py | 6163 | # Create a text-only request (no mm_features). | COMMENT |
| MEDIUM | tests/v1/cudagraph/test_cudagraph_dispatch.py | 56 | # Create a real LoRAConfig with specialize_active_lora enabled | COMMENT |
| MEDIUM | tests/v1/kv_connector/unit/test_nixl_connector.py | 3272 | # Create a request that triggers do_remote_decode so that | COMMENT |
| MEDIUM⚡ | tests/v1/kv_connector/unit/test_lmcache_connector.py | 218 | # Create a mock object that is not LMCacheKVEvents | COMMENT |
| MEDIUM | tests/v1/kv_connector/unit/test_moriio_connector.py | 286 | # Define a fake remote engine id for testing | COMMENT |
| MEDIUM | …ts/v1/kv_connector/unit/test_decode_bench_connector.py | 168 | # Create a request with multiple blocks worth of tokens | COMMENT |
| MEDIUM | …ts/v1/kv_connector/unit/test_decode_bench_connector.py | 567 | # Create a request | COMMENT |
| MEDIUM | …ts/v1/kv_connector/unit/test_decode_bench_connector.py | 589 | # Create a request with just 1 token | COMMENT |
| MEDIUM | …ts/v1/kv_connector/unit/test_decode_bench_connector.py | 607 | # Create a request with 2 tokens | COMMENT |
| MEDIUM | …ts/v1/kv_connector/unit/test_decode_bench_connector.py | 633 | # Create a request with many blocks | COMMENT |
| MEDIUM | …ts/v1/kv_connector/unit/test_decode_bench_connector.py | 716 | # Create a request that doesn't align to block boundaries | COMMENT |
| MEDIUM | tests/v1/kv_connector/unit/test_nixl_connector_hma.py | 562 | # Create a mock worker with just the required attributes | COMMENT |
| MEDIUM | tests/v1/kv_connector/unit/test_example_connector.py | 142 | # Create the LLM instance | COMMENT |
| MEDIUM | …r/extract_hidden_states_integration/test_extraction.py | 45 | # Create a minimal Llama config with small dimensions | COMMENT |
| MEDIUM | …r/extract_hidden_states_integration/test_extraction.py | 60 | # Create a simple tokenizer | COMMENT |
| MEDIUM | tests/v1/determinism/test_batch_invariance.py | 119 | # Create a batch of size `max_batch_size` and insert the needle at | COMMENT |
| MEDIUM | tests/v1/distributed/test_external_lb_dp.py | 162 | # Create a client for each server | COMMENT |
| MEDIUM | tests/v1/distributed/test_hybrid_lb_dp.py | 182 | # Create a client for each node (each node has its own API endpoint) | COMMENT |
| MEDIUM | tests/v1/streaming_input/test_async_llm_streaming.py | 20 | # Create a minimal mock without initializing the full engine | COMMENT |
| MEDIUM | tests/v1/streaming_input/test_async_llm_streaming.py | 50 | # Create a mock queue with outputs | COMMENT |
| MEDIUM | …/v1/streaming_input/test_gpu_model_runner_streaming.py | 33 | # Create a real InputBatch for e2e testing | COMMENT |
| 276 more matches not shown… | ||||
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| MEDIUM | setup.py | 152 | def find_tcmalloc() -> Path | None: | CODE |
| MEDIUM | setup.py | 1211 | def get_base_commit_in_main_branch() -> str: | CODE |
| LOW⚡ | setup.py | 660 | except Exception: | CODE |
| LOW⚡ | setup.py | 678 | except Exception as e: | CODE |
| LOW | setup.py | 160 | except Exception: | CODE |
| LOW | setup.py | 249 | except Exception as e: | CODE |
| LOW | setup.py | 622 | except Exception: | CODE |
| LOW | setup.py | 633 | except Exception: | CODE |
| LOW | setup.py | 744 | except Exception: | CODE |
| LOW | setup.py | 861 | except Exception as e: | CODE |
| LOW | setup.py | 1037 | except Exception as e: | CODE |
| LOW | setup.py | 1257 | except Exception as err: | CODE |
| LOW | setup.py | 1328 | except Exception: | CODE |
| LOW | setup.py | 1429 | except Exception: | CODE |
| MEDIUM | tools/benchmark_helion_kernels.py | 71 | print(f"Error importing vLLM: {e}") | CODE |
| LOW | tools/benchmark_helion_kernels.py | 285 | except Exception as e: | CODE |
| LOW | tools/benchmark_helion_kernels.py | 406 | except Exception as e: # noqa: BLE001 - any failure is recorded, not fatal | CODE |
| MEDIUM | tools/benchmark_helion_kernels.py | 278 | def cleanup_gpu_resources() -> None: | CODE |
| MEDIUM | tools/autotune_helion_kernels.py | 44 | print(f"Error importing vLLM: {e}") | CODE |
| MEDIUM | tools/generate_cmake_presets.py | 167 | print(f"Error writing file: {e}") | CODE |
| LOW | tools/install_nixl_from_source_ubuntu.py | 30 | except Exception: | CODE |
| LOW | tools/ci/kernel_symbol_map.py | 368 | except Exception as e: # never fail the image build | CODE |
| LOW | tools/recipes/recipe_json_to_vllm_config.py | 1093 | except Exception as exc: | CODE |
| LOW | tools/recipes/sweep/runtime_tuning.py | 171 | except Exception as exc: | CODE |
| LOW | tools/recipes/sweep/sweep_generation.py | 280 | except Exception: | CODE |
| LOW | tools/recipes/sweep/sweep_generation.py | 926 | except Exception as exc: | CODE |
| LOW | tools/profiler/nsys_profile_tools/gputrc2graph.py | 234 | except Exception: | CODE |
| LOW | tools/vllm-rocm/pin_rocm_dependencies.py | 78 | except Exception as e: | CODE |
| LOW | rust/src/chat/tests/grammar_replay.py | 102 | except Exception as error: | CODE |
| LOW | tests/conftest.py | 1399 | except Exception: | CODE |
| LOW | tests/conftest.py | 1668 | except Exception as e: | CODE |
| MEDIUM | tests/utils.py | 76 | def _nvml(): | CODE |
| LOW | tests/utils.py | 294 | except Exception: | CODE |
| LOW | tests/utils.py | 586 | except Exception as e: | CODE |
| LOW | tests/utils.py | 675 | except Exception: | CODE |
| LOW | tests/utils.py | 1791 | except Exception as e: | CODE |
| LOW | tests/utils.py | 1814 | except Exception: | CODE |
| LOW | tests/utils.py | 2012 | except Exception as e: | CODE |
| LOW | tests/utils.py | 2143 | except Exception: | CODE |
| LOW | tests/v1/test_tensor_ipc_queue.py | 133 | except Exception as e: | CODE |
| LOW | tests/v1/test_tensor_ipc_queue.py | 176 | except Exception as e: | CODE |
| LOW | tests/v1/test_tensor_ipc_queue.py | 377 | except Exception as e: | CODE |
| LOW | tests/v1/test_tensor_ipc_queue.py | 470 | except Exception as e: | CODE |
| LOW | tests/v1/test_tensor_ipc_queue.py | 510 | except Exception as e: | CODE |
| LOW | tests/v1/test_tensor_ipc_queue.py | 611 | except Exception as e: | CODE |
| LOW | tests/v1/test_tensor_ipc_queue.py | 655 | except Exception as e: | CODE |
| LOW | tests/v1/test_tensor_ipc_queue.py | 817 | except Exception as e: | CODE |
| LOW | tests/v1/utils.py | 65 | except Exception as e: | CODE |
| LOW⚡ | tests/v1/kv_offload/cpu/test_shared_offload_region.py | 237 | except Exception as e: | CODE |
| LOW | tests/v1/kv_offload/cpu/test_shared_offload_region.py | 147 | except Exception as e: | CODE |
| LOW | tests/v1/kv_offload/cpu/test_shared_offload_region.py | 186 | except Exception as e: | CODE |
| LOW | tests/v1/kv_offload/tiering/p2p/p2p_connector_proxy.py | 300 | except Exception as e: | CODE |
| LOW | tests/v1/kv_offload/tiering/p2p/p2p_connector_proxy.py | 346 | except Exception as exc: | CODE |
| LOW | tests/v1/kv_offload/tiering/p2p/p2p_connector_proxy.py | 359 | except Exception as exc: | CODE |
| LOW | tests/v1/kv_offload/tiering/p2p/p2p_connector_proxy.py | 379 | except Exception as e: | CODE |
| MEDIUM | tests/v1/kv_offload/tiering/p2p/p2p_connector_proxy.py | 340 | def _run_decode(): | CODE |
| LOW | tests/v1/shutdown/test_forward_error.py | 82 | except Exception as e: | CODE |
| LOW | tests/v1/shutdown/test_processor_error.py | 44 | except Exception as e: | CODE |
| LOW⚡ | tests/v1/kv_connector/unit/test_hf3fs_client.py | 27 | except Exception: | CODE |
| LOW | …ts/v1/kv_connector/unit/test_nixl_rocm_gpu_mem_diag.py | 39 | except Exception: | CODE |
| 877 more matches not shown… | ||||
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| HIGH | setup.py | 622 | except Exception: | CODE |
| HIGH | setup.py | 633 | except Exception: | CODE |
| HIGH | setup.py | 660 | except Exception: | CODE |
| HIGH | tools/build_nixl_with_torch215_for_rubin.py | 89 | except OSError: | CODE |
| HIGH | tests/utils.py | 420 | except subprocess.TimeoutExpired: | CODE |
| HIGH | tests/utils.py | 525 | except (ProcessLookupError, OSError): | CODE |
| HIGH | …1/kv_connector/unit/test_invalid_blocks_correctness.py | 270 | except KeyError: | CODE |
| HIGH | …_connector/nixl_integration/nixl_side_channel_probe.py | 28 | except ValueError: | CODE |
| HIGH | …worker/test_gpu_model_runner_v2_cudagraph_profiling.py | 195 | except RuntimeError: | CODE |
| HIGH | tests/v1/engine/test_abort_final_step.py | 78 | except Exception: | CODE |
| HIGH | tests/kernels/moe/modular_kernel_tools/common.py | 332 | except NotImplementedError: | CODE |
| HIGH | tests/kernels/moe/modular_kernel_tools/common.py | 373 | except NotImplementedError: | CODE |
| HIGH | tests/kernels/moe/modular_kernel_tools/common.py | 214 | except NotImplementedError: | CODE |
| HIGH | tests/distributed/test_mnnvl_alltoall.py | 42 | except Exception: | CODE |
| HIGH | tests/distributed/test_multiproc_executor.py | 493 | except Exception: | CODE |
| HIGH | tests/models/language/generation/test_hybrid.py | 87 | except ValueError: | CODE |
| HIGH | tests/models/language/generation/test_hybrid.py | 130 | except ValueError: | CODE |
| HIGH | tests/models/language/generation/test_hybrid.py | 337 | except ValueError: | CODE |
| HIGH | tests/models/language/generation/test_hybrid.py | 383 | except ValueError: | CODE |
| HIGH | tests/models/language/generation/test_hybrid.py | 486 | except ValueError: | CODE |
| HIGH | tests/models/language/generation/test_hybrid.py | 550 | except ValueError: | CODE |
| HIGH | tests/models/language/generation/test_hybrid.py | 631 | except ValueError: | CODE |
| HIGH | tests/models/language/generation/test_hybrid.py | 700 | except ValueError: | CODE |
| HIGH | tests/models/language/generation/test_hybrid.py | 785 | except ValueError: | CODE |
| HIGH | tests/vllm_test_utils/vllm_test_utils/monitor.py | 65 | except NameError: | CODE |
| HIGH | tests/vllm_test_utils/vllm_test_utils/blame.py | 46 | except NameError: | CODE |
| HIGH | tests/compile/fusions_e2e/common.py | 65 | except ImportError: | CODE |
| HIGH | tests/compile/fusions_e2e/common.py | 76 | except ImportError: | CODE |
| HIGH | …transcription/test_transcription_validation_whisper.py | 48 | except Exception: | CODE |
| HIGH | tests/entrypoints/launchers/test_shutdown.py | 288 | except TimeoutError: | CODE |
| HIGH | tests/entrypoints/launchers/test_shutdown.py | 466 | except TimeoutError: | CODE |
| HIGH | tests/entrypoints/launchers/test_shutdown.py | 595 | except TimeoutError: | CODE |
| HIGH | tests/entrypoints/launchers/test_shutdown.py | 58 | except psutil.NoSuchProcess: | CODE |
| HIGH | …ts/serve/sagemaker/test_sagemaker_handler_overrides.py | 112 | except ImportError: | CODE |
| HIGH | tests/lora/test_layers.py | 140 | except Exception: | CODE |
| HIGH | …sts/basic_correctness/models/test_basic_correctness.py | 63 | except Exception: | CODE |
| HIGH | vllm/env_override.py | 913 | except ImportError: | CODE |
| HIGH | vllm/env_override.py | 637 | except Exception: | CODE |
| HIGH | vllm/logger.py | 399 | except NameError: | CODE |
| HIGH | vllm/v1/attention/backends/turboquant_attn.py | 494 | except Exception: # noqa: BLE001 | CODE |
| HIGH | vllm/v1/attention/backends/cpu_attn.py | 512 | except Exception: | CODE |
| HIGH | vllm/v1/attention/backends/fa_utils.py | 341 | except (ImportError, AssertionError): | CODE |
| HIGH | vllm/v1/attention/ops/flydsl_ultraquant_decode.py | 186 | except Exception: # noqa: BLE001 | CODE |
| HIGH | vllm/v1/attention/ops/flydsl_turboquant_decode.py | 165 | except Exception: # noqa: BLE001 | CODE |
| HIGH | vllm/v1/spec_decode/vocab_mapping.py | 29 | except Exception: | CODE |
| HIGH | vllm/v1/sample/ops/topk_topp_sampler.py | 71 | except Exception: | CODE |
| HIGH | vllm/v1/engine/async_llm.py | 215 | except RuntimeError: | CODE |
| HIGH | vllm/v1/engine/core_client.py | 1147 | except RuntimeError: | CODE |
| HIGH | vllm/v1/engine/core_client.py | 1462 | except RuntimeError: | CODE |
| HIGH | vllm/v1/pool/flash_maxsim/_common.py | 47 | except Exception: | CODE |
| HIGH | vllm/tool_parsers/hy_v4_tool_parser.py | 286 | except Exception: | CODE |
| HIGH | vllm/tool_parsers/hy_v4_tool_parser.py | 290 | except Exception: | CODE |
| HIGH | vllm/tool_parsers/hy_v4_tool_parser.py | 331 | except (json.JSONDecodeError, ValueError): | CODE |
| HIGH | vllm/tool_parsers/hunyuan_a13b_tool_parser.py | 221 | except json.JSONDecodeError: | CODE |
| HIGH | vllm/tool_parsers/apertus_tool_parser.py | 353 | except Exception: | CODE |
| HIGH | vllm/tool_parsers/hy_v3_tool_parser.py | 198 | except Exception: | CODE |
| HIGH | vllm/tool_parsers/hy_v3_tool_parser.py | 202 | except Exception: | CODE |
| HIGH | vllm/tool_parsers/hy_v3_tool_parser.py | 245 | except (json.JSONDecodeError, ValueError): | CODE |
| HIGH | vllm/tool_parsers/poolside_v1_tool_parser.py | 108 | except json.JSONDecodeError: | CODE |
| HIGH | vllm/tool_parsers/poolside_v1_tool_parser.py | 113 | except (ValueError, SyntaxError): | CODE |
| 70 more matches not shown… | ||||
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| LOW | setup.py | 47 | logger = logging.getLogger(__name__) | CODE |
| LOW | tools/profiler/nsys_profile_tools/gputrc2graph.py | 14 | logger = logging.getLogger(__name__) | CODE |
| LOW⚡ | tests/test_request_input_bounds.py | 327 | def _set_max_n(monkeypatch: pytest.MonkeyPatch, value: int) -> None: | STRING |
| LOW | tests/v1/kv_offload/tiering/p2p/p2p_connector_proxy.py | 31 | logger = logging.getLogger(__name__) | CODE |
| LOW⚡ | tests/v1/kv_connector/unit/test_multi_connector.py | 134 | def update_state_after_alloc(self, request, blocks, num_tokens) -> None: | CODE |
| LOW | tests/v1/kv_connector/unit/test_multi_connector.py | 99 | def update_state_after_alloc(self, request, blocks, num_tokens) -> None: | CODE |
| LOW | …/v1/kv_connector/unit/offloading_connector/conftest.py | 7 | __all__ = ["request_runner"] | CODE |
| LOW | …s/v1/kv_connector/nixl_integration/toy_proxy_server.py | 15 | logger = logging.getLogger(__name__) | CODE |
| LOW | tests/v1/worker/test_gpu_model_runner_v2_eplb.py | 48 | def update_mapping(self, model_config: Any, mapping: torch.Tensor) -> None: | CODE |
| LOW | tests/v1/worker/test_gpu_worker_weight_transfer.py | 39 | def update_weights(self, update_info: dict) -> None: | CODE |
| LOW | tests/v1/worker/test_gpu_trace_replay.py | 63 | def _set_lens(state: TraceReplayState, total_len, prompt_len) -> None: | CODE |
| LOW | tests/v1/engine/test_core_engine_actor_manager.py | 48 | def _set_visible_devices(self, vllm_config: Any, local_dp_rank: int) -> None: | CODE |
| LOW | tests/kernels/moe/test_marlin_vs_trtllm_mxint4.py | 79 | __all__ = [ | CODE |
| LOW | tests/distributed/test_weight_transfer.py | 1231 | def update_weights(self, update_info: dict) -> None: | CODE |
| LOW⚡ | tests/distributed/test_eplb_migration_scheduler.py | 35 | def set_transfer_context(self, old_indices: np.ndarray, layer_idx: int) -> None: | CODE |
| LOW⚡ | tests/distributed/eplb_utils.py | 74 | def set_env_vars_and_device(env: dict[str, str]) -> None: | CODE |
| LOW | tests/models/language/generation/test_hybrid.py | 62 | def _set_conv_state_layout(monkeypatch, layout: str) -> None: | CODE |
| LOW | tests/models/qwen4_exp/test_ple.py | 211 | def _set_test_embedding_weight_loader(embedding) -> None: | CODE |
| LOW | tests/vllm_test_utils/vllm_test_utils/__init__.py | 10 | __all__ = ["blame", "BlameResult", "monitor", "MonitoredValues"] | CODE |
| LOW | tests/parser/engine/replay_harness.py | 83 | def set_vocab(self, vocab: dict[str, int]) -> None: | CODE |
| LOW | tests/quantization/test_gfx950_moe.py | 39 | def set_rocm_aiter(monkeypatch: pytest.MonkeyPatch, enabled: bool) -> None: | CODE |
| LOW | tests/entrypoints/launchers/test_dp_supervisor.py | 644 | async def _set_healthy(port: int, use_ssl: bool = False) -> None: | CODE |
| LOW | tests/entrypoints/launchers/test_dp_supervisor.py | 654 | async def _set_unhealthy(port: int, use_ssl: bool = False) -> None: | CODE |
| LOW | tests/entrypoints/openai/responses/conftest.py | 15 | logger = logging.getLogger(__name__) | CODE |
| LOW | …/entrypoints/openai/responses/test_parsable_context.py | 21 | logger = logging.getLogger(__name__) | CODE |
| LOW | tests/entrypoints/openai/responses/test_harmony.py | 30 | logger = logging.getLogger(__name__) | CODE |
| LOW⚡ | tests/rocm/test_moe_padding.py | 82 | def _set_padding_env(monkeypatch: pytest.MonkeyPatch, padding: bool) -> None: | CODE |
| LOW⚡ | tests/rocm/test_moe_padding.py | 87 | def _set_padding_env_with_aiter(monkeypatch: pytest.MonkeyPatch, padding: bool) -> None: | CODE |
| LOW | tests/rocm/test_moe_weight_replay.py | 64 | def _set_backend_env(monkeypatch: pytest.MonkeyPatch, backend: str) -> None: | CODE |
| LOW | vllm/envs.py | 575 | logger = logging.getLogger(__name__) | CODE |
| LOW | vllm/env_override.py | 503 | def _update_scheduler_patched(self) -> None: | CODE |
| LOW | vllm/sampling_params.py | 802 | def update_from_tokenizer(self, tokenizer: TokenizerLike) -> None: | CODE |
| LOW | vllm/__init__.py | 76 | __all__ = [ | CODE |
| LOW | vllm/logger.py | 54 | def set_vllm_process_name(process_name: str, *, skip_if_set: bool = False) -> None: | CODE |
| LOW | vllm/v1/request.py | 285 | def update_block_hashes(self) -> None: | CODE |
| LOW | vllm/v1/kv_offload/base.py | 107 | def set_state(self, val: Any) -> None: | CODE |
| LOW | vllm/v1/kv_offload/base.py | 113 | def set_offload_key_position(self, key: OffloadKey, end_token: int) -> None: | CODE |
| LOW | vllm/v1/kv_offload/cpu/policies/lru.py | 60 | def _update_recency(self, keys: Iterable[OffloadKey]) -> None: | CODE |
| LOW | vllm/v1/kv_offload/tiering/example/manager.py | 30 | logger = logging.getLogger(__name__) | CODE |
| LOW | vllm/v1/kv_offload/tiering/p2p/data/__init__.py | 6 | __all__ = [ | CODE |
| LOW | vllm/v1/kv_offload/tiering/p2p/control/__init__.py | 12 | __all__ = [ | CODE |
| LOW | vllm/v1/kv_offload/tiering/p2p/session/__init__.py | 11 | __all__ = [ | CODE |
| LOW | vllm/v1/kv_offload/tiering/fs/io.py | 23 | logger = logging.getLogger(__name__) | CODE |
| LOW | vllm/v1/metrics/stats.py | 447 | def update_from_output(self, prefill_stats: PrefillStats) -> None: | CODE |
| LOW | vllm/v1/attention/backend.py | 751 | def update_draft_decode_metadata(self, metadata: M) -> None: | CODE |
| LOW | vllm/v1/attention/backends/flash_attn.py | 1077 | def update_draft_decode_metadata(self, metadata: FlashAttentionMetadata) -> None: | CODE |
| LOW | vllm/v1/attention/backends/triton_attn.py | 306 | def update_draft_decode_metadata(self, _metadata: TritonAttentionMetadata) -> None: | CODE |
| LOW | vllm/v1/attention/backends/zentorch_sdpa.py | 17 | __all__ = ["should_use_zentorch_sdpa", "zentorch_sdpa_attn"] | CODE |
| LOW | vllm/v1/attention/backends/flash_attn_diffkv.py | 40 | def set_head_size_v(cls, head_size_v: int) -> None: | CODE |
| LOW | vllm/v1/attention/backends/triton_attn_diffkv.py | 81 | def set_head_size_v(cls, head_size_v: int) -> None: | CODE |
| LOW | vllm/v1/attention/backends/flashinfer.py | 1925 | def update_draft_decode_metadata(self, _metadata: FlashInferMetadata) -> None: | CODE |
| LOW | vllm/v1/attention/backends/cpu_attn.py | 173 | def _set_isa_to_layers(self, layer_names: list[str]) -> None: | CODE |
| LOW | vllm/v1/attention/backends/mla/flashattn_mla_sparse.py | 141 | def update_draft_decode_metadata(self, metadata: SparseMLACommonMetadata) -> None: | CODE |
| LOW | vllm/v1/attention/backends/mla/index_group.py | 59 | def set_logical_topk_ready(self, layer_index: int) -> None: | CODE |
| LOW⚡ | vllm/v1/attention/backends/mla/triton_mla.py | 64 | def update_draft_decode_metadata(self, _metadata: MLACommonMetadata) -> None: | CODE |
| LOW | vllm/v1/attention/backends/mla/prefill/__init__.py | 10 | __all__ = [ | CODE |
| LOW | vllm/v1/attention/ops/triton_decode_attention.py | 41 | logger = logging.getLogger(__name__) | CODE |
| LOW | vllm/v1/attention/ops/ultraquant/__init__.py | 11 | __all__ = ["ultraquant_store"] | CODE |
| LOW | vllm/v1/core/sched/async_scheduler.py | 25 | def _update_after_schedule(self, scheduler_output: SchedulerOutput) -> None: | CODE |
| LOW | vllm/v1/core/sched/interface.py | 114 | def update_draft_token_ids(self, draft_token_ids: "DraftTokenIds") -> None: | CODE |
| 482 more matches not shown… | ||||
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| HIGH | setup.py | 1052 | "build_tag": null, | STRING |
| HIGH | setup.py | 1056 | "variant": null, | STRING |
| HIGH | csrc/cpu/generate_cpu_attn_dispatch.py | 170 | (__riscv_v_min_vlen == 128 || __riscv_v_min_vlen == 256) | CODE |
| HIGH | csrc/cpu/generate_cpu_attn_dispatch.py | 247 | "&& (__riscv_v_min_vlen == 128 || __riscv_v_min_vlen == 256)", | CODE |
| HIGH⚡ | csrc/libtorch_stable/quantization/machete/generate.py | 512 | "M > 256 && K <= 16384 && N <= 4096": ((128, 128), (2, 1, 1)), | STRING |
| HIGH⚡ | csrc/libtorch_stable/quantization/machete/generate.py | 515 | "M > 128 && K <= 4096 && N <= 4096": ((128, 64), (2, 1, 1)), | STRING |
| HIGH⚡ | csrc/libtorch_stable/quantization/machete/generate.py | 516 | "M > 128 && K <= 8192 && N <= 8192": ((128, 128), (2, 1, 1)), | STRING |
| HIGH⚡ | csrc/libtorch_stable/quantization/machete/generate.py | 519 | "M > 64 && K <= 4069 && N <= 4069": ((128, 32), (2, 1, 1)), | STRING |
| HIGH⚡ | csrc/libtorch_stable/quantization/machete/generate.py | 520 | "M > 64 && K <= 4069 && N <= 8192": ((128, 64), (2, 1, 1)), | STRING |
| HIGH⚡ | csrc/libtorch_stable/quantization/machete/generate.py | 521 | "M > 64 && K >= 8192 && N >= 12288": ((256, 128), (2, 1, 1)), | STRING |
| HIGH⚡ | csrc/libtorch_stable/quantization/machete/generate.py | 524 | "M > 32 && K <= 6144 && N <= 6144": ((128, 16), (1, 1, 1)), | STRING |
| HIGH⚡ | csrc/libtorch_stable/quantization/machete/generate.py | 525 | "M > 32 && K >= 16384 && N >= 12288": ((256, 64), (2, 1, 1)), | STRING |
| HIGH⚡ | csrc/libtorch_stable/quantization/machete/generate.py | 528 | "M > 16 && K <= 12288 && N <= 8192": ((128, 32), (2, 1, 1)), | STRING |
| HIGH | tools/pre_commit/check_test_tethering.py | 483 | ``"pytest a.py && torchrun b.py"`` passed as an argument to a runner | STRING |
| HIGH | tests/v1/attention/test_mm_prefix.py | 425 | # null, so is_mm_prefix_lm is False out of the box. Tests that exercise the | COMMENT |
| HIGH | tests/v1/attention/test_mla_backends.py | 603 | # Permute the context blocks (excluding block 0 which is null) | COMMENT |
| HIGH | tests/v1/attention/test_attention_backends.py | 213 | # Permute the context blocks (excluding block 0 which is null) | COMMENT |
| HIGH | tests/v1/core/test_prefix_replay.py | 147 | # The window group never hits: blocks below the window are null, the | COMMENT |
| HIGH⚡ | tests/v1/core/test_scheduler.py | 4246 | num_blocks=5, # Can hold 64 tokens (first block is null) | CODE |
| HIGH | tests/v1/core/test_scheduler.py | 5121 | num_blocks=15, # can hold 244 tokens with 14 blocks (first block is null) | CODE |
| HIGH | tests/v1/core/test_scheduler.py | 5361 | num_blocks=11, # Can hold 160 tokens (first block is null) | CODE |
| HIGH⚡ | tests/v1/kv_connector/unit/test_moriio_hma_scheduler.py | 599 | # Mamba group: [null, running state, scratch, scratch]. | COMMENT |
| HIGH | …_connector/unit/offloading_connector/test_scheduler.py | 4349 | # 4 GPU blocks: block 0 is null, blocks 1-3 are usable. | COMMENT |
| HIGH⚡ | …_connector/unit/offloading_connector/test_scheduler.py | 532 | are similar -- neither is required to retain the whole prefix, so a non-null, | STRING |
| HIGH | tests/tool_parsers/test_granite_tool_parser.py | 44 | "null_field": null, | CODE |
| HIGH⚡ | tests/tool_parsers/test_structural_tag_registry.py | 184 | '{"nested":[true,null,1.5]}</｜DSML｜ parameter>\n' | CODE |
| HIGH⚡ | tests/tool_parsers/test_structural_tag_registry.py | 190 | grammar, output.replace('{"nested":[true,null,1.5]}', "invalid") | CODE |
| HIGH | tests/tool_parsers/test_granite_20b_fc_tool_parser.py | 36 | "null_field": null, | CODE |
| HIGH⚡ | tests/tool_parsers/test_olmo3_tool_parser.py | 34 | "role=null, " | CODE |
| HIGH⚡ | tests/tool_parsers/test_olmo3_tool_parser.py | 43 | '"role": null, ' | CODE |
| HIGH⚡ | tests/tool_parsers/test_pythonic_tool_parser.py | 35 | '"role": null, ' | CODE |
| HIGH | tests/tool_parsers/test_deepseekv3_tool_parser.py | 43 | """"bool_field": true, "null_field": null, """ | STRING |
| HIGH | tests/tool_parsers/test_internlm2_tool_parser.py | 56 | "null_field": null, | CODE |
| HIGH | tests/tool_parsers/test_llama4_pythonic_tool_parser.py | 35 | '"role": null, ' | CODE |
| HIGH | tests/tool_parsers/test_phi4mini_tool_parser.py | 50 | "null_field": null, | CODE |
| HIGH⚡ | tests/tool_parsers/test_lfm2_tool_parser.py | 38 | '"role": null, ' | CODE |
| HIGH | tests/tool_parsers/test_lfm2_tool_parser.py | 75 | "bash(command='cd /testbed && python3 -c \"from sympy import latex\n" | CODE |
| HIGH | tests/tool_parsers/test_lfm2_tool_parser.py | 81 | '{"command": "cd /testbed && python3 -c \\"from sympy import latex\\n' | CODE |
| HIGH | tests/tool_parsers/test_longcat_tool_parser.py | 54 | "null_field": null, | CODE |
| HIGH⚡ | tests/tool_parsers/test_hunyuan_a13b_tool_parser.py | 45 | '<tool_calls>[{"name": "get_weather", "arguments": {"city": "San Francisco", "metric": "celsius"}}, {"name": | CODE |
| HIGH | tests/tool_parsers/test_qwen3coder_tool_parser.py | 594 | # Multi non-null: anyOf[string, integer, null] → first non-null is string | COMMENT |
| HIGH⚡ | tests/renderers/test_chat_utils.py | 3118 | bad = '{"cmd": "mkdir -p /tmp/mmadtest && cat > /tmp/mmadtest' | CODE |
| HIGH | tests/tools/test_check_test_tethering.py | 172 | ("pytest kernels/test_a.py && pytest lora/test_b.py", "lora/test_b.py", True), | CODE |
| HIGH | tests/tools/test_check_test_tethering.py | 257 | "find kernels -name 'test_*.py' -delete && pytest test_regression.py", | CODE |
| HIGH | tests/parser/test_response_template.py | 531 | "enabled:false,ratio:5.0,label:null,value:null}" | CODE |
| HIGH | tests/parser/test_chat_parsing.py | 478 | 'null_value:null,number_value:1,string_value:<|"|>foo<|"|>,' | CODE |
| HIGH | tests/parser/engine/test_deepseek_v41.py | 138 | '{"a": [true, null]}</｜DSML｜ parameter>' | CODE |
| HIGH | tests/parser/engine/test_qwen3.py | 263 | ' "multiSelect": false, "answer": null}]' | CODE |
| HIGH | tests/parser/engine/test_qwen3.py | 274 | ' "multiSelect": false, "answer": null}]', | CODE |
| HIGH | tests/parser/engine/test_qwen3.py | 1143 | ' "multiSelect": false, "answer": null}]' | CODE |
| HIGH | tests/parser/engine/test_qwen3.py | 1164 | ' "multiSelect": false, "answer": null}]', | CODE |
| HIGH⚡ | tests/parser/engine/test_parser_engine.py | 695 | result = engine._fix_arg_types('{"val": null}', "f") | CODE |
| HIGH⚡ | tests/parser/engine/test_gemma4_streaming_reasoning.py | 1282 | text = "<|tool_call>call:configure{value:null}<tool_call|>" | CODE |
| HIGH⚡ | tests/parser/engine/test_gemma4_streaming_reasoning.py | 1290 | text = "<|tool_call>call:configure{label:null}<tool_call|>" | CODE |
| HIGH⚡ | tests/parser/mistral/test_tool_calls.py | 2770 | _REAL_LEAK_CONTENT = 'bash[TOOL_CALLS]{"command": "cd /testbed && git diff xarray/core/weighted.py"}ometracediff --git a | CODE |
| HIGH⚡ | tests/parser/mistral/test_tool_calls.py | 2783 | {"command": "cd /testbed && git diff xarray/core/weighted.py"} | CODE |
| HIGH | …ntrypoints/rl/return_routed_experts/test_openai_api.py | 44 | '{"sliding_window": null}', | CODE |
| HIGH | …ypoints/rl/return_routed_experts/test_scale_out_api.py | 34 | '{"sliding_window": null}', | CODE |
| HIGH | vllm/_custom_ops.py | 1686 | assert k_times_2 % 2 == 0, "input width must be even (gate || up layout)" | CODE |
| HIGH | vllm/_custom_ops.py | 1794 | assert k_times_2 % 2 == 0, "input width must be even (gate || up layout)" | CODE |
| 17 more matches not shown… | ||||
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| LOW | tests/conftest.py | 374 | CODE | |
| LOW | tests/conftest.py | 412 | CODE | |
| LOW | tests/conftest.py | 976 | CODE | |
| LOW | tests/test_config.py | 151 | CODE | |
| LOW | tests/v1/kv_offload/test_factory.py | 38 | CODE | |
| LOW | tests/v1/kv_offload/tiering/test_kvcr_tier.py | 171 | CODE | |
| LOW | tests/v1/kv_offload/tiering/p2p/test_manager.py | 638 | CODE | |
| LOW | tests/v1/metrics/test_perf_metrics.py | 83 | CODE | |
| LOW | tests/v1/attention/utils.py | 182 | CODE | |
| LOW | tests/v1/attention/test_sparse_mla_backends.py | 1951 | CODE | |
| LOW | tests/v1/attention/test_rocm_aiter_mla_mtp_split.py | 89 | CODE | |
| LOW | tests/v1/attention/test_attention_backends.py | 415 | CODE | |
| LOW | tests/v1/core/test_kv_cache_utils.py | 499 | CODE | |
| LOW | tests/v1/core/test_scheduler.py | 3358 | CODE | |
| LOW | tests/v1/core/test_scheduler.py | 3482 | CODE | |
| LOW | tests/v1/core/utils.py | 54 | CODE | |
| LOW | tests/v1/core/utils.py | 249 | CODE | |
| LOW | tests/v1/kv_connector/unit/test_tp_mapping.py | 32 | CODE | |
| LOW | tests/v1/kv_connector/unit/test_nixl_connector.py | 4014 | CODE | |
| LOW | …sts/v1/kv_connector/unit/test_mooncake_store_worker.py | 144 | CODE | |
| LOW | tests/v1/kv_connector/unit/test_moriio_connector.py | 304 | CODE | |
| LOW | tests/v1/kv_connector/unit/test_nixl_push_connector.py | 418 | CODE | |
| LOW | tests/v1/kv_connector/unit/utils.py | 106 | CODE | |
| LOW | tests/v1/kv_connector/unit/utils.py | 217 | CODE | |
| LOW | …/kv_connector/unit/offloading_connector/test_config.py | 303 | CODE | |
| LOW | tests/v1/sample/test_rejection_sampler.py | 71 | CODE | |
| LOW | tests/v1/e2e/general/test_mamba_prefix_cache.py | 182 | CODE | |
| LOW | tests/v1/simple_kv_offload/test_scheduler.py | 2345 | CODE | |
| LOW | tests/v1/simple_kv_offload/test_kv_events.py | 142 | CODE | |
| LOW | tests/kernels/test_gate_linear_rocm_dispatch.py | 23 | CODE | |
| LOW | tests/kernels/utils.py | 154 | CODE | |
| LOW | tests/kernels/attention/test_mha_attn.py | 121 | CODE | |
| LOW | tests/kernels/attention/test_attention.py | 55 | CODE | |
| LOW | …sts/kernels/attention/test_triton_prefill_attention.py | 76 | CODE | |
| LOW | tests/kernels/mamba/test_mamba_ssm.py | 115 | CODE | |
| LOW | tests/kernels/mamba/utils.py | 82 | CODE | |
| LOW | tests/kernels/quantization/test_nvfp4_quant.py | 81 | CODE | |
| LOW | tests/kernels/moe/test_deepep_v2_moe.py | 213 | CODE | |
| LOW | tests/kernels/moe/test_moe_permute_unpermute.py | 117 | CODE | |
| LOW | tests/kernels/moe/test_moe.py | 2072 | CODE | |
| LOW | tests/kernels/moe/utils.py | 59 | CODE | |
| LOW | tests/kernels/moe/utils.py | 120 | CODE | |
| LOW | tests/kernels/moe/test_moe_layer.py | 1019 | CODE | |
| LOW | tests/kernels/moe/test_moe_layer.py | 1122 | CODE | |
| LOW | tests/evals/mrcr/mrcr_eval.py | 161 | CODE | |
| LOW | tests/evals/gsm8k/gsm8k_eval.py | 236 | CODE | |
| LOW | tests/distributed/test_multiproc_executor.py | 70 | CODE | |
| LOW | tests/config/test_aux_output_config.py | 16 | CODE | |
| LOW | tests/models/test_dsv41_mono_numerics.py | 492 | CODE | |
| LOW | tests/models/kimi_k3/test_kda_metadata.py | 91 | CODE | |
| LOW | …/models/multimodal/generation/vlm_utils/model_utils.py | 503 | CODE | |
| LOW | tests/models/transformers/fusers/test_moe.py | 188 | CODE | |
| LOW | tests/quantization/reference_mxfp4.py | 88 | CODE | |
| LOW | tests/quantization/reference_mxfp4.py | 235 | CODE | |
| LOW | tests/compile/test_graph_partition.py | 101 | CODE | |
| LOW | tests/compile/fullgraph/test_toy_llama.py | 154 | CODE | |
| LOW | tests/compile/passes/test_silu_mul_quant_fusion.py | 132 | CODE | |
| LOW | tests/compile/passes/test_fusion.py | 610 | CODE | |
| LOW | tests/benchmarks/test_random_dataset.py | 179 | CODE | |
| LOW | tests/entrypoints/cohere/test_serving_streaming.py | 76 | CODE | |
| 361 more matches not shown… | ||||
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| LOW | setup.py | 1233 | # Check if the upstream_main_commit exists in the local repo | COMMENT |
| LOW | tools/install_torchcodec_rocm.sh | 19 | # Check if torchcodec is already installed and working | COMMENT |
| LOW | tools/pre_commit/check_forbidden_imports.py | 172 | # Check if it's allowed | COMMENT |
| LOW | tools/pre_commit/update-dockerfile-graph.sh | 10 | # Check if docker/Dockerfile is among the provided files | COMMENT |
| LOW | tools/pre_commit/update-dockerfile-graph.sh | 14 | # Check if Docker is installed and running | COMMENT |
| LOW | tools/pre_commit/update-dockerfile-graph.sh | 72 | # Check if the graph has changed | COMMENT |
| LOW | tools/vllm-rocm/pin_rocm_dependencies.py | 146 | # Check if this line is for one of our custom packages | COMMENT |
| LOW | tests/conftest.py | 388 | # Set this to avoid hanging issue | COMMENT |
| LOW | tests/conftest.py | 992 | # Set this to avoid hanging issue | COMMENT |
| LOW | tests/utils.py | 523 | os.kill(spid, 0) # Check if still alive | CODE |
| LOW | tests/test_config.py | 2445 | # Check if LONGCHAT_ROPE_PARAMETERS entries are in longchat_model_config | COMMENT |
| LOW | tests/v1/attention/test_mla_backends.py | 1679 | # Set num_speculative_tokens to query_len - 1 | COMMENT |
| LOW | tests/v1/attention/test_sparse_mla_backends.py | 972 | # Set some to -1 to test masking | COMMENT |
| LOW | tests/v1/attention/test_sparse_mla_backends.py | 976 | # Set some to out of bounds | COMMENT |
| LOW | tests/v1/attention/test_sparse_mla_backends.py | 1101 | # Set some to -1 to test masking | COMMENT |
| LOW | tests/v1/attention/test_sparse_mla_backends.py | 1105 | # Set some to out of bounds | COMMENT |
| LOW | tests/v1/core/test_scheduler.py | 3546 | # Verify if position length is identical | COMMENT |
| LOW | tests/v1/core/test_scheduler.py | 4510 | # Check if scheduled_encoder_inputs is empty as expected | COMMENT |
| LOW | tests/v1/core/test_scheduler.py | 4946 | # Set up to test different encoder cache existence scenario after preemption | COMMENT |
| LOW | tests/v1/core/test_scheduler.py | 5274 | # Set up to test different encoder cache existence scenario after preemption | COMMENT |
| LOW | tests/v1/core/utils.py | 304 | # Verify if position length is identical | COMMENT |
| LOW | …extract_hidden_states_integration/predictable_llama.py | 82 | # Check if we need auxiliary hidden states | COMMENT |
| LOW | …nnector/nixl_integration/config_sweep_accuracy_test.sh | 98 | # Check if cross-layers is enabled (non-empty) | COMMENT |
| LOW | tests/v1/determinism/test_batch_invariance.py | 280 | # Check if tokens match first | COMMENT |
| LOW | tests/v1/determinism/test_batch_invariance.py | 631 | # Check if tokens match first | COMMENT |
| LOW | tests/v1/determinism/test_batch_invariance.py | 831 | # Check if tokens match | COMMENT |
| LOW | tests/v1/determinism/test_batch_invariance.py | 849 | # Check if logprobs match bitwise | COMMENT |
| LOW | tests/v1/spec_decode/test_acceptance_length.py | 107 | # Check if get_valid_backends is actually defined in the platform class | COMMENT |
| LOW | …1/ec_connector/integration/run_epd_correctness_test.sh | 28 | # Set 1 to use multimodal prompts; else to use text-only | COMMENT |
| LOW | …ts/v1/ec_connector/integration/test_epd_correctness.py | 262 | # Check if server is ready | COMMENT |
| LOW | tests/v1/engine/test_engine_core_client.py | 548 | # Check if all request IDs in outputs have finished | COMMENT |
| LOW | tests/v1/engine/utils.py | 133 | # Check if the sampled_token_id occurs in choice_tensor[1:] | COMMENT |
| LOW | tests/utils_/test_network_utils.py | 185 | # Check if IPv6 is supported by trying to create an IPv6 socket | COMMENT |
| LOW | tests/kernels/mamba/test_cpu_short_conv.py | 106 | # Check if KV cache was updated | COMMENT |
| LOW | tests/kernels/mamba/test_cpu_short_conv.py | 159 | # Check if KV cache was updated | COMMENT |
| LOW | …/kernels/quantization/test_rocm_aiter_grouped_quant.py | 21 | # Check if aiter package is installed | COMMENT |
| LOW | tests/kernels/moe/test_moe_layer.py | 2098 | # Check if enough GPUs available | COMMENT |
| LOW⚡ | tests/kernels/moe/test_rocm_aiter_topk.py | 26 | # Check if aiter package is installed | COMMENT |
| LOW⚡ | tests/kernels/moe/test_rocm_aiter_topk.py | 35 | # Check if the op exists in torch.ops.vllm | COMMENT |
| LOW⚡ | tests/kernels/moe/test_rocm_aiter_topk.py | 38 | # Check if the op is callable | COMMENT |
| LOW⚡ | tests/kernels/moe/test_rocm_aiter_topk.py | 44 | # Check if the op exists in torch.ops.vllm | COMMENT |
| LOW⚡ | tests/kernels/moe/test_rocm_aiter_topk.py | 47 | # Check if the op is callable | COMMENT |
| LOW | tests/evals/gsm8k/gsm8k_eval.py | 476 | # Print results to terminal | COMMENT |
| LOW | tests/distributed/test_eplb_execute.py | 180 | # Check if the weights are correct | COMMENT |
| LOW | tests/model_executor/test_qwen3_omni.py | 33 | # Check if it's a special token that should be compressed | COMMENT |
| LOW | …model_loader/runai_streamer_loader/test_runai_utils.py | 56 | # Read the file in chunks to handle large files efficiently | COMMENT |
| LOW⚡ | tests/models/multimodal/generation/test_maverick.py | 57 | # Print the outputs | COMMENT |
| LOW⚡ | tests/multimodal/media/test_audio.py | 139 | # Set limit to 1 MiB — should reject | COMMENT |
| LOW | tests/compile/test_config.py | 54 | # Check if get_raw_stream exists in builtins | COMMENT |
| LOW | tests/compile/fusions_e2e/conftest.py | 59 | # Print the outputs. | COMMENT |
| LOW | tests/compile/fullgraph/test_full_graph.py | 197 | # Print the outputs. | COMMENT |
| LOW | tests/entrypoints/llm/offline_mode/test_offline_mode.py | 75 | # Set HF to offline mode and ensure we can still construct an LLM | COMMENT |
| LOW | tests/entrypoints/llm/offline_mode/test_offline_mode.py | 145 | # Set HF to offline mode and ensure we can still construct an LLM | COMMENT |
| LOW | tests/entrypoints/launchers/test_shutdown.py | 101 | # Check if any choice has finish_reason='abort' | COMMENT |
| LOW | tests/entrypoints/serve/utils/test_request_logger.py | 97 | # Set max_log_len to 10 | COMMENT |
| LOW | …rve/sagemaker/test_sagemaker_middleware_integration.py | 326 | # Check if environment variable middleware was applied | COMMENT |
| LOW | tests/lora/test_transformers_model.py | 39 | # Print the outputs. | COMMENT |
| LOW | tests/lora/test_gptoss_tp.py | 95 | # Print the outputs. | COMMENT |
| LOW | tests/lora/test_qwen3moe_tp.py | 57 | # Print the outputs. | COMMENT |
| LOW | tests/lora/test_chatglm3_tp.py | 46 | # Print the outputs. | COMMENT |
| 189 more matches not shown… | ||||
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| LOW⚡ | csrc/cpu/cpu_types_vsx.hpp | 340 | // Step 1: pairwise sum of the two 4-wide halves | COMMENT |
| LOW⚡ | csrc/cpu/cpu_types_vsx.hpp | 342 | // Step 2: rotate by 8 bytes (2 floats) and add | COMMENT |
| LOW⚡ | csrc/cpu/cpu_types_vsx.hpp | 344 | // Step 3: rotate by 4 bytes (1 float) and add => all lanes hold total | COMMENT |
| LOW⚡ | csrc/cpu/sgl-kernels/gemm_fp8_w8a8.cpp | 729 | // Step 1: Handle case of overflow | COMMENT |
| LOW⚡ | csrc/cpu/sgl-kernels/gemm_fp8_w8a8.cpp | 736 | // Step 2: Handle small numbers (denormals) | COMMENT |
| LOW | csrc/cpu/sgl-kernels/gemm_fp8_w8a8.cpp | 748 | // Step 3: Handle normal numbers | COMMENT |
| LOW | csrc/cpu/sgl-kernels/gemm_fp8_w8a8.cpp | 833 | // Step 1: Compute channel-wise max using AVX512 | COMMENT |
| LOW⚡ | csrc/cpu/sgl-kernels/gemm_fp8_w8a8.cpp | 860 | // Step 2: Apply quant_max and EPS to compute scale | COMMENT |
| LOW⚡ | csrc/cpu/sgl-kernels/gemm_fp8_w8a8.cpp | 865 | // Step 3: Scale and clamp using AVX512 (reuse the same loop structure) | COMMENT |
| LOW⚡ | tests/v1/kv_offload/tiering/test_async_lookup.py | 237 | # Step 1: lookup key 1, flush | COMMENT |
| LOW⚡ | tests/v1/kv_offload/tiering/test_async_lookup.py | 243 | # Step 2: lookup keys 2 and 3, flush | COMMENT |
| LOW⚡ | tests/v1/kv_offload/tiering/p2p/test_sessions.py | 886 | # Step 1: probe hA, hB. | COMMENT |
| LOW⚡ | tests/v1/kv_offload/tiering/p2p/test_sessions.py | 896 | # Step 2: a new key is discovered for the same request. The | COMMENT |
| LOW | tests/v1/logits_processors/test_correctness.py | 1294 | # Step 1: think-start token appears. | COMMENT |
| LOW⚡ | tests/v1/core/test_scheduler.py | 6915 | # Step 1: A's load is admitted; B's is held back by the reservation (B never | COMMENT |
| LOW⚡ | tests/v1/core/test_scheduler.py | 6925 | # Step 2: nothing changes until A's recv lands. | COMMENT |
| LOW⚡ | tests/v1/core/test_scheduler.py | 6934 | # Step 3: A makes forward progress straight to RUNNING - no preemption was | COMMENT |
| LOW⚡ | tests/v1/kv_connector/unit/test_lmcache_connector.py | 569 | # Step 1: Get events from lmcache engine | COMMENT |
| LOW⚡ | tests/v1/kv_connector/unit/test_lmcache_connector.py | 578 | # Step 2: Update connector output (simulate receiving from worker) | COMMENT |
| LOW⚡ | tests/v1/kv_connector/unit/test_lmcache_connector.py | 584 | # Step 3: Take events | COMMENT |
| LOW⚡ | tests/v1/kv_connector/unit/test_moriio_connector.py | 812 | # Step 1: first chunk (4 of 8 tokens) is buffered, not saved. | COMMENT |
| LOW⚡ | tests/v1/kv_connector/unit/test_moriio_connector.py | 820 | # Step 2: final chunk (computed 4 + scheduled 4 >= 8) assembles the full | COMMENT |
| LOW⚡ | …/v1/kv_connector/unit/test_remote_prefill_lifecycle.py | 436 | # Step 2: 5 blocks are in use (2 new for remote blocks). | COMMENT |
| LOW⚡ | …/v1/kv_connector/unit/test_remote_prefill_lifecycle.py | 444 | # Step 3: finish recving (5 blocks in use) | COMMENT |
| LOW⚡ | …/v1/kv_connector/unit/test_remote_prefill_lifecycle.py | 453 | # Step 4: try to schedule, remote request is put to running list | COMMENT |
| LOW⚡ | …/v1/kv_connector/unit/test_remote_prefill_lifecycle.py | 463 | # Step 5: Remote request will be put back to waiting list | COMMENT |
| LOW⚡ | …/v1/kv_connector/unit/test_remote_prefill_lifecycle.py | 471 | # Step 6: finish the request, free it. | COMMENT |
| LOW⚡ | …/v1/kv_connector/unit/test_remote_prefill_lifecycle.py | 480 | # Step 7: now we can schedule (with 2 blocks computed), | COMMENT |
| LOW⚡ | …/v1/kv_connector/unit/test_remote_prefill_lifecycle.py | 552 | # Step 3: finish the request, free it. | COMMENT |
| LOW⚡ | …/v1/kv_connector/unit/test_remote_prefill_lifecycle.py | 561 | # Step 4: now we can initiate KV transfer (with 2 blocks computed). | COMMENT |
| LOW⚡ | …/v1/kv_connector/unit/test_remote_prefill_lifecycle.py | 569 | # Step 5: finish recving (5 blocks in use) | COMMENT |
| LOW⚡ | …/v1/kv_connector/unit/test_remote_prefill_lifecycle.py | 578 | # Step 6: schedule remote request | COMMENT |
| LOW⚡ | …/v1/kv_connector/unit/test_remote_prefill_lifecycle.py | 585 | # Step 7: free everything. | COMMENT |
| LOW | …/v1/kv_connector/unit/test_remote_prefill_lifecycle.py | 496 | # Step 8: free everything. | COMMENT |
| LOW | …/v1/kv_connector/unit/test_remote_prefill_lifecycle.py | 541 | # Step 2: 3 blocks are in use, | COMMENT |
| LOW | …/v1/kv_connector/unit/test_remote_prefill_lifecycle.py | 863 | # Step 1: req_a's lookup is deferred; req_b is admitted as an async | COMMENT |
| LOW | tests/v1/determinism/test_batch_invariance.py | 763 | # Step 1: Run decode and collect logprobs | COMMENT |
| LOW | tests/v1/determinism/test_batch_invariance.py | 782 | # Step 2: For each token position, run prefill and compare. | COMMENT |
| LOW⚡ | tests/v1/streaming_input/test_scheduler_streaming.py | 474 | # Step 2: Schedule creates NewRequestData | COMMENT |
| LOW⚡ | tests/v1/streaming_input/test_scheduler_streaming.py | 548 | # Step 7: Schedule again - now request uses cached state | COMMENT |
| LOW⚡ | tests/v1/streaming_input/test_scheduler_streaming.py | 617 | # Step 12: Add new streaming request with seq_id=1 | COMMENT |
| LOW | tests/v1/streaming_input/test_scheduler_streaming.py | 489 | # Step 3: Simulate model runner caching the prompt_token_ids | COMMENT |
| LOW | tests/v1/streaming_input/test_scheduler_streaming.py | 522 | # Step 6: Verify request state after Cycle 1 | COMMENT |
| LOW | tests/v1/streaming_input/test_scheduler_streaming.py | 560 | # Step 8: Calculate num_tokens like gpu_model_runner.py:1284 does | COMMENT |
| LOW | tests/v1/streaming_input/test_scheduler_streaming.py | 596 | # Step 11: Verify request transitioned to WAITING_FOR_STREAMING_REQ | COMMENT |
| LOW | tests/v1/streaming_input/test_scheduler_streaming.py | 629 | # Step 13: Scheduler schedules the updated session | COMMENT |
| LOW | tests/v1/streaming_input/test_scheduler_streaming.py | 647 | # Step 14: Model runner caches NEW prompt_token_ids reference | COMMENT |
| LOW | tests/v1/streaming_input/test_scheduler_streaming.py | 660 | # Step 15: FINAL CRITICAL VERIFICATION | COMMENT |
| LOW | …/streaming_input/test_gpu_model_runner_v2_streaming.py | 85 | # Step 1: Add initial request with 3 prompt tokens, all computed | COMMENT |
| LOW | …/streaming_input/test_gpu_model_runner_v2_streaming.py | 101 | # Step 2: Create streaming update with extended prompt | COMMENT |
| LOW | …/streaming_input/test_gpu_model_runner_v2_streaming.py | 117 | # Step 3: Verify no free_indices leak (old slot recycled) | COMMENT |
| LOW | …/streaming_input/test_gpu_model_runner_v2_streaming.py | 156 | # Step 1: Add initial request with one audio feature | COMMENT |
| LOW | …/streaming_input/test_gpu_model_runner_v2_streaming.py | 176 | # Step 2: Create streaming update with additional multimodal feature | COMMENT |
| LOW | …/streaming_input/test_gpu_model_runner_v2_streaming.py | 193 | # Step 3: Verify no free_indices leak | COMMENT |
| LOW | …/v1/streaming_input/test_gpu_model_runner_streaming.py | 59 | # Step 1: Create initial request state with some computed tokens | COMMENT |
| LOW | …/v1/streaming_input/test_gpu_model_runner_streaming.py | 77 | # Step 2: Create new request data with extended prompt | COMMENT |
| LOW⚡ | …/v1/streaming_input/test_gpu_model_runner_streaming.py | 96 | # Step 3: Update the request | COMMENT |
| LOW⚡ | …/v1/streaming_input/test_gpu_model_runner_streaming.py | 101 | # Step 4: Verify the request state was updated correctly | COMMENT |
| LOW | …/v1/streaming_input/test_gpu_model_runner_streaming.py | 131 | # Step 1: Create initial request state with one multimodal feature | COMMENT |
| LOW | …/v1/streaming_input/test_gpu_model_runner_streaming.py | 156 | # Step 2: Create new request data with additional multimodal feature | COMMENT |
| 142 more matches not shown… | ||||
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| LOW⚡ | csrc/cpu/cpu_types_vsx.hpp | 340 | // Step 1: pairwise sum of the two 4-wide halves | COMMENT |
| LOW⚡ | csrc/cpu/cpu_types_vsx.hpp | 342 | // Step 2: rotate by 8 bytes (2 floats) and add | COMMENT |
| LOW⚡ | csrc/cpu/cpu_types_vsx.hpp | 344 | // Step 3: rotate by 4 bytes (1 float) and add => all lanes hold total | COMMENT |
| LOW⚡ | csrc/cpu/sgl-kernels/gemm_fp8_w8a8.cpp | 729 | // Step 1: Handle case of overflow | COMMENT |
| LOW⚡ | csrc/cpu/sgl-kernels/gemm_fp8_w8a8.cpp | 736 | // Step 2: Handle small numbers (denormals) | COMMENT |
| LOW | csrc/cpu/sgl-kernels/gemm_fp8_w8a8.cpp | 748 | // Step 3: Handle normal numbers | COMMENT |
| LOW | csrc/cpu/sgl-kernels/gemm_fp8_w8a8.cpp | 833 | // Step 1: Compute channel-wise max using AVX512 | COMMENT |
| LOW⚡ | csrc/cpu/sgl-kernels/gemm_fp8_w8a8.cpp | 860 | // Step 2: Apply quant_max and EPS to compute scale | COMMENT |
| LOW⚡ | csrc/cpu/sgl-kernels/gemm_fp8_w8a8.cpp | 865 | // Step 3: Scale and clamp using AVX512 (reuse the same loop structure) | COMMENT |
| LOW⚡ | tests/v1/kv_offload/tiering/test_async_lookup.py | 237 | # Step 1: lookup key 1, flush | COMMENT |
| LOW⚡ | tests/v1/kv_offload/tiering/test_async_lookup.py | 243 | # Step 2: lookup keys 2 and 3, flush | COMMENT |
| LOW⚡ | tests/v1/kv_offload/tiering/p2p/test_sessions.py | 886 | # Step 1: probe hA, hB. | COMMENT |
| LOW⚡ | tests/v1/kv_offload/tiering/p2p/test_sessions.py | 896 | # Step 2: a new key is discovered for the same request. The | COMMENT |
| LOW | tests/v1/logits_processors/test_correctness.py | 1294 | # Step 1: think-start token appears. | COMMENT |
| LOW⚡ | tests/v1/core/test_scheduler.py | 6915 | # Step 1: A's load is admitted; B's is held back by the reservation (B never | COMMENT |
| LOW⚡ | tests/v1/core/test_scheduler.py | 6925 | # Step 2: nothing changes until A's recv lands. | COMMENT |
| LOW⚡ | tests/v1/core/test_scheduler.py | 6934 | # Step 3: A makes forward progress straight to RUNNING - no preemption was | COMMENT |
| LOW⚡ | tests/v1/kv_connector/unit/test_lmcache_connector.py | 569 | # Step 1: Get events from lmcache engine | COMMENT |
| LOW⚡ | tests/v1/kv_connector/unit/test_lmcache_connector.py | 578 | # Step 2: Update connector output (simulate receiving from worker) | COMMENT |
| LOW⚡ | tests/v1/kv_connector/unit/test_lmcache_connector.py | 584 | # Step 3: Take events | COMMENT |
| LOW⚡ | tests/v1/kv_connector/unit/test_moriio_connector.py | 812 | # Step 1: first chunk (4 of 8 tokens) is buffered, not saved. | COMMENT |
| LOW⚡ | tests/v1/kv_connector/unit/test_moriio_connector.py | 820 | # Step 2: final chunk (computed 4 + scheduled 4 >= 8) assembles the full | COMMENT |
| LOW⚡ | …/v1/kv_connector/unit/test_remote_prefill_lifecycle.py | 436 | # Step 2: 5 blocks are in use (2 new for remote blocks). | COMMENT |
| LOW⚡ | …/v1/kv_connector/unit/test_remote_prefill_lifecycle.py | 444 | # Step 3: finish recving (5 blocks in use) | COMMENT |
| LOW⚡ | …/v1/kv_connector/unit/test_remote_prefill_lifecycle.py | 453 | # Step 4: try to schedule, remote request is put to running list | COMMENT |
| LOW⚡ | …/v1/kv_connector/unit/test_remote_prefill_lifecycle.py | 463 | # Step 5: Remote request will be put back to waiting list | COMMENT |
| LOW⚡ | …/v1/kv_connector/unit/test_remote_prefill_lifecycle.py | 471 | # Step 6: finish the request, free it. | COMMENT |
| LOW⚡ | …/v1/kv_connector/unit/test_remote_prefill_lifecycle.py | 480 | # Step 7: now we can schedule (with 2 blocks computed), | COMMENT |
| LOW⚡ | …/v1/kv_connector/unit/test_remote_prefill_lifecycle.py | 552 | # Step 3: finish the request, free it. | COMMENT |
| LOW⚡ | …/v1/kv_connector/unit/test_remote_prefill_lifecycle.py | 561 | # Step 4: now we can initiate KV transfer (with 2 blocks computed). | COMMENT |
| LOW⚡ | …/v1/kv_connector/unit/test_remote_prefill_lifecycle.py | 569 | # Step 5: finish recving (5 blocks in use) | COMMENT |
| LOW⚡ | …/v1/kv_connector/unit/test_remote_prefill_lifecycle.py | 578 | # Step 6: schedule remote request | COMMENT |
| LOW⚡ | …/v1/kv_connector/unit/test_remote_prefill_lifecycle.py | 585 | # Step 7: free everything. | COMMENT |
| LOW | …/v1/kv_connector/unit/test_remote_prefill_lifecycle.py | 496 | # Step 8: free everything. | COMMENT |
| LOW | …/v1/kv_connector/unit/test_remote_prefill_lifecycle.py | 541 | # Step 2: 3 blocks are in use, | COMMENT |
| LOW | …/v1/kv_connector/unit/test_remote_prefill_lifecycle.py | 863 | # Step 1: req_a's lookup is deferred; req_b is admitted as an async | COMMENT |
| LOW | tests/v1/determinism/test_batch_invariance.py | 763 | # Step 1: Run decode and collect logprobs | COMMENT |
| LOW | tests/v1/determinism/test_batch_invariance.py | 782 | # Step 2: For each token position, run prefill and compare. | COMMENT |
| LOW⚡ | tests/v1/streaming_input/test_scheduler_streaming.py | 474 | # Step 2: Schedule creates NewRequestData | COMMENT |
| LOW⚡ | tests/v1/streaming_input/test_scheduler_streaming.py | 548 | # Step 7: Schedule again - now request uses cached state | COMMENT |
| LOW⚡ | tests/v1/streaming_input/test_scheduler_streaming.py | 617 | # Step 12: Add new streaming request with seq_id=1 | COMMENT |
| LOW | tests/v1/streaming_input/test_scheduler_streaming.py | 489 | # Step 3: Simulate model runner caching the prompt_token_ids | COMMENT |
| LOW | tests/v1/streaming_input/test_scheduler_streaming.py | 522 | # Step 6: Verify request state after Cycle 1 | COMMENT |
| LOW | tests/v1/streaming_input/test_scheduler_streaming.py | 560 | # Step 8: Calculate num_tokens like gpu_model_runner.py:1284 does | COMMENT |
| LOW | tests/v1/streaming_input/test_scheduler_streaming.py | 596 | # Step 11: Verify request transitioned to WAITING_FOR_STREAMING_REQ | COMMENT |
| LOW | tests/v1/streaming_input/test_scheduler_streaming.py | 629 | # Step 13: Scheduler schedules the updated session | COMMENT |
| LOW | tests/v1/streaming_input/test_scheduler_streaming.py | 647 | # Step 14: Model runner caches NEW prompt_token_ids reference | COMMENT |
| LOW | tests/v1/streaming_input/test_scheduler_streaming.py | 660 | # Step 15: FINAL CRITICAL VERIFICATION | COMMENT |
| LOW | …/streaming_input/test_gpu_model_runner_v2_streaming.py | 85 | # Step 1: Add initial request with 3 prompt tokens, all computed | COMMENT |
| LOW | …/streaming_input/test_gpu_model_runner_v2_streaming.py | 101 | # Step 2: Create streaming update with extended prompt | COMMENT |
| LOW | …/streaming_input/test_gpu_model_runner_v2_streaming.py | 117 | # Step 3: Verify no free_indices leak (old slot recycled) | COMMENT |
| LOW | …/streaming_input/test_gpu_model_runner_v2_streaming.py | 156 | # Step 1: Add initial request with one audio feature | COMMENT |
| LOW | …/streaming_input/test_gpu_model_runner_v2_streaming.py | 176 | # Step 2: Create streaming update with additional multimodal feature | COMMENT |
| LOW | …/streaming_input/test_gpu_model_runner_v2_streaming.py | 193 | # Step 3: Verify no free_indices leak | COMMENT |
| LOW | …/v1/streaming_input/test_gpu_model_runner_streaming.py | 59 | # Step 1: Create initial request state with some computed tokens | COMMENT |
| LOW | …/v1/streaming_input/test_gpu_model_runner_streaming.py | 77 | # Step 2: Create new request data with extended prompt | COMMENT |
| LOW⚡ | …/v1/streaming_input/test_gpu_model_runner_streaming.py | 96 | # Step 3: Update the request | COMMENT |
| LOW⚡ | …/v1/streaming_input/test_gpu_model_runner_streaming.py | 101 | # Step 4: Verify the request state was updated correctly | COMMENT |
| LOW | …/v1/streaming_input/test_gpu_model_runner_streaming.py | 131 | # Step 1: Create initial request state with one multimodal feature | COMMENT |
| LOW | …/v1/streaming_input/test_gpu_model_runner_streaming.py | 156 | # Step 2: Create new request data with additional multimodal feature | COMMENT |
| 114 more matches not shown… | ||||
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| MEDIUM | csrc/cpu/sgl-kernels/decode.cpp | 1434 | // use smaller BLOCK_H when batches is small to utilize all cores | COMMENT |
| LOW | csrc/libtorch_stable/quantization/machete/generate.py | 435 | # For now, we can just use the first accumulator type seen since | STRING |
| MEDIUM⚡ | docker/entrypoints/test_vllm_nonroot_entrypoint.sh | 145 | # matching our UID instead for a robust check. | COMMENT |
| MEDIUM | tests/v1/logits_processors/test_custom_offline.py | 30 | # Create a mixture of requests which do and don't utilize the dummy logitproc | COMMENT |
| LOW | tests/v1/cudagraph/test_breakable_cudagraph.py | 306 | # Outside capture: decorator should just call through. | COMMENT |
| MEDIUM | …_connector/unit/offloading_connector/test_scheduler.py | 2969 | # freed blocks before the harness records its deferred stores. | COMMENT |
| LOW | tests/v1/ec_connector/integration/test_nixl_failure.py | 198 | # the fault, not just pass a frontend health check or hit a local cache. | COMMENT |
| MEDIUM⚡ | tests/renderers/test_sparse_tensor_validation.py | 65 | # explicitly so this fixture is robust to process-wide invariant-check state | COMMENT |
| MEDIUM | tests/kernels/attention/test_attention_selector.py | 607 | # binding, so patch is_xpu there (not on flash_attn's) to stay robust to | COMMENT |
| MEDIUM | tests/kernels/mamba/test_mamba_mixer2.py | 106 | # - utilize mock patching to disable TP when | COMMENT |
| MEDIUM | tests/distributed/test_pynccl.py | 442 | # Essentially this is an all-gather operation. | COMMENT |
| MEDIUM | tests/distributed/test_context_parallel.py | 45 | # .buildkite/lm-eval-harness/configs/DeepSeek-V2-Lite-Chat.yaml | COMMENT |
| MEDIUM | tests/distributed/test_context_parallel.py | 47 | # .buildkite/lm-eval-harness/configs/Qwen2.5-1.5B-Instruct.yaml | COMMENT |
| MEDIUM | tests/distributed/test_sharded_rdt_trainer.py | 400 | assert server.live_count == 1 # num_consumers=1 in the harness | CODE |
| LOW | tests/model_executor/test_qwen3_omni.py | 45 | # Regular token, just add it | COMMENT |
| MEDIUM | tests/models/registry.py | 1076 | # Keep the init/schema harness eager (these tests never run forward). | COMMENT |
| MEDIUM⚡ | tests/parser/engine/test_replay.py | 70 | # token, so it does not fit this TOOL_END-based replay harness. | COMMENT |
| MEDIUM⚡ | tests/parser/engine/test_replay.py | 80 | # harness; their direct and delegating coverage lives in dedicated | COMMENT |
| MEDIUM⚡ | tests/parser/engine/test_replay.py | 85 | # it does not fit this token-terminal harness; its replay coverage | COMMENT |
| MEDIUM | tests/parser/engine/test_delegating_replay.py | 90 | # token, so it does not fit this TOOL_END-based replay harness. | COMMENT |
| MEDIUM⚡ | tests/entrypoints/rl/conftest.py | 65 | # Server harness | COMMENT |
| LOW | vllm/env_override.py | 417 | # functions just return True. | COMMENT |
| LOW | vllm/v1/attention/backends/flash_attn.py | 616 | # but for now just set it to `UNIFORM_BATCH` to get use to drop down | COMMENT |
| LOW | vllm/v1/attention/backends/utils.py | 527 | # then we can simply use a cdiv for the rest. | COMMENT |
| MEDIUM | vllm/v1/attention/ops/triton_decode_attention.py | 382 | # explicitly facilitate overlapping load/compute | COMMENT |
| LOW | vllm/v1/core/kv_cache_coordinator.py | 584 | # Single group; useless but just set ``use_eagle`` for consistency regardless. | COMMENT |
| MEDIUM | vllm/v1/core/encoder_cache_manager.py | 368 | # utilize the cache and this class will fold into EncoderCacheManager, as | COMMENT |
| LOW | vllm/v1/spec_decode/llm_base_proposer.py | 619 | # KV cache in sync, so just return an empty tensor. | COMMENT |
| LOW | vllm/v1/spec_decode/llm_base_proposer.py | 1829 | # Therefore, we can just return the logits. | COMMENT |
| MEDIUM | vllm/v1/worker/gpu_worker.py | 954 | # CUDAGraph memory size and may not utilize all gpu memory. | COMMENT |
| MEDIUM | vllm/v1/worker/gpu_worker.py | 1352 | # Generate the trace name by combining prefix with comprehensive rank suffix | COMMENT |
| LOW | vllm/v1/worker/gpu/model_runner.py | 2000 | # For piecewise and eager mode, just call model(). | COMMENT |
| LOW | vllm/v1/engine/parallel_sampling.py | 118 | # If streaming, just return the current output | COMMENT |
| MEDIUM | vllm/tool_parsers/llama_tool_parser.py | 261 | # re-set stuff pertaining to progress in the current tool | COMMENT |
| MEDIUM | vllm/tool_parsers/jamba_tool_parser.py | 218 | # re-set stuff pertaining to progress in the current tool | COMMENT |
| LOW | vllm/tool_parsers/xlam_tool_parser.py | 560 | # If we encounter an error, just return the delta text as regular content | COMMENT |
| MEDIUM | vllm/tool_parsers/granite_20b_fc_tool_parser.py | 202 | # re-set stuff pertaining to progress in the current tool | COMMENT |
| MEDIUM | vllm/tool_parsers/step3_tool_parser.py | 50 | # Explicit state flags for robust streaming | COMMENT |
| LOW | vllm/tokenizers/mistral.py | 534 | # if underlying tokenizer is sentencepiece, we just add "�". | COMMENT |
| LOW | vllm/platforms/cuda.py | 722 | # users can just use IR op priority directly | COMMENT |
| MEDIUM | vllm/distributed/utils.py | 331 | """A robust barrier to synchronize all ranks. | STRING |
| MEDIUM | …ibuted/kv_transfer/kv_connector/v1/nixl/base_worker.py | 1185 | # we can leverage host_buffer for permute. | COMMENT |
| MEDIUM | …ransfer/kv_connector/v1/mooncake/mooncake_connector.py | 1199 | # Tasks can await async events, so a surplus (2x is a robust heuristic) | COMMENT |
| MEDIUM | vllm/config/model.py | 389 | """Enable the custom cumem allocator to leverage advanced GPU memory | STRING |
| MEDIUM | vllm/config/parallel.py | 714 | # To make the initialization more robust we retry a few times | COMMENT |
| LOW | vllm/model_executor/layers/mamba/ops/causal_conv1d.py | 194 | # first chunk and does not have prior-token, so just set to 0 | COMMENT |
| MEDIUM | vllm/model_executor/layers/fused_moe/modular_kernel.py | 53 | # The goal is to be able to utilize different communication mechanisms with | COMMENT |
| MEDIUM | …xecutor/layers/fused_moe/prepare_finalize/deepep_ll.py | 38 | # TODO (varun) : Optimize leverage num_tokens_per_expert counts | COMMENT |
| MEDIUM | …executor/layers/fused_moe/experts/fused_humming_moe.py | 466 | # Neighboring nodes are required to utilize distinct workspaces. | COMMENT |
| MEDIUM | vllm/model_executor/layers/quantization/fp8.py | 247 | # For GPUs that lack FP8 hardware support, we can leverage the Marlin | COMMENT |
| MEDIUM | vllm/model_executor/layers/quantization/modelopt.py | 357 | # Normalize quant_algo for robust matching (ModelOpt may emit lowercase). | COMMENT |
| MEDIUM | vllm/model_executor/layers/quantization/fbgemm_fp8.py | 53 | # For GPUs that lack FP8 hardware support, we can leverage the Marlin | COMMENT |
| MEDIUM | …executor/layers/quantization/utils/marlin_utils_fp4.py | 110 | # to fully utilize the E4M3 dynamic range (e.g., global_scale=1). | COMMENT |
| MEDIUM | …executor/layers/quantization/utils/marlin_utils_fp4.py | 170 | # For GPUs that lack FP4 hardware support, we can leverage the | COMMENT |
| MEDIUM | …executor/layers/quantization/utils/marlin_utils_fp8.py | 59 | # For GPUs that lack FP8 hardware support, we can leverage the | COMMENT |
| LOW | …odel_executor/layers/quantization/utils/quant_utils.py | 529 | # Unquantized layer: just return base weights | COMMENT |
| MEDIUM | vllm/model_executor/models/deepseek_ocr.py | 141 | """Example of overriding the wrapper class `__init__()` in order to utilize | STRING |
| MEDIUM | vllm/model_executor/models/deepseek_ocr.py | 141 | """Example of overriding the wrapper class `__init__()` in order to utilize | STRING |
| LOW | vllm/model_executor/models/qwen3_asr.py | 565 | # No audio features, just return linear positions | COMMENT |
| MEDIUM | vllm/models/hy_v4/nvidia/hc.py | 172 | # present; getattr keeps this robust if that ever changes. | COMMENT |
| 59 more matches not shown… | ||||
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| HIGH | vllm/_xpu_ops.py | 319 | XPU FP8 batched GEMM implementation for ``torch.ops.vllm.xpu_fp8_bmm``. Computes batched matrix multiplication over | STRING |
| HIGH | vllm/v1/kv_offload/tiering/factory.py | 101 | Get a secondary tier class by type. Args: tier_config: Tier configuration dict. Must contain 'type' | STRING |
| HIGH | vllm/v1/metrics/buckets.py | 211 | Return the bucket boundaries for a histogram family. Args: family: Canonical bucket family key. max | STRING |
| HIGH | vllm/v1/attention/backends/registry.py | 345 | Register or override a backend implementation. Args: backend: The AttentionBackendEnum/MambaAttentionBacken | STRING |
| HIGH | vllm/v1/attention/backends/mla/prefill/registry.py | 147 | Register or override an MLA prefill backend implementation. Args: backend: The MLAPrefillBackendEnum member | STRING |
| HIGH | vllm/v1/attention/backends/mla/prefill/selector.py | 116 | Auto-select the best available MLA prefill backend. Args: selector_config: Hashable configuration for backe | STRING |
| HIGH | vllm/v1/core/single_type_kv_cache_manager.py | 1317 | For chunked local attention, we need to find the longest cache hit prefix of the blocks that is not longer than | STRING |
| HIGH | vllm/v1/structured_output/utils.py | 407 | Check if grammar appears to use Lark syntax. Args: grammar_str: Input grammar string Returns: | STRING |
| HIGH | vllm/v1/worker/utils.py | 520 | Select a block size that is supported by all backends and is a factor of kv_manager_block_size. If kv_manager_b | STRING |
| HIGH | vllm/tool_parsers/hy_v4_tool_parser.py | 92 | Detect the per-checkpoint suffix used by Hunyuan structural tokens. Args: tokenizer: Tokenizer of the serve | STRING |
| HIGH | vllm/tool_parsers/apertus_tool_parser.py | 135 | Buffers incoming delta chunks to prevent fragmentation of multi-token special tags. If a chunk ends wit | STRING |
| HIGH | vllm/tool_parsers/apertus_tool_parser.py | 179 | Extract tool calls from a complete model response (non-streaming). Args: model_output: The full com | STRING |
| HIGH | vllm/tool_parsers/apertus_tool_parser.py | 281 | Handles streaming chunks. Args: previous_text: The complete model text generated prior to this chun | STRING |
| HIGH | vllm/tool_parsers/apertus_tool_parser.py | 492 | Calculates the exact string difference to safely append new tool parameters. This ensures characters like `{`, | STRING |
| HIGH | vllm/renderers/chat_utils.py | 1694 | Parses a given multi-modal content part based on its type. Args: part: A dict containing the content part, | STRING |
| HIGH | …ed/kv_transfer/kv_connector/v1/lmcache_mp_connector.py | 729 | Get number of new tokens that can be loaded from the external KV cache beyond the num_computed_tokens. | STRING |
| HIGH | vllm/distributed/kv_transfer/kv_connector/v1/base.py | 506 | Get number of new tokens that can be loaded from the external KV cache beyond the num_computed_tokens. | STRING |
| HIGH | …ed/kv_transfer/kv_connector/v1/moriio/moriio_engine.py | 287 | Get remote allocation info for a request. Args: transfer_id:TransferId The request ID Retu | STRING |
| HIGH | …nector/v1/lmcache_integration/multi_process_adapter.py | 197 | Submit a new lookup request to LMCache if there is no ongoing request. Supports both token-based and hash-based | STRING |
| HIGH | …nector/v1/lmcache_integration/multi_process_adapter.py | 602 | Check and get the finished store and retrieve requests. Args: finished_req_ids_from_engine: the set | STRING |
| HIGH | vllm/distributed/weight_transfer/factory.py | 92 | Create a weight transfer engine instance. Args: config: Weight transfer configuration containing th | STRING |
| HIGH | vllm/distributed/weight_transfer/base.py | 422 | Construct typed init info from dict with validation. Args: init_dict: Dictionary containing backend | STRING |
| HIGH | vllm/distributed/weight_transfer/base.py | 442 | Construct typed update info from dict with validation. Args: update_dict: Dictionary containing bac | STRING |
| HIGH | vllm/logging_utils/formatter.py | 24 | Shortens a file path for logging display: - Removes leading 'vllm' folder if present. - If path | STRING |
| HIGH | vllm/model_executor/kernels/linear/__init__.py | 645 | Choose a _KernelT that can implement the given config for the given compute capability. Attempts to choose the best | STRING |
| HIGH | vllm/model_executor/kernels/linear/__init__.py | 823 | Choose an MPLinearKernel that can implement the given config for the given compute capability. Attempts to choose t | STRING |
| HIGH | vllm/model_executor/layers/fused_moe/utils.py | 60 | Resolve whether AITER fused shared-expert execution is enabled. Args: quant_config: Model quantization conf | STRING |
| HIGH | vllm/model_executor/layers/fused_moe/moe_output.py | 45 | Normalize the two FlashInfer TRTLLM MoE return layouts. Args: flashinfer_output: Tensor returned by the Fla | STRING |
| HIGH | …/model_executor/layers/fused_moe/expert_map_manager.py | 339 | Map global expert ID to local expert ID. Args: global_id: Global expert ID (0 to global_num_experts | STRING |
| HIGH | …cutor/layers/fused_moe/experts/ocp_mx_emulation_moe.py | 40 | Activation dtype `moe_kernel_quantize_input` should fake-quantize to. Args: ocp_mx_scheme: The OCP MX schem | STRING |
| HIGH | vllm/model_executor/models/keye_vl1_5.py | 79 | Return num_patches per video. Args: grid_thw: Tensor with shape [N, 3] containing temporal, height, width | STRING |
| HIGH | vllm/model_executor/models/isaac.py | 245 | Apply pixel shuffle to a packed vision sequence without unpacking per image. Args: x (`torch.Tensor`): | STRING |
| HIGH | vllm/models/hy_v4/nvidia/attention.py | 61 | Return the backbone layers that reuse a previous layer's top-k indices. A "shared" indexer layer performs sparse at | STRING |
| HIGH | vllm/models/hy_v4/nvidia/flashmla_sparse.py | 124 | Return the sink laid out for the kernel's query head count. Args: q: Query tensor, before any head | STRING |
| HIGH | …ls/glm5next/amd/ops/third_party/kda/fused_recurrent.py | 593 | Args: q (torch.Tensor): queries of shape `[B, T, H, K]`. k (torch.Tensor): keys | STRING |
| HIGH | …glm5next/nvidia/ops/third_party/kda/fused_recurrent.py | 595 | Args: q (torch.Tensor): queries of shape `[B, T, H, K]`. k (torch.Tensor): keys | STRING |
| HIGH | vllm/reasoning/hy_v4_reasoning_parser.py | 36 | Detect the per-checkpoint suffix used by Hunyuan structural tokens. Args: tokenizer: Tokenizer of the serve | STRING |
| HIGH | vllm/multimodal/gpu_ipc_memory.py | 160 | Return KV-cache memory remaining after frontend multimodal reservations. The reservation covers: * The total ` | STRING |
| HIGH | vllm/multimodal/audio.py | 104 | Normalize audio to the specified format. This function handles channel reduction for multi-channel audio, suppo | STRING |
| HIGH | vllm/multimodal/audio.py | 407 | Split audio into chunks with intelligent split points. Splits long audio into smaller chunks at low-energy regions | STRING |
| HIGH | vllm/third_party/flash_linear_attention/ops/chunk.py | 153 | Args: q (torch.Tensor): Queries of shape `[B, T, H, K]`. k (torch.Tensor): | STRING |
| HIGH | …rd_party/flash_linear_attention/ops/fused_recurrent.py | 537 | Args: q (torch.Tensor): queries of shape `[B, T, H, K]`. k (torch.Tensor): | STRING |
| HIGH | vllm/benchmarks/lib/ready_checker.py | 25 | Wait for an endpoint to become available before starting benchmarks. Args: request_func: The async request | STRING |
| HIGH | vllm/entrypoints/speech_to_text/base/utils.py | 18 | Read an uploaded file enforcing a size limit *before* full materialization. The function first checks the Conte | STRING |
| HIGH | vllm/transformers_utils/processors/isaac.py | 200 | Convert normalized images into flattened ViT-style patches. Args: image (`torch.Tensor`): Tenso | STRING |
| HIGH | vllm/lora/resolver.py | 75 | Get a registered resolver instance by name. Args: resolver_name: Name of the resolver to get. | STRING |
| HIGH | …ications/chatbot/streamlit_openai_chatbot_webserver.py | 112 | Generate and stream LLM response with optional reasoning process. Args: messages (list): List of conversati | STRING |
| HIGH | benchmarks/benchmark_long_document_qa_throughput.py | 67 | Repeat each prompt in the list for a specified number of times. The order of prompts in the output list depends on t | STRING |
| HIGH | benchmarks/attention_benchmarks/batch_spec.py | 74 | Parse batch specification string into list of BatchRequest objects. Grammar: (<count>?) q<q_len>(k?) (s<seq_len>(k? | STRING |
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| CRITICAL | rust/src/cmd/src/cli/unsupported.rs | 141 | /// - `vllm.entrypoints.cli.serve.ServeSubcommand.subparser_init(...)` | COMMENT |
| CRITICAL | tests/v1/e2e/general/test_mamba_prefix_cache.py | 931 | assert engine.llm_engine.engine_core.engine_core.scheduler.reset_prefix_cache() | CODE |
| CRITICAL | tests/v1/e2e/general/test_mamba_prefix_cache.py | 1194 | engine.llm_engine.engine_core.engine_core.scheduler.reset_prefix_cache() | CODE |
| CRITICAL | tests/v1/e2e/general/test_mamba_prefix_cache.py | 1219 | assert engine.llm_engine.engine_core.engine_core.scheduler.reset_prefix_cache() | CODE |
| CRITICAL | tests/distributed/test_torchrun_example_moe.py | 81 | llm.llm_engine.model_executor.driver_worker.worker.model_runner.model.parameters() | CODE |
| CRITICAL | tests/distributed/test_torchrun_example.py | 72 | llm.llm_engine.model_executor.driver_worker.worker.model_runner.model.parameters() | CODE |
| CRITICAL | tests/models/language/generation/test_gemma.py | 231 | self.model_runner.model.language_model.model.normalizer.cpu().item() | CODE |
| CRITICAL | tests/models/language/generation/test_gemma.py | 237 | lambda self: self.model_runner.model.model.normalizer.cpu().item() | CODE |
| CRITICAL | vllm/v1/spec_decode/llm_base_proposer.py | 1452 | self.model.model.embed_tokens.weight.cpu(), | CODE |
| CRITICAL | vllm/third_party/flash_linear_attention/ops/utils.py | 170 | triton.runtime.driver.active.utils.get_device_properties(i)[ | CODE |
| CRITICAL | vllm/entrypoints/cli/main.py | 73 | vllm.entrypoints.cli.benchmark.main.maybe_exec_rust_bench() | CODE |
| CRITICAL | docs/training/layerwise.md | 124 | model = llm.llm_engine.engine_core.engine_core.model_executor.driver_worker.worker.get_model() | CODE |
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| MEDIUM⚡ | tests/v1/kv_offload/tiering/p2p/test_manager.py | 1394 | # _poll_once orchestration | COMMENT |
| MEDIUM | tests/v1/core/test_scheduler.py | 1297 | # happen when max_tokens is clamped to the context window after | COMMENT |
| MEDIUM | tests/v1/kv_connector/unit/test_offloading_connector.py | 167 | # Use a long prompt that fits within the model's context window. | COMMENT |
| MEDIUM | tests/v1/engine/test_async_llm.py | 1235 | # Wait for generation to start (event-driven) | COMMENT |
| MEDIUM | tests/v1/engine/test_async_llm.py | 1277 | # Wait for some tokens (event-driven, handles slow token generation) | COMMENT |
| MEDIUM | tests/v1/engine/test_async_llm.py | 1341 | # Wait for at least one token across any request (event-driven) | COMMENT |
| MEDIUM | tests/tool_parsers/test_lfm2_tool_parser.py | 281 | # Messy agentic shell commands: literal bracket inside a string argument | COMMENT |
| MEDIUM | tests/renderers/test_cohere.py | 438 | # chain-of-thought across the rendered prompt. | COMMENT |
| MEDIUM | tests/renderers/test_completions.py | 521 | # 50-token prompt with a 100-token context window and max_tokens=60. | COMMENT |
| MEDIUM | tests/kernels/ultraquant/test_seam_contract.py | 57 | # call the inherited orchestration makes stays valid for UltraQuant. | COMMENT |
| MEDIUM | tests/evals/mrcr/mrcr_eval.py | 32 | # Skip chain-of-thought on reasoning models; ignored by non-reasoning templates. | COMMENT |
| MEDIUM | tests/compile/fullgraph/test_basic_correctness.py | 39 | # embedding model | COMMENT |
| MEDIUM | tests/entrypoints/cohere/test_serving_conversion.py | 419 | # Cohere's ``tool_plan`` is the older chain-of-thought field; it | COMMENT |
| MEDIUM | tests/entrypoints/openai/responses/test_harmony.py | 603 | # NOTE: chain-of-thought should be removed. | COMMENT |
| MEDIUM | …entrypoints/openai/responses/test_serving_responses.py | 519 | # context window, so the prompt budget is the full max_model_len minus the | COMMENT |
| MEDIUM | vllm/_xpu_ops.py | 1115 | window_size=(-1, -1), # -1 means infinite context window | CODE |
| MEDIUM | vllm/v1/attention/ops/vit_attn_wrappers.py | 281 | # Without it, hallucinations occur with the backend | COMMENT |
| MEDIUM | vllm/v1/executor/abstract.py | 61 | uses_ray: bool = False # whether the executor uses Ray for orchestration. | CODE |
| MEDIUM | vllm/platforms/cpu.py | 168 | # the shared MLA scaffolding so that DeepSeek-style models can | COMMENT |
| MEDIUM⚡ | …ibuted/kv_transfer/kv_connector/v1/nixl/push_worker.py | 290 | # progress is event-driven (see module docstring). | COMMENT |
| MEDIUM | …d/kv_transfer/kv_connector/v1/mooncake/store/worker.py | 4 | # The transfer-thread scaffolding (KVTransferThread, KVCacheStoreSendingThread, | COMMENT |
| MEDIUM | vllm/model_executor/models/granite_speech.py | 81 | # work pretty well with zero shot. | COMMENT |
| MEDIUM | vllm/model_executor/models/phi3v.py | 582 | # initialized as an embedding model | COMMENT |
| MEDIUM | vllm/models/minimax_m3/amd/model.py | 1686 | # NVIDIA copy — it only orchestrates the shared vision tower + the per-platform | COMMENT |
| MEDIUM | vllm/reasoning/hy_v4_reasoning_parser.py | 262 | # chain-of-thought (and a stray ``</think:SUF>``) as content. | COMMENT |
| MEDIUM | vllm/reasoning/gemma4_utils.py | 27 | print(result["thinking"]) # chain-of-thought or None | STRING |
| MEDIUM | vllm/reasoning/gemma4_utils.py | 76 | >>> print(result["thinking"]) # chain-of-thought reasoning or None | STRING |
| MEDIUM | vllm/entrypoints/cohere/serving.py | 277 | # Controls how the assistant's chain-of-thought is surfaced on | COMMENT |
| MEDIUM | vllm/entrypoints/cohere/serving.py | 463 | # ``tool_plan`` is Cohere's chain-of-thought emitted alongside tool | COMMENT |
| MEDIUM | vllm/entrypoints/speech_to_text/base/serving.py | 407 | # which is a strong sign of hallucination in outputs. | COMMENT |
| MEDIUM | vllm/vllm_flash_attn/flash_attn_interface.py | 138 | window_size=(-1, -1), # -1 means infinite context window | CODE |
| MEDIUM | docs/serving/integrations/claude_code.md | 3 | [Claude Code](https://code.claude.com/docs/en/quickstart) is Anthropic's official agentic coding tool that lives in your | CODE |
| MEDIUM | docs/serving/integrations/codex.md | 3 | [Codex](https://github.com/openai/codex) is OpenAI's official agentic coding tool that lives in your terminal. It can un | CODE |
| MEDIUM | docs/features/tool_calling.md | 266 | The tool calling that is supported is the [JSON-based tool calling](https://llama.meta.com/docs/model-cards-and-prompt-f | CODE |
| MEDIUM | docs/deployment/frameworks/dify.md | 3 | [Dify](https://github.com/langgenius/dify) is an open-source LLM app development platform. Its intuitive interface combi | CODE |
| MEDIUM | .buildkite/scripts/hardware_ci/run-amd-test.sh | 1535 | # Keep AMD CI orchestration variables out of vLLM's runtime environment. | COMMENT |
| MEDIUM | .buildkite/amd-disagg/vllm_disagg.sh | 88 | # node-mode orchestration knobs | COMMENT |
| MEDIUM⚡ | .buildkite/amd-disagg/vllm_disagg.sh | 414 | # node mode — SLURM-native rank-based self-select + rank-0 orchestration | COMMENT |
| MEDIUM | .buildkite/amd-disagg/vllm_disagg.sh | 600 | # shell's stdout stays free for orchestration logs the srun/CI captures). | COMMENT |
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| LOW | tests/v1/attention/test_mla_backends.py | 1986 | ["placeholder"], | CODE |
| LOW | tests/v1/attention/test_dspark_noncausal_sparse_mla.py | 334 | vllm_config.compilation_config.static_forward_context["placeholder"] = ( | CODE |
| LOW | tests/v1/attention/test_dspark_noncausal_sparse_mla.py | 339 | kv_cache_spec, ["placeholder"], vllm_config, device | CODE |
| LOW | tests/v1/attention/test_sparse_mla_backends.py | 739 | vllm_config.compilation_config.static_forward_context["placeholder"] = ( | CODE |
| LOW | tests/v1/attention/test_sparse_mla_backends.py | 744 | builder = builder_cls(kv_cache_spec, ["placeholder"], vllm_config, device) | CODE |
| LOW | tests/v1/attention/test_sparse_mla_backends.py | 1581 | vllm_config.compilation_config.static_forward_context["placeholder"] = ( | CODE |
| LOW | tests/v1/attention/test_sparse_mla_backends.py | 1586 | builder = builder_cls(kv_cache_spec, ["placeholder"], vllm_config, device) | CODE |
| LOW | tests/v1/attention/test_sparse_mla_backends.py | 3059 | vllm_config.compilation_config.static_forward_context["placeholder"] = ( | CODE |
| LOW | tests/v1/attention/test_sparse_mla_backends.py | 3064 | builder = builder_cls(kv_cache_spec, ["placeholder"], vllm_config, device) | CODE |
| LOW | tests/v1/attention/test_sparse_mla_backends.py | 4017 | vllm_config.compilation_config.static_forward_context["placeholder"] = ( | CODE |
| LOW | tests/v1/attention/test_sparse_mla_backends.py | 4048 | builder_cls(kv_cache_spec, ["placeholder"], vllm_config, device) | CODE |
| LOW | tests/v1/attention/test_sparse_mla_backends.py | 4050 | builder = builder_cls(kv_cache_spec, ["placeholder"], vllm_config, device) | CODE |
| LOW | tests/v1/attention/test_attention_backends.py | 698 | ["placeholder"], | CODE |
| LOW | tests/v1/attention/test_attention_backends.py | 1615 | "placeholder": PerLayerParameters( | CODE |
| LOW | tests/v1/attention/test_attention_backends.py | 1663 | kv_cache_spec, ["placeholder"], vllm_config, device | CODE |
| LOW | tests/v1/attention/test_attention_backends.py | 1769 | ["placeholder"], | CODE |
| LOW⚡ | tests/v1/kv_connector/unit/test_nixl_push_connector.py | 1260 | payload = _registration_data("placeholder") | CODE |
| LOW | tests/tool_parsers/test_kimi_k2_tool_parser.py | 85 | '{"to": "user@example.com", "subject": "Daily Update"}', | CODE |
| LOW | tests/tool_parsers/test_kimi_k2_tool_parser.py | 92 | {"to": "user@example.com", "subject": "Daily Update"}, | CODE |
| LOW⚡ | tests/tool_parsers/test_olmo3_tool_parser.py | 23 | "register_user(name='John Doe', " | CODE |
| LOW⚡ | tests/tool_parsers/test_olmo3_tool_parser.py | 31 | "register_user(name='John Doe', " | CODE |
| LOW⚡ | tests/tool_parsers/test_olmo3_tool_parser.py | 40 | arguments='{"name": "John Doe", ' | CODE |
| LOW⚡ | tests/tool_parsers/test_pythonic_tool_parser.py | 23 | "register_user(name='John Doe', " | CODE |
| LOW⚡ | tests/tool_parsers/test_pythonic_tool_parser.py | 32 | arguments='{"name": "John Doe", ' | CODE |
| LOW⚡ | tests/tool_parsers/test_lfm2_tool_parser.py | 26 | "register_user(name='John Doe', " | CODE |
| LOW⚡ | tests/tool_parsers/test_lfm2_tool_parser.py | 35 | arguments='{"name": "John Doe", ' | CODE |
| LOW | tests/tool_parsers/test_lfm2_tool_parser.py | 537 | "deliveryAddress='123 Main St')]" | CODE |
| LOW⚡ | tests/tool_parsers/test_hunyuan_a13b_tool_parser.py | 45 | '<tool_calls>[{"name": "get_weather", "arguments": {"city": "San Francisco", "metric": "celsius"}}, {"name": | CODE |
| LOW⚡ | tests/tool_parsers/test_hunyuan_a13b_tool_parser.py | 53 | "name": "John Doe", | CODE |
| LOW | tests/kernels/helion/test_pattern_matching.py | 126 | input_node = next(n for n in gm.graph.nodes if n.op == "placeholder") | CODE |
| LOW | tests/kernels/helion/test_pattern_matching.py | 201 | input_node = next(n for n in gm.graph.nodes if n.op == "placeholder") | CODE |
| LOW | …ls/attention/test_rocm_aiter_mla_causal_verify_mask.py | 129 | layer_name = "placeholder" | CODE |
| LOW | …rnels/attention/test_rocm_aiter_mla_decode_metadata.py | 118 | layer_name = "placeholder" | CODE |
| LOW⚡ | …ts/models/multimodal/processing/test_audioflamingo3.py | 131 | dummy_data = builder.get_dummy_mm_data(100, mm_counts, MultiModalDummyOptions()) | CODE |
| LOW⚡ | …ts/models/multimodal/processing/test_audioflamingo3.py | 133 | assert "audio" in dummy_data | CODE |
| LOW⚡ | …ts/models/multimodal/processing/test_audioflamingo3.py | 134 | assert len(dummy_data["audio"]) == 2 | CODE |
| LOW⚡ | …ts/models/multimodal/processing/test_audioflamingo3.py | 137 | assert len(dummy_data["audio"][0]) == expected_len | CODE |
| LOW | tests/parser/mistral/test_tool_calls.py | 407 | "name": "John Doe", | CODE |
| LOW | tests/parser/mistral/test_tool_calls.py | 1115 | "name": "John Doe", | CODE |
| LOW | tests/parser/mistral/test_tool_calls.py | 1577 | "name": "John Doe", | CODE |
| LOW | tests/parser/mistral/test_tool_calls.py | 400 | """[TOOL_CALLS] [{"arguments":{"name": "John Doe"}, "name": "get_age"}]""", # noqa: E501 | STRING |
| LOW | tests/parser/mistral/test_tool_calls.py | 1108 | """[TOOL_CALLS] [{"arguments": {"name": "John Doe"}, "name": "get_age"}]""", # noqa: E501 | STRING |
| LOW | tests/parser/mistral/test_tool_calls.py | 1570 | """[TOOL_CALLS] [{"arguments": {"name": "John Doe"}, "name": "get_age"}]""", # noqa: E501 | STRING |
| LOW | tests/compile/test_graph_partition.py | 63 | and node.args[0].op == "placeholder" | CODE |
| LOW | tests/compile/test_graph_partition.py | 123 | and node.args[0].op == "placeholder" | CODE |
| LOW | tests/compile/test_graph_partition.py | 188 | assert [node.op for node in splitting_gm.graph.nodes] == ["placeholder"] + 2 * [ | CODE |
| LOW | tests/compile/test_graph_partition.py | 430 | if node.op == "placeholder" | CODE |
| LOW | tests/compile/test_graph_partition.py | 495 | if n.op == "placeholder" | CODE |
| LOW | tests/compile/test_graph_partition.py | 501 | if n.op == "placeholder" | CODE |
| LOW | tests/compile/test_graph_partition.py | 608 | if n.op != "placeholder": | CODE |
| LOW⚡ | tests/compile/passes/ir/test_clone_cleanup.py | 224 | x_node = [n for n in graph_module.graph.nodes if n.op == "placeholder"][0] | CODE |
| LOW⚡ | tests/compile/passes/ir/test_clone_cleanup.py | 236 | placeholders = [n for n in graph_module.graph.nodes if n.op == "placeholder"] | CODE |
| LOW | tests/compile/passes/ir/test_clone_cleanup.py | 254 | placeholders = [n for n in graph_module.graph.nodes if n.op == "placeholder"] | CODE |
| LOW⚡ | tests/compile/passes/ir/test_clone_cleanup.py | 273 | placeholders = [n for n in graph_module.graph.nodes if n.op == "placeholder"] | CODE |
| LOW⚡ | tests/compile/passes/ir/test_clone_cleanup.py | 291 | x_node = [n for n in graph_module.graph.nodes if n.op == "placeholder"][0] | CODE |
| LOW⚡ | tests/compile/passes/ir/test_clone_cleanup.py | 310 | x_node = [n for n in graph_module.graph.nodes if n.op == "placeholder"][0] | CODE |
| LOW⚡ | tests/benchmarks/test_txt_slices_dataset.py | 20 | Lorem ipsum dolor sit amet, consectetur adipiscing elit, sed do eiusmod tempor | CODE |
| LOW⚡ | tests/benchmarks/test_txt_slices_dataset.py | 20 | Lorem ipsum dolor sit amet, consectetur adipiscing elit, sed do eiusmod tempor | CODE |
| LOW⚡ | tests/entrypoints/launchers/test_run_batch.py | 835 | def _make_aiohttp_mocks(response_data: bytes = b"fake-data", status: int = 200): | CODE |
| LOW | vllm/model_executor/models/transformers/fx_utils.py | 126 | if kind == "placeholder": | CODE |
| 21 more matches not shown… | ||||
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| LOW | tools/setup_deepgemm_pythons.sh | 6 | # Usage: | COMMENT |
| LOW | tools/vllm-rocm/generate-rocm-wheels-root-index.sh | 11 | # Usage: | COMMENT |
| LOW | docker/docker-bake.hcl | 5 | # Usage: | COMMENT |
| LOW | docker/docker-bake-rocm.hcl | 6 | # Usage: | COMMENT |
| LOW | docker/docker-bake-rocm.hcl | 131 | # Usage: | COMMENT |
| LOW | docker/entrypoints/test_vllm_nonroot_entrypoint.sh | 9 | # Usage: | COMMENT |
| LOW | …/nixl_integration/run_multi_connector_accuracy_test.sh | 15 | # Usage: | COMMENT |
| LOW | …nector/nixl_integration/spec_decode_acceptance_test.sh | 13 | # Usage: | COMMENT |
| LOW | …nixl_integration/run_multi_connector_edge_case_test.sh | 15 | # Usage: | COMMENT |
| LOW | …/scale_out/ec_integration/run_scale_out_ec_e2e_test.sh | 22 | # Usage: | COMMENT |
| LOW | examples/tool_calling/chat_with_tools_offline.py | 44 | # Usage: | COMMENT |
| LOW | examples/ray_serving/run_cluster.sh | 8 | # Usage: | COMMENT |
| LOW | examples/ray_serving/multi-node-serving.sh | 11 | # Example usage: | COMMENT |
| LOW | examples/generate/multimodal/mistral-small_offline.py | 51 | # Usage: | COMMENT |
| LOW | benchmarks/kv_cache_watermark.sh | 29 | # Usage: | COMMENT |
| LOW | …n_benchmarks/configs/mla_sparse_masked_mha_vs_mqa.yaml | 9 | # Usage: | COMMENT |
| LOW | …ttention_benchmarks/configs/mla_sparse_mha_vs_mqa.yaml | 3 | # Usage: | COMMENT |
| LOW | …s/attention_benchmarks/configs/mla_fa4_fp8_output.yaml | 6 | # Usage: | COMMENT |
| LOW | …nchmarks/attention_benchmarks/configs/mla_prefill.yaml | 14 | # Usage: | COMMENT |
| LOW | .buildkite/scripts/cache-rocm-base-wheels.sh | 10 | # Usage: | COMMENT |
| LOW | .buildkite/scripts/ci-bake-rocm.sh | 9 | # Usage: | COMMENT |
| LOW | .buildkite/scripts/ci-fetch-log.sh | 4 | # Usage: | COMMENT |
| LOW | .buildkite/scripts/tool_call/run-bfcl-eval.sh | 5 | # Usage: | COMMENT |
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| HIGH | docs/deployment/frameworks/crusoe.md | 13 | export CRUSOE_API_KEY="your-api-key" | CODE |
| HIGH | …les/pooling/embed/openai_embedding_long_text/client.py | 21 | --api-key your-api-key | STRING |
| HIGH | …les/pooling/embed/openai_embedding_long_text/client.py | 31 | --api-key your-api-key | STRING |
| HIGH | …les/pooling/embed/openai_embedding_long_text/client.py | 43 | API_KEY = "your-api-key" # Replace with your actual API key | CODE |
| HIGH | …es/pooling/embed/openai_embedding_long_text/service.sh | 19 | API_KEY=${API_KEY:-"your-api-key"} | CODE |
| 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 | csrc/cpu/sgl-kernels/common.h | 426 | // TODO: implement reverse order of [MB / cache_blocks_mb, NB, cache_blocks_mb] | COMMENT |
| LOW | …nector/v1/lmcache_integration/multi_process_adapter.py | 633 | # TODO: add error handling here | COMMENT |
| LOW | …m/model_executor/kernels/linear/mixed_precision/xpu.py | 178 | # TODO: implement asym case | COMMENT |
| LOW | vllm/entrypoints/openai/responses/protocol.py | 853 | # TODO: implement the other reason for incomplete_details, | COMMENT |
| LOW | vllm/lora/punica_wrapper/punica_base.py | 315 | # TODO: implement it based on torch ops | COMMENT |
| LOW | vllm/lora/punica_wrapper/punica_base.py | 347 | # TODO: implement it based on torch ops | COMMENT |
| LOW | vllm/lora/punica_wrapper/punica_base.py | 371 | # TODO: implement it based on torch ops | COMMENT |
| LOW | vllm/lora/punica_wrapper/punica_base.py | 408 | # TODO: implement it based on torch ops | COMMENT |
| LOW | vllm/lora/punica_wrapper/punica_base.py | 438 | # TODO: implement it based on torch ops | COMMENT |
| LOW | vllm/lora/punica_wrapper/punica_base.py | 456 | # TODO: implement it based on torch ops | COMMENT |
| LOW | vllm/lora/punica_wrapper/punica_base.py | 482 | # TODO: implement it based on torch ops | COMMENT |
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| MEDIUM | tests/models/multimodal/generation/test_common.py | 61 | # model arch happens to be a substring of another one, you can add a | COMMENT |
| MEDIUM | tests/models/multimodal/generation/test_common.py | 82 | # NOTE you can add --collect-only to any of the above commands to see | COMMENT |
| LOW | …ibuted/kv_transfer/kv_connector/v1/nixl/pull_worker.py | 93 | # while processing the next batch, we make sure to only set an | COMMENT |
| MEDIUM | vllm/model_executor/models/interfaces.py | 312 | as a language model component. | STRING |
| MEDIUM | …ache/disagg_prefill_lmcache_v1/disagg_vllm_launcher.sh | 24 | # secure random value. This is set to a fixed value for demonstration purposes only. | COMMENT |
| MEDIUM | .buildkite/test-amd.yaml | 1 | # In this file, you can add more tests to run either by adding a new step or | COMMENT |
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| HIGH | tests/v1/sample/test_logprobs.py | 1060 | # Based on user's example: "In this example," | COMMENT |
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| HIGH | …tic_prefix_caching/automatic_prefix_caching_offline.py | 25 | # A prompt containing a large markdown table. The table is randomly generated by GPT-4. | COMMENT |
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| MEDIUM | tests/v1/streaming_input/test_async_llm_streaming.py | 190 | CODE | |
| MEDIUM | tests/v1/e2e/general/test_streaming_input.py | 502 | CODE |
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| LOW⚡ | …gated/disaggregated_serving/moriio_toy_proxy_server.py | 277 | async def handle_request(api: str, request: Request): | CODE |