areal-project/AReaL

Pattern Findings

The scanner identified 7579 distinct pattern matches across 23 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.

Hyper-Verbose Identifiers3878 hits · 3682 pts

Severity	File	Line	Snippet	Context
LOW	.opencode/skills/add-reward/SKILL.md	122	def test_reward_correct_answer():	CODE
LOW	.opencode/skills/add-archon-model/SKILL.md	369	def test_state_dict_roundtrip():	CODE
LOW	.opencode/skills/add-unit-tests/SKILL.md	52	def test_function_under_condition_returns_expected():	CODE
LOW⚡	.opencode/skills/add-unit-tests/SKILL.md	104	def test_slow_but_required_in_ci():	CODE
LOW⚡	.opencode/skills/add-unit-tests/SKILL.md	115	def test_distributed_function(monkeypatch):	CODE
LOW	tests/test_train_controller.py	159	def create_mock_distributed_batch(size=4, seq_len=10):	CODE
LOW	tests/test_train_controller.py	254	def test_destroy_handles_errors(self, train_controller, ft_spec):	CODE
LOW	tests/test_train_controller.py	273	def test_destroy_requests_reverse_order(self, train_controller, ft_spec):	CODE
LOW	tests/test_train_controller.py	293	def test_merge_tensors_reorders_results(self):	CODE
LOW	tests/test_train_controller.py	304	def test_merge_tensors_with_tensor_results(self):	CODE
LOW	tests/test_train_controller.py	431	def test_custom_function_call_with_distributed_batch(	CODE
LOW	tests/test_train_controller.py	455	def test_custom_function_call_with_regular_args(self, train_controller, ft_spec):	CODE
LOW	tests/test_train_controller.py	476	def test_custom_function_call_filters_dp_heads(self, train_controller, ft_spec):	CODE
LOW	tests/test_train_controller.py	491	def test_create_process_group_is_dummy(self, train_controller, parallel_strategy):	CODE
LOW	tests/test_train_controller.py	515	def test_connect_engine_sets_rollout(self, train_controller, ft_spec):	CODE
LOW	tests/test_train_controller.py	529	def test_connect_engine_warns_on_change(self, train_controller, ft_spec):	CODE
LOW	tests/test_train_controller.py	545	def test_connect_engine_same_rollout_no_warning(self, train_controller, ft_spec):	CODE
LOW	tests/test_train_controller.py	560	def test_check_rollout_engine_connected_raises_when_not_connected(	CODE
LOW	tests/test_train_controller.py	572	def test_check_rollout_engine_connected_passes_when_connected(	CODE
LOW	tests/test_train_controller.py	589	def test_prepare_batch_delegates_to_rollout(	CODE
LOW	tests/test_train_controller.py	627	def test_rollout_batch_delegates_to_rollout(	CODE
LOW	tests/test_train_controller.py	667	def test_update_weights_raises_when_not_connected(self, train_controller, ft_spec):	CODE
LOW	tests/test_train_controller.py	710	def test_prepare_dispatch_splits_distributed_batch(self, train_controller, ft_spec):	CODE
LOW	tests/test_train_controller.py	724	def test_prepare_dispatch_partitions_without_duplication(	CODE
LOW	tests/test_train_controller.py	751	def test_prepare_dispatch_replicates_non_batch_args(	CODE
LOW	tests/test_train_controller.py	768	def test_prepare_dispatch_handles_kwargs(self, train_controller, ft_spec):	CODE
LOW	tests/test_train_controller.py	787	def test_storage_clear_surfaces_per_node_failures_and_stats(self):	CODE
LOW	tests/test_train_controller.py	828	def test_failed_storage_clear_is_retried_on_next_call(self):	CODE
LOW	tests/test_train_controller.py	849	def test_second_storage_clear_failure_cleans_workers_then_raises(self):	CODE
LOW	tests/test_train_controller.py	878	def test_worker_cleanup_failure_preserves_exhausted_storage_state(self):	CODE
LOW	tests/test_train_controller.py	917	def test_storage_clear_propagates_cancellation(self):	CODE
LOW	tests/test_train_controller.py	932	def test_storage_clear_cancellation_keeps_batch_state_atomic(self):	CODE
LOW⚡	tests/test_slurm_terminal_state.py	14	def test_node_fail_maps_to_failed(self):	CODE
LOW⚡	tests/test_slurm_terminal_state.py	20	def test_terminal_states_are_not_active(self, state):	CODE
LOW⚡	tests/test_slurm_terminal_state.py	24	def test_live_states_are_active(self, state):	CODE
LOW	tests/test_slurm_terminal_state.py	41	def test_maps_sacct_output_to_a_job_state(self, output, expected):	CODE
LOW	tests/test_slurm_terminal_state.py	45	def test_returns_none_when_sacct_has_no_record(self):	CODE
LOW	tests/test_slurm_terminal_state.py	56	def test_returns_none_when_sacct_is_unavailable(self, error):	CODE
LOW	tests/test_slurm_terminal_state.py	60	def test_reads_only_the_first_record(self):	CODE
LOW	tests/test_incomplete_rollout_groups.py	50	def test_rollout_group_size_contract_rejects_invalid_threshold(	CODE
LOW	tests/test_incomplete_rollout_groups.py	68	async def test_grouped_rollout_keeps_each_usable_member_once(rollout_stats):	CODE
LOW	tests/test_incomplete_rollout_groups.py	89	async def test_grouped_rollout_marks_empty_group_unusable(rollout_stats):	CODE
LOW	tests/test_incomplete_rollout_groups.py	111	async def test_grouped_rollout_applies_estimator_minimum(rollout_stats):	CODE
LOW	tests/test_incomplete_rollout_groups.py	134	async def test_grouped_rollout_keeps_singleton_by_default(rollout_stats):	CODE
LOW	tests/test_incomplete_rollout_groups.py	148	async def test_group_relative_tensor_slot_retains_split_rows(rollout_stats):	CODE
LOW	tests/test_incomplete_rollout_groups.py	162	async def test_group_relative_interaction_slot_retains_split_rows(rollout_stats):	CODE
LOW	tests/test_incomplete_rollout_groups.py	183	def test_executor_propagates_group_contract_error_without_retry(dynamic_bs):	CODE
LOW	tests/test_incomplete_rollout_groups.py	220	def test_group_training_metrics_report_actual_size_and_token_weight():	CODE
LOW	tests/test_incomplete_rollout_groups.py	241	def test_actor_loss_keeps_existing_token_weighted_group_reduction():	CODE
LOW⚡	tests/test_incomplete_rollout_groups.py	266	def test_min_usable_group_size_defaults_to_estimator_minimum(normalization):	CODE
LOW⚡	tests/test_incomplete_rollout_groups.py	273	def test_batch_relative_estimator_keeps_usable_singleton():	CODE
LOW⚡	tests/test_incomplete_rollout_groups.py	279	def test_explicit_min_usable_group_size_overrides_derivation():	CODE
LOW⚡	tests/test_incomplete_rollout_groups.py	288	def test_group_statistics_reject_explicit_singleton_minimum():	CODE
LOW⚡	tests/test_incomplete_rollout_groups.py	296	def test_batch_relative_estimator_accepts_explicit_singleton_minimum():	CODE
LOW⚡	tests/test_incomplete_rollout_groups.py	305	def test_dynamic_collection_backfills_from_ready_groups():	CODE
LOW	tests/test_incomplete_rollout_groups.py	320	def test_dynamic_collection_aborts_after_consecutive_empty_rounds():	CODE
LOW	tests/test_incomplete_rollout_groups.py	334	def test_dynamic_collection_empty_streak_resets_on_progress():	CODE
LOW	tests/test_incomplete_rollout_groups.py	351	def test_dynamic_collection_tolerates_fast_all_reject_bursts():	CODE
LOW	tests/test_incomplete_rollout_groups.py	366	def test_fixed_collection_rejects_undersized_batch():	CODE
LOW⚡	tests/test_zigzag_indices.py	67	def test_vectorized_matches_loop_reference(cp_size, seq_lens, builder):	CODE
3818 more matches not shown…

Decorative Section Separators977 hits · 3376 pts

Severity	File	Line	Snippet	Context
MEDIUM	pyproject.toml	218	# =============================================================================	COMMENT
MEDIUM	pyproject.toml	220	# =============================================================================	COMMENT
MEDIUM	pyproject.toml	325	# =============================================================================	COMMENT
MEDIUM	pyproject.toml	327	# =============================================================================	COMMENT
MEDIUM	pyproject.toml	348	# =============================================================================	COMMENT
MEDIUM	pyproject.toml	350	# =============================================================================	COMMENT
MEDIUM	pyproject.vllm.toml	229	# =============================================================================	COMMENT
MEDIUM	pyproject.vllm.toml	231	# =============================================================================	COMMENT
MEDIUM	pyproject.vllm.toml	324	# =============================================================================	COMMENT
MEDIUM	pyproject.vllm.toml	326	# =============================================================================	COMMENT
MEDIUM	pyproject.vllm.toml	347	# =============================================================================	COMMENT
MEDIUM	pyproject.vllm.toml	349	# =============================================================================	COMMENT
MEDIUM	tests/test_per_layer_optim_step.py	22	# ---------------------------------------------------------------------------	COMMENT
MEDIUM	tests/test_per_layer_optim_step.py	24	# ---------------------------------------------------------------------------	COMMENT
MEDIUM⚡	tests/test_per_layer_optim_step.py	119	# ---------------------------------------------------------------------------	COMMENT
MEDIUM⚡	tests/test_per_layer_optim_step.py	121	# ---------------------------------------------------------------------------	COMMENT
MEDIUM	tests/test_per_layer_optim_step.py	253	# ---------------------------------------------------------------------------	COMMENT
MEDIUM	tests/test_per_layer_optim_step.py	255	# ---------------------------------------------------------------------------	COMMENT
MEDIUM⚡	tests/test_seqpack.py	463	# =============================================================================	COMMENT
MEDIUM⚡	tests/test_seqpack.py	465	# =============================================================================	COMMENT
MEDIUM⚡	tests/test_seqpack.py	981	# =============================================================================	COMMENT
MEDIUM⚡	tests/test_seqpack.py	983	# =============================================================================	COMMENT
MEDIUM	tests/test_seqpack.py	27	# =============================================================================	COMMENT
MEDIUM	tests/test_seqpack.py	29	# =============================================================================	COMMENT
MEDIUM⚡	tests/test_slurm_scheduler.py	30	# ============================================================================	COMMENT
MEDIUM⚡	tests/test_slurm_scheduler.py	32	# ============================================================================	COMMENT
MEDIUM⚡	tests/test_slurm_scheduler.py	81	# ============================================================================	COMMENT
MEDIUM⚡	tests/test_slurm_scheduler.py	83	# ============================================================================	COMMENT
MEDIUM⚡	tests/test_slurm_scheduler.py	219	# ============================================================================	COMMENT
MEDIUM⚡	tests/test_slurm_scheduler.py	221	# ============================================================================	COMMENT
MEDIUM⚡	tests/test_slurm_scheduler.py	242	# ============================================================================	COMMENT
MEDIUM⚡	tests/test_slurm_scheduler.py	244	# ============================================================================	COMMENT
MEDIUM⚡	tests/test_slurm_scheduler.py	301	# ============================================================================	COMMENT
MEDIUM⚡	tests/test_slurm_scheduler.py	303	# ============================================================================	COMMENT
MEDIUM⚡	tests/test_slurm_scheduler.py	354	# ============================================================================	COMMENT
MEDIUM⚡	tests/test_slurm_scheduler.py	356	# ============================================================================	COMMENT
MEDIUM	tests/test_slurm_scheduler.py	154	# ============================================================================	COMMENT
MEDIUM	tests/test_slurm_scheduler.py	156	# ============================================================================	COMMENT
MEDIUM⚡	tests/test_kk_allocate.py	223	# ---------------------------------------------------------------------------	COMMENT
MEDIUM⚡	tests/test_kk_allocate.py	225	# ---------------------------------------------------------------------------	COMMENT
MEDIUM⚡	tests/test_kk_allocate.py	461	# ---------------------------------------------------------------------------	COMMENT
MEDIUM⚡	tests/test_kk_allocate.py	463	# ---------------------------------------------------------------------------	COMMENT
MEDIUM⚡	tests/test_kk_allocate.py	550	# ---------------------------------------------------------------------------	COMMENT
MEDIUM⚡	tests/test_kk_allocate.py	552	# ---------------------------------------------------------------------------	COMMENT
MEDIUM	tests/test_kk_allocate.py	34	# ---------------------------------------------------------------------------	COMMENT
MEDIUM	tests/test_kk_allocate.py	36	# ---------------------------------------------------------------------------	COMMENT
MEDIUM	tests/test_kk_allocate.py	62	# ---------------------------------------------------------------------------	COMMENT
MEDIUM	tests/test_kk_allocate.py	64	# ---------------------------------------------------------------------------	COMMENT
MEDIUM	tests/test_kk_allocate.py	114	# ---------------------------------------------------------------------------	COMMENT
MEDIUM	tests/test_kk_allocate.py	116	# ---------------------------------------------------------------------------	COMMENT
MEDIUM	tests/test_kk_allocate.py	158	# ---------------------------------------------------------------------------	COMMENT
MEDIUM	tests/test_kk_allocate.py	160	# ---------------------------------------------------------------------------	COMMENT
MEDIUM	tests/test_kk_allocate.py	373	# ---------------------------------------------------------------------------	COMMENT
MEDIUM	tests/test_kk_allocate.py	375	# ---------------------------------------------------------------------------	COMMENT
MEDIUM	tests/test_kk_allocate.py	426	# ---------------------------------------------------------------------------	COMMENT
MEDIUM	tests/test_kk_allocate.py	428	# ---------------------------------------------------------------------------	COMMENT
MEDIUM	tests/test_local_scheduler.py	39	# ============================================================================	COMMENT
MEDIUM	tests/test_local_scheduler.py	41	# ============================================================================	COMMENT
MEDIUM	tests/test_inference_engines.py	254	# ============================================================================	COMMENT
MEDIUM	tests/test_inference_engines.py	256	# ============================================================================	COMMENT
917 more matches not shown…

