20.5

Adjusted Signal

20.5

Raw Pattern Score

100%

Time Factor

2026-10-31

Last Push

30.1K

Stars

Mojo

Language

2.1M

Lines of Code

7.3K

Files

33.3K

Pattern Hits

2026-10-31

Scan Date

0.22

HC Hit Rate

Risk Review · Payload v1

Use each vector to guide manual verification of the underlying findings.

Payload schema: 1.0.0 · Engine: 2.0.0 · Scoring: heuristic-density-v1 · Scan date: 2026-10-31.

Cognitive Ownership Risk 46.8/100 · MEDIUM

Can maintainers explain intent, assumptions, and trade-offs? 24396 finding(s).

Hallucinated Dependency & Supply Chain Risk 11.8/100 · LOW

Are imports declared and package/API names verified? 5212 finding(s).

Agent Over-Engineering Ratio 43.7/100 · MEDIUM

Are wrappers, guards, and repeated structures necessary? 17461 finding(s).

PR Review Velocity Impact 66.3/100 · HIGH

Which changes need focused human review? 33349 finding(s).

What These Metrics Mean

Adjusted Signal: Legacy pattern score normalized per 1,000 scanned lines and multiplied by the repository-level time factor. It is not an estimate of AI authorship.
Raw Pattern Score: Sum of severity- and context-weighted pattern findings before applying the time factor.
Time Factor: A coarse multiplier based on repository last-push metadata. It does not identify when individual code was written.
Pattern Hits: Count of configured pattern findings. Review each location and surrounding code before acting.
HC Hit Rate: High- and critical-severity findings per scanned file. Severity is a configured detector weight, not proof of a vulnerability or authorship.
Lines / Files: Scan coverage denominators used by the legacy normalized score.

Pattern-Signal History

This chart tracks the legacy adjusted pattern score across scan runs. Changes can reflect source changes, scan coverage, or detector/scoring versions; they do not measure AI adoption or prove that a team refactored code. Compare runs only when their scoring versions are compatible. The dashed line tracks the raw pattern-hit count.

Severity Breakdown

Groups findings by configured severity weight. Severity is not a confidence estimate, security rating, or proof that a package is missing. Verify imports against the project's manifests and inspect each finding in its local design context.

CRITICAL 1HIGH 1613MEDIUM 2223LOW 29512

Directory Score Breakdown

This horizontal bar chart shows where configured pattern scores accumulate across directories. Treat the bars as locations to inspect, not as evidence of authorship or code quality. Use the file-level findings to decide whether a directory needs a focused ownership, dependency, complexity, or review check.

Pattern Findings

The scanner identified 33349 distinct pattern matches across 27 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 Identifiers13528 hits · 12771 pts

