radixark/miles

33.9

Adjusted Score

33.9

Raw Score

100%

Time Factor

2026-08-09

Last Push

1.9K

Stars

Python

Language

220.8K

Lines of Code

1.4K

Files

5.1K

Pattern Hits

2026-08-09

Scan Date

0.05

HC Hit Rate

What These Metrics Mean

Adjusted Score: Primary synthetic code indicator. Raw score normalised per 1,000 lines of code and multiplied by the temporal discount factor. This is the definitive comparative metric — use it to rank repositories by AI authorship density.
Raw Score: The unmodified sum of all severity-weighted, context-multiplied pattern match scores before temporal discounting. Reflects the absolute signal strength independent of when the repository was last active.
Time Factor: The temporal discount multiplier (0–100%) applied to the raw score. Repositories last updated before ChatGPT's launch (Nov 2022) receive a 5% factor. Full signal is only assigned to repositories active in the post-adoption era (Jan 2024+).
Pattern Hits: Total count of individual pattern matches across all files and categories. A high hit count with a low score may indicate a very large codebase with isolated AI snippets; a low count with a high score indicates dense, concentrated AI signatures.
HC Hit Rate: High+Critical pattern hits per file, averaged across the repository. This orthogonal signal catches repositories where a few files are densely packed with high-severity AI tells — a strong indicator even when the normalised score appears moderate due to codebase size.
Lines of Code / Files: Total lines and files analysed. The scanner examines 94 file extensions. These denominators are used to normalise the score, enabling fair comparison between repositories of vastly different sizes.

Score History

Longitudinal tracking requires multiple scan runs. Once this repository is re-scanned after new commits land, this chart will visualise how the synthetic code signal evolves over time — enabling you to detect whether AI authorship is growing, stabilising, or being actively corrected by human engineers.

No multi-scan history yet — run the scanner again to build trend data.

Severity Breakdown

Classifies detected patterns by their diagnostic confidence and structural impact. CRITICAL patterns (coefficient 10) represent definitive synthetic signatures — hallucinated imports, explicit LLM attribution metadata — virtually never produced by human authors. HIGH (5) indicates strong structural tells such as cross-file repetition or cross-linguistic idioms. MEDIUM (2) covers recognisable conversational padding and AI-specific vocabulary. LOW (1) captures subtle indicators like tautological comments and generic boilerplate that require density to carry independent signal.

CRITICAL 6HIGH 68MEDIUM 705LOW 4323

Directory Score Breakdown

This horizontal bar chart decomposes the repository's raw synthetic code score by top-level directory, allowing you to pinpoint precisely which modules or components carry the highest AI authorship density. Directories with disproportionately high scores relative to their size warrant targeted manual review: concentrated AI signatures often trace back to mass-generated configuration layers, auto-ported test suites, LLM-scaffolded boilerplate classes, or entire subsystems authored under heavy copilot assistance. Use this view to prioritise your human code-review effort.

Pattern Findings

The scanner identified 5102 distinct pattern matches across 22 syntactic categories. Each entry below represents a discrete location in the source code where the engine recorded a statistically significant AI authorship indicator. Expand any category row to inspect the individual file paths, line numbers, code snippets, and the lexical context (CODE, COMMENT, or STRING) in which each match was detected.

Reading the findings table: The Severity column indicates the diagnostic confidence level (CRITICAL / HIGH / MEDIUM / LOW). The Context column identifies whether the match occurred inside executable code, an inline comment, or a string literal — comment-context matches receive a ×1.5 weight because LLMs systematically over-annotate. The ⚡ bolt icon marks clustered matches: three or more patterns within a 10-line window, each receiving an additional ×1.5 density multiplier as dense clusters constitute far stronger evidence of synthetic authorship than isolated hits.

Hyper-Verbose Identifiers3092 hits · 3128 pts

Severity	File	Line	Snippet	Context
LOW	tools/convert_hf_to_nvfp4.py	60	def _is_moe_expert_weight_name(name: str) -> bool:	CODE
LOW	tools/convert_hf_to_nvfp4.py	157	def _update_quantization_config(cfg: dict, ignore_list: list[str]) -> None:	CODE
LOW	tools/convert_hf_to_nvfp4.py	231	def _collect_gated_pair_locations(	CODE
LOW	tools/convert_torch_dist_to_hf_ray.py	375	def compute_dcp_load_accounting(storage_data: dict[Any, Any], plan: LoadPlan) -> tuple[int, int, int]:	STRING
LOW	tools/convert_torch_dist_to_hf_ray.py	385	def prepare_cached_metadata_for_reader(	STRING
LOW	tools/convert_torch_dist_to_hf_ray.py	475	def is_supported_moe_read_item(read_item: Any, source_key: str, tensor_size: torch.Size) -> bool:	STRING
LOW	tools/convert_torch_dist_to_hf_ray.py	501	def create_full_tensor_read_items(source_key: str, metadata: dist_cp.metadata.Metadata) -> list[Any]:	STRING
LOW	tools/convert_torch_dist_to_hf_ray.py	508	def create_direct_moe_read_items(block: MoeBlockSpec, md: dist_cp.metadata.TensorStorageMetadata) -> list[ReadItem]:	STRING
LOW	tools/convert_torch_dist_to_hf_ray.py	573	def converted_moe_tensors_from_chunk(	STRING
LOW	tools/convert_torch_dist_to_hf_ray.py	693	def tensor_metadata_from_checkpoint_metadata(	STRING
LOW	tools/convert_torch_dist_to_hf_ray.py	772	def collect_moe_blocks_by_file(	STRING
LOW	tools/convert_torch_dist_to_hf_ray.py	958	def prepare_moe_block_task_tensors(	STRING
LOW	tools/convert_torch_dist_to_hf_ray.py	976	def prepare_whole_source_task_tensors(	STRING
LOW	tools/convert_torch_dist_to_hf_ray.py	1001	def write_prepared_tensor_groups(	STRING
LOW	tools/convert_torch_dist_to_hf_ray.py	1049	def initialize_worker_cuda_device(actor_id: int, quantization_config: dict[str, Any] \| None) -> int \| None:	STRING
LOW	tools/convert_torch_dist_to_hf_ray.py	1295	def convert_torch_dist_to_hf_ray(args: Args) -> str:	STRING
LOW	tools/convert_hf_to_int4_direct.py	76	def round_to_quantized_type_dtype(	CODE
LOW	…ch/latest/sglang_attn_bridge/hf_sglang_triton_patch.py	117	def apply_sglang_triton_attention_patch(model):	CODE
LOW	…amd_patch/latest/sglang_attn_bridge/triton_attn_bwd.py	339	def triton_attention_backward(q, k, v, o, do, B, S, sm_scale=None):	CODE
LOW	miles_plugins/mbridge/deepseek_v32.py	49	def _hf_config_with_rope_fields(self):	CODE
LOW	miles_plugins/mbridge/deepseek_v32.py	56	def _using_hf_config_with_rope_fields(self):	CODE
LOW	miles_plugins/mbridge/mimo.py	49	def _weight_name_mapping_mcore_to_hf(self, mcore_weights_name: str) -> list[str]:	CODE
LOW	miles_plugins/mbridge/deepseekv4.py	73	def _weight_name_mapping_mcore_to_hf(self, mcore_weights_name: str) -> list[str]:	CODE
LOW	miles_plugins/mbridge/inkling.py	80	def _adjust_mapping_for_shared_weights(self):	CODE
LOW	miles_plugins/mbridge/inkling.py	83	def _get_hf_shared_weight_keys(self):	CODE
LOW⚡	miles_plugins/mbridge/qwen3_5.py	154	def _adjust_mapping_for_shared_weights(self):	CODE
LOW⚡	miles_plugins/mbridge/qwen3_5.py	160	def _get_hf_shared_weight_keys(self) -> list[str]:	CODE
LOW⚡	miles_plugins/mbridge/qwen3_5.py	165	def _supports_transformer_config_kwarg(self, kwarg_name: str) -> bool:	CODE
LOW	miles_plugins/mbridge/qwen3_5.py	181	def _get_transformer_layer_spec(self, vp_stage=None):	CODE
LOW	miles_plugins/mbridge/qwen3_5.py	267	def _weight_name_mapping_mtp_mlp(self, name: str) -> list[str]:	CODE
LOW	miles_plugins/mbridge/qwen3_5.py	286	def _weight_name_mapping_mcore_to_hf(self, mcore_weights_name: str) -> list[str]:	CODE
LOW	miles_plugins/mbridge/glm4.py	64	def _get_transformer_layer_spec(self):	CODE
LOW	miles_plugins/mbridge/glm4.py	83	def _weight_name_mapping_mcore_to_hf(self, mcore_weights_name: str) -> list[str]:	CODE
LOW	miles_plugins/mbridge/glm4moe.py	60	def _weight_name_mapping_mcore_to_hf(self, mcore_weights_name: str) -> list[str]:	CODE
LOW	miles_plugins/mbridge/qwen3_next.py	108	def _weight_name_mapping_mcore_to_hf(self, mcore_weights_name: str) -> list[str]:	CODE
LOW	miles_plugins/megatron_bridge/nemotron_h.py	151	def _install_nemotronh_hybrid_layer_shims() -> None:	CODE
LOW	miles_plugins/megatron_bridge/nemotron_h.py	184	def _install_mamba_model_loss_mask_shim() -> None:	CODE
LOW	miles_plugins/megatron_bridge/__init__.py	23	def _install_bridge_pp_group_unwrap() -> None:	CODE
LOW	miles_plugins/megatron_bridge/__init__.py	43	def broadcast_obj_from_pp_rank(self, obj, name=None):	CODE
LOW	miles_plugins/models/dsa_topk.py	47	def _flashinfer_tie_break_value() -> int:	CODE
LOW	miles_plugins/models/qwen_gdn_backend.py	7	def get_chunk_gated_delta_rule(backend: str):	CODE
LOW	miles_plugins/models/qwen3_vl.py	30	def _patch_allgather_vision_embeddings_kwarg() -> None:	CODE
LOW	miles_plugins/models/qwen3_vl.py	76	def _patch_model_forward_and_rope_index() -> None:	CODE
LOW	miles_plugins/models/qwen3_vl.py	134	def _patch_preprocess_packed_seqs_identity(model_mod) -> None:	CODE
LOW	miles_plugins/models/qwen3_vl.py	182	def _prepare_cp_local_context(parsed):	CODE
LOW	miles_plugins/models/hf_attention.py	33	def _zigzag_to_packed_shard_impl(hidden_states, cu_seqlens, cp_group, cp_rank, cp_size):	CODE
LOW	miles_plugins/models/hf_attention.py	59	def _packed_shard_to_zigzag_impl(hidden_states, cu_seqlens, cp_group, cp_rank, cp_size):	CODE
LOW	miles_plugins/models/qwen3_next.py	80	def fix_query_key_value_ordering(self, mixed_qkvz, mixed_ba):	CODE
LOW	miles_plugins/models/cp_utils.py	11	def detect_and_setup_hybrid_cp(model: nn.Module, cp_group: dist.ProcessGroup, cp_rank: int, cp_world_size: int) -> int:	CODE
LOW	miles_plugins/models/glm5/glm5.py	513	def get_absorb_query_key_value_tensors(	CODE
LOW	miles_plugins/models/glm5/glm5.py	693	def get_query_key_value_tensors(self):	CODE
LOW	miles_plugins/models/glm5/glm5.py	716	def set_for_recompute_input_layernorm(self):	CODE
LOW	miles_plugins/models/glm5/ops/indexer.py	10	def pytorch_extract_topk_scores(logits, topk_indices, dim=-1):	CODE
LOW	miles_plugins/models/glm5/ops/indexer.py	65	def generate_varlen_mask_params(cu_seqlens):	CODE
LOW	miles_plugins/models/inkling/mm_processor.py	117	def _reasoning_effort_from_env() -> float \| None:	CODE
LOW	miles_plugins/models/inkling/mm_processor.py	126	def render_inkling_messages_to_ids(	CODE
LOW	miles_plugins/models/inkling/__init__.py	13	def inkling_mm_model_provider(args, *kwargs):	CODE
LOW	miles_plugins/models/inkling/model.py	151	def get_inkling_dense_layer_spec(config) -> ModuleSpec:	CODE
LOW	miles_plugins/models/inkling/model.py	300	def inkling_mm_model_provider(pre_process=True, post_process=True, vp_stage=None):	CODE
LOW	miles_plugins/models/inkling/ops.py	486	def get_inkling_relative_attention_score_mod(rel_extent: int) -> Callable:	CODE
3032 more matches not shown…

Decorative Section Separators636 hits · 2272 pts

Severity	File	Line	Snippet	Context
MEDIUM	miles_plugins/models/glm5/glm5.py	218	# =====================	COMMENT
MEDIUM	miles_plugins/models/glm5/glm5.py	220	# =====================	COMMENT
MEDIUM	miles_plugins/models/glm5/glm5.py	317	# =================	COMMENT
MEDIUM	miles_plugins/models/glm5/glm5.py	319	# =================	COMMENT
MEDIUM⚡	miles_plugins/models/glm5/glm5.py	531	# =========================================	COMMENT
MEDIUM⚡	miles_plugins/models/glm5/glm5.py	533	# =========================================	COMMENT
MEDIUM⚡	miles_plugins/models/glm5/glm5.py	543	# =========================================	COMMENT
MEDIUM⚡	miles_plugins/models/glm5/glm5.py	545	# =========================================	COMMENT
MEDIUM	miles_plugins/models/glm5/glm5.py	559	# =========================================	COMMENT
MEDIUM	miles_plugins/models/glm5/glm5.py	561	# =========================================	COMMENT
MEDIUM	miles_plugins/models/glm5/glm5.py	626	# =========================================	COMMENT
MEDIUM	miles_plugins/models/glm5/glm5.py	628	# =========================================	COMMENT
MEDIUM	miles_plugins/models/deepseek_v4/ops/v4_indexer.py	85	# =========================================	COMMENT
MEDIUM	miles_plugins/models/deepseek_v4/ops/v4_indexer.py	87	# =========================================	COMMENT
MEDIUM	tests/test_attention_output_gate_tp.py	17	# ---------------------------------------------------------------------------	COMMENT
MEDIUM	tests/test_attention_output_gate_tp.py	19	# ---------------------------------------------------------------------------	COMMENT
MEDIUM⚡	tests/test_attention_output_gate_tp.py	84	# ---------------------------------------------------------------------------	COMMENT
MEDIUM⚡	tests/test_attention_output_gate_tp.py	86	# ---------------------------------------------------------------------------	COMMENT
MEDIUM⚡	tests/test_attention_output_gate_tp.py	121	# ---------------------------------------------------------------------------	COMMENT
MEDIUM⚡	tests/test_attention_output_gate_tp.py	123	# ---------------------------------------------------------------------------	COMMENT
MEDIUM⚡	tests/test_attention_output_gate_tp.py	170	# ---------------------------------------------------------------------------	COMMENT
MEDIUM⚡	tests/test_attention_output_gate_tp.py	172	# ---------------------------------------------------------------------------	COMMENT
MEDIUM	tests/test_fused_experts_backward.py	12	# ============================================================================	COMMENT
MEDIUM	tests/test_fused_experts_backward.py	14	# ============================================================================	COMMENT
MEDIUM⚡	tests/test_fused_experts_backward.py	259	# ============================================================================	COMMENT
MEDIUM⚡	tests/test_fused_experts_backward.py	261	# ============================================================================	COMMENT
MEDIUM⚡	tests/test_fused_experts_backward.py	266	# ============================================================================	COMMENT
MEDIUM⚡	tests/test_fused_experts_backward.py	268	# ============================================================================	COMMENT
MEDIUM	tests/test_fused_experts_backward.py	313	# ============================================================================	COMMENT
MEDIUM	tests/test_fused_experts_backward.py	315	# ============================================================================	COMMENT
MEDIUM	tests/test_fused_experts_backward.py	531	# ============================================================================	COMMENT
MEDIUM	tests/test_fused_experts_backward.py	533	# ============================================================================	COMMENT
MEDIUM⚡	tests/ci/test/test_log_groups.py	64	# ---------------------------------------------------------------------------	COMMENT
MEDIUM⚡	tests/ci/test/test_log_groups.py	66	# ---------------------------------------------------------------------------	COMMENT
MEDIUM⚡	tests/ci/test/test_log_groups.py	107	# ---------------------------------------------------------------------------	COMMENT
MEDIUM⚡	tests/ci/test/test_log_groups.py	109	# ---------------------------------------------------------------------------	COMMENT
MEDIUM⚡	tests/ci/test/test_log_groups.py	144	# ---------------------------------------------------------------------------	COMMENT
MEDIUM⚡	tests/ci/test/test_log_groups.py	146	# ---------------------------------------------------------------------------	COMMENT
MEDIUM⚡	tests/ci/test/test_log_groups.py	174	# ---------------------------------------------------------------------------	COMMENT
MEDIUM⚡	tests/ci/test/test_log_groups.py	176	# ---------------------------------------------------------------------------	COMMENT
MEDIUM⚡	tests/ci/test/test_log_groups.py	212	# ---------------------------------------------------------------------------	COMMENT
MEDIUM⚡	tests/ci/test/test_log_groups.py	214	# ---------------------------------------------------------------------------	COMMENT
MEDIUM⚡	tests/ci/test/test_log_groups.py	334	# ---------------------------------------------------------------------------	COMMENT
MEDIUM⚡	tests/ci/test/test_log_groups.py	336	# ---------------------------------------------------------------------------	COMMENT
MEDIUM⚡	tests/ci/test/test_log_groups.py	380	# ---------------------------------------------------------------------------	COMMENT
MEDIUM⚡	tests/ci/test/test_log_groups.py	382	# ---------------------------------------------------------------------------	COMMENT
MEDIUM	tests/ci/test/test_log_groups.py	30	# ---------------------------------------------------------------------------	COMMENT
MEDIUM	tests/ci/test/test_log_groups.py	32	# ---------------------------------------------------------------------------	COMMENT
MEDIUM⚡	tests/ci/test/test_log_groups.py	448	# ---------------------------------------------------------------------------	STRING
MEDIUM⚡	tests/ci/test/test_log_groups.py	450	# ---------------------------------------------------------------------------	STRING
MEDIUM⚡	tests/fast/backends/megatron_utils/test_lora_utils.py	24	# ---------------------------------------------------------------------------	COMMENT
MEDIUM⚡	tests/fast/backends/megatron_utils/test_lora_utils.py	26	# ---------------------------------------------------------------------------	COMMENT
MEDIUM⚡	tests/fast/backends/megatron_utils/test_lora_utils.py	46	# ---------------------------------------------------------------------------	COMMENT
MEDIUM⚡	tests/fast/backends/megatron_utils/test_lora_utils.py	48	# ---------------------------------------------------------------------------	COMMENT
MEDIUM⚡	tests/fast/backends/megatron_utils/test_lora_utils.py	133	# ---------------------------------------------------------------------------	COMMENT
MEDIUM⚡	tests/fast/backends/megatron_utils/test_lora_utils.py	135	# ---------------------------------------------------------------------------	COMMENT
MEDIUM⚡	tests/fast/backends/megatron_utils/test_lora_utils.py	192	# ---------------------------------------------------------------------------	COMMENT
MEDIUM⚡	tests/fast/backends/megatron_utils/test_lora_utils.py	194	# ---------------------------------------------------------------------------	COMMENT
MEDIUM⚡	tests/fast/backends/megatron_utils/test_lora_utils.py	219	# ---------------------------------------------------------------------------	COMMENT
MEDIUM⚡	tests/fast/backends/megatron_utils/test_lora_utils.py	221	# ---------------------------------------------------------------------------	COMMENT
576 more matches not shown…

Unused Imports273 hits · 264 pts

Severity	File	Line	Context
LOW	tools/convert_hf_to_torch_dist.py	12	CODE
LOW	tools/convert_torch_dist_to_hf_ray.py	75	CODE
LOW	docker/amd_patch/latest/sglang_attn_bridge/__init__.py	3	CODE
LOW	miles_plugins/mbridge/__init__.py	1	CODE
LOW	miles_plugins/mbridge/__init__.py	2	CODE
LOW	miles_plugins/mbridge/__init__.py	3	CODE
LOW	miles_plugins/mbridge/__init__.py	4	CODE
LOW	miles_plugins/mbridge/__init__.py	5	CODE
LOW	miles_plugins/mbridge/__init__.py	6	CODE
LOW	miles_plugins/mbridge/__init__.py	7	CODE
LOW	miles_plugins/mbridge/__init__.py	8	CODE
LOW	miles_plugins/mbridge/__init__.py	9	CODE
LOW	miles_plugins/mbridge/__init__.py	10	CODE
LOW	miles_plugins/megatron_bridge/nemotron_h.py	23	CODE
LOW	miles_plugins/megatron_bridge/__init__.py	16	CODE
LOW	miles_plugins/megatron_bridge/__init__.py	75	CODE
LOW	miles_plugins/models/qwen3_vl.py	9	CODE
LOW	miles_plugins/models/inkling/mm_towers.py	1	CODE
LOW	miles_plugins/models/inkling/mm_processor.py	1	CODE
LOW	miles_plugins/models/inkling/audio_processing.py	1	CODE
LOW	miles_plugins/models/inkling/vision_encoder.py	1	CODE
LOW	miles_plugins/models/inkling/model.py	4	CODE
LOW	miles_plugins/models/inkling/lora.py	1	CODE
LOW	miles_plugins/models/inkling/layers.py	4	CODE
LOW	miles_plugins/models/inkling/image_processing.py	1	CODE
LOW	tests/session_parity_utils.py	3	CODE
LOW	tests/ci/test/test_neon_store.py	15	CODE
LOW	tests/ci/test/test_metric_history_store.py	9	CODE
LOW	tests/ci/test/test_gate_integration.py	20	CODE
LOW	tests/ci/test/test_history_gate.py	8	CODE
LOW	tests/ci/test/test_metric_selection.py	4	CODE
LOW	tests/ci/test/test_log_groups.py	18	CODE
LOW	tests/ci/metric_history/register.py	31	CODE
LOW	tests/ci/metric_history/__init__.py	15	CODE
LOW	tests/ci/metric_history/__init__.py	15	CODE
LOW	tests/ci/metric_history/__init__.py	16	CODE
LOW	tests/ci/metric_history/__init__.py	16	CODE
LOW	tests/ci/metric_history/__init__.py	17	CODE
LOW	tests/ci/metric_history/__init__.py	18	CODE
LOW	tests/ci/metric_history/__init__.py	18	CODE
LOW	tests/ci/metric_history/__init__.py	18	CODE
LOW	tests/ci/metric_history/__init__.py	18	CODE
LOW	tests/ci/metric_history/gate.py	29	CODE
LOW	tests/ci/metric_history/constraints.py	24	CODE
LOW	tests/ci/metric_history/selection.py	27	CODE
LOW	tests/ci/metric_history/storage/store.py	22	CODE
LOW	tests/ci/metric_history/storage/__init__.py	9	CODE
LOW	tests/ci/metric_history/storage/__init__.py	9	CODE
LOW	tests/ci/metric_history/storage/__init__.py	10	CODE
LOW	tests/ci/metric_history/storage/__init__.py	11	CODE
LOW	tests/ci/metric_history/storage/__init__.py	11	CODE
LOW	tests/ci/metric_history/storage/__init__.py	11	CODE
LOW	tests/ci/metric_history/storage/__init__.py	11	CODE
LOW	tests/ci/metric_history/storage/sqlite_store.py	17	CODE
LOW	tests/ci/metric_history/storage/neon_store.py	35	CODE
LOW	tests/fast-gpu/test_run_megatron_worker_main.py	9	CODE
LOW	tests/fast-gpu/test_run_megatron_worker_main.py	48	CODE
LOW	tests/fast/backends/test_fsdp_routing_replay.py	151	CODE
LOW	tests/fast/backends/test_fsdp_routing_replay.py	165	CODE
LOW	…ckends/training_utils/test_reduce_gathered_log_dict.py	3	CODE
213 more matches not shown…

Modern Structural Boilerplate236 hits · 242 pts

Severity	File	Line	Snippet	Context
LOW	train_multi_lora_async.py	20	logger = logging.getLogger(__name__)	CODE
LOW	train_async.py	18	logger = logging.getLogger(__name__)	CODE
LOW	train.py	19	logger = logging.getLogger(__name__)	CODE
LOW	tools/cpu_memory_profiler.py	23	logger = logging.getLogger(__name__)	CODE
LOW	tools/convert_hf_to_nvfp4.py	157	def _update_quantization_config(cfg: dict, ignore_list: list[str]) -> None:	CODE
LOW	tools/convert_torch_dist_to_hf_ray.py	274	def _set_postfix(self) -> None:	STRING
LOW	tools/visualize/cpu_memory_profiler_visualize.py	12	logger = logging.getLogger(__name__)	CODE
LOW	docker/amd_patch/latest/sglang_attn_bridge/__init__.py	5	__all__ = ["apply_sglang_triton_attention_patch"]	CODE
LOW	miles_plugins/mbridge/__init__.py	12	__all__ = [	CODE
LOW	miles_plugins/megatron_bridge/nemotron_h.py	29	logger = logging.getLogger(__name__)	CODE
LOW	miles_plugins/megatron_bridge/__init__.py	20	logger = logging.getLogger(__name__)	CODE
LOW	miles_plugins/models/qwen3_vl.py	18	logger = logging.getLogger(__name__)	CODE
LOW	miles_plugins/models/cp_utils.py	8	logger = logging.getLogger(__name__)	CODE
LOW	miles_plugins/models/inkling/__init__.py	19	__all__ = ["get_inkling_spec", "inkling_mm_model_provider", "inkling_model_provider"]	CODE
LOW	miles_plugins/models/inkling/model.py	25	logger = logging.getLogger(__name__)	CODE
LOW	miles_plugins/models/inkling/lora.py	13	logger = logging.getLogger(__name__)	CODE
LOW	miles_plugins/optimizers/nvme_stream.py	42	logger = logging.getLogger(__name__)	CODE
LOW	tests/ci/ci_utils.py	71	logger = logging.getLogger(__name__)	CODE
LOW	tests/ci/ci_register.py	9	__all__ = [	CODE
LOW	tests/ci/metric_history/__init__.py	20	__all__ = [	CODE
LOW	tests/ci/metric_history/storage/__init__.py	13	__all__ = [	CODE
LOW	tests/fast/utils/test_health_checker.py	8	async def _settle(clock: FakeClock) -> None:	CODE
LOW⚡	tests/fast/ray/train/test_group.py	831	async def _set_all_train_return(group: RayTrainGroup, value: TrainStepOutcome) -> None:	CODE
LOW	tests/fast/ray/train/dummy_actor.py	25	def set_fail_methods(self, methods: list[str]) -> None:	CODE
LOW	tests/fast/ray/train/dummy_actor.py	28	def set_train_return_value(self, value: Any) -> None:	CODE
LOW	tests/fast/ray/train/dummy_actor.py	52	def set_rollout_manager(self, args: Any, *kwargs: Any) -> None:	CODE
LOW⚡	tests/fast/ray/train/dummy_actor.py	67	def update_weights(self) -> None:	CODE
LOW⚡	tests/fast/ray/train/dummy_actor.py	70	def set_heartbeat_fail(self, fail: bool) -> None:	CODE
LOW⚡	tests/fast/ray/train/dummy_actor.py	73	def set_last_active_timestamp(self, ts: float) -> None:	CODE
LOW	tests/e2e/sglang/utils/session_tool_agent.py	36	logger = logging.getLogger(__name__)	CODE
LOW	tests/e2e/sglang/utils/router_equivalence_generate.py	27	logger = logging.getLogger(__name__)	CODE
LOW	tests/e2e/ft/conftest_ft/fault_injection.py	14	logger = logging.getLogger(__name__)	CODE
LOW	miles/backends/megatron_utils/checkpoint.py	95	logger = logging.getLogger(__name__)	CODE
LOW	miles/backends/megatron_utils/checkpoint.py	97	__all__ = ["save_checkpoint", "save_checkpoint_with_lora", "load_checkpoint"]	CODE
LOW	miles/backends/megatron_utils/initialize.py	19	logger = logging.getLogger(__name__)	CODE
LOW	miles/backends/megatron_utils/bridge_lora_helpers.py	20	logger = logging.getLogger(__name__)	CODE
LOW⚡	miles/backends/megatron_utils/arguments.py	7	__all__ = ["validate_args", "parse_args", "set_default_megatron_args"]	CODE
LOW⚡	miles/backends/megatron_utils/arguments.py	9	logger = logging.getLogger(__name__)	CODE
LOW	miles/backends/megatron_utils/local_weight_checksum.py	24	logger = logging.getLogger(__name__)	CODE
LOW	miles/backends/megatron_utils/fp32_param_utils.py	7	logger = logging.getLogger(__name__)	CODE
LOW	miles/backends/megatron_utils/model.py	57	logger = logging.getLogger(__name__)	CODE
LOW	miles/backends/megatron_utils/multi_lora_optimizer.py	17	logger = logging.getLogger(__name__)	CODE
LOW	miles/backends/megatron_utils/model_provider.py	24	logger = logging.getLogger(__name__)	CODE
LOW	miles/backends/megatron_utils/multi_lora_utils.py	17	logger = logging.getLogger(__name__)	CODE
LOW	miles/backends/megatron_utils/ci_utils.py	15	logger = logging.getLogger(__name__)	CODE
LOW⚡	miles/backends/megatron_utils/actor.py	70	logger = logging.getLogger(__name__)	CODE
LOW⚡	miles/backends/megatron_utils/actor.py	73	def _setup_disk_offload_reclaim(disk_dir: str) -> None:	CODE
LOW	miles/backends/megatron_utils/actor.py	356	def _set_replay_stage(self, stage: str) -> None:	CODE
LOW	miles/backends/megatron_utils/actor.py	723	def update_weights(self, info: "EnginesAndLock") -> None:	CODE
LOW	miles/backends/megatron_utils/parallel.py	15	logger = logging.getLogger(__name__)	CODE
LOW	miles/backends/megatron_utils/rematerialize_utils.py	13	logger = logging.getLogger(__name__)	CODE
LOW	miles/backends/megatron_utils/sglang.py	27	__all__ = [	CODE
LOW⚡	miles/backends/megatron_utils/lora_utils.py	16	logger = logging.getLogger(__name__)	CODE
LOW	miles/backends/megatron_utils/hf_export.py	28	logger = logging.getLogger(__name__)	CODE
LOW	miles/backends/megatron_utils/multi_lora_scheduler.py	9	logger = logging.getLogger(__name__)	CODE
LOW	…atron_utils/update_weight/update_weight_from_tensor.py	34	logger = logging.getLogger(__name__)	CODE
LOW	…atron_utils/update_weight/update_weight_from_tensor.py	237	def update_weights(self) -> None:	CODE
LOW	miles/backends/megatron_utils/update_weight/common.py	18	logger = logging.getLogger(__name__)	CODE
LOW	…ls/update_weight/update_weight_from_distributed/p2p.py	33	logger = logging.getLogger(__name__)	CODE
LOW	…/update_weight/update_weight_from_distributed/mixin.py	29	logger = logging.getLogger(__name__)	CODE
176 more matches not shown…

Structural Annotation Overuse129 hits · 237 pts

Severity	File	Line	Snippet	Context
LOW	…amd_patch/latest/sglang_attn_bridge/triton_attn_bwd.py	410	# Step 1: preprocess delta	COMMENT
LOW	…amd_patch/latest/sglang_attn_bridge/triton_attn_bwd.py	427	# Step 2: compute LSE	COMMENT
LOW	…amd_patch/latest/sglang_attn_bridge/triton_attn_bwd.py	448	# Step 3: compute dK, dV	COMMENT
LOW	…amd_patch/latest/sglang_attn_bridge/triton_attn_bwd.py	482	# Step 4: compute dQ	COMMENT
LOW	tests/ci/README.md	11	### Step 0: Prepare `/data/miles_ci`	COMMENT
LOW⚡	tests/ci/README.md	28	### Step 1: Env	COMMENT
LOW⚡	tests/ci/README.md	38	### Step 2: Prepare `/home/runner/externals`	COMMENT
LOW⚡	tests/ci/README.md	48	### Step 3: Run	COMMENT
LOW⚡	…st/backends/megatron_utils/test_checkpoint_transfer.py	97	# Step 1: sender — wrap state_dict for transport	COMMENT
LOW⚡	…st/backends/megatron_utils/test_checkpoint_transfer.py	100	# Step 2: sender — flatten via pytree (what PGTransport does internally)	COMMENT
LOW⚡	…st/backends/megatron_utils/test_checkpoint_transfer.py	103	# Step 3: sender — pickle treespec + non-tensor leaves; "send" tensor leaves over the wire	COMMENT
LOW⚡	…st/backends/megatron_utils/test_checkpoint_transfer.py	110	# Step 4: receiver — unpickle metadata + interleave received tensors back into leaf order	COMMENT
LOW	…st/backends/megatron_utils/test_checkpoint_transfer.py	124	# Step 5: receiver — unwrap into iteration + state_dict	COMMENT
LOW⚡	tests/fast/utils/test_process_group_utils.py	267	# Step 1: single group	COMMENT
LOW⚡	tests/fast/utils/test_process_group_utils.py	273	# Step 2: two groups = global sum	COMMENT
LOW⚡	tests/fast/utils/test_process_group_utils.py	278	# Step 3: bitwise equality across all ranks	COMMENT
LOW⚡	tests/fast/utils/test_process_group_utils.py	286	# Step 4: empty groups = no-op	COMMENT
LOW⚡	tests/fast/utils/test_process_group_utils.py	291	# Step 5: MAX op	COMMENT
LOW⚡	tests/fast/utils/test_process_group_utils.py	310	# Step 1: single group gather	COMMENT
LOW	tests/fast/utils/test_process_group_utils.py	324	# Step 2: two group gather — global rank 0 gets everything	COMMENT
LOW⚡	tests/fast/utils/test_env_report.py	341	# Step 1: Run the full collection (no mocks)	COMMENT
LOW⚡	tests/fast/utils/test_env_report.py	348	# Step 2: Verify the package appears in editable_packages	COMMENT
LOW⚡	tests/fast/utils/test_env_report.py	355	# Step 3: Verify git info — clean repo	COMMENT
LOW⚡	tests/fast/utils/test_env_report.py	365	# Step 4: Verify single-line JSON output is parseable and contains this package	COMMENT
LOW⚡	tests/fast/utils/test_env_report.py	376	# Step 5: Verify package also in full_pip_list	COMMENT
LOW⚡	tests/fast/utils/test_env_report.py	386	# Step 1: Stage an uncommitted file	COMMENT
LOW⚡	tests/fast/utils/test_env_report.py	390	# Step 2: Run collection	COMMENT
LOW⚡	tests/fast/utils/test_env_report.py	397	# Step 3: Verify dirty + diff_stat mentions the file	COMMENT
LOW⚡	tests/fast/utils/test_env_report.py	412	# Step 1: Modify a tracked file without staging	COMMENT
LOW⚡	tests/fast/utils/test_env_report.py	416	# Step 2: Run collection	COMMENT
LOW⚡	tests/fast/utils/test_env_report.py	423	# Step 3: Verify dirty	COMMENT
LOW⚡	tests/fast/utils/test_health_checker.py	83	# Step 1: first_wait=0, so first check runs immediately after task starts	COMMENT
LOW⚡	tests/fast/utils/test_health_checker.py	87	# Step 2: Elapse less than interval — no second check	COMMENT
LOW⚡	tests/fast/utils/test_health_checker.py	91	# Step 3: Elapse to interval — second check	COMMENT
LOW⚡	tests/fast/utils/test_health_checker.py	107	# Step 1: Elapse 100s — still in first_wait	COMMENT
LOW⚡	tests/fast/utils/test_health_checker.py	111	# Step 2: Elapse to 300s — first_wait completes, first check runs	COMMENT
LOW⚡	tests/fast/utils/test_health_checker.py	115	# Step 3: Elapse interval — second check	COMMENT
LOW⚡	tests/fast/utils/test_health_checker.py	262	# Step 1: Initial first_wait (100s)	COMMENT
LOW⚡	tests/fast/utils/test_health_checker.py	268	# Step 2: Normal interval (5s)	COMMENT
LOW⚡	tests/fast/utils/test_health_checker.py	272	# Step 3: Pause + resume resets first_wait	COMMENT
LOW⚡	tests/fast/utils/test_health_checker.py	276	# Step 4: Need to elapse past the pending interval sleep first,	COMMENT
LOW⚡	tests/fast/utils/test_clock.py	239	# Step 1: first_wait=10	COMMENT
LOW⚡	tests/fast/utils/test_clock.py	244	# Step 2: interval=5	COMMENT
LOW⚡	tests/fast/utils/test_clock.py	248	# Step 3: interval=5	COMMENT
LOW⚡	tests/fast/utils/test_mini_ft_controller.py	382	# Step 1: First poll → heal attempt fails, sets next_attempt_at	COMMENT
LOW⚡	tests/fast/utils/test_mini_ft_controller.py	387	# Step 2: Poll again without advancing clock → should skip heal	COMMENT
LOW⚡	tests/fast/utils/test_mini_ft_controller.py	392	# Step 3: Advance clock past backoff → should attempt heal again	COMMENT
LOW⚡	tests/fast/utils/test_mini_ft_controller.py	421	# Step 1: First attempt fails	COMMENT
LOW⚡	tests/fast/utils/test_mini_ft_controller.py	428	# Step 2: Advance clock past backoff, second attempt succeeds	COMMENT
LOW⚡	tests/fast/utils/test_mini_ft_controller.py	498	# Step 1: Unhealthy → heal fails → backoff entry created	COMMENT
LOW⚡	tests/fast/utils/test_mini_ft_controller.py	503	# Step 2: Cell becomes NOT_APPLICABLE → backoff entry cleaned	COMMENT
LOW⚡	tests/fast/utils/test_mini_ft_controller.py	524	# Step 1: Unhealthy → heal fails → backoff entry created	COMMENT
LOW⚡	tests/fast/utils/test_mini_ft_controller.py	529	# Step 2: Cell becomes HEALTHY → backoff entry cleaned	COMMENT
LOW⚡	tests/fast/utils/test_witness/test_module.py	71	# Step 1: Compute loss without witness	COMMENT
LOW⚡	tests/fast/utils/test_witness/test_module.py	81	# Step 2: Compute loss with witness	COMMENT
LOW⚡	…t_analyzer/rules/test_cross_replica_weight_checksum.py	80	# Step 0: match	COMMENT
LOW⚡	…t_analyzer/rules/test_cross_replica_weight_checksum.py	83	# Step 1: mismatch	COMMENT
LOW⚡	…t_analyzer/rules/test_cross_replica_weight_checksum.py	86	# Step 2: match	COMMENT
LOW⚡	tests/fast/ray/train/test_group.py	239	# Step 1: Stop cell 1	COMMENT
LOW⚡	tests/fast/ray/train/test_group.py	242	# Step 2: Refresh	COMMENT
69 more matches not shown…

Excessive Try-Catch Wrapping197 hits · 219 pts

Severity	File	Line	Snippet	Context
LOW	tools/convert_hf_to_hf_int4.py	39	except Exception as e:	CODE
LOW	tools/convert_hf_to_torch_dist.py	44	except Exception:	CODE
LOW	tools/convert_hf_to_int4.py	39	except Exception as e:	CODE
LOW	miles_plugins/megatron_bridge/nemotron_h.py	216	except Exception as e: # best-effort; avoid breaking unrelated models	CODE
LOW⚡	miles_plugins/megatron_bridge/__init__.py	59	except Exception as _e: # best-effort	CODE
LOW⚡	miles_plugins/megatron_bridge/__init__.py	67	except Exception as _e: # best-effort; Qwen3-VL may be unavailable in some envs	CODE
LOW⚡	miles_plugins/megatron_bridge/__init__.py	76	except Exception as _e: # pragma: no cover - defensive	CODE
LOW	miles_plugins/models/qwen3_vl.py	200	except Exception:	CODE
MEDIUM	miles_plugins/models/qwen3_vl.py	76	def _patch_model_forward_and_rope_index() -> None:	CODE
LOW	miles_plugins/models/inkling/ops.py	21	except Exception: # pragma: no cover	CODE
LOW	miles_plugins/models/inkling/ops.py	476	except Exception as _import_error:	CODE
LOW	miles_plugins/models/inkling/lora.py	645	except Exception:	CODE
LOW	tests/ci/ci_utils.py	142	except Exception as e:	CODE
LOW	tests/ci/ci_utils.py	159	except Exception as e:	CODE
LOW	tests/ci/ci_utils.py	260	except Exception as e: # noqa: BLE001 -- never let store setup fail CI	CODE
LOW	tests/ci/ci_utils.py	341	except Exception as e:	CODE
LOW	tests/ci/ci_utils.py	527	except Exception as e: # noqa: BLE001 -- gate infra must not affect pass/fail	CODE
LOW	tests/ci/ci_utils.py	571	except Exception:	CODE
MEDIUM	tests/ci/ci_utils.py	156	def _target_func():	CODE
LOW	tests/ci/metric_history/storage/neon_store.py	138	except Exception:	STRING
LOW	tests/ci/metric_history/storage/neon_store.py	192	except Exception:	STRING
LOW⚡	tests/fast/dashboard/test_gpu_sampler.py	177	except Exception:	CODE
LOW⚡	tests/fast/ray/rollout/conftest.py	218	except Exception:	CODE
LOW	tests/fast/ray/rollout/conftest.py	249	except Exception:	CODE
MEDIUM	tests/fast/ray/rollout/conftest.py	215	def _count():	CODE
LOW	tests/fast/ray/rollout/real_ray/conftest.py	32	except Exception:	CODE
LOW	tests/fast/ray/rollout/real_ray/test_fault_tolerance.py	50	except Exception:	CODE
LOW	tests/manual/session/_bench_load_generator.py	222	except Exception:	CODE
LOW	tests/e2e/short/test_qwen2.5_0.5B_opd_sglang.py	78	except Exception:	CODE
LOW	tests/e2e/ft/conftest_ft/fault_injection.py	78	except Exception:	CODE
LOW	tests/e2e/ft/conftest_ft/fault_injection.py	109	except Exception:	CODE
LOW	miles/backends/megatron_utils/__init__.py	25	except Exception as _e: # best-effort; not every environment uses megatron.bridge	CODE
LOW	miles/backends/megatron_utils/model_provider.py	273	except Exception as e:	CODE
LOW	miles/backends/megatron_utils/lora_utils.py	497	except Exception as hf_export_err:	CODE
LOW	miles/backends/megatron_utils/hf_export.py	169	except Exception as e:	CODE
LOW	miles/backends/megatron_utils/hf_export.py	185	except Exception as e:	CODE
LOW	…atron_utils/update_weight/update_weight_from_tensor.py	511	except Exception as _unload_err:	CODE
LOW	…t/update_weight_from_distributed/p2p_transfer_utils.py	179	except Exception as e:	CODE
LOW	miles/backends/megatron_utils/megatron_to_hf/inkling.py	56	except Exception:	CODE
MEDIUM	miles/backends/megatron_utils/megatron_to_hf/inkling.py	51	def _qkv_tp_size(args):	CODE
LOW	miles/backends/megatron_utils/ft/indep_dp.py	142	except Exception:	CODE
LOW	miles/backends/fsdp_utils/checkpoint.py	118	except Exception as e:	CODE
LOW	miles/backends/fsdp_utils/checkpoint.py	130	except Exception as e:	CODE
LOW	miles/backends/fsdp_utils/checkpoint.py	143	except Exception as e:	CODE
LOW⚡	miles/backends/fsdp_utils/models/nemotron_h.py	32	except Exception: # pragma: no cover	CODE
LOW	miles/backends/fsdp_utils/models/nemotron_h.py	56	except Exception:	CODE
LOW	miles/backends/fsdp_utils/models/nemotron_h.py	119	except Exception: # pragma: no cover	CODE
LOW	miles/backends/fsdp_utils/models/qwen3_5.py	96	except Exception:	CODE
LOW	miles/backends/training_utils/mm_data.py	191	except Exception:	CODE
LOW	miles/backends/training_utils/loss_hub/math_utils.py	97	except Exception:	CODE
LOW	miles/backends/sglang_utils/sglang_engine.py	66	except Exception:	CODE
LOW	miles/backends/sglang_utils/sglang_engine.py	458	except Exception as e:	CODE
LOW	miles/backends/sglang_utils/sglang_engine.py	492	except Exception as e:	CODE
LOW	miles/utils/dumper_utils.py	315	except Exception:	CODE
LOW	miles/utils/misc.py	96	except Exception as e:	CODE
LOW	miles/utils/misc.py	247	except Exception:	CODE
LOW	miles/utils/reloadable_process_group.py	278	except Exception as e:	CODE
MEDIUM	miles/utils/reloadable_process_group.py	275	def _wrap_low_level_call():	CODE
LOW	miles/utils/retry_utils.py	31	except Exception:	CODE
LOW	miles/utils/processing_utils.py	37	except Exception as e:	CODE
137 more matches not shown…

Cross-Language Confusion37 hits · 215 pts

Severity	File	Line	Snippet	Context
HIGH⚡	tests/ci/test/test_run_suite.py	323	assert "github.event.schedule \|\| github.run_id" in workflow	CODE
HIGH⚡	tests/ci/test/test_run_suite.py	329	"group: ${{ github.workflow }}-${{ github.event.number \|\| github.event.schedule \|\| github.run_id }}"	CODE
HIGH	tests/ci/test/test_run_suite.py	360	assert "github.event.schedule \|\| github.run_id" in workflow	CODE
HIGH	…st/backends/megatron_utils/test_checkpoint_transfer.py	411	# Empty view shares its base's storage (data_ptr is non-null).	COMMENT
HIGH	tests/fast/ray/multi_lora/test_controller_http.py	179	unknown names report null; COMPLETED is retained after free_slot."""	STRING
HIGH⚡	tests/fast/router/test_session_samples_op_v2.py	671	response = app_client.post(f"/sessions/{uuid.uuid4().hex}/samples", content=b'{"max_seq_len":null}')	CODE
HIGH⚡	tests/fast/router/test_session_samples_op_v2.py	681	response = app_client.post(f"/sessions/{sid}/samples", content=b'{"max_seq_len":null}')	CODE
HIGH⚡	tests/fast/router/test_session_samples_op.py	297	response = app_client.post(f"/sessions/{uuid.uuid4().hex}/samples", content=b'{"max_seq_len":null}')	CODE
HIGH⚡	tests/fast/router/test_session_samples_op.py	307	response = app_client.post(f"/sessions/{sid}/samples", content=b'{"max_seq_len":null}')	CODE
HIGH	tests/e2e/megatron/test_qwen3_5_35b_a3b_lora_ci.py	54	U.exec_command("ray stop --force \|\| true; pkill -9 sglang \|\| true; sleep 10")	CODE
HIGH	…/model_scripts/test_glm5_2_744b_a40b_5layer_lora_ci.py	60	U.exec_command("ray stop --force \|\| true; pkill -9 sglang \|\| true; sleep 10")	CODE
HIGH	…/model_scripts/test_glm5_1_744b_a40b_6layer_lora_ci.py	60	U.exec_command("ray stop --force \|\| true; pkill -9 sglang \|\| true; sleep 10")	CODE
HIGH	…megatron/model_scripts/test_gpt_oss_20b_moe_lora_ci.py	126	U.exec_command("ray stop --force \|\| true; pkill -9 sglang \|\| true; sleep 10")	CODE
HIGH	…al-refactor-verify/mechanical_refactor_verify_utils.py	30	exec_command(f"git add -A && git commit -m {shlex.quote(message)}", cwd=cwd)	CODE
HIGH	miles/utils/external_utils/command_utils.py	199	f"export no_proxy=127.0.0.1 && export PYTHONBUFFERED=16 && "	CODE
HIGH	miles/utils/external_utils/command_utils.py	312	"pkill -x mooncake_master >/dev/null 2>&1 \|\| true; "	CODE
HIGH	miles/utils/external_utils/command_utils.py	319	exec_command("pkill -x mooncake_master >/dev/null 2>&1 \|\| true")	CODE
HIGH	miles/rollout/session/samples/codec.py	184	setattr(sample, field, copy(spec.null))	CODE
HIGH⚡	examples/infra_features/p2p_weight_transfer/run.py	769	pkill_python = f"pgrep -x 'python\|python3' \| grep -v -w {my_pid} \| grep -v -w {my_ppid} \| xargs -r kill -9 \|\| true"	CODE
HIGH⚡	examples/infra_features/p2p_weight_transfer/run.py	770	run_cmd("pkill -9 sglang \|\| true", check=False)	CODE
HIGH⚡	examples/infra_features/p2p_weight_transfer/run.py	772	run_cmd("ray stop --force \|\| true", check=False)	CODE
HIGH⚡	examples/infra_features/p2p_weight_transfer/run.py	773	run_cmd("pkill -9 ray \|\| true", check=False)	CODE
HIGH⚡	examples/infra_features/p2p_weight_transfer/run.py	776	run_cmd("pkill -9 ray \|\| true", check=False)	CODE
HIGH⚡	examples/infra_features/p2p_weight_transfer/run.py	778	run_cmd("pkill -9 redis \|\| true", check=False)	CODE
HIGH	…/infra_features/random_async/run_random_async_3node.py	31	f'test "$(cat {args.model_dir}/{args.model_name}_torch_dist/latest_checkpointed_iteration.txt 2>/dev/null)" = re	CODE
HIGH	examples/experimental/nemo-gym/run.py	68	f"pgrep -f '{t}' \| {exclude} \| xargs -r kill 2>/dev/null \|\| true",	CODE
HIGH	examples/experimental/openenv/openenv_launch_common.py	61	f"pgrep -f '{t}' \| {exclude} \| xargs -r kill 2>/dev/null \|\| true",	CODE
HIGH	examples/experimental/openenv/openenv_launch_common.py	206	provision_hint="mkdir -p ~/.config/daytona && echo dtn_... > ~/.config/daytona/api_key",	CODE
HIGH	examples/experimental/openenv/openenv_launch_common.py	217	provision_hint="mkdir -p ~/.config/e2b && echo <key> > ~/.config/e2b/api_key"	CODE
HIGH	examples/experimental/openenv/tb2_sandbox_recipe.py	264	f"mkdir -p /opt/tb2-tasks && curl -fsSL {url} \| "	CODE
HIGH	examples/experimental/openenv/tb2_sandbox_recipe.py	327	"apt-get update && DEBIAN_FRONTEND=noninteractive apt-get install -y "	CODE
HIGH	examples/experimental/openenv/tb2_sandbox_recipe.py	341	f"mkdir -p /opt/src && echo {_env_src_tar_b64()} \| base64 -d \| tar xz -C /opt/src",	CODE
HIGH	examples/experimental/openenv/openenv_agent_function.py	221	"! -name repo_cache -exec rm -rf {} + 2>/dev/null \|\| true"	CODE
HIGH	examples/experimental/openenv/scan_golden.py	67	cmds.append("mkdir -p /solution && base64 -d /tmp/solution.b64 \| tar xz -C /solution && rm -f /tmp/solution.b64")	CODE
HIGH	…xperimental/formal_math/single_round/kimina_wrapper.py	104	'[ -n "$ids" ] && docker stop $ids && docker rm $ids; '	CODE
HIGH	…e-agent-harbor-docker/run-glm47-flash-agentic-async.py	119	f"pgrep -f '{pattern}' \| {exclude} \| xargs -r kill 2>/dev/null \|\| true",	CODE
HIGH	examples/swe-agent-harbor-docker/run.py	82	f"pgrep -f '{t}' \| {exclude} \| xargs -r kill 2>/dev/null \|\| true",	CODE

Verbosity Indicators112 hits · 207 pts

Severity	File	Line	Snippet	Context
LOW	…amd_patch/latest/sglang_attn_bridge/triton_attn_bwd.py	410	# Step 1: preprocess delta	COMMENT
LOW	…amd_patch/latest/sglang_attn_bridge/triton_attn_bwd.py	427	# Step 2: compute LSE	COMMENT
LOW	…amd_patch/latest/sglang_attn_bridge/triton_attn_bwd.py	448	# Step 3: compute dK, dV	COMMENT
LOW	…amd_patch/latest/sglang_attn_bridge/triton_attn_bwd.py	482	# Step 4: compute dQ	COMMENT
LOW⚡	…st/backends/megatron_utils/test_checkpoint_transfer.py	97	# Step 1: sender — wrap state_dict for transport	COMMENT
LOW⚡	…st/backends/megatron_utils/test_checkpoint_transfer.py	100	# Step 2: sender — flatten via pytree (what PGTransport does internally)	COMMENT
LOW⚡	…st/backends/megatron_utils/test_checkpoint_transfer.py	103	# Step 3: sender — pickle treespec + non-tensor leaves; "send" tensor leaves over the wire	COMMENT
LOW⚡	…st/backends/megatron_utils/test_checkpoint_transfer.py	110	# Step 4: receiver — unpickle metadata + interleave received tensors back into leaf order	COMMENT
LOW	…st/backends/megatron_utils/test_checkpoint_transfer.py	124	# Step 5: receiver — unwrap into iteration + state_dict	COMMENT
LOW⚡	tests/fast/utils/test_process_group_utils.py	267	# Step 1: single group	COMMENT
LOW⚡	tests/fast/utils/test_process_group_utils.py	273	# Step 2: two groups = global sum	COMMENT
LOW⚡	tests/fast/utils/test_process_group_utils.py	278	# Step 3: bitwise equality across all ranks	COMMENT
LOW⚡	tests/fast/utils/test_process_group_utils.py	286	# Step 4: empty groups = no-op	COMMENT
LOW⚡	tests/fast/utils/test_process_group_utils.py	291	# Step 5: MAX op	COMMENT
LOW⚡	tests/fast/utils/test_process_group_utils.py	310	# Step 1: single group gather	COMMENT
LOW	tests/fast/utils/test_process_group_utils.py	324	# Step 2: two group gather — global rank 0 gets everything	COMMENT
LOW⚡	tests/fast/utils/test_env_report.py	341	# Step 1: Run the full collection (no mocks)	COMMENT
LOW⚡	tests/fast/utils/test_env_report.py	348	# Step 2: Verify the package appears in editable_packages	COMMENT
LOW⚡	tests/fast/utils/test_env_report.py	355	# Step 3: Verify git info — clean repo	COMMENT
LOW⚡	tests/fast/utils/test_env_report.py	365	# Step 4: Verify single-line JSON output is parseable and contains this package	COMMENT
LOW⚡	tests/fast/utils/test_env_report.py	376	# Step 5: Verify package also in full_pip_list	COMMENT
LOW⚡	tests/fast/utils/test_env_report.py	386	# Step 1: Stage an uncommitted file	COMMENT
LOW⚡	tests/fast/utils/test_env_report.py	390	# Step 2: Run collection	COMMENT
LOW⚡	tests/fast/utils/test_env_report.py	397	# Step 3: Verify dirty + diff_stat mentions the file	COMMENT
LOW⚡	tests/fast/utils/test_env_report.py	412	# Step 1: Modify a tracked file without staging	COMMENT
LOW⚡	tests/fast/utils/test_env_report.py	416	# Step 2: Run collection	COMMENT
LOW⚡	tests/fast/utils/test_env_report.py	423	# Step 3: Verify dirty	COMMENT
LOW⚡	tests/fast/utils/test_health_checker.py	83	# Step 1: first_wait=0, so first check runs immediately after task starts	COMMENT
LOW⚡	tests/fast/utils/test_health_checker.py	87	# Step 2: Elapse less than interval — no second check	COMMENT
LOW⚡	tests/fast/utils/test_health_checker.py	91	# Step 3: Elapse to interval — second check	COMMENT
LOW⚡	tests/fast/utils/test_health_checker.py	107	# Step 1: Elapse 100s — still in first_wait	COMMENT
LOW⚡	tests/fast/utils/test_health_checker.py	111	# Step 2: Elapse to 300s — first_wait completes, first check runs	COMMENT
LOW⚡	tests/fast/utils/test_health_checker.py	115	# Step 3: Elapse interval — second check	COMMENT
LOW⚡	tests/fast/utils/test_health_checker.py	262	# Step 1: Initial first_wait (100s)	COMMENT
LOW⚡	tests/fast/utils/test_health_checker.py	268	# Step 2: Normal interval (5s)	COMMENT
LOW⚡	tests/fast/utils/test_health_checker.py	272	# Step 3: Pause + resume resets first_wait	COMMENT
LOW⚡	tests/fast/utils/test_health_checker.py	276	# Step 4: Need to elapse past the pending interval sleep first,	COMMENT
LOW⚡	tests/fast/utils/test_clock.py	239	# Step 1: first_wait=10	COMMENT
LOW⚡	tests/fast/utils/test_clock.py	244	# Step 2: interval=5	COMMENT
LOW⚡	tests/fast/utils/test_clock.py	248	# Step 3: interval=5	COMMENT
LOW⚡	tests/fast/utils/test_mini_ft_controller.py	382	# Step 1: First poll → heal attempt fails, sets next_attempt_at	COMMENT
LOW⚡	tests/fast/utils/test_mini_ft_controller.py	387	# Step 2: Poll again without advancing clock → should skip heal	COMMENT
LOW⚡	tests/fast/utils/test_mini_ft_controller.py	392	# Step 3: Advance clock past backoff → should attempt heal again	COMMENT
LOW⚡	tests/fast/utils/test_mini_ft_controller.py	421	# Step 1: First attempt fails	COMMENT
LOW⚡	tests/fast/utils/test_mini_ft_controller.py	428	# Step 2: Advance clock past backoff, second attempt succeeds	COMMENT
LOW⚡	tests/fast/utils/test_mini_ft_controller.py	498	# Step 1: Unhealthy → heal fails → backoff entry created	COMMENT
LOW⚡	tests/fast/utils/test_mini_ft_controller.py	503	# Step 2: Cell becomes NOT_APPLICABLE → backoff entry cleaned	COMMENT
LOW⚡	tests/fast/utils/test_mini_ft_controller.py	524	# Step 1: Unhealthy → heal fails → backoff entry created	COMMENT
LOW⚡	tests/fast/utils/test_mini_ft_controller.py	529	# Step 2: Cell becomes HEALTHY → backoff entry cleaned	COMMENT
LOW⚡	tests/fast/utils/test_witness/test_module.py	71	# Step 1: Compute loss without witness	COMMENT
LOW⚡	tests/fast/utils/test_witness/test_module.py	81	# Step 2: Compute loss with witness	COMMENT
LOW⚡	…t_analyzer/rules/test_cross_replica_weight_checksum.py	80	# Step 0: match	COMMENT
LOW⚡	…t_analyzer/rules/test_cross_replica_weight_checksum.py	83	# Step 1: mismatch	COMMENT
LOW⚡	…t_analyzer/rules/test_cross_replica_weight_checksum.py	86	# Step 2: match	COMMENT
LOW⚡	tests/fast/ray/train/test_group.py	239	# Step 1: Stop cell 1	COMMENT
LOW⚡	tests/fast/ray/train/test_group.py	242	# Step 2: Refresh	COMMENT
LOW⚡	tests/fast/ray/train/test_group.py	245	# Step 3: Quorum bumped (init was quorum 0, this is first reconfigure)	COMMENT
LOW⚡	tests/fast/ray/train/test_group.py	248	# Step 4: Remaining alive cells have updated indep_dp_info	COMMENT
LOW⚡	tests/fast/ray/train/test_group.py	257	# Step 5: Stopped cell untouched	COMMENT
LOW⚡	tests/fast/ray/train/test_group.py	260	# Step 6: Actors received reconfigure_indep_dp	COMMENT
52 more matches not shown…

Deep Nesting182 hits · 179 pts

Severity	File	Line	Context
LOW	train_async.py	22	CODE
LOW	train.py	22	CODE
LOW	tools/convert_hf_to_nvfp4.py	157	CODE
LOW	tools/convert_hf_to_nvfp4.py	208	CODE
LOW	tools/convert_hf_to_nvfp4.py	231	CODE
LOW	tools/convert_hf_to_nvfp4.py	277	CODE
LOW	tools/convert_hf_to_mxfp8.py	161	CODE
LOW	tools/convert_kimi_int4_to_bf16.py	101	CODE
LOW	tools/convert_torch_dist_to_hf_ray.py	573	CODE
LOW	tools/convert_torch_dist_to_hf_ray.py	325	CODE
LOW	tools/fp8_cast_bf16.py	43	CODE
LOW	tools/convert_to_hf.py	29	CODE
LOW	docker/build.py	98	CODE
LOW	miles_plugins/mbridge/deepseek_v32.py	72	CODE
LOW	miles_plugins/mbridge/deepseek_v32.py	106	CODE
LOW	miles_plugins/mbridge/glm4.py	83	CODE
LOW	miles_plugins/mbridge/glm4moe.py	35	CODE
LOW	miles_plugins/models/glm5/ops/tilelang_indexer_bwd.py	18	CODE
LOW	miles_plugins/models/glm5/ops/tilelang_indexer_bwd.py	46	CODE
LOW	miles_plugins/models/glm5/ops/tilelang_indexer_fwd.py	91	CODE
LOW	miles_plugins/models/glm5/ops/tilelang_indexer_fwd.py	102	CODE
LOW	…les_plugins/models/glm5/ops/tilelang_sparse_mla_bwd.py	87	CODE
LOW	…les_plugins/models/glm5/ops/tilelang_sparse_mla_bwd.py	137	CODE
LOW	miles_plugins/models/inkling/mm_processor.py	51	CODE
LOW	miles_plugins/models/inkling/mm_processor.py	79	CODE
LOW	miles_plugins/models/inkling/mm_processor.py	211	CODE
LOW	miles_plugins/models/inkling/image_processing.py	101	CODE
LOW	miles_plugins/models/deepseek_v4/deepseek_v4.py	47	CODE
LOW	…les_plugins/models/deepseek_v4/ops/kernel/act_quant.py	48	CODE
LOW	…les_plugins/models/deepseek_v4/ops/kernel/act_quant.py	64	CODE
LOW	…/models/deepseek_v4/ops/kernel/tilelang_indexer_bwd.py	18	CODE
LOW	…/models/deepseek_v4/ops/kernel/tilelang_indexer_bwd.py	46	CODE
LOW	…/models/deepseek_v4/ops/kernel/tilelang_indexer_fwd.py	95	CODE
LOW	…/models/deepseek_v4/ops/kernel/tilelang_indexer_fwd.py	106	CODE
LOW	…dels/deepseek_v4/ops/kernel/tilelang_sparse_mla_bwd.py	87	CODE
LOW	…dels/deepseek_v4/ops/kernel/tilelang_sparse_mla_bwd.py	134	CODE
LOW	tests/ci/ci_utils.py	39	CODE
LOW	tests/ci/ci_utils.py	275	CODE
LOW	tests/ci/ci_utils.py	383	CODE
LOW	tests/ci/ci_register.py	309	CODE
LOW	tests/ci/test/test_ci_register.py	607	CODE
LOW	tests/ci/metric_history/register.py	282	CODE
LOW	…s/fast/backends/training_utils/loss/loss_test_utils.py	266	CODE
LOW	tests/fast/dashboard/dummy_telemetry.py	121	CODE
LOW	tests/fast/ray/rollout/conftest.py	234	CODE
LOW	tests/fast/ray/rollout/test_addr_allocator.py	41	CODE
LOW	tests/fast/ray/rollout/real_ray/test_rollout_manager.py	385	CODE
LOW	tests/fast/router/test_session_race_conditions.py	194	CODE
LOW	tests/fast/router/test_session_race_conditions.py	234	CODE
LOW	tests/fast/router/test_session_race_conditions.py	282	CODE
LOW	tests/fast/router/test_session_race_conditions.py	351	CODE
LOW	tests/fast/router/test_tree_assembly_invariants.py	144	CODE
LOW	tests/fast/router/test_sessions_v2.py	131	CODE
LOW	…nual/models/deepseek_v4/test_v4_tilelang_sparse_mla.py	138	CODE
LOW	…/manual/models/deepseek_v4/test_v4_tilelang_indexer.py	232	CODE
LOW	…/manual/models/deepseek_v4/test_v4_tilelang_indexer.py	474	CODE
LOW	tests/e2e/short/test_qwen2.5_0.5B_opd_sglang.py	38	CODE
LOW	tests/e2e/precision/test_qwen3_0.6B_parallel_check.py	32	CODE
LOW	tests/e2e/megatron/test_qwen3_4B_offload_disk_stream.py	51	CODE
LOW	tests/e2e/megatron/test_qwen3_4B_offload_disk.py	43	CODE
122 more matches not shown…

Cross-File Repetition28 hits · 140 pts

Severity	File	Snippet	Context
HIGH	tests/e2e/fsdp/test_qwen3_0.6B_megatron_fsdp_align.py	--train-env-vars '{"pytorch_cuda_alloc_conf":"expandable_segments:true"}'	STRING
HIGH	scripts/run_qwen3_4b.py	--train-env-vars '{"pytorch_cuda_alloc_conf":"expandable_segments:true"}'	STRING
HIGH	scripts/run_mcore_fsdp.py	--train-env-vars '{"pytorch_cuda_alloc_conf":"expandable_segments:true"}'	STRING
HIGH	scripts/run_qwen3_30b_a3b_fsdp.py	--train-env-vars '{"pytorch_cuda_alloc_conf":"expandable_segments:true"}'	STRING
HIGH	scripts/run_qwen3_0_6b_fsdp.py	--train-env-vars '{"pytorch_cuda_alloc_conf":"expandable_segments:true"}'	STRING
HIGH	scripts/run_nemotron_3_nano_4b_fsdp.py	--train-env-vars '{"pytorch_cuda_alloc_conf":"expandable_segments:true"}'	STRING
HIGH	examples/experimental/nemo-gym/run.py	kill old ray jobs and stale processes to free gpu resources.	STRING
HIGH	examples/experimental/openenv/openenv_launch_common.py	kill old ray jobs and stale processes to free gpu resources.	STRING
HIGH	…e-agent-harbor-docker/run-glm47-flash-agentic-async.py	kill old ray jobs and stale processes to free gpu resources.	STRING
HIGH	examples/swe-agent-harbor-docker/run.py	kill old ray jobs and stale processes to free gpu resources.	STRING
HIGH	examples/experimental/nemo-gym/run.py	convert hf checkpoint to torch_dist format if not already done.	STRING
HIGH	examples/experimental/openenv/run-openenv-tbench2.py	convert hf checkpoint to torch_dist format if not already done.	STRING
HIGH	examples/swe-agent-harbor-docker/run.py	convert hf checkpoint to torch_dist format if not already done.	STRING
HIGH	scripts/run_deepseek_v32.py	use_tis: true use_rs: {"true" if args.tis_use_rs else "false"} tis_level: "token" rs_level: "token" tis_mode: "truncate"	STRING
HIGH	scripts/run_joy_ai_llm_flash.py	use_tis: true use_rs: {"true" if args.tis_use_rs else "false"} tis_level: "token" rs_level: "token" tis_mode: "truncate"	STRING
HIGH	scripts/run_qwen3_4b.py	use_tis: true use_rs: {"true" if args.tis_use_rs else "false"} tis_level: "token" rs_level: "token" tis_mode: "truncate"	STRING
HIGH	scripts/run_qwen3_30b_a3b.py	use_tis: true use_rs: {"true" if args.tis_use_rs else "false"} tis_level: "token" rs_level: "token" tis_mode: "truncate"	STRING
HIGH	scripts/run_glm45_355b_a32b.py	use_tis: true use_rs: {"true" if args.tis_use_rs else "false"} tis_level: "token" rs_level: "token" tis_mode: "truncate"	STRING
HIGH	scripts/amd/run_qwen3_30b_a3b.py	use_tis: true use_rs: {"true" if args.tis_use_rs else "false"} tis_level: "token" rs_level: "token" tis_mode: "truncate"	STRING
HIGH	scripts/run_glm5_1_744b_a40b_lora.py	run grpo lora training (assumes the dataset is already prepared).	STRING
HIGH	scripts/run_glm5_2_744b_a40b_lora.py	run grpo lora training (assumes the dataset is already prepared).	STRING
HIGH	scripts/run_qwen3_5_35b_a3b_lora.py	run grpo lora training (assumes the dataset is already prepared).	STRING
HIGH	scripts/run_glm5_1_744b_a40b_lora.py	download the model checkpoint + dataset, then run grpo lora training.	STRING
HIGH	scripts/run_glm5_2_744b_a40b_lora.py	download the model checkpoint + dataset, then run grpo lora training.	STRING
HIGH	scripts/run_qwen3_5_35b_a3b_lora.py	download the model checkpoint + dataset, then run grpo lora training.	STRING
HIGH	scripts/run_inkling.py	return parallel config args for tested gpu configurations. only includes configurations that have been verified to work.	STRING
HIGH	scripts/run_deepseek_v4.py	return parallel config args for tested gpu configurations. only includes configurations that have been verified to work.	STRING
HIGH	scripts/amd/run_deepseek_v4.py	return parallel config args for tested gpu configurations. only includes configurations that have been verified to work.	STRING

Hallucination Indicators6 hits · 90 pts

Severity	File	Line	Snippet	Context
CRITICAL⚡	tests/fast/utils/test_witness/test_module.py	218	model.local_head_witness.witness.weight.data.fill_(42.0)	CODE
CRITICAL⚡	tests/fast/utils/test_witness/test_module.py	246	assert 5 in model.local_head_witness.witness.weight.grad.squeeze(-1).nonzero(as_tuple=True)[0].tolist()	CODE
CRITICAL⚡	tests/fast/utils/test_witness/test_module.py	402	chunk.module.local_head_witness.witness.weight.data.fill_(1.0)	CODE
CRITICAL⚡	tests/fast/utils/test_witness/test_module.py	403	chunk.module.local_tail_witness.witness.weight.data.fill_(1.0)	CODE
CRITICAL⚡	tests/fast/utils/test_witness/test_module.py	425	chunk.module.local_head_witness.witness.weight.data.fill_(1.0)	CODE
CRITICAL⚡	tests/fast/utils/test_witness/test_module.py	426	chunk.module.local_tail_witness.witness.weight.data.fill_(1.0)	CODE

AI Slop Vocabulary24 hits · 63 pts

Severity	File	Line	Snippet	Context
MEDIUM	tests/ci/test/test_history_gate.py	611	# --- dual-register files + harness-supplied registry ------------------------	STRING
MEDIUM	tests/ci/test/test_history_gate.py	636	# single-registry reparse raise (ambiguous). With the harness passing the	STRING
MEDIUM	tests/ci/test/test_ci_history.py	201	# This covers the backend's current file naming detail, not a CI harness	COMMENT
MEDIUM	tests/ci/metric_history/gate.py	269	# never raise on a dual-register / no-register file. Use the harness's	COMMENT
MEDIUM	tests/ci/metric_history/gate.py	285	# The harness already selected one register_*_ci() call; use it directly and	COMMENT
MEDIUM⚡	…fast/utils/chat_template_utils/test_fixed_templates.py	49	# stay append-only regardless of which roles the harness appends.	COMMENT
MEDIUM⚡	…nual/models/deepseek_v4/test_v4_tilelang_sparse_mla.py	388	"""Print a comprehensive diff summary across all forward configs."""	STRING
MEDIUM⚡	…/manual/models/deepseek_v4/test_v4_tilelang_indexer.py	526	"""Print a comprehensive diff summary across all configurations."""	STRING
MEDIUM	tests/e2e/short/test_run_megatron.py	156	# The CI harness invokes this file as bare `python3 <file>` with no args.	COMMENT
MEDIUM	…s/e2e/sglang/test_session_server_multi_role/_common.py	60	# This harness produces one rollout batch, so its train-side batch divisor	COMMENT
LOW	…ron_utils/megatron_to_hf/processors/quantizer_nvfp4.py	95	# for other parameters, we just return the original converted_named_params	COMMENT
LOW	…atron_utils/megatron_to_hf/processors/quantizer_fp8.py	110	# for other parameters, we just return the original converted_named_params	COMMENT
LOW	…ron_utils/megatron_to_hf/processors/quantizer_mxfp8.py	108	# for other parameters, we just return the original converted_named_params	COMMENT
LOW	miles/backends/fsdp_utils/lr_scheduler.py	98	# If the learning rate is constant, just return the initial value.	COMMENT
MEDIUM	miles/utils/arguments.py	2663	# On iff the CI harness injected MILES_CI_GATE_RECORD_DIR (the same env var	COMMENT
MEDIUM	miles/utils/tracking_utils/ci_history.py	58	# Env var naming the directory the harness assigns for this run's records.	COMMENT
MEDIUM	miles/utils/tracking_utils/ci_history.py	92	# No harness-assigned directory: nothing to collect into. Leaving	COMMENT
MEDIUM	miles/rollout/rm_hub/math_dapo_utils.py	14	# Adapted from https://github.com/EleutherAI/lm-evaluation-harness/blob/main/lm_eval/tasks/hendrycks_math/utils.py	COMMENT
MEDIUM	…les/experimental/openenv/openenv_e2b_agent_function.py	55	# older install outright, and the shared agent loop's harness-marker guard	COMMENT
MEDIUM	…experimental/openenv/openenv_daytona_agent_function.py	63	# older install outright, and the shared agent loop's harness-marker guard	COMMENT
MEDIUM	examples/experimental/openenv/openenv_agent_function.py	329	# - harness marker absent: the server scored, but not through the	COMMENT
MEDIUM⚡	examples/experimental/openenv/openenv_agent_function.py	443	# No canonical verdict (infra/harness failure or a non-canonical server,	COMMENT
MEDIUM⚡	examples/experimental/openenv/openenv_agent_function.py	452	# the scalar reward (plus the harness marker). The detailed pytest CTRF	COMMENT
MEDIUM	examples/experimental/openenv/scan_golden.py	114	# server-side scoring failure or a non-canonical harness is	COMMENT

Over-Commented Block57 hits · 55 pts

Severity	File	Line	Snippet	Context
LOW	tools/cpu_memory_profiler.py	1	# CPU Memory Profiler - standalone tool for profiling system memory during training.	COMMENT
LOW	tests/ci/github_runner/docker-compose.yml	1	# Please refer to `README.md` for how to setup this GitHub action runner.	COMMENT
LOW	tests/fast/utils/test_http_utils.py	101		COMMENT
LOW	…/fast/utils/chat_template_utils/test_tito_tokenizer.py	141	chat_template_kwargs={"clear_thinking": False},	COMMENT
LOW	…utils/chat_template_utils/test_token_seq_comparator.py	141		COMMENT
LOW	…utils/chat_template_utils/test_token_seq_comparator.py	241	#	COMMENT
LOW	…utils/chat_template_utils/test_token_seq_comparator.py	401	# ---------------------------------------------------------------------------	COMMENT
LOW	tests/manual/session/bench_session_server_overhead.py	301	json.dump(payload, f, indent=2)	COMMENT
LOW	tests/e2e/conftest_dumper.py	21	# attn_k disabled: megatron dumps k pre-RoPE (after adjust_key_value) while	COMMENT
LOW	…ang/test_session_server_multi_role/test_minimax_m27.py	21	register_ci_gate(metric_key="rollout/tito_session_mismatch_rate/v2/assistant_text")	COMMENT
LOW	tests/e2e/ft/conftest_ft/scenario_deterministic.py	81		COMMENT
LOW	tests/e2e/ft/conftest_ft/scenario_with_failure.py	21		COMMENT
LOW	.claude/skills/setup-ci-host/setup-host.sh	1	#!/usr/bin/env bash	COMMENT
LOW	.claude/skills/manage-gh-runners/manage-runners.sh	1	#!/usr/bin/env bash	COMMENT
LOW	miles/utils/seqlen_balancing.py	1	# Copied from https://github.com/volcengine/verl/blob/468adf22c43b744348051fccd7a5d830c6c3c36a/verl/utils/seqlen_balanci	COMMENT
LOW	miles/utils/test_utils/chat_template_verify.py	141	# * ``APPEND_ROLES: frozenset[str]`` — non-assistant roles that appear after	COMMENT
LOW	miles/utils/test_utils/chat_template_verify.py	401	)	COMMENT
LOW	miles/rollout/inference_rollout/__init__.py	1	# This is a refactor of the portions above generate-function in sglang_rollout.py,	COMMENT
LOW	miles/rollout/rm_hub/math_dapo_utils.py	1	# Copyright 2024 Bytedance Ltd. and/or its affiliates	COMMENT
LOW	miles/dashboard/static/carpet.js	21		COMMENT
LOW	…es/infra_features/train_infer_mismatch_helper/mis.yaml	1	# Enable importance sampling, details refer to the comments of compute_mis_weights in mis.py	COMMENT
LOW	…tures/low_precision/run-qwen3-30b-a3b-fp8-two-nodes.sh	1	#!/bin/bash	COMMENT
LOW	…_features/p2p_weight_transfer/prepare-qwen3-30B-A3B.sh	1	#!/bin/bash	COMMENT
LOW	…p2p_weight_transfer/run-qwen3-30B-A3B-4node-profile.sh	1	#!/bin/bash	COMMENT
LOW	…a_features/p2p_weight_transfer/prepare-glm4.7-flash.sh	1	#!/bin/bash	COMMENT
LOW	…fra_features/p2p_weight_transfer/prepare-glm4.5-air.sh	1	#!/bin/bash	COMMENT
LOW	…les/infra_features/p2p_weight_transfer/prepare-glm5.sh	1	#!/bin/bash	COMMENT
LOW	…/p2p_weight_transfer/run-glm4.7-flash-2node-profile.sh	1	#!/bin/bash	COMMENT
LOW	…es/p2p_weight_transfer/run-glm4.5-air-8node-profile.sh	1	#!/bin/bash	COMMENT
LOW	…eatures/p2p_weight_transfer/run-glm5-disagg-profile.sh	1	#!/bin/bash	COMMENT
LOW	…_weight_transfer/run-qwen3-235B-A22B-16node-profile.sh	1	#!/bin/bash	COMMENT
LOW	…ures/p2p_weight_transfer/run-kimi-k2-64node-profile.sh	1	#!/bin/bash	COMMENT
LOW	…/infra_features/p2p_weight_transfer/prepare-kimi-k2.sh	1		COMMENT
LOW	examples/experimental/openenv/tb2_sandbox_recipe.py	321	# state is part of the task environment — solutions and agents run	COMMENT
LOW	examples/experimental/openenv/run-openenv-tbench2.py	81	# within the limit is terminated and scored reward 0, bounding long-trajectory	COMMENT
LOW	examples/experimental/openenv/openenv_agent_function.py	61	# (CAPACITY_REACHED) and the docker-mode tbench2 server caps concurrent envs	COMMENT
LOW	examples/experimental/openenv/openenv_agent_function.py	81	# what _multi_turn parses: exactly one shell command per turn in a single	COMMENT
LOW	examples/experimental/openenv/openenv_agent_function.py	181	# (openenv_daytona_agent_function).	COMMENT
LOW	examples/experimental/openenv/openenv_agent_function.py	201	# (TB2_OUTPUT_DIR=/tmp/tbench2_env_runs) leaves a per-episode trial dir	COMMENT
LOW	examples/experimental/openenv/openenv_agent_function.py	321		COMMENT
LOW	examples/experimental/search-r1/qa_em_format.py	1	# Adapt from https://github.com/PeterGriffinJin/Search-R1/blob/ceee7b89655ed52f205b9beb98e1190c3eedcfb0/verl/utils/rewar	COMMENT
LOW	…al/search-r1/local_dense_retriever/retrieval_server.py	1	# Copyright 2024 Bytedance Ltd. and/or its affiliates	COMMENT
LOW	…perimental/search-r1/local_dense_retriever/download.py	1	# Copyright 2024 Bytedance Ltd. and/or its affiliates	COMMENT
LOW	…n-qwen2.5-3B-megatron-lora-disaggregated-multi-node.sh	1	#!/bin/bash	COMMENT
LOW	…n-qwen2.5-3B-megatron-lora-disaggregated-multi-node.sh	21	# Run the script once per node, with NODE_RANK 0 on the head node and a distinct	COMMENT
LOW	…n-qwen2.5-3B-megatron-lora-disaggregated-multi-node.sh	41	# env ${ENV} bash ...-multi-node.sh broadcast 3 # worker 3 (rollout node 1)	COMMENT
LOW	…_policy_distillation/run-qwen3-8B-opd-multi-teacher.sh	1	#!/bin/bash	COMMENT
LOW	…llation/qwen3_5_35b_selfdistill/phase1_rlvr_teacher.sh	1	#!/bin/bash	COMMENT
LOW	…tion/qwen3_5_35b_selfdistill/phase2_opd_selfdistill.sh	1	#!/bin/bash	COMMENT
LOW	…y_distillation/qwen3_5_35b_selfdistill/phase2_gb200.sh	1	#!/bin/bash	COMMENT
LOW	scripts/run_glm45_355b_a32b.py	361	# "NCCL_PXN_DISABLE": "0",	COMMENT
LOW	scripts/models/nemotron-3-ultra-550b-a55b-4layer.sh	1	# 4-layer slice of NVIDIA Nemotron-3-Ultra-550B-A55B, for single-node (8 GPU) CI.	COMMENT
LOW	scripts/models/qwen3.5-35B-A3B_lora.sh	61	# (identical to qwen3.5-35B-A3B.sh); every LoRA / run-mode flag lives in the runner,	COMMENT
LOW	scripts/models/nemotron-3-super-120b-a12b.sh	1	# NVIDIA Nemotron-3-Super-120B-A12B (BF16, MoE nemotron_h = hybrid Mamba + Attention + MoE).	COMMENT
LOW	scripts/models/qwen3.6-35B-A3B_lora.sh	61	# (identical to qwen3.5-35B-A3B.sh); every LoRA / run-mode flag lives in the runner,	COMMENT
LOW	scripts/models/nemotron-3-ultra-550b-a55b.sh	1	# NVIDIA Nemotron-3-Ultra-550B-A55B (BF16, MoE nemotron_h = hybrid Mamba2 + Attention + MoE).	COMMENT
LOW	.github/workflows/pr-test-rocm.yml	1	# doc-dev: docs/ci/00-stage.md	COMMENT

Modern AI Meta-Vocabulary15 hits · 42 pts

Severity	File	Line	Snippet	Context
MEDIUM⚡	…nds/megatron_utils/test_lora_weight_sync_validation.py	355	# - DistBucketedWeightUpdateMixin._update_lora_weights → shared orchestration	COMMENT
MEDIUM	tests/fast/dashboard/dummy_dump.py	168	# every third sample is agentic-shaped: multi-turn (mixed weight versions	COMMENT
MEDIUM	miles/utils/test_utils/session_verify_agent.py	114	# English-only on purpose: matches the production agentic flows tokenization	COMMENT
MEDIUM	miles/rollout/fully_async_data_buffer.py	24	# returns multiple samples per trajectory (e.g. multi-agent).	COMMENT
MEDIUM	miles/rollout/base_types.py	84	# One generate may lead to multiple samples, such as multi-agent, tree-like exploration, or	COMMENT
MEDIUM	miles/dashboard/store.py	704	# per-turn spans (multi_turn / agentic) supersede the coarse	COMMENT
MEDIUM	docs/models/gpt-oss/gpt-oss.md	8	[GPT-OSS](https://huggingface.co/openai/gpt-oss-20b) is OpenAI's open-weight language model, designed for reasoning, age	CODE
MEDIUM	docs/models/glm/glm5.md	7	[GLM-5](https://huggingface.co/zai-org/GLM-5) is the most powerful language model in Zhipu AI's GLM series, scaling to 7	CODE
MEDIUM	docs/models/kimi/kimi-k2.5.md	9	[Kimi-K2.5](https://huggingface.co/moonshotai/Kimi-K2.5) is an open-source, natively multimodal agentic model from Moons	CODE
MEDIUM	…erimental/multi_agent/run-qwen3-30B-A3B-multi-agent.sh	57	# multi-agent do not support eval for now	COMMENT
MEDIUM	examples/experimental/tau-bench/generate_with_tau.py	45	"agent": "tool-calling", # Select between ["tool-calling", "act", "react", "few-shot"]	CODE
MEDIUM	examples/experimental/nemo-gym/run.py	183	# Gates the dynamic registration of the agentic flags above; without	COMMENT
MEDIUM	examples/experimental/openenv/openenv_agent_function.py	105	# total episode time: a long agentic trajectory can loop for turns * (long	COMMENT
MEDIUM	examples/experimental/openenv/openenv_agent_function.py	315	# text protocol the Harbor mini-swe-agent scaffold uses.	COMMENT
MEDIUM	examples/experimental/strands_sglang/README.md	3	This example connects `miles` with [`strands-sglang`](https://github.com/horizon-rl/strands-sglang) (SGLang extension fo	CODE

Self-Referential Comments13 hits · 40 pts

Severity	File	Line	Snippet	Context
MEDIUM	train_multi_lora_async.py	43	# Create a controller nclusing MultiLoRAController and MultiLoRAHTTPServer to manage lora	COMMENT
MEDIUM	miles_plugins/mbridge/mimo.py	82	# Create a proxy name to use with parent class methods	COMMENT
MEDIUM	miles_plugins/models/qwen3_5.py	231	# Define the decoder block spec	COMMENT
MEDIUM	miles_plugins/models/qwen3_next.py	232	# Define the decoder block spec	COMMENT
MEDIUM	miles_plugins/models/glm5/glm5.py	751	# Define the decoder block spec	COMMENT
MEDIUM	tests/test_attention_output_gate_tp.py	138	# Create a full gate where each head has a distinct value	COMMENT
MEDIUM	miles/backends/megatron_utils/model_provider.py	230	# Define the decoder block spec	COMMENT
MEDIUM	miles/backends/megatron_utils/model_provider.py	238	# Define the decoder layer spec	COMMENT
MEDIUM	miles/backends/fsdp_utils/lr_scheduler.py	77	# Initialize parent class	COMMENT
MEDIUM⚡	miles/utils/async_utils.py	16	# Create a background event loop thread	COMMENT
MEDIUM	miles/utils/http_utils.py	313	# Define the async actor	COMMENT
MEDIUM	miles/rollout/rm_hub/__init__.py	57	# This function is intended for remote or time-consuming reward model evaluation.	COMMENT
MEDIUM	examples/experimental/tau-bench/trainable_agents.py	212	# Initialize tracking variables	COMMENT

Example Usage Blocks14 hits · 25 pts

Severity	File	Line	Snippet	Context
LOW⚡	…_features/p2p_weight_transfer/prepare-qwen3-30B-A3B.sh	7	# Usage:	COMMENT
LOW	…p2p_weight_transfer/run-qwen3-30B-A3B-4node-profile.sh	8	# Usage:	COMMENT
LOW⚡	…a_features/p2p_weight_transfer/prepare-glm4.7-flash.sh	8	# Usage:	COMMENT
LOW⚡	…fra_features/p2p_weight_transfer/prepare-glm4.5-air.sh	10	# Usage:	COMMENT
LOW	…les/infra_features/p2p_weight_transfer/prepare-glm5.sh	11	# Usage:	COMMENT
LOW⚡	…eatures/p2p_weight_transfer/prepare-qwen3-235b-A22B.sh	8	# Usage:	COMMENT
LOW	…/p2p_weight_transfer/run-glm4.7-flash-2node-profile.sh	5	# Usage:	COMMENT
LOW	…es/p2p_weight_transfer/run-glm4.5-air-8node-profile.sh	11	# Usage:	COMMENT
LOW	…eatures/p2p_weight_transfer/run-glm5-disagg-profile.sh	11	# Usage:	COMMENT
LOW	…_weight_transfer/run-qwen3-235B-A22B-16node-profile.sh	5	# Usage:	COMMENT
LOW	…ures/p2p_weight_transfer/run-kimi-k2-64node-profile.sh	9	# Usage:	COMMENT
LOW⚡	…/infra_features/p2p_weight_transfer/prepare-kimi-k2.sh	11	# Usage:	COMMENT
LOW	…n-qwen2.5-3B-megatron-lora-disaggregated-multi-node.sh	15	# Usage:	COMMENT
LOW	examples/geo3k_vlm/run_geo3k_vlm.sh	5	# Usage:	COMMENT

AI Structural Patterns23 hits · 22 pts

Severity	File	Line	Context
LOW	tools/convert_hf_to_int4_direct.py	118	CODE
LOW	miles_plugins/models/qwen3_5.py	182	CODE
LOW	miles_plugins/models/hf_attention.py	175	CODE
LOW	miles_plugins/models/qwen3_next.py	193	CODE
LOW	miles_plugins/models/glm5/glm5.py	193	CODE
LOW	…les_plugins/models/glm5/ops/tilelang_sparse_mla_bwd.py	87	CODE
LOW	miles_plugins/models/inkling/model.py	64	CODE
LOW	miles_plugins/models/deepseek_v4/deepseek_v4.py	211	CODE
LOW	…ast/backends/training_utils/loss/test_loss_snapshot.py	201	CODE
LOW	tests/fast/utils/test_health_checker.py	15	CODE
LOW	tests/fast/utils/test_arguments.py	75	CODE
LOW	tests/fast/rollout/generate_hub/test_single_turn.py	68	CODE
LOW	tests/fast/rollout/generate_utils/test_sample_utils.py	17	CODE
LOW	tests/fast/dashboard/dummy_dump.py	60	CODE
LOW	…atron_utils/update_weight/update_weight_from_tensor.py	234	CODE
LOW	…/update_weight/update_weight_from_distributed/mixin.py	337	CODE
LOW	…atron_to_hf/processors/quantizer_compressed_tensors.py	223	CODE
LOW	miles/utils/http_utils.py	368	CODE
LOW	miles/utils/distributed_utils.py	40	CODE
LOW	miles/utils/data.py	187	CODE
LOW	examples/experimental/tau-bench/openai_tool_adapter.py	109	CODE
LOW	…al/search-r1/local_dense_retriever/retrieval_server.py	290	CODE
LOW	examples/retool_v2/tool_sandbox.py	351	CODE

Redundant / Tautological Comments12 hits · 17 pts

Severity	File	Line	Snippet	Context
LOW	miles_plugins/mbridge/mimo.py	51	# Check if this is an MTP layer weight	COMMENT
LOW	miles_plugins/mbridge/mimo.py	80	# Check if this is a transformer_layer component	COMMENT
LOW	tests/ci/ci_utils.py	539	# Check if we should retry	COMMENT
LOW	miles/backends/megatron_utils/model_provider.py	147	# Check if the custom provider supports vp_stage parameter	COMMENT
LOW	miles/backends/megatron_utils/model_provider.py	270	# Check if fp8_model_init supports preserve_high_precision_init_val	COMMENT
LOW	miles/backends/megatron_utils/lora_utils.py	273	# Check if already in Megatron format (standard / canonical / Kimi MLA linear_*).	COMMENT
LOW	miles/backends/fsdp_utils/arguments.py	75	if hasattr(f.type, "__args__"): # Check if it's a Union type	CODE
LOW	miles/backends/training_utils/log_utils.py	374	if val: # Check if val is not empty	CODE
LOW	miles/utils/hf_config.py	29	# Set True to override transformers' native config.	COMMENT
LOW	examples/experimental/tau-bench/trainable_agents.py	321	# Check if done	COMMENT
LOW	examples/experimental/search-r1/qa_em_format.py	86	# Check if this is a tag	COMMENT
LOW	examples/retool_v2/tool_sandbox.py	235	# Return result	COMMENT

Docstring Block Structure3 hits · 15 pts

Severity	File	Line	Snippet	Context
HIGH	miles/backends/sglang_utils/sglang_engine.py	295	Run /health_generate on the underlying SGLang HTTP server. Args: timeout: Timeout for the health re	STRING
HIGH	examples/experimental/tau-bench/openai_tool_adapter.py	52	Parse sglang response to OpenAI compatible format Args: response: Raw response text from s	STRING
HIGH	examples/experimental/tau-bench/generate_with_tau.py	118	Generate a complete agent-environment interaction trajectory for tau-bench. This is the main entry point for m	STRING

Fake / Example Data10 hits · 10 pts

Severity	File	Line	Snippet	Context
LOW	tests/fast/rollout/generate_hub/test_single_turn.py	208	"process_fn_kwargs": {"routed_experts": "placeholder"},	CODE
LOW⚡	tests/fast/ray/rollout/test_config_matrix.py	82	@pytest.mark.parametrize("wt", ["regular", "prefill", "decode", "placeholder"])	CODE
LOW	tests/fast/ray/rollout/real_ray/test_server_group.py	23	engines = [ServerEngine() for _ in range(num_engines)] if worker_type != "placeholder" else []	CODE
LOW⚡	tests/fast/ray/rollout/real_ray/test_server_group.py	53	group = _build_group(pg_tuple=pg, num_engines=0, worker_type="placeholder")	CODE
LOW⚡	miles/backends/sglang_utils/sglang_config.py	16	worker_type: One of "regular", "prefill", "decode", or "placeholder".	STRING
LOW⚡	miles/backends/sglang_utils/sglang_config.py	17	"placeholder" reserves GPU slots without creating engines.	STRING
LOW	miles/backends/sglang_utils/sglang_config.py	32	valid_types = {"regular", "prefill", "decode", "placeholder"}	CODE
LOW	miles/utils/debug_utils/send_to_sglang.py	53	client = AsyncOpenAI(api_key="dummy_key", base_url=url)	CODE
LOW	miles/ray/rollout/rollout_server.py	65	[ServerEngine() for _ in range(num_engines)] if group_cfg.worker_type != "placeholder" else []	CODE
LOW	miles/ray/rollout/server_group.py	68	if self.args.debug_train_only or self.worker_type == "placeholder":	CODE

Slop Phrases2 hits · 5 pts

Severity	File	Line	Snippet	Context
MEDIUM	miles/rollout/on_policy_distillation.py	433	# If you have task rewards, you can add them here.	COMMENT
MEDIUM	examples/on_policy_distillation/run-qwen3-8B-opd.sh	9	TEACHER_IP="127.0.0.1" # Use localhost here, you can change it to your IP	CODE

TODO Padding1 hit · 2 pts

Severity	File	Line	Snippet	Context
LOW	…atron_utils/update_weight/update_weight_from_tensor.py	245	# TODO: implement lora weight checker	COMMENT

Analysis Overview

What These Metrics Mean

Score History

Severity Breakdown

Directory Score Breakdown

Pattern Findings