Empty Error Handlers155 hits · 775 pts

Severity	File	Line	Snippet	Context
HIGH	tests/test_rtensor.py	115	except Exception:	CODE
HIGH	tests/test_local_scheduler.py	1620	except EngineCallError:	CODE
HIGH	tests/test_local_scheduler.py	2178	except (	CODE
HIGH	tests/test_local_scheduler.py	2234	except (	CODE
HIGH	tests/test_local_scheduler.py	2088	except (	CODE
HIGH	tests/test_local_scheduler.py	2109	except Exception:	CODE
HIGH	tests/test_local_scheduler.py	2292	except (	CODE
HIGH	tests/test_local_scheduler.py	2538	except (	CODE
HIGH	tests/test_local_scheduler.py	2419	except (	CODE
HIGH	tests/test_mopd_gpu.py	82	except ProcessLookupError:	CODE
HIGH	tests/test_cuda_deterministic.py	338	except Exception:	CODE
HIGH	tests/test_sglang_pp_distributed.py	222	except ValueError:	CODE
HIGH	tests/test_sglang_pp_distributed.py	230	except (ProcessLookupError, PermissionError, OSError):	CODE
HIGH	tests/test_sglang_pp_distributed.py	234	except OSError:	CODE
HIGH	tests/test_sglang_pp_distributed.py	133	except TimeoutError:	CODE
HIGH	tests/test_examples.py	95	except TimeoutError:	CODE
HIGH	tests/experimental/archon/test_weight_sync.py	162	except Exception:	CODE
HIGH	tests/experimental/archon/test_weight_sync.py	378	except Exception:	CODE
HIGH	tests/experimental/archon/test_async_checkpoint.py	264	except RuntimeError:	CODE
HIGH	tests/experimental/openai/test_proxy_gateway.py	386	except asyncio.CancelledError:	CODE
HIGH	tests/experimental/openai/test_proxy_gateway.py	434	except asyncio.CancelledError:	CODE
HIGH	tests/experimental/openai/test_proxy_gateway.py	691	except (asyncio.CancelledError, Exception):	CODE
HIGH	tests/experimental/openai/test_proxy_gateway.py	750	except asyncio.CancelledError:	CODE
HIGH	tests/infra/test_process_supervisor.py	50	except json.JSONDecodeError:	CODE
HIGH	tests/infra/test_process_supervisor.py	111	except psutil.NoSuchProcess:	CODE
HIGH	tests/infra/test_process_supervisor.py	165	except psutil.NoSuchProcess:	CODE
HIGH	tests/infra/test_process_supervisor.py	241	except psutil.NoSuchProcess:	CODE
HIGH	tests/infra/data_service/test_epoch_crossing.py	65	except httpx.HTTPError:	CODE
HIGH	tests/infra/data_service/test_data_service_e2e.py	200	except httpx.HTTPError:	CODE
HIGH	tests/infra/data_service/test_data_service_e2e.py	63	except httpx.HTTPError:	CODE
HIGH	tests/v2/weight_update/test_sglang_integration.py	83	except (httpx.ConnectError, httpx.ReadTimeout):	CODE
HIGH	tests/fp8/engine_utils.py	150	except Exception:	CODE
HIGH	examples/agent_service/run_agent_service.py	36	except httpx.ConnectError:	CODE
HIGH	examples/scaffolding/controllers.py	848	except Exception:	CODE
HIGH	…/scaffolding/tests/test_scaffolding_llm_integration.py	999	except Exception:	CODE
HIGH	…/scaffolding/tests/test_scaffolding_llm_integration.py	1054	except Exception:	CODE
HIGH	examples/terminal_bench/agent/chat_agent_trace.py	214	except ImportError:	CODE
HIGH	examples/terminal_bench/agent/chat_agent_trace.py	50	except Exception:	CODE
HIGH	examples/terminal_bench/agent/camel_terminal_agent.py	335	except Exception:	CODE
HIGH	examples/swe/arena_agent.py	390	except asyncio.CancelledError:	CODE
HIGH	examples/swe/arena_agent.py	461	except FileExistsError:	CODE
HIGH	examples/swe/preprocessors.py	145	except (json.JSONDecodeError, TypeError):	CODE
HIGH	examples/math/boba_grpo.py	55	except Exception:	CODE
HIGH	areal/version.py	87	except (subprocess.CalledProcessError, FileNotFoundError):	CODE
HIGH	areal/experimental/openai/tool_call_parser.py	591	except ModuleNotFoundError:	CODE
HIGH	areal/experimental/openai/tool_call_parser.py	604	except ModuleNotFoundError:	CODE
HIGH	areal/experimental/openai/client.py	1991	except ImportError:	CODE
HIGH	areal/experimental/openai/client.py	664	except (json.JSONDecodeError, TypeError):	CODE
HIGH	areal/experimental/openai/proxy/proxy_rollout_server.py	1171	except asyncio.CancelledError:	CODE
HIGH	areal/infra/async_task_runner.py	684	except RuntimeError:	CODE
HIGH	areal/infra/launcher/local.py	85	except psutil.NoSuchProcess:	CODE
HIGH	areal/infra/launcher/ray.py	180	except (AttributeError, OSError, ValueError):	CODE
HIGH	areal/infra/launcher/ray.py	186	except Exception:	CODE
HIGH	areal/infra/launcher/ray.py	215	except OSError:	CODE
HIGH	areal/infra/launcher/ray.py	297	except ray.exceptions.GetTimeoutError:	CODE
HIGH	areal/infra/launcher/ray.py	220	except Exception:	CODE
HIGH	areal/infra/platforms/rocm.py	71	except ImportError:	CODE
HIGH	areal/infra/platforms/cuda.py	40	except ImportError:	CODE
HIGH	areal/infra/platforms/unknown.py	39	except ImportError:	CODE
HIGH	areal/infra/scheduler/slurm.py	74	except ValueError:	CODE
95 more matches not shown…

Unused Imports784 hits · 746 pts

Severity	File	Line	Context
LOW	tests/test_awex_delta_common.py	9	CODE
LOW	tests/test_warmup_process_groups.py	8	CODE
LOW	tests/test_mopd_teacher_manager.py	3	CODE
LOW	tests/test_filter_function.py	3	CODE
LOW	tests/test_megatron_mopd_teacher.py	3	CODE
LOW	tests/test_prm_runner.py	3	CODE
LOW	tests/test_token_rewards_advantage.py	3	CODE
LOW	tests/test_swe_preprocessors.py	1	CODE
LOW	tests/test_perf_trace_converter.py	1	CODE
LOW	tests/test_prm_config.py	3	CODE
LOW	tests/test_megatron_async_save.py	17	CODE
LOW	tests/utils.py	7	CODE
LOW	tests/utils.py	7	CODE
LOW	tests/test_interaction_token_rewards.py	3	CODE
LOW	tests/test_mopd_gpu.py	5	CODE
LOW	tests/test_virl39k.py	3	CODE
LOW	tests/test_geometry3k_agent.py	3	CODE
LOW	tests/test_reassemble_cp_logprobs.py	15	CODE
LOW	tests/test_examples.py	737	CODE
LOW	tests/test_examples.py	817	CODE
LOW	tests/test_profile_example.py	1	CODE
LOW	tests/test_mopd_rtensor.py	3	CODE
LOW	tests/test_grouped_rollout_workflow.py	3	CODE
LOW	tests/experimental/archon/test_hf_parity_qwen2.py	1	CODE
LOW	tests/experimental/archon/test_moe_common.py	25	CODE
LOW	tests/experimental/archon/conftest.py	107	CODE
LOW	tests/experimental/archon/conftest.py	115	CODE
LOW	tests/experimental/archon/test_hf_parity_qwen3.py	1	CODE
LOW	tests/experimental/archon/test_hf_parity_qwen3_moe.py	1	CODE
LOW	tests/experimental/archon/test_hf_parity_qwen3_5_moe.py	1	CODE
LOW	tests/experimental/archon/test_async_checkpoint.py	11	CODE
LOW	tests/experimental/archon/utils.py	22	CODE
LOW	tests/experimental/archon/utils.py	22	CODE
LOW	tests/experimental/archon/utils.py	22	CODE
LOW	tests/experimental/archon/test_qwen3_5.py	14	CODE
LOW	tests/experimental/archon/test_moe_hf_parity.py	23	CODE
LOW	tests/experimental/archon/test_hf_parity_qwen3_5.py	1	CODE
LOW	…experimental/archon/torchrun/run_pp_gradient_verify.py	12	CODE
LOW	tests/experimental/archon/torchrun/run_pp_tests.py	21	CODE
LOW	…s/experimental/archon/torchrun/run_checkpoint_tests.py	24	CODE
LOW	tests/experimental/archon/fp8/test_fp8_scale_layout.py	8	CODE
LOW	tests/experimental/archon/fp8/test_conversion.py	11	CODE
LOW	tests/experimental/archon/fp8/test_moe_dispatch.py	23	CODE
LOW	tests/experimental/archon/fp8/test_fp8_linear.py	20	CODE
LOW	…sts/experimental/archon/fp8/torchrun/run_checkpoint.py	2	CODE
LOW	…xperimental/archon/fp8/torchrun/run_sharded_dequant.py	16	CODE
LOW	tests/experimental/openai/test_proxy_gateway.py	3	CODE
LOW	tests/experimental/openai/test_proxy_integration.py	1	CODE
LOW	tests/experimental/openai/test_proxy_rollout_server.py	3	CODE
LOW	tests/experimental/openai/test_online_agent.py	3	CODE
LOW	…/experimental/openai/test_context_overflow_recovery.py	3	CODE
LOW	tests/experimental/openai/test_vlm_trajectory.py	3	CODE
LOW	…experimental/openai/test_streaming_chat_completions.py	10	CODE
LOW	tests/experimental/openai/test_tensor_reference.py	3	CODE
LOW	tests/experimental/openai/test_retry_orphan.py	9	CODE
LOW	tests/infra/test_proc.py	1	CODE
LOW	tests/infra/test_remote_inf_engine.py	1	CODE
LOW	tests/infra/test_local_launcher.py	1	CODE
LOW	tests/infra/test_launcher_utils.py	1	CODE
LOW	tests/infra/sandbox/test_runner.py	5	CODE
724 more matches not shown…

Excessive Try-Catch Wrapping586 hits · 605 pts

Severity	File	Line	Snippet	Context
LOW	.opencode/skills/add-reward/SKILL.md	61	except Exception:	CODE
MEDIUM	tests/test_swe_sft_cache.py	590	def build(label, source, attempt_id):	CODE
LOW	tests/test_swe_sft_cache.py	601	except Exception as error:	CODE
LOW	tests/test_rtensor.py	115	except Exception:	CODE
LOW	tests/test_slurm_scheduler.py	60	except Exception as e:	CODE
LOW⚡	tests/test_ray_scheduler.py	86	except Exception as e:	CODE
LOW	tests/test_local_scheduler.py	2109	except Exception:	CODE
LOW⚡	tests/test_offload.py	93	except Exception as e:	CODE
LOW	tests/test_cuda_deterministic.py	307	except Exception as e:	CODE
LOW	tests/test_cuda_deterministic.py	338	except Exception:	CODE
LOW	tests/test_sglang_pp_distributed.py	296	except Exception:	CODE
LOW	tests/test_adv_norm_config.py	1073	except Exception as e:	CODE
LOW⚡	tests/experimental/archon/conftest.py	32	except Exception:	CODE
LOW	tests/experimental/archon/conftest.py	118	except Exception:	CODE
LOW⚡	tests/experimental/archon/conftest.py	174	except Exception:	CODE
LOW	tests/experimental/archon/test_weight_sync.py	162	except Exception:	CODE
LOW	tests/experimental/archon/test_weight_sync.py	220	except Exception:	CODE
LOW	tests/experimental/archon/test_weight_sync.py	282	except Exception:	CODE
LOW	tests/experimental/archon/test_weight_sync.py	378	except Exception:	CODE
LOW⚡	tests/experimental/archon/test_grpo.py	447	except Exception as e:	CODE
LOW	…/experimental/archon/torchrun/run_qwen3_parallelize.py	551	except Exception as e:	CODE
LOW	…experimental/archon/torchrun/run_pp_gradient_verify.py	578	except Exception as e:	CODE
LOW	…s/experimental/archon/torchrun/run_archon_engine_pp.py	167	except Exception as e:	CODE
LOW	…s/experimental/archon/torchrun/run_archon_engine_pp.py	197	except Exception as e:	CODE
LOW	…s/experimental/archon/torchrun/run_archon_engine_pp.py	231	except Exception as e:	CODE
LOW	…s/experimental/archon/torchrun/run_archon_engine_pp.py	286	except Exception as e:	CODE
LOW	tests/experimental/archon/torchrun/run_parallel_dims.py	222	except Exception as e:	CODE
LOW	tests/experimental/archon/torchrun/run_parallel_dims.py	396	except Exception as e:	CODE
LOW	tests/experimental/archon/torchrun/run_ep_tests.py	818	except Exception as e:	CODE
LOW	tests/experimental/archon/torchrun/run_pp_tests.py	901	except Exception as e:	CODE
LOW	…s/experimental/archon/torchrun/run_checkpoint_tests.py	149	except Exception as e:	CODE
LOW	…s/experimental/archon/torchrun/run_checkpoint_tests.py	294	except Exception as e:	CODE
LOW	…s/experimental/archon/torchrun/run_checkpoint_tests.py	461	except Exception as e:	CODE
LOW	…s/experimental/archon/torchrun/run_checkpoint_tests.py	585	except Exception as e:	CODE
LOW	…s/experimental/archon/torchrun/run_checkpoint_tests.py	778	except Exception as e:	CODE
LOW	…s/experimental/archon/torchrun/run_checkpoint_tests.py	1008	except Exception as e:	CODE
LOW	…s/experimental/archon/torchrun/run_checkpoint_tests.py	1197	except Exception as e:	CODE
LOW	…s/experimental/archon/torchrun/run_checkpoint_tests.py	1359	except Exception as e:	CODE
LOW	…s/experimental/archon/torchrun/run_checkpoint_tests.py	1467	except Exception as e:	CODE
LOW	…ts/experimental/archon/torchrun/run_pp_combinations.py	558	except Exception as e:	CODE
LOW	…ts/experimental/archon/torchrun/run_pp_combinations.py	629	except Exception as e:	CODE
LOW	…sts/experimental/archon/fp8/torchrun/run_checkpoint.py	140	except Exception as e:	CODE
LOW	…sts/experimental/archon/fp8/torchrun/run_checkpoint.py	191	except Exception as e:	CODE
LOW	…xperimental/archon/fp8/torchrun/run_sharded_dequant.py	244	except Exception as e:	CODE
LOW	tests/infra/data_service/test_data_service_e2e.py	35	except Exception as exc:	CODE
LOW⚡	tests/torchrun/run_sglang_pp_weight_sync.py	275	except Exception as e:	CODE
LOW⚡	tests/torchrun/run_sglang_pp_weight_sync.py	443	except Exception as e:	CODE
LOW	tests/torchrun/run_fsdp_memory_efficient_lora.py	116	except Exception as e:	CODE
LOW	tests/torchrun/run_lock.py	72	except Exception:	CODE
LOW	tests/v2/weight_update/test_kv_store.py	99	except Exception as exc:	CODE
LOW	…/v2/weight_update/torchrun/run_nccl_weight_transfer.py	65	except Exception as e:	CODE
LOW	…/v2/weight_update/torchrun/run_nccl_weight_transfer.py	110	except Exception as e:	CODE
LOW	…/v2/weight_update/torchrun/run_nccl_weight_transfer.py	233	except Exception as e:	CODE
LOW	…/v2/weight_update/torchrun/run_nccl_weight_transfer.py	283	except Exception as e:	CODE
LOW	…sts/v2/training_service/test_controller_integration.py	173	except Exception as exc:	CODE
LOW	tests/v2/inference_service/integration_utils.py	62	except Exception as e:	CODE
MEDIUM	tests/fp8/model_hooks.py	308	def hook(module, input, output):	CODE
MEDIUM	tests/fp8/model_hooks.py	327	def input_hook(module, input):	CODE
MEDIUM	tests/fp8/model_hooks.py	346	def backward_hook(module, grad_input, grad_output):	CODE
MEDIUM	tests/fp8/model_hooks.py	512	def backward_hook(module, grad_input, grad_output):	CODE
526 more matches not shown…

Structural Annotation Overuse245 hits · 419 pts

Severity	File	Line	Snippet	Context
LOW	.opencode/skills/add-reward/SKILL.md	20	### Step 1: Create Reward File	COMMENT
LOW	.opencode/skills/add-reward/SKILL.md	80	### Step 2: Register in __init__.py	COMMENT
LOW	.opencode/skills/add-reward/SKILL.md	99	### Step 3: Handle Blocking Operations	COMMENT
LOW	.opencode/skills/add-reward/SKILL.md	114	### Step 4: Add Tests	COMMENT
LOW	.opencode/skills/add-dataset/SKILL.md	20	### Step 1: Create Dataset File	COMMENT
LOW	.opencode/skills/add-dataset/SKILL.md	107	### Step 2: Register in __init__.py	COMMENT
LOW	.opencode/skills/add-dataset/SKILL.md	129	### Step 3: Add Config (Optional)	COMMENT
LOW	.opencode/skills/add-dataset/SKILL.md	140	### Step 4: Add Tests	COMMENT
LOW	.opencode/skills/add-archon-model/SKILL.md	29	### Step 1: Analyze the Target Model Architecture	COMMENT
LOW	.opencode/skills/add-archon-model/SKILL.md	66	### Step 2: Select the Reference Model	COMMENT
LOW	.opencode/skills/add-archon-model/SKILL.md	90	### Step 3: Implement `args.py`	COMMENT
LOW	.opencode/skills/add-archon-model/SKILL.md	131	### Step 4: Implement `model.py`	COMMENT
LOW	.opencode/skills/add-archon-model/SKILL.md	178	### Step 5: Implement `rope.py`	COMMENT
LOW	.opencode/skills/add-archon-model/SKILL.md	200	### Step 6: Implement `state_dict_adapter.py`	COMMENT
LOW	.opencode/skills/add-archon-model/SKILL.md	243	### Step 7: Implement `parallelize.py`	COMMENT
LOW	.opencode/skills/add-archon-model/SKILL.md	282	### Step 8: Create `spec.py` and Register	COMMENT
LOW	.opencode/skills/add-archon-model/SKILL.md	315	### Step 9: Register in `__init__.py`	COMMENT
LOW	.opencode/skills/add-archon-model/SKILL.md	325	### Step 10: Verify and Test	COMMENT
LOW	.opencode/skills/add-workflow/SKILL.md	28	### Step 1: Create Workflow File	COMMENT
LOW	.opencode/skills/add-workflow/SKILL.md	105	### Step 2: Register in __init__.py	COMMENT
LOW	.opencode/skills/add-workflow/SKILL.md	118	### Step 3: Update Entry Script	COMMENT
LOW	.opencode/skills/add-workflow/SKILL.md	129	### Step 4: Add Tests	COMMENT
LOW	.opencode/skills/add-unit-tests/SKILL.md	21	### Step 1: Understand Test Types	COMMENT
LOW	.opencode/skills/add-unit-tests/SKILL.md	33	### Step 2: Create Test File Structure	COMMENT
LOW	.opencode/skills/add-unit-tests/SKILL.md	47	### Step 3: Write Test Functions	COMMENT
LOW	.opencode/skills/add-unit-tests/SKILL.md	65	### Step 4: Add Pytest Markers and CI Strategy	COMMENT
LOW⚡	.opencode/skills/add-unit-tests/SKILL.md	108	### Step 5: Mock Distributed Environment	COMMENT
LOW⚡	.opencode/skills/add-unit-tests/SKILL.md	122	### Step 6: Handle GPU Dependencies	COMMENT
LOW⚡	tests/experimental/openai/test_proxy_gateway.py	358	# Step 2: populate queue again for the post-refresh session.	COMMENT
LOW⚡	tests/experimental/openai/test_proxy_gateway.py	368	# Step 3: refresh — same api_key triggers auto-end + new session.	COMMENT
LOW	tests/experimental/openai/test_proxy_gateway.py	339	# Step 1: populate ready queue and start initial session.	COMMENT
LOW	tests/fp8/test_fp8_tensor.py	409	# Step 1: View to 4D	COMMENT
LOW	tests/fp8/test_fp8_tensor.py	423	# Step 2: Split into Q, K, V along dim=1	COMMENT
LOW	tests/fp8/test_fp8_tensor.py	451	# Step 3: Reshape to 2D	COMMENT
LOW⚡	tests/fp8/model_hooks.py	127	# Step 1: Prepare micro-batches	COMMENT
LOW⚡	tests/fp8/model_hooks.py	130	# Step 2: Define loss functions	COMMENT
LOW⚡	tests/fp8/model_hooks.py	157	# Step 3: Compute total loss weight	COMMENT
LOW⚡	tests/fp8/model_hooks.py	162	# Step 4: Forward-backward using Megatron's pipeline function	COMMENT
LOW	tests/fp8/model_hooks.py	195	# Step 5: Collect gradients before optimizer.step()	COMMENT
LOW⚡	.agents/skills/translate-doc-zh/SKILL.md	16	### Step 1: Validate the source path	COMMENT
LOW⚡	.agents/skills/translate-doc-zh/SKILL.md	23	### Step 2: Resolve the target path	COMMENT
LOW⚡	.agents/skills/translate-doc-zh/SKILL.md	28	### Step 3: Choose translation mode	COMMENT
LOW	.agents/skills/upgrade-deps/SKILL.md	172	### Step 0: Parse Input and Record Baseline	COMMENT
LOW	.agents/skills/upgrade-deps/SKILL.md	239	### Step 1: Update Pyproject Files	COMMENT
LOW	.agents/skills/upgrade-deps/SKILL.md	253	### Step 2: Lock Dependencies (Variant-Aware)	COMMENT
LOW	.agents/skills/upgrade-deps/SKILL.md	280	### Step 3: Resolve Dependency Conflicts	COMMENT
LOW	.agents/skills/upgrade-deps/SKILL.md	312	### Step 4: Update Dockerfile (If Required)	COMMENT
LOW	.agents/skills/upgrade-deps/SKILL.md	354	### Step 5: Identify Updated Focused Packages	COMMENT
LOW	.agents/skills/upgrade-deps/SKILL.md	383	### Step 6: API Compatibility Audit	COMMENT
LOW	.agents/skills/upgrade-deps/SKILL.md	466	### Step 7: Run Pre-Commit	COMMENT
LOW	.agents/skills/upgrade-deps/SKILL.md	472	### Step 8: Generate Upgrade Summary	COMMENT
LOW	.agents/skills/upgrade-deps/SKILL.md	523	### Step 9: Create PR and Trigger CI (Optional)	COMMENT
LOW	.agents/skills/add-reward/SKILL.md	20	### Step 1: Create Reward File	COMMENT
LOW	.agents/skills/add-reward/SKILL.md	80	### Step 2: Register in __init__.py	COMMENT
LOW	.agents/skills/add-reward/SKILL.md	99	### Step 3: Handle Blocking Operations	COMMENT
LOW	.agents/skills/add-reward/SKILL.md	114	### Step 4: Add Tests	COMMENT
LOW	.agents/skills/add-dataset/SKILL.md	20	### Step 1: Create Dataset File	COMMENT
LOW	.agents/skills/add-dataset/SKILL.md	107	### Step 2: Register in __init__.py	COMMENT
LOW	.agents/skills/add-dataset/SKILL.md	129	### Step 3: Add Config (Optional)	COMMENT
LOW	.agents/skills/add-dataset/SKILL.md	140	### Step 4: Add Tests	COMMENT
185 more matches not shown…

Cross-File Repetition69 hits · 345 pts

Severity	File	Line	Snippet	Context
HIGH	.opencode/skills/add-reward/SKILL.md	0	compute reward for a single completion. args: prompt: prompt string completions: completion string (model output) prompt	STRING
HIGH	.agents/skills/add-reward/SKILL.md	0	compute reward for a single completion. args: prompt: prompt string completions: completion string (model output) prompt	STRING
HIGH	.claude/skills/add-reward/SKILL.md	0	compute reward for a single completion. args: prompt: prompt string completions: completion string (model output) prompt	STRING
HIGH	.opencode/skills/add-reward/SKILL.md	0	extract the answer from a completion string. implement your extraction logic here.	STRING
HIGH	.agents/skills/add-reward/SKILL.md	0	extract the answer from a completion string. implement your extraction logic here.	STRING
HIGH	.claude/skills/add-reward/SKILL.md	0	extract the answer from a completion string. implement your extraction logic here.	STRING
HIGH	.opencode/skills/add-dataset/SKILL.md	0	load dataset for sft training. args: path: path to dataset (huggingface hub or local path) split: dataset split (train/v	STRING
HIGH	.agents/skills/add-dataset/SKILL.md	0	load dataset for sft training. args: path: path to dataset (huggingface hub or local path) split: dataset split (train/v	STRING
HIGH	.claude/skills/add-dataset/SKILL.md	0	load dataset for sft training. args: path: path to dataset (huggingface hub or local path) split: dataset split (train/v	STRING
HIGH	.opencode/skills/add-dataset/SKILL.md	0	load dataset for rl training. args: path: path to dataset split: dataset split tokenizer: tokenizer for length filtering	STRING
HIGH	.agents/skills/add-dataset/SKILL.md	0	load dataset for rl training. args: path: path to dataset split: dataset split tokenizer: tokenizer for length filtering	STRING
HIGH	.claude/skills/add-dataset/SKILL.md	0	load dataset for rl training. args: path: path to dataset split: dataset split tokenizer: tokenizer for length filtering	STRING
HIGH	.opencode/skills/add-workflow/SKILL.md	0	run a single episode. must be async and non-blocking.	STRING
HIGH	.agents/skills/add-workflow/SKILL.md	0	run a single episode. must be async and non-blocking.	STRING
HIGH	.claude/skills/add-workflow/SKILL.md	0	run a single episode. must be async and non-blocking.	STRING
HIGH	.opencode/skills/add-unit-tests/SKILL.md	0	test that function returns expected value under condition.	STRING
HIGH	.agents/skills/add-unit-tests/SKILL.md	0	test that function returns expected value under condition.	STRING
HIGH	.claude/skills/add-unit-tests/SKILL.md	0	test that function returns expected value under condition.	STRING
HIGH	tests/test_train_engine.py	0	create mock padded input data (same format for huggingface) for testing. returns a dict with input_ids, attention_mask,	STRING
HIGH	tests/test_megatron_engine.py	0	create mock padded input data (same format for huggingface) for testing. returns a dict with input_ids, attention_mask,	STRING
HIGH	tests/torchrun/run_megatron_engine_distributed.py	0	create mock padded input data (same format for huggingface) for testing. returns a dict with input_ids, attention_mask,	STRING
HIGH	tests/torchrun/run_fsdp_ulysses_forward.py	0	create mock padded input data (same format for huggingface) for testing. returns a dict with input_ids, attention_mask,	STRING
HIGH	tests/torchrun/run_fsdp_ulysses_train_batch.py	0	create mock padded input data (same format for huggingface) for testing. returns a dict with input_ids, attention_mask,	STRING
HIGH	…sts/v2/inference_service/test_data_proxy_standalone.py	0	mock arealopenai client that returns a valid chatcompletion.	STRING
HIGH	tests/v2/inference_service/test_data_proxy_version.py	0	mock arealopenai client that returns a valid chatcompletion.	STRING
HIGH	tests/v2/inference_service/test_data_proxy_pause.py	0	mock arealopenai client that returns a valid chatcompletion.	STRING
HIGH	areal/experimental/engine/archon_engine.py	0	drain this worker's client-side rtensor fetch buffer. called via rpc by ``traincontroller.clear_batches`` at step end so	STRING
HIGH	areal/engine/fsdp_engine.py	0	drain this worker's client-side rtensor fetch buffer. called via rpc by ``traincontroller.clear_batches`` at step end so	STRING
HIGH	areal/engine/megatron_engine.py	0	drain this worker's client-side rtensor fetch buffer. called via rpc by ``traincontroller.clear_batches`` at step end so	STRING
HIGH	areal/engine/sglang_remote.py	0	drain this worker's client-side rtensor fetch buffer. called via rpc by ``traincontroller.clear_batches`` at step end so	STRING
HIGH	areal/engine/vllm_remote.py	0	drain this worker's client-side rtensor fetch buffer. called via rpc by ``traincontroller.clear_batches`` at step end so	STRING
HIGH	areal/experimental/engine/archon_engine.py	0	expose local fetch-buffer stats for post-step drain verification.	STRING
HIGH	areal/engine/fsdp_engine.py	0	expose local fetch-buffer stats for post-step drain verification.	STRING
HIGH	areal/engine/megatron_engine.py	0	expose local fetch-buffer stats for post-step drain verification.	STRING
HIGH	areal/engine/sglang_remote.py	0	expose local fetch-buffer stats for post-step drain verification.	STRING
HIGH	areal/engine/vllm_remote.py	0	expose local fetch-buffer stats for post-step drain verification.	STRING
HIGH	areal/experimental/engine/archon_engine.py	0	temporarily onload parameters for offload-unsafe operations. reentrant: nested calls increment depth; only the outermost	STRING
HIGH	areal/engine/fsdp_engine.py	0	temporarily onload parameters for offload-unsafe operations. reentrant: nested calls increment depth; only the outermost	STRING
HIGH	areal/engine/megatron_engine.py	0	temporarily onload parameters for offload-unsafe operations. reentrant: nested calls increment depth; only the outermost	STRING
HIGH	areal/experimental/openai/client.py	0	set reward for a specific completion/response by its id.	STRING
HIGH	areal/experimental/openai/cache.py	0	set reward for a specific completion/response by its id.	STRING
HIGH	areal/experimental/openai/proxy/client_session.py	0	set reward for a specific completion/response by its id.	STRING
HIGH	areal/experimental/openai/client.py	0	set reward for the most recent completion/response.	STRING
HIGH	areal/experimental/openai/cache.py	0	set reward for the most recent completion/response.	STRING
HIGH	areal/experimental/openai/proxy/client_session.py	0	set reward for the most recent completion/response.	STRING
HIGH	areal/infra/scheduler/slurm.py	0	fork a single worker asynchronously. parameters ---------- command : str, optional custom module path to run instead of	STRING
HIGH	areal/infra/scheduler/local.py	0	fork a single worker asynchronously. parameters ---------- command : str, optional custom module path to run instead of	STRING
HIGH	areal/infra/scheduler/ray.py	0	fork a single worker asynchronously. parameters ---------- command : str, optional custom module path to run instead of	STRING
HIGH	areal/infra/scheduler/slurm.py	0	kill a single forked worker via its parent's rpc server.	STRING
HIGH	areal/infra/scheduler/local.py	0	kill a single forked worker via its parent's rpc server.	STRING
HIGH	areal/infra/scheduler/ray.py	0	kill a single forked worker via its parent's rpc server.	STRING
HIGH	areal/infra/scheduler/slurm.py	0	cleanup forked workers by calling kill endpoint on parent workers.	STRING
HIGH	areal/infra/scheduler/local.py	0	cleanup forked workers by calling kill endpoint on parent workers.	STRING
HIGH	areal/infra/scheduler/ray.py	0	cleanup forked workers by calling kill endpoint on parent workers.	STRING
HIGH	areal/infra/scheduler/slurm.py	0	create forked workers concurrently using async requests. parameters ---------- command : str, optional custom module pat	STRING
HIGH	areal/infra/scheduler/local.py	0	create forked workers concurrently using async requests. parameters ---------- command : str, optional custom module pat	STRING
HIGH	areal/infra/scheduler/ray.py	0	create forked workers concurrently using async requests. parameters ---------- command : str, optional custom module pat	STRING
HIGH	areal/infra/scheduler/slurm.py	0	fork new worker processes from existing workers. creates new worker processes by forking from existing workers of the ta	STRING
HIGH	areal/infra/scheduler/local.py	0	fork new worker processes from existing workers. creates new worker processes by forking from existing workers of the ta	STRING
HIGH	areal/infra/scheduler/ray.py	0	fork new worker processes from existing workers. creates new worker processes by forking from existing workers of the ta	STRING
9 more matches not shown…

Deep Nesting347 hits · 290 pts

Severity	File	Line	Context
LOW	tests/test_train_controller.py	78	CODE
LOW	tests/test_tree_training.py	108	CODE
LOW	tests/test_tree_training.py	134	CODE
LOW	tests/test_vlm_sft_dataset.py	19	CODE
LOW	tests/test_allocation_mode.py	203	CODE
LOW	tests/test_training_diagnostics_distributed.py	17	CODE
LOW	tests/test_broadcast_tensor_container.py	58	CODE
LOW	tests/test_rollout_controller.py	90	CODE
LOW	tests/test_rollout_controller.py	678	CODE
LOW	tests/test_swe_arena_multi_stream.py	1348	CODE
LOW	tests/test_local_scheduler.py	2031	CODE
LOW	tests/test_local_scheduler.py	2243	CODE
LOW	tests/test_local_scheduler.py	2364	CODE
LOW	tests/test_local_scheduler.py	2491	CODE
LOW	tests/test_packed_vs_padded_consistency.py	132	CODE
LOW	tests/test_packed_vs_padded_consistency.py	230	CODE
LOW	tests/test_sglang_pp_distributed.py	77	CODE
LOW	tests/test_fsdp_transport.py	59	CODE
LOW	tests/test_examples.py	26	CODE
LOW	tests/test_examples.py	935	CODE
LOW	tests/test_examples.py	952	CODE
LOW	tests/test_eval_rtensor_cleanup.py	195	CODE
LOW	tests/experimental/archon/test_weight_sync.py	263	CODE
LOW	tests/experimental/archon/test_weight_sync.py	440	CODE
LOW	tests/experimental/archon/test_forward.py	300	CODE
LOW	…experimental/archon/torchrun/run_pp_gradient_verify.py	308	CODE
LOW	tests/experimental/archon/torchrun/run_pp_tests.py	251	CODE
LOW	tests/experimental/archon/torchrun/run_pp_tests.py	277	CODE
LOW	tests/experimental/archon/torchrun/run_pp_tests.py	321	CODE
LOW	tests/experimental/archon/torchrun/run_pp_tests.py	569	CODE
LOW	tests/experimental/archon/torchrun/run_tp_forward.py	110	CODE
LOW	tests/experimental/archon/torchrun/dist_utils.py	155	CODE
LOW	…s/experimental/archon/torchrun/run_checkpoint_tests.py	620	CODE
LOW	…s/experimental/archon/torchrun/run_checkpoint_tests.py	810	CODE
LOW	…s/experimental/archon/torchrun/run_checkpoint_tests.py	1040	CODE
LOW	…s/experimental/archon/torchrun/run_checkpoint_tests.py	1233	CODE
LOW	…s/experimental/archon/torchrun/run_checkpoint_tests.py	1405	CODE
LOW	…ts/experimental/archon/torchrun/run_pp_combinations.py	465	CODE
LOW	tests/experimental/openai/test_vision_tensor_sharing.py	45	CODE
LOW	…/experimental/openai/test_context_overflow_recovery.py	572	CODE
LOW	tests/torchrun/run_fsdp_optimizer_dtype.py	71	CODE
LOW	tests/torchrun/run_sglang_pp_weight_sync.py	287	CODE
LOW	tests/torchrun/run_perf_tracer.py	12	CODE
LOW	tests/torchrun/run_qwen_mtp_sft.py	34	CODE
LOW	tests/torchrun/run_mopd_teacher_residency.py	67	CODE
LOW	tests/torchrun/run_megatron_engine_distributed.py	667	CODE
LOW	tests/torchrun/run_megatron_engine_vlm_distributed.py	481	CODE
LOW	…/v2/weight_update/torchrun/run_nccl_weight_transfer.py	123	CODE
LOW	tests/v2/inference_service/test_multimodal_processor.py	102	CODE
LOW	tests/fp8/test_fp8_rmsnorm.py	174	CODE
LOW	tests/fp8/test_fp8_rmsnorm.py	281	CODE
LOW	tests/fp8/model_hooks.py	102	CODE
LOW	tests/fp8/model_hooks.py	216	CODE
LOW	tests/fp8/model_hooks.py	266	CODE
LOW	tests/fp8/engine_utils.py	35	CODE
LOW	tests/megatron/torchrun/run_grad_norm_tp.py	56	CODE
LOW	docs/generate_cli_docs.py	38	CODE
LOW	docs/generate_cli_docs.py	143	CODE
LOW	docs/generate_cli_docs.py	186	CODE
LOW	examples/tir/tools/python_tool.py	120	CODE
287 more matches not shown…

Verbosity Indicators103 hits · 190 pts

Severity	File	Line	Snippet	Context
LOW⚡	tests/experimental/openai/test_proxy_gateway.py	358	# Step 2: populate queue again for the post-refresh session.	COMMENT
LOW⚡	tests/experimental/openai/test_proxy_gateway.py	368	# Step 3: refresh — same api_key triggers auto-end + new session.	COMMENT
LOW	tests/experimental/openai/test_proxy_gateway.py	339	# Step 1: populate ready queue and start initial session.	COMMENT
LOW	tests/fp8/test_fp8_tensor.py	409	# Step 1: View to 4D	COMMENT
LOW	tests/fp8/test_fp8_tensor.py	423	# Step 2: Split into Q, K, V along dim=1	COMMENT
LOW	tests/fp8/test_fp8_tensor.py	451	# Step 3: Reshape to 2D	COMMENT
LOW⚡	tests/fp8/model_hooks.py	127	# Step 1: Prepare micro-batches	COMMENT
LOW⚡	tests/fp8/model_hooks.py	130	# Step 2: Define loss functions	COMMENT
LOW⚡	tests/fp8/model_hooks.py	157	# Step 3: Compute total loss weight	COMMENT
LOW⚡	tests/fp8/model_hooks.py	162	# Step 4: Forward-backward using Megatron's pipeline function	COMMENT
LOW	tests/fp8/model_hooks.py	195	# Step 5: Collect gradients before optimizer.step()	COMMENT
LOW⚡	examples/countdown/countdown.py	452	# Step 1: 1+2=3	COMMENT
LOW⚡	examples/countdown/countdown.py	453	# Step 2: 3*3=9	COMMENT
LOW⚡	examples/countdown/countdown.py	454	# Step 3: 9*3=27	COMMENT
LOW⚡	examples/countdown/countdown.py	455	# Step 4: 27+3=30	COMMENT
LOW⚡	examples/countdown/countdown.py	459	# Step 1: 24/2 = 12	COMMENT
LOW⚡	examples/countdown/countdown.py	460	# Step 2: 12 + 2 = 14	COMMENT
LOW⚡	examples/scaffolding/controllers.py	500	# Step 1: Run generation (yields task lists for worker execution)	COMMENT
LOW⚡	examples/scaffolding/controllers.py	503	# Step 2: Create reward tasks and compute rewards locally	COMMENT
LOW⚡	…eal/experimental/models/archon/moe_weight_converter.py	249	# Step 1: Store metadata for from_hf reconstruction	COMMENT
LOW⚡	…eal/experimental/models/archon/moe_weight_converter.py	255	# Step 2: Calculate local expert indices from placements	COMMENT
LOW⚡	…eal/experimental/models/archon/moe_weight_converter.py	267	# Step 3: Store indices for from_hf reconstruction	COMMENT
LOW⚡	…eal/experimental/models/archon/moe_weight_converter.py	270	# Step 4: Create new placements for individual expert weights (2D)	COMMENT
LOW	…eal/experimental/models/archon/moe_weight_converter.py	290	# Step 5: Extract local tensor and split into individual experts	COMMENT
LOW	areal/experimental/models/archon/expert_parallel.py	129	# Step 1: Exchange token counts via all-to-all	COMMENT
LOW⚡	areal/experimental/models/archon/expert_parallel.py	162	# Step 2: All-to-all to dispatch tokens	COMMENT
LOW⚡	areal/experimental/models/archon/expert_parallel.py	170	# Step 3: Permute for grouped_mm alignment	COMMENT
LOW⚡	areal/experimental/models/archon/expert_parallel.py	203	# Step 1: Unpermute to pre-permutation order	COMMENT
LOW⚡	areal/experimental/models/archon/expert_parallel.py	206	# Step 2: All-to-all to combine (reverse direction)	COMMENT
LOW	areal/utils/save_load.py	24	# Step 1: Identify if the input is a Hugging Face repo ID or local path	COMMENT
LOW⚡	areal/utils/save_load.py	39	# Step 2: Download the repo if it's a Hugging Face repo ID	COMMENT
LOW⚡	areal/utils/save_load.py	47	# Step 3: Load all .safetensors and .bin files	COMMENT
LOW	areal/utils/data.py	2199	# Step 1: Compute mean	COMMENT
LOW	areal/utils/data.py	2241	# Step 2: Compute std	COMMENT
LOW⚡	areal/utils/functional/vocab_parallel.py	135	# Step 1: Numerical stability - subtract max	COMMENT
LOW⚡	areal/utils/functional/vocab_parallel.py	142	# Step 2: Compute exp in-place and sum across all ranks	COMMENT
LOW⚡	areal/utils/functional/vocab_parallel.py	147	# Step 3: Get the logit value at labels position	COMMENT
LOW⚡	areal/utils/functional/vocab_parallel.py	161	# Step 4: Compute log probability	COMMENT
LOW⚡	areal/utils/functional/vocab_parallel.py	165	# Step 5: Compute softmax in-place for backward (reuse exp_logits memory)	COMMENT
LOW⚡	areal/utils/functional/vocab_parallel.py	265	# Step 1: Numerical stability - subtract max (shared)	COMMENT
LOW⚡	areal/utils/functional/vocab_parallel.py	272	# Step 2: Compute exp and sum_exp (shared)	COMMENT
LOW⚡	areal/utils/functional/vocab_parallel.py	278	# Step 3: Compute softmax in-place (shared)	COMMENT
LOW⚡	areal/utils/functional/vocab_parallel.py	282	# Step 4: For logprobs - get labels logit	COMMENT
LOW⚡	areal/utils/functional/vocab_parallel.py	296	# Step 5: For entropy - compute sum(softmax * logits).	COMMENT
LOW⚡	areal/utils/functional/vocab_parallel.py	304	# Step 6: Compute final results	COMMENT
LOW	areal/utils/functional/vocab_parallel.py	346	# Step 1: Compute entropy gradient contribution	COMMENT
LOW	areal/utils/functional/vocab_parallel.py	362	# Step 2: Add logprobs gradient contribution	COMMENT
LOW	areal/utils/functional/functional.py	212	# Step 0: Validate input shapes.	COMMENT
LOW⚡	areal/utils/functional/functional.py	229	# Step 1: Compute log ratio = log(π_proximal / π_behave)	COMMENT
LOW⚡	areal/utils/functional/functional.py	235	# Step 2: Compute metric value (reuse existing KLEstimator sign conventions)	COMMENT
LOW⚡	areal/utils/functional/functional.py	262	# Step 3: Compute behave_imp_weight (needed for both modes)	COMMENT
LOW⚡	areal/utils/functional/functional.py	267	# Step 4: Aggregate and filter	COMMENT
LOW	areal/utils/functional/functional.py	413	# Step 5: Update loss_mask or keep it based on action mode	COMMENT
LOW	areal/models/transformers/vision_sp_shard.py	291	# Step 1: Get image assignment based on patch counts	COMMENT
LOW⚡	areal/models/transformers/vision_sp_shard.py	310	# Step 2: Extract local inputs (use CPU grid_thw to avoid GPU→CPU syncs	COMMENT
LOW⚡	areal/models/transformers/vision_sp_shard.py	317	# Step 3: Process local images	COMMENT
LOW⚡	areal/models/transformers/vision_sp_shard.py	342	# Step 4: Unpack deepstack if present (Qwen3-VL returns (embeddings, list[Tensor]))	COMMENT
LOW⚡	areal/models/transformers/vision_sp_shard.py	347	# Step 5: All-gather (contiguous assignment, no reordering needed)	COMMENT
LOW	areal/models/transformers/vision_sp_shard.py	363	# Step 6: All-gather deepstack embeddings (all ranks must participate)	COMMENT
LOW⚡	areal/v2/training_service/controller/controller.py	174	# Step 0: Create world_size guards via scheduler (one per GPU rank)	COMMENT
43 more matches not shown…

Modern Structural Boilerplate166 hits · 162 pts

Severity	File	Line	Snippet	Context
LOW	tests/test_staleness_manager.py	24	def set_version(self, version: int) -> None:	CODE
LOW	tests/utils.py	12	__all__ = ["get_dataset_path", "get_model_path"]	CODE
LOW	tests/experimental/archon/utils.py	34	__all__ = [	CODE
LOW	tests/infra/data_service/test_trainer_compat.py	16	def set_epoch(self, epoch: int) -> None:	CODE
LOW⚡	tests/torchrun/run_sglang_pp_weight_sync.py	45	logger = logging.getLogger(__name__)	CODE
LOW⚡	tests/v2/inference_service/test_data_proxy_pause.py	142	async def _set_paused(pause_state: PauseState, value: bool) -> None:	CODE
LOW	examples/tir/tools/__init__.py	18	__all__ = [	CODE
LOW	…perimental/inference_service/human_in_the_loop_demo.py	105	def _set_reward(gateway_addr: str, api_key: str, reward: float) -> None:	CODE
LOW	examples/scaffolding/__init__.py	35	__all__ = [	CODE
LOW	examples/scaffolding/_compat.py	36	__all__ = [	CODE
LOW	examples/scaffolding/core/controller.py	16	logger = logging.getLogger(__name__)	CODE
LOW	examples/scaffolding/core/__init__.py	36	__all__ = [	CODE
LOW	examples/terminal_bench/agent/__init__.py	5	__all__ = [	CODE
LOW	examples/terminal_bench/workflow/__init__.py	4	__all__ = ["CamelRLVRWorkflow", "build_docker_image"]	CODE
LOW	areal/__init__.py	32	__all__ = [	CODE
LOW	areal/experimental/models/archon/model_spec.py	131	__all__ = [	CODE
LOW	…eal/experimental/models/archon/moe_weight_converter.py	446	__all__ = ["MoEConversionState", "MoEWeightConverter"]	CODE
LOW	areal/experimental/models/archon/ulysses.py	18	__all__ = [	CODE
LOW	…al/experimental/models/archon/activation_checkpoint.py	310	__all__ = [	CODE
LOW	areal/experimental/models/archon/__init__.py	34	__all__ = [	CODE
LOW	areal/experimental/models/archon/utils.py	161	__all__ = [	CODE
LOW	areal/experimental/models/archon/parallel_dims.py	21	__all__ = [	CODE
LOW	areal/experimental/models/archon/expert_parallel.py	508	__all__ = [	CODE
LOW	areal/experimental/models/archon/pipeline_parallel.py	40	__all__ = [	CODE
LOW	areal/experimental/models/archon/base.py	168	__all__ = [	CODE
LOW	areal/experimental/models/archon/attention/__init__.py	7	__all__ = [	CODE
LOW	areal/experimental/models/archon/attention/varlen.py	13	__all__ = ["varlen_attn", "VarlenAttentionWrapper"]	CODE
LOW	areal/experimental/models/archon/attention/varlen.py	162	def _setup_context(ctx: Any, inputs: tuple[Any, ...], output: Any) -> None:	CODE
LOW	areal/experimental/models/archon/attention/sdpa.py	15	__all__ = ["SDPAWrapper", "create_block_causal_mask_2d"]	CODE
LOW	areal/experimental/models/archon/qwen2/__init__.py	12	__all__ = [	CODE
LOW	areal/experimental/models/archon/qwen2/spec.py	25	__all__ = ["QWEN2_SPEC"]	CODE
LOW	…/experimental/models/archon/qwen2/infra/parallelize.py	361	__all__ = [	CODE
LOW	areal/experimental/models/archon/qwen2/model/model.py	379	__all__ = [	CODE
LOW	areal/experimental/models/archon/qwen2/model/rope.py	155	__all__ = [	CODE
LOW	areal/experimental/models/archon/qwen2/model/args.py	70	__all__ = ["Qwen2ModelArgs"]	CODE
LOW	areal/experimental/models/archon/qwen3/__init__.py	12	__all__ = [	CODE
LOW	areal/experimental/models/archon/qwen3/spec.py	25	__all__ = ["QWEN3_SPEC"]	CODE
LOW	…/experimental/models/archon/qwen3/infra/parallelize.py	436	def _setup_fsdp_prefetch(model: nn.Module) -> None:	CODE
LOW	…/experimental/models/archon/qwen3/infra/parallelize.py	751	__all__ = [	CODE
LOW	areal/experimental/models/archon/qwen3/model/model.py	493	__all__ = [	CODE
LOW	areal/experimental/models/archon/qwen3/model/rope.py	12	__all__ = [	CODE
LOW	areal/experimental/models/archon/qwen3/model/args.py	111	__all__ = ["Qwen3ModelArgs"]	CODE
LOW	areal/experimental/models/archon/qwen3_5/__init__.py	12	__all__ = [	CODE
LOW	areal/experimental/models/archon/qwen3_5/spec.py	29	__all__ = ["QWEN3_5_SPEC"]	CODE
LOW	…xperimental/models/archon/qwen3_5/infra/parallelize.py	298	__all__ = [	CODE
LOW	areal/experimental/models/archon/qwen3_5/model/model.py	671	__all__ = [	CODE
LOW	areal/experimental/models/archon/qwen3_5/model/rope.py	149	__all__ = [	CODE
LOW	areal/experimental/models/archon/qwen3_5/model/args.py	182	__all__ = ["Qwen3_5ModelArgs"]	CODE
LOW	areal/experimental/models/archon/moe/token_reorderer.py	73	__all__ = ["TokenReorderer"]	CODE
LOW	areal/experimental/models/archon/moe/__init__.py	17	__all__ = [	CODE
LOW	areal/experimental/models/archon/moe/grouped_experts.py	288	__all__ = ["GroupedExperts"]	CODE
LOW	areal/experimental/models/archon/moe/utils.py	290	__all__ = [	CODE
LOW	areal/experimental/models/archon/moe/kernels.py	9	__all__ = ["generate_permute_indices", "fill_indices_wrapper"]	CODE
LOW	areal/experimental/models/archon/moe/args.py	109	__all__ = ["MoEArgs"]	CODE
LOW	areal/experimental/models/archon/moe/router.py	308	__all__ = ["TokenChoiceTopKRouter"]	CODE
LOW	areal/experimental/models/archon/moe/moe.py	269	__all__ = ["MoE", "FeedForward"]	CODE
LOW	areal/experimental/engine/archon_weight_sync.py	31	logger = logging.getLogger(__name__)	CODE
LOW	areal/experimental/openai/client.py	1903	def set_reward(self, id: str, reward: float) -> None:	CODE
LOW	areal/experimental/openai/client.py	1909	def set_last_reward(self, reward: float) -> None:	CODE
LOW	areal/experimental/openai/cache.py	89	def set_reward(self, interaction_id: str, reward: float) -> None:	CODE
106 more matches not shown…

Docstring Block Structure19 hits · 95 pts

Severity	File	Line	Snippet	Context
HIGH	areal/experimental/models/archon/fp8_checkpoint.py	234	Dequantize all FP8 weights in hf_state_dict to target_dtype. Detects FP8 weights by their ``float8_e4m3fn`` dty	STRING
HIGH	…eal/experimental/models/archon/moe_weight_converter.py	71	Calculate start/end indices for [StridedShard(dim=i), Shard(dim=i)] placement. GPU Layout (strided_shard_rank,	STRING
HIGH	areal/experimental/models/archon/qwen3/model/model.py	42	Determine if a layer should use MoE instead of dense FFN. Args: layer_id: The layer index (0-based).	STRING
HIGH	areal/experimental/openai/client.py	344	Validate that ``value`` is a list of dictionaries or BaseModel objects. Args: name: Name of the argument be	STRING
HIGH	areal/experimental/openai/client.py	553	Convert a Responses API input item into Chat Completions message dicts. Handles ``output_text``, ``input_text``, ``	STRING
HIGH	areal/infra/platforms/platform.py	69	Fallback attribute accessor for device-specific Torch modules. This method is called when the requested attribut	STRING
HIGH	areal/utils/seqpack.py	168	Return the allocation function for the given algorithm name. Args: algorithm: One of ``"ffd"`` or ``"kk"``.	STRING
HIGH	areal/utils/seqpack.py	453	Partition values into groups using the Karmarkar-Karp differencing method. This is a drop-in replacement for	STRING
HIGH	areal/utils/seqpack.py	541	Splits `nums` into K groups such that the maximum difference between group sums is minimized. Returns indices	STRING
HIGH	areal/utils/name_resolve.py	644	Get the value for a key. Args: name: Key to retrieve Returns: The value as a s	STRING
HIGH	areal/utils/name_resolve.py	660	Get a value with lock already acquired. Args: name: Key to retrieve Returns: T	STRING
HIGH	areal/utils/name_resolve.py	1057	Get the value for a key. Args: name: Key to retrieve Returns: The value as a s	STRING
HIGH	areal/utils/network.py	13	Find the local IP address for outbound route to `probe_host:probe_port`. Args: probe_host: Remote addr	STRING
HIGH	areal/utils/network.py	140	Find multiple free ports within a specified range. Args: count: Number of free ports to find p	STRING
HIGH	areal/api/alloc_mode.py	1151	Convert parsed result to _AllocationMode object. Args: result: Parsed result (list of ModelAllocati	STRING
HIGH	areal/engine/megatron_utils/fp8/config.py	9	Extract block size from quantization configuration. Args: quantization_config: Quantization configuration d	STRING
HIGH	areal/engine/megatron_utils/fp8/tensor_helper.py	313	Convert Transformer Engine Float8BlockwiseQTensor to FP8BlockwiseTensorHelper. Args: tensor: Transf	STRING
HIGH	areal/engine/fsdp_utils/optimizer.py	20	Convert string to corresponding torch dtype, only supports bfloat16 and float32. Args: dtype_str: Data	STRING
HIGH	areal/trainer/ppo/actor.py	1462	Resolve the proximal policy log-probabilities based on the method. This function determines the final proximal	STRING

Redundant / Tautological Comments64 hits · 94 pts

Severity	File	Line	Snippet	Context
LOW	tests/test_tree_training.py	258	# Check if results match with detailed error reporting	COMMENT
LOW	tests/test_tree_training.py	456	# Check if gradients are close:	COMMENT
LOW⚡	tests/test_slurm_scheduler.py	23	# Check if 'srun' exists in the system PATH	COMMENT
LOW	tests/test_fsdp_grad.py	17	# Check if CUDA is available	COMMENT
LOW	tests/test_local_scheduler.py	2093	# Check if process died	COMMENT
LOW	tests/test_examples.py	90	# Check if process has terminated	COMMENT
LOW	tests/experimental/archon/test_checkpoint_e2e.py	391	# Check if weight tying is enabled	COMMENT
LOW	tests/experimental/archon/test_qwen3_5.py	891	# Set wq to output zeros for the gate half.	COMMENT
LOW	tests/experimental/openai/test_client_with_tool.py	281	# Check if tool call was made (might depend on model capability)	COMMENT
LOW	tests/experimental/openai/test_client.py	367	# Set rewards to leaf nodes only, which should be c_a1, c_a2, c_b1	COMMENT
LOW	tests/experimental/openai/test_client.py	571	# Set rewards to leaf nodes only, which should be c_a1	COMMENT
LOW	tests/fp8/test_fp8_rmsnorm.py	296	# Check if it's a combined file or individual file	COMMENT
LOW	tests/fp8/engine_utils.py	69	# Check if this is a CUDA kernel event	COMMENT
LOW	tests/fp8/engine_utils.py	81	# Check if this is a gemm-related kernel	COMMENT
LOW	tests/fp8/engine_utils.py	128	# Check if this is a gemm-related operation (may be at higher level)	COMMENT
LOW	tests/fp8/engine_utils.py	130	# Check if we already have this in gemm_events	COMMENT
LOW	tests/fp8/test_fp8_conversion.py	81	# Check if tensor can be quantized (needs to satisfy block size requirements)	COMMENT
LOW	.claude/hooks/check-expert-update.sh	5	# Check if jq is available	COMMENT
LOW	docs/generate_cli_docs.py	147	# Check if it's a Union type (includes both Union[Type, None] and Type \| None syntax)	COMMENT
LOW	docs/generate_cli_docs.py	150	# Check if it's a union with None (optional type)	COMMENT
LOW	examples/tir/tir_workflow.py	203	# Check if tool start marker is detected	COMMENT
LOW⚡	examples/countdown/reward_score.py	31	# Check if all numbers in equation are available	COMMENT
LOW	examples/terminal_bench/agent/chat_agent_trace.py	94	# Check if we've reached max parse errors	STRING
LOW	examples/terminal_bench/agent/chat_agent_trace.py	217	# Check if this call is from a RegisteredAgentToolkit to prevent tool	COMMENT
LOW	examples/terminal_bench/agent/chat_agent_trace.py	348	# Check if we've reached max parse errors	COMMENT
LOW	.github/workflows/build-docker-image.yml	279	# Check if instance is running	COMMENT
LOW	.github/workflows/tag-release-image.yml	217	# Check if instance is running	COMMENT
LOW	areal/experimental/models/archon/qwen3/model/args.py	62	# Check if MoE is enabled	COMMENT
LOW	…mental/models/archon/qwen3/model/state_dict_adapter.py	465	# Check if all experts collected	COMMENT
LOW	areal/experimental/models/archon/qwen3_5/model/args.py	118	# Check if MoE is enabled	COMMENT
LOW	areal/experimental/openai/cache.py	399	# Assign rewards to interactions in cache based on their creation order	COMMENT
LOW	areal/infra/async_task_runner.py	522	# Check if thread is still alive	COMMENT
LOW	areal/infra/remote_inf_engine.py	233	# Check if results are InteractionWithTokenLogpReward dicts	COMMENT
LOW	areal/infra/remote_inf_engine.py	904	# Check if it's a RolloutWorkflow class	COMMENT
LOW	areal/infra/remote_inf_engine.py	915	# Check if it's a RolloutWorkflow instance	COMMENT
LOW	areal/infra/workflow_executor.py	169	# Check if all values are InteractionWithTokenLogpReward	COMMENT
LOW	areal/infra/scheduler/slurm.py	1801	# Check if retryable	COMMENT
LOW	areal/infra/scheduler/slurm.py	1137	# Check if fork mode is enabled	STRING
LOW	areal/infra/scheduler/local.py	752	# Check if fork mode is enabled	COMMENT
LOW	areal/infra/scheduler/local.py	1743	# Check if worker died (forked workers have process=None)	COMMENT
LOW	areal/infra/scheduler/local.py	1793	# Check if process has exited (forked workers have process=None)	COMMENT
LOW	areal/infra/scheduler/local.py	1831	# Check if worker died (forked workers have process=None)	COMMENT
LOW	areal/infra/scheduler/ray.py	1574	# Check if fork mode is enabled	COMMENT
LOW	areal/infra/utils/concurrent.py	71	# Check if we're already in an async context	COMMENT
LOW	areal/infra/utils/proc.py	114	# Check if stdbuf is available (not present on macOS by default)	COMMENT
LOW	areal/tools/profile_fsdp.py	383	# Print results - sorted by CUDA time	COMMENT
LOW	areal/tools/profile_archon.py	341	# Print results - sorted by CUDA time	COMMENT
LOW	areal/dataset/swe_sft/messages.py	396	# Check if current segment has any tool errors	COMMENT
LOW	areal/utils/fs.py	80	# Check if path is on network filesystem	COMMENT
LOW	areal/utils/name_resolve.py	517	# Check if key exists when replace=False	COMMENT
LOW	areal/utils/name_resolve.py	764	# Check if this is a delete event	COMMENT
LOW	areal/utils/name_resolve.py	821	# Check if key matches prefix (exact match or starts with prefix/)	COMMENT
LOW	areal/utils/name_resolve.py	970	# Check if key exists when replace=False	COMMENT
LOW	areal/utils/name_resolve.py	1024	# Check if key exists	COMMENT
LOW	areal/models/mcore/hf_load.py	522	# Check if any HF weight is FP8 (has _scale_inv suffix)	COMMENT
LOW	areal/models/transformers/qwen3_vl.py	109	# Check if hidden_states is a DTensor (TP is enabled)	COMMENT
LOW	areal/api/cli_args.py	367	# Check if equal to the default value	COMMENT
LOW	areal/api/alloc_mode.py	790	# Check if name is provided	COMMENT
LOW	areal/engine/megatron_utils/megatron.py	148	# Check if this param is truly NOT TP-sharded.	COMMENT
LOW	areal/engine/megatron_utils/fp8/tensor_helper.py	171	# Check if dim 0 is one of the last two dimensions	COMMENT
4 more matches not shown…

Self-Referential Comments26 hits · 74 pts

Severity	File	Line	Snippet	Context
MEDIUM	tests/test_train_engine.py	128	# Create the engine using from_pretrained method, without TrainEngineConfig	COMMENT
MEDIUM	tests/test_rollout_controller.py	36	# Create a mutable SchedulingSpec that can be modified by the controller	COMMENT
MEDIUM⚡	tests/experimental/archon/test_state_dict_adapter.py	63	# Create a minimal HF state dict	COMMENT
MEDIUM	tests/experimental/archon/test_weight_sync.py	189	# Create a fake tensor with correct shape on CPU	COMMENT
MEDIUM	tests/experimental/openai/test_proxy_integration.py	234	# Create a session	COMMENT
MEDIUM	tests/fp8/test_fp8_rmsnorm.py	231	# Create a dummy loss (sum of output)	COMMENT
MEDIUM⚡	examples/countdown/reward_score.py	44	# Define a regex pattern that only allows numbers, operators, parentheses, and whitespace	COMMENT
MEDIUM	examples/scaffolding/worker.py	201	# Create a worker	STRING
MEDIUM	examples/scaffolding/worker.py	239	# Create an async OpenAI client pointing to the SGLang server	COMMENT
MEDIUM	examples/terminal_bench/agent/chat_agent_trace.py	131	# Create an error tool calling record	STRING
MEDIUM	examples/terminal_bench/agent/chat_agent_trace.py	144	# Create the error record	STRING
MEDIUM	areal/infra/remote_inf_engine.py	1053	# Create a shallow copy of the input request	COMMENT
MEDIUM⚡	areal/infra/rpc/guard/engine_blueprint.py	48	# Define the constraint	COMMENT
MEDIUM	areal/tools/plot_session_trace.py	1020	# Create a separate figure for each rank	COMMENT
MEDIUM	areal/tools/plot_session_trace.py	1030	# Create a figure for this rank	COMMENT
MEDIUM	areal/tools/validate_docker_installation.py	202	# Create an FP8 recipe	COMMENT
MEDIUM	areal/tools/validate_docker_installation.py	223	# Create a simple model and optimizer	COMMENT
MEDIUM	areal/utils/recover.py	614	# This method is called by check_if_recover to check if the experiment should	COMMENT
MEDIUM	areal/models/tree_attn/tree.py	382	# Create an empty dummy trie	COMMENT
MEDIUM	areal/models/tree_attn/tree.py	773	# Create a minimal valid block mask for empty trees	COMMENT
MEDIUM	areal/v2/inference_service/sglang/scheduler.py	227	# Create a scheduler and run the event loop	COMMENT
MEDIUM	areal/workflow/openai_agent/math_agent.py	67	# Create the main orchestrator agent with handoffs	COMMENT
MEDIUM	areal/api/reward_api.py	47	"""This function is a placeholder for the reward function that will be used in the RLVR pipeline.	STRING
MEDIUM	areal/engine/megatron_utils/fp8/__init__.py	5	# This module provides FP8 quantization and dequantization utilities for	COMMENT
MEDIUM	areal/engine/megatron_utils/fp8/tensor_helper.py	32	# Create a tensor subclass that wraps rowwise_data	COMMENT
MEDIUM	areal/trainer/rl_trainer.py	1605	# Create a working copy of config	COMMENT

AI Structural Patterns66 hits · 57 pts

Severity	File	Line	Context
LOW	tests/test_rollout_controller.py	987	CODE
LOW	tests/test_dte_topology_gating.py	20	CODE
LOW	tests/experimental/archon/torchrun/dist_utils.py	21	CODE
LOW	tests/experimental/archon/torchrun/dist_utils.py	131	CODE
LOW	tests/experimental/openai/test_cache.py	28	CODE
LOW	tests/v2/inference_service/test_inf_bridge.py	63	CODE
LOW	.agents/skills/review-pr/sync_review_pr_refs.py	48	CODE
LOW	examples/tir/tir_workflow.py	273	CODE
LOW	examples/tir/tools/daytona_python_tool.py	74	CODE
LOW	examples/terminal_bench/agent/camel_terminal_agent.py	38	CODE
LOW	examples/terminal_bench/workflow/camel_rlvr_workflow.py	25	CODE
LOW	examples/swe/arena_preflight.py	233	CODE
LOW	examples/swe/arena_preflight.py	266	CODE
LOW	areal/experimental/camel/openai_model.py	194	CODE
LOW	areal/experimental/models/archon/expert_parallel.py	214	CODE
LOW	…/experimental/models/archon/qwen3/infra/parallelize.py	302	CODE
LOW	areal/experimental/models/archon/moe/utils.py	248	CODE
LOW	areal/experimental/models/archon/moe/utils.py	285	CODE
LOW	areal/experimental/engine/archon_engine.py	733	CODE
LOW	areal/experimental/openai/client.py	997	CODE
LOW	areal/experimental/openai/client.py	1024	CODE
LOW	areal/experimental/openai/client.py	1050	CODE
LOW	areal/experimental/openai/client.py	1539	CODE
LOW	areal/experimental/openai/proxy/workflow.py	110	CODE
LOW	areal/infra/dist_rollout.py	324	CODE
LOW	areal/infra/remote_inf_engine.py	1223	CODE
LOW	areal/infra/remote_inf_engine.py	1362	CODE
LOW	areal/infra/remote_inf_engine.py	1529	CODE
LOW	areal/infra/scheduler/slurm.py	138	CODE
LOW	areal/infra/scheduler/local.py	112	CODE
LOW	areal/infra/scheduler/ray.py	588	CODE
LOW	areal/infra/controller/rollout_controller.py	1161	CODE
LOW	areal/infra/controller/rollout_controller.py	1260	CODE
LOW	areal/infra/sandbox/runner.py	103	CODE
LOW	areal/dataset/__init__.py	65	CODE
LOW	areal/dataset/swe_sft/pipeline.py	506	CODE
LOW	areal/dataset/swe_sft/pipeline.py	609	CODE
LOW	areal/dataset/swe_sft/pipeline.py	673	CODE
LOW	areal/models/mcore/kda_attention.py	366	CODE
LOW	areal/models/mcore/kda_attention.py	679	CODE
LOW	areal/models/mcore/vocab_parallel_head.py	390	CODE
LOW	areal/models/mcore/lightning_attention.py	151	CODE
LOW	areal/models/mcore/lightning_attention.py	174	CODE
LOW	areal/models/mcore/lightning_attention.py	375	CODE
LOW	areal/models/transformers/vision_sp_shard.py	196	CODE
LOW	areal/v2/weight_update/nccl_group.py	18	CODE
LOW	areal/v2/weight_update/controller/controller.py	106	CODE
LOW	areal/v2/inference_service/controller/controller.py	1221	CODE
LOW	areal/v2/inference_service/controller/controller.py	1351	CODE
LOW	areal/v2/inference_service/controller/workflow.py	46	CODE
LOW	areal/api/engine_api.py	240	CODE
LOW	areal/api/engine_api.py	744	CODE
LOW	areal/api/engine_api.py	907	CODE
LOW	areal/engine/fsdp_engine.py	626	CODE
LOW	areal/engine/megatron_engine.py	1171	CODE
LOW	areal/engine/sglang_remote.py	665	CODE
LOW	areal/engine/sglang_remote.py	742	CODE
LOW	areal/engine/vllm_remote.py	431	CODE
LOW	areal/engine/vllm_remote.py	508	CODE
LOW	areal/engine/core/distributed.py	79	CODE
6 more matches not shown…

Modern AI Meta-Vocabulary17 hits · 56 pts

Severity	File	Line	Snippet	Context
MEDIUM	examples/scaffolding/search_agent_controller.py	152	# the SGLang context window.	COMMENT
MEDIUM⚡	examples/scaffolding/README.md	63	# Step 4: Create ScaffoldingLlm (orchestrates controllers with workers)	COMMENT
MEDIUM	examples/scaffolding/core/worker.py	2	# Vendored from tensorrt_llm.scaffolding.worker	COMMENT
MEDIUM	examples/scaffolding/core/task.py	2	# Vendored from tensorrt_llm.scaffolding.task	COMMENT
MEDIUM⚡	examples/scaffolding/core/math_utils.py	2	# Vendored from tensorrt_llm.scaffolding.math_utils	COMMENT
MEDIUM	examples/scaffolding/core/controller.py	2	# Vendored from tensorrt_llm.scaffolding.controller	COMMENT
MEDIUM	examples/scaffolding/core/__init__.py	2	# Vendored from TensorRT-LLM scaffolding framework.	COMMENT
MEDIUM	examples/scaffolding/core/__init__.py	3	# Core scaffolding primitives adapted for standalone use in AReaL.	COMMENT
MEDIUM	examples/scaffolding/core/scaffolding_llm.py	2	# Vendored from tensorrt_llm.scaffolding.scaffolding_llm	COMMENT
MEDIUM	examples/scaffolding/core/result.py	2	# Vendored from tensorrt_llm.scaffolding.result	COMMENT
MEDIUM	examples/scaffolding/core/task_collection.py	2	# Vendored from tensorrt_llm.scaffolding.task_collection	COMMENT
MEDIUM⚡	areal/experimental/openai/client.py	632	# prompt + generation past the backend model's context window.	COMMENT
MEDIUM	areal/utils/logging.py	34	# - white: Controllers, RPC, Inference wrappers (orchestration)	COMMENT
MEDIUM	areal/utils/logging.py	131	# Inference service - white (orchestration)	COMMENT
MEDIUM	areal/v2/weight_update/awex/delta_detect.py	472	# orchestration offloads the optimizer to CPU; consumed (or discarded)	COMMENT
MEDIUM	areal/v2/cli/inference/client.py	15	# the scaffold names directly.	COMMENT
MEDIUM	areal/v2/inference_service/vllm/bridge.py	37	# Compute effective max_new_tokens (cap by remaining context window)	COMMENT

Cross-Language Confusion7 hits · 54 pts

Severity	File	Line	Snippet	Context
HIGH⚡	tests/test_inference_targets.py	87	@pytest.mark.parametrize("contents", ["invalid JSON", "{}", '[null, {"labels": null}]'])	CODE
HIGH⚡	tests/test_adv_norm_config.py	599	# With single element, unbiased std should return zero (undefined)	COMMENT
HIGH⚡	areal/infra/utils/slurm.py	81	if kill -0 "$pid" 2>/dev/null; then	CODE
HIGH⚡	areal/infra/utils/slurm.py	83	kill "$pid" 2>/dev/null \|\| true	CODE
HIGH⚡	areal/infra/utils/slurm.py	90	if kill -0 "$pid" 2>/dev/null; then	CODE
HIGH⚡	areal/infra/utils/slurm.py	91	kill -9 "$pid" 2>/dev/null \|\| true	CODE
HIGH	areal/infra/utils/slurm.py	143	if ! kill -0 "$pid" 2>/dev/null; then	CODE

Over-Commented Block37 hits · 32 pts

Severity	File	Line	Snippet	Context
LOW	pyproject.vllm.toml	1	# vLLM variant of the AReaL project configuration.	COMMENT
LOW	tests/test_megatron_engine_distributed.py	301	# bridge_type=megatron-bridge via the runner's override map; the padded BSHD	COMMENT
LOW	tests/test_adv_norm_config.py	1161		COMMENT
LOW	tests/experimental/archon/torchrun/run_cp_forward.py	301	# Run with `torchrun --nproc_per_node=N` to test with multiple GPUs	COMMENT
LOW	tests/torchrun/run_megatron_engine_distributed.py	41	# attention layers.	COMMENT
LOW	examples/environ_ref.yaml	21	NCCL_SET_THREAD_NAME: "1"	COMMENT
LOW	examples/countdown/countdown.py	441	type=str,	COMMENT
LOW	…/swe/qwen38_flash_next/sbatch_sft_qwen38_flash_next.sh	1	#!/usr/bin/env bash	COMMENT
LOW	…lm/qwen3_5_2b_megatron_geometry3k_grpo_mtp_sglang.yaml	1	# Qwen3.5-2B GRPO on geometry3k with SGLang inference + MTP training.	COMMENT
LOW	examples/search_agent/tongyi_deepresearch/prompt.py	1	# SYSTEM_PROMPT = """You are a deep research assistant. Your core function is to conduct thorough, multi-source investig	COMMENT
LOW	examples/search_agent/tongyi_deepresearch/prompt.py	21	# import numpy as np	COMMENT
LOW	areal/infra/launcher/ray.py	1001	# The 'force' is passed to ray.cancel(future, force=force).	COMMENT
LOW	areal/infra/launcher/ray.py	1041	# usage: python -m areal.infra.launcher.ray \	COMMENT
LOW	areal/infra/platforms/platform.py	41	# use "CPU" as a fallback for platforms not registered in PyTorch	COMMENT
LOW	areal/infra/rpc/rtensor.py	321	return _backend	COMMENT
LOW	areal/infra/rpc/guard/engine_blueprint.py	621	),	COMMENT
LOW	areal/dataset/swe_sft/tokenization.py	21		COMMENT
LOW	areal/dataset/swe_sft/messages.py	221		COMMENT
LOW	areal/utils/logging.py	21	)	COMMENT
LOW	areal/utils/functional/vocab_parallel.py	341	# Total gradient = grad_logprobs * (one_hot - softmax) + grad_entropy * softmax * (E[x] - x)	COMMENT
LOW	areal/models/mcore/bailing_moe.py	101	qk_layernorm=True,	COMMENT
LOW	areal/models/mcore/hf_save.py	281	# EP all-gather: stack local-merged on dim 0, all-gather, split.	COMMENT
LOW	areal/v2/inference_service/data_proxy/app.py	1141	#	COMMENT
LOW	areal/api/cli_args.py	2241	# values (True). Using no_enable_=True correctly passes --no-enable- flags	COMMENT
LOW	areal/engine/fsdp_engine.py	541	self._per_layer_optim_wrapper = None	COMMENT
LOW	areal/engine/megatron_engine.py	661	self._apply_megatron_bridge_lora()	COMMENT
LOW	areal/engine/megatron_engine.py	1201	# Colocate mode flow (mirrors the AWEX reference integration):	COMMENT
LOW	areal/engine/megatron_engine.py	1701	)	COMMENT
LOW	areal/engine/megatron_engine.py	2681	# NOTE: Processes launched with torchrun will set the following env var to True,	COMMENT
LOW	areal/engine/megatron_engine.py	2761	else:	COMMENT
LOW	areal/engine/sglang_remote.py	321	pp_size = gen_parallel.pp_size	COMMENT
LOW	areal/engine/megatron_utils/checkpointer.py	1	# SPDX-License-Identifier: Apache-2.0	COMMENT
LOW	areal/engine/megatron_utils/checkpointer.py	421	model = model.module	COMMENT
LOW	areal/engine/megatron_utils/packed_context_parallel.py	521	# `attention_mask`: dense torch.Tensor (flex attention with Megatron) or None.	COMMENT
LOW	areal/engine/megatron_utils/fp8/__init__.py	1	# SPDX-License-Identifier: Apache-2.0	COMMENT
LOW	areal/engine/awex/colocate_writer.py	361	# peer mappings closed after our last collect (belt-and-braces with the	COMMENT
LOW	areal/engine/awex/colocate_writer.py	401	# convert_param returns the ORIGINAL tensor (shared storage)	COMMENT

AI Slop Vocabulary9 hits · 21 pts

Severity	File	Line	Snippet	Context
MEDIUM	tests/test_math_verify_reward.py	523	"""Advanced LaTeX formula tests for robust mathematical notation."""	STRING
MEDIUM	tests/experimental/archon/torchrun/run_cp_forward.py	259	# Use cosine similarity for comparison (more robust to numerical errors)	COMMENT
MEDIUM	examples/swe/qwen38_flash_next/train_rl.py	191	# Training pauses can outlast the harness default. Evaluation keeps its	COMMENT
MEDIUM	examples/search_agent/tongyi_deepresearch/prompt.py	1	# SYSTEM_PROMPT = """You are a deep research assistant. Your core function is to conduct thorough, multi-source investig	COMMENT
MEDIUM	examples/search_agent/tongyi_deepresearch/prompt.py	1	# SYSTEM_PROMPT = """You are a deep research assistant. Your core function is to conduct thorough, multi-source investig	COMMENT
MEDIUM	examples/search_agent/tongyi_deepresearch/prompt.py	38	SYSTEM_PROMPT = """You are a deep research assistant. Your core function is to conduct thorough, multi-source investigat	CODE
LOW	areal/tools/perf_trace_converter.py	919	# Replace .jsonl extension with .json, or just add .json	COMMENT
MEDIUM⚡	areal/utils/seqpack.py	387	"""Compute comprehensive packing quality metrics for a given partition.	STRING
MEDIUM	areal/v2/agent_service/worker/app.py	127	# Validate outside the harness so caught emitter errors cannot	COMMENT

Magic Placeholder Names2 hits · 10 pts

Severity	File	Line	Snippet	Context
HIGH	docs/zh/best_practices/debugging.md	72	asyncio.run(MyAgent().run(data, base_url="https://api.openai.com/v1", api_key="YOUR_API_KEY"))	CODE
HIGH	docs/en/best_practices/debugging.md	79	asyncio.run(MyAgent().run(data, base_url="https://api.openai.com/v1", api_key="YOUR_API_KEY"))	CODE

Fake / Example Data13 hits · 8 pts

Severity	File	Line	Snippet	Context
LOW	tests/v2/inference_service/test_inf_bridge.py	605	"image_url": {"url": "placeholder"},	CODE
LOW	tests/v2/inference_service/test_image_input.py	101	assert vllm_content[1]["image_url"]["url"] == "placeholder"	CODE
LOW	tests/v2/inference_service/test_image_input.py	371	"image_url": {"url": "placeholder"},	CODE
LOW	tests/v2/inference_service/test_image_input.py	408	{"type": "image_url", "image_url": {"url": "placeholder"}},	CODE
LOW	tests/v2/inference_service/test_image_input.py	409	{"type": "image_url", "image_url": {"url": "placeholder"}},	CODE
LOW	tests/v2/inference_service/test_image_input.py	446	{"type": "image_url", "image_url": {"url": "placeholder"}},	CODE
LOW	tests/v2/inference_service/test_image_input.py	447	{"type": "image_url", "image_url": {"url": "placeholder"}},	CODE
LOW	tests/v2/inference_service/test_image_input.py	479	{"type": "image_url", "image_url": {"url": "placeholder"}},	CODE
LOW	tests/v2/inference_service/test_image_input.py	1086	{"type": "image_url", "image_url": {"url": "placeholder"}}	CODE
LOW⚡	tests/v2/inference_service/test_data_proxy_chat.py	631	headers={"Authorization": "Bearer fake-key"},	CODE
LOW	examples/profile/README.md	79	The fake-data generator in `train_sft_profile.py` follows three rules:	CODE
LOW	examples/profile/train_sft_profile.py	97	"""SFT config extended with profile fake-data options."""	STRING
LOW	areal/experimental/openai/client.py	480	"image_url": {"url": "placeholder"},	CODE

Overly Generic Function Names5 hits · 4 pts

Severity	File	Line	Snippet	Context
LOW	tests/test_swe_arena.py	1340	async def run_task() -> float:	CODE
LOW	examples/scaffolding/core/worker.py	21	async def run_task(self, task: Task) -> TaskStatus:	CODE
LOW	areal/v2/agent_service/gateway/bridge.py	34	async def handle_request(self, request: Request) -> Any: ...	CODE
LOW	areal/v2/agent_service/gateway/bridge.py	48	async def handle_request(self, request: Request) -> Any:	CODE
LOW	areal/v2/agent_service/gateway/bridge.py	336	async def handle_request(self, request: Request) -> Any:	CODE

Example Usage Blocks3 hits · 4 pts

Severity	File	Line	Snippet	Context
LOW	examples/vlm/qwen3_6_27b_geometry3k_grpo.yaml	3	# Usage:	COMMENT
LOW	examples/vlm/qwen3_6_35b_a3b_geometry3k_grpo.yaml	3	# Usage:	COMMENT
LOW	scripts/uv_sync.sh	4	# Usage:	COMMENT

TODO Padding1 hit · 1 pts

Severity	File	Line	Snippet	Context
LOW	areal/models/mcore/registry.py	634	share_embeddings_and_output_weights=False, # TODO: implement share output weights	CODE

Maintenance and Governance Review

Risk Review · v3

What These Metrics Mean

Pattern-Signal History

Severity Breakdown

Directory Score Breakdown

Pattern Findings