Severity	File	Line	Snippet	Context
LOW	…/integration-test/python/test_signal_handler_python.py	19	def test_python_stack_trace_in_signal_handler() -> None:	CODE
LOW	…egration-test/python/test_signal_handler_python_gil.py	20	def test_signal_handler_with_gil_held() -> None:	CODE
LOW⚡	max/kernels/test/gpu/fuzz/test_fuzz.py	38	def test_oracle_preserves_device_list(oracle: str) -> None:	CODE
LOW⚡	max/kernels/test/gpu/fuzz/test_fuzz.py	48	def test_collective_requires_multiple_devices(	CODE
LOW⚡	max/kernels/test/gpu/fuzz/test_fuzz.py	56	def test_replay_with_insufficient_devices(	CODE
LOW	max/kernels/benchmarks/misc/utils/bench.py	334	def bench_kineto_with_cupti_warmup(	CODE
LOW⚡	…ernels/benchmarks/misc/comparison/bench_amd_mx_gemm.py	390	def _compile_dense_preshuffle(	CODE
LOW	…kernels/benchmarks/misc/comparison/test_moe_routing.py	30	def test_each_token_picks_distinct_experts(	CODE
LOW	…kernels/benchmarks/misc/comparison/test_moe_routing.py	44	def test_balanced_spreads_experts_and_ranks(num_tokens: int) -> None:	CODE
LOW⚡	…kernels/benchmarks/misc/comparison/test_moe_routing.py	56	def test_random_routings_are_seeded() -> None:	CODE
LOW⚡	…kernels/benchmarks/misc/comparison/test_moe_routing.py	63	def test_skewed_routing_has_hot_experts() -> None:	CODE
LOW⚡	…kernels/benchmarks/misc/comparison/test_moe_routing.py	72	def test_top_k_above_num_experts_raises() -> None:	CODE
LOW⚡	…kernels/benchmarks/misc/comparison/test_moe_routing.py	77	def test_rank_counts_needs_an_even_split() -> None:	CODE
LOW	…kernels/benchmarks/misc/comparison/test_moe_routing.py	91	def test_dispatch_layout_starts_each_expert_on_a_fresh_block() -> None:	CODE
LOW	…kernels/benchmarks/misc/comparison/test_moe_routing.py	109	def test_scale_rows_capacity_pads_one_block_per_expert() -> None:	CODE
LOW	…ernels/benchmarks/misc/comparison/bench_ep_baseline.py	68	def _make_dispatch_quant_config(dispatch_dtype: DType) -> QuantConfig \| None:	CODE
LOW	…/benchmarks/misc/comparison/test_sparse_mla_indices.py	25	def test_topk_positions_are_causal_distinct_and_padded(cache_len: int) -> None:	CODE
LOW	…/benchmarks/misc/comparison/test_sparse_mla_indices.py	45	def test_topk_positions_are_seeded() -> None:	CODE
LOW	…/benchmarks/misc/comparison/test_sparse_mla_indices.py	51	def test_copies_to_exceed_counts_rows_per_request() -> None:	CODE
LOW	…/benchmarks/misc/comparison/test_sparse_mla_indices.py	63	def test_copies_to_exceed_ignores_pads() -> None:	CODE
LOW	…/benchmarks/misc/comparison/test_sparse_mla_indices.py	68	def test_physical_rows_follow_the_page_table() -> None:	CODE
LOW	…hmarks/misc/comparison/bench_blackwell_grouped_gemm.py	93	def generate_contiguous_inputs(	CODE
LOW	…x/kernels/benchmarks/misc/comparison/install_sglang.py	44	def _matching_torchvision_version() -> str:	CODE
LOW	…nchmarks/misc/comparison/bench_blackwell_mla_decode.py	138	def calculate_mla_memory_bytes(	CODE
LOW	…hmarks/misc/comparison/bench_allreduce_subgraph_max.py	114	def bench_allreduce_modular_max(	CODE
LOW	…benchmarks/misc/comparison/bench_allreduce_subgraph.py	128	def _init_sglang_parallel_state() -> None:	CODE
LOW	max/kernels/benchmarks/misc/comparison/bench_amd_mla.py	903	def _build_mla_persistent_metadata(	CODE
LOW	max/kernels/benchmarks/sweep/test_sweeptime.py	36	def test_parse_lines_skips_other_output() -> None:	STRING
LOW	max/kernels/benchmarks/sweep/test_sweeptime.py	42	def test_collect_sums_across_processes(tmp_path: Path) -> None:	STRING
LOW	max/kernels/benchmarks/sweep/test_sweeptime.py	65	def test_format_table_reports_gaps(tmp_path: Path) -> None:	STRING
LOW	max/kernels/benchmarks/sweep/test_sweeptime.py	80	def test_main_without_lines_fails(tmp_path: Path) -> None:	STRING
LOW	max/kernels/benchmarks/autotune/kdiff.py	70	def check_argo_workflow_exists(git_sha): # noqa: ANN001, ANN201	CODE
LOW	max/kernels/benchmarks/autotune/kbench_model.py	1648	def schedule_unique_build_items(	CODE
LOW	max/kernels/benchmarks/autotune/utils.py	71	def _get_visible_device_prefix(target_accelerator: str = "") -> str:	CODE
LOW	max/kernels/benchmarks/autotune/utils.py	222	def get_target_accelerator_helpstr() -> str:	CODE
LOW	max/kernels/benchmarks/autotune/utils.py	230	def check_valid_target_accelerator(target_accelerator: str) -> bool:	CODE
LOW	max/kernels/benchmarks/autotune/kplot.py	57	def _get_max_groups_per_chart(x_labels, width_px, font_size_pt): # noqa: ANN001, ANN202	CODE
LOW	max/kernels/benchmarks/autotune/tests/test_bench.py	41	def test_met_label_indicates_seconds() -> None:	CODE
LOW	max/kernels/benchmarks/autotune/tests/test_bench.py	65	def test_dump_report_time_and_throughput() -> None:	CODE
LOW	…kernels/benchmarks/autotune/tests/test_kbench_sweep.py	54	def test_sweep_collapses_each_group_to_one_process() -> None:	STRING
LOW	…kernels/benchmarks/autotune/tests/test_kbench_sweep.py	99	def test_sweep_rejects_multi_valued_comptime_param() -> None:	STRING
LOW	…kernels/benchmarks/autotune/tests/test_kbench_sweep.py	110	def test_sweep_rejects_empty_axis_and_missing_groups() -> None:	STRING
LOW	…kernels/benchmarks/autotune/tests/test_kbench_sweep.py	144	def test_sweep_filter_that_empties_the_run_warns(	STRING
LOW	…kernels/benchmarks/autotune/tests/test_kbench_sweep.py	161	def test_sweep_param_restricts_axes_and_pins_arms() -> None:	STRING
LOW	…kernels/benchmarks/autotune/tests/test_kbench_sweep.py	194	def test_sweep_build_adds_the_benchmark_directory(	STRING
LOW	…kernels/benchmarks/autotune/tests/test_kbench_sweep.py	218	def test_sweep_multi_row_csv_loads_one_row_per_benchmark(	STRING
LOW	…kernels/benchmarks/autotune/tests/test_kbench_sweep.py	253	def test_sweep_join_requires_matching_mode() -> None:	STRING
LOW⚡	max/kernels/benchmarks/autotune/tests/test_kbench.py	515	def test_get_gpu_count_nvidia_no_smi(mocker: MockerFixture) -> None:	CODE
LOW⚡	max/kernels/benchmarks/autotune/tests/test_kbench.py	521	def test_get_gpu_count_unknown_accelerator() -> None:	CODE
LOW⚡	max/kernels/benchmarks/autotune/tests/test_kbench.py	527	def test_get_gpu_count_nvidia_smi_failure(mocker: MockerFixture) -> None:	CODE
LOW⚡	max/kernels/benchmarks/autotune/tests/test_kbench.py	570	def test_get_gpu_count_amd_no_smi(mocker: MockerFixture) -> None:	CODE
LOW⚡	max/kernels/benchmarks/autotune/tests/test_kbench.py	576	def test_get_gpu_count_amd_smi_failure(mocker: MockerFixture) -> None:	CODE
LOW⚡	max/kernels/benchmarks/autotune/tests/test_kbench.py	586	def test_cli_rejects_num_gpu_exceeding_available(mocker: MockerFixture) -> None:	CODE
LOW	max/kernels/benchmarks/autotune/tests/test_kbench.py	83	def test_check_valid_target_accelerator() -> None:	CODE
LOW	max/kernels/benchmarks/autotune/tests/test_kbench.py	285	def test_scheduler_init_validates_num_gpu(tmp_path: Path) -> None:	CODE
LOW	max/kernels/benchmarks/autotune/tests/test_kbench.py	316	def test_scheduler_get_chunksize(tmp_path: Path) -> None:	CODE
LOW	max/kernels/benchmarks/autotune/tests/test_kbench.py	337	def test_scheduler_kbench_mkdir_creates_directory(tmp_path: Path) -> None:	CODE
LOW	max/kernels/benchmarks/autotune/tests/test_kbench.py	355	def test_scheduler_kbench_mkdir_run_only_requires_existing(	CODE
LOW	max/kernels/benchmarks/autotune/tests/test_kbench.py	372	def test_kbench_cache_basic_operations(tmp_path: Path) -> None:	CODE
LOW	max/kernels/benchmarks/autotune/tests/test_kbench.py	403	def test_param_define_regular() -> None:	CODE
13468 more matches not shown…

Over-Commented Block8700 hits · 8558 pts

Severity	File	Line	Snippet	Context
LOW	.pre-commit-config.yaml	1	##===----------------------------------------------------------------------===##	COMMENT
LOW	Init/include/Init/DevelopmentSignalHandler.h	1	//===----------------------------------------------------------------------===//	COMMENT
LOW	Init/include/Init/DevelopmentSignalHandler.h	21	namespace M::Init {	COMMENT
LOW	Init/include/Init/Init.h	1	//===----------------------------------------------------------------------===//	COMMENT
LOW	Init/unittests/DevelopmentSignalHandlerTests.cpp	1	//===----------------------------------------------------------------------===//	COMMENT
LOW	Init/integration-test/python/InitBindings.cpp	1	//===----------------------------------------------------------------------===//	COMMENT
LOW	Init/integration-test/python/test_script.py	1	# ===----------------------------------------------------------------------=== #	COMMENT
LOW	…/integration-test/python/test_signal_handler_python.py	1	# ===----------------------------------------------------------------------=== #	COMMENT
LOW	Init/integration-test/python/test_script_gil.py	1	# ===----------------------------------------------------------------------=== #	COMMENT
LOW	…egration-test/python/test_signal_handler_python_gil.py	1	# ===----------------------------------------------------------------------=== #	COMMENT
LOW	Init/lib/DevelopmentSignalHandler.cpp	1	//===----------------------------------------------------------------------===//	COMMENT
LOW	Init/lib/Init.cpp	1	//===----------------------------------------------------------------------===//	COMMENT
LOW	Kernels/tools/max_license/gpu-query/gpu_query.cpp	1	//===----------------------------------------------------------------------===//	COMMENT
LOW	…ls/lib/max_license/WeightsRegistry/WeightsRegistry.cpp	1	//===----------------------------------------------------------------------===//	COMMENT
LOW	…nels/lib/max_license/WeightsRegistry/WeightsRegistry.h	1	//===----------------------------------------------------------------------===//	COMMENT
LOW	Kernels/lib/max_license/Tracing/Tracing.cpp	1	//===----------------------------------------------------------------------===//	COMMENT
LOW	Kernels/lib/max_license/Tracing/Tracing.h	1	//===----------------------------------------------------------------------===//	COMMENT
LOW	Kernels/lib/max_license/Tracing/Tracing.h	61	/// Does nothing if the trace event had an error.	COMMENT
LOW	Cache/tools/cache-mgr/cache-mgr.cpp	1	//===----------------------------------------------------------------------===//	COMMENT
LOW	Cache/tools/cache-mgr/cache-mgr.cpp	21	#include "Support/ADT/SmartVariant.h"	COMMENT
LOW	Cache/include/Cache/CachedTransform.h	1	//===----------------------------------------------------------------------===//	COMMENT
LOW	Cache/include/Cache/CachedTransform.h	21	#include "Cache/BlobCache.h"	COMMENT
LOW	Cache/include/Cache/CachedTransform.h	61	/// The most basic function that performs a transformation, writing the	COMMENT
LOW	Cache/include/Cache/BlobCache.h	1	//===----------------------------------------------------------------------===//	COMMENT
LOW	Cache/include/Cache/BlobCache.h	21	#include "Support/Buffer.h"	COMMENT
LOW	Cache/include/Cache/BlobCache.h	101	virtual ErrorOr<std::optional<BufferRef>> findSyncImpl(StringRef keyHash) = 0;	COMMENT
LOW	Cache/include/Cache/CacheTelemetryContext.h	1	//===----------------------------------------------------------------------===//	COMMENT
LOW	Cache/include/Cache/Support/Keys.h	1	//===----------------------------------------------------------------------===//	COMMENT
LOW	Cache/unittests/CachedTransformTest.cpp	1	//===----------------------------------------------------------------------===//	COMMENT
LOW	Cache/unittests/BlobCacheTest.cpp	1	//===----------------------------------------------------------------------===//	COMMENT
LOW	Cache/unittests/KeyTest.cpp	1	//===----------------------------------------------------------------------===//	COMMENT
LOW	Cache/lib/CacheTelemetryContext.cpp	1	//===----------------------------------------------------------------------===//	COMMENT
LOW	Cache/lib/CachedTransform.cpp	1	//===----------------------------------------------------------------------===//	COMMENT
LOW	Cache/lib/CachedTransform.cpp	21	#include "Support/HashUtils.h"	COMMENT
LOW	Cache/lib/BlobCache.cpp	1	//===----------------------------------------------------------------------===//	COMMENT
LOW	Cache/lib/BlobCache.cpp	21	#include "Support/Buffer.h"	COMMENT
LOW	Cache/lib/BlobCache.cpp	41	#include <cstring>	COMMENT
LOW	Cache/lib/BlobCache.cpp	601	// If we have write access, do a little cache pruning on the host system in	COMMENT
LOW	Config/include/Config/Version.h	1	//===----------------------------------------------------------------------===//	COMMENT
LOW	Config/lib/Version.cpp	1	//===----------------------------------------------------------------------===//	COMMENT
LOW	Target/test/HIPPluginTest.cpp	1	//===----------------------------------------------------------------------===//	COMMENT
LOW	Target/test/DriverPluginLoadTest.cpp	1	//===----------------------------------------------------------------------===//	COMMENT
LOW	Target/test/SupportTest.cpp	1	//===----------------------------------------------------------------------===//	COMMENT
LOW	Target/test/CUDAPluginTest.cpp	1	//===----------------------------------------------------------------------===//	COMMENT
LOW	Target/test/Fixture.h	1	//===----------------------------------------------------------------------===//	COMMENT
LOW	Target/test/Fixture.h	21	#include <iostream>	COMMENT
LOW	Target/test/PluginTest.cpp	1	//===----------------------------------------------------------------------===//	COMMENT
LOW	Target/test/LegacyStubPlugin.cpp	1	//===----------------------------------------------------------------------===//	COMMENT
LOW	Target/test/HostMemoryTest.cpp	1	//===----------------------------------------------------------------------===//	COMMENT
LOW	Target/test/MetalPluginTest.cpp	1	//===----------------------------------------------------------------------===//	COMMENT
LOW	Target/test/CPUPluginTest.cpp	1	//===----------------------------------------------------------------------===//	COMMENT
LOW	Target/include/Target/DriverPlugin.h	1	//===----------------------------------------------------------------------===//	COMMENT
LOW	Target/lib/Metal/Memory.cpp	1	//===----------------------------------------------------------------------===//	COMMENT
LOW	Target/lib/Metal/Common.cpp	1	//===----------------------------------------------------------------------===//	COMMENT
LOW	Target/lib/Metal/Device.h	1	//===----------------------------------------------------------------------===//	COMMENT
LOW	Target/lib/Metal/M_driver_tma.cpp	1	//===----------------------------------------------------------------------===//	COMMENT
LOW	Target/lib/Metal/Event.h	1	//===----------------------------------------------------------------------===//	COMMENT
LOW	Target/lib/Metal/Event.h	21	#include <atomic>	COMMENT
LOW	Target/lib/Metal/Context.cpp	1	//===----------------------------------------------------------------------===//	COMMENT
LOW	Target/lib/Metal/Context.h	1	//===----------------------------------------------------------------------===//	COMMENT
8640 more matches not shown…

Decorative Section Separators1796 hits · 5909 pts

Severity	File	Line	Snippet	Context
MEDIUM	…ernels/benchmarks/misc/comparison/bench_amd_mx_gemm.py	128	# ----------------------------------------------------------------------------	COMMENT
MEDIUM	…ernels/benchmarks/misc/comparison/bench_amd_mx_gemm.py	130	# ----------------------------------------------------------------------------	COMMENT
MEDIUM⚡	…ernels/benchmarks/misc/comparison/bench_amd_mx_gemm.py	385	# ----------------------------------------------------------------------------	COMMENT
MEDIUM⚡	…ernels/benchmarks/misc/comparison/bench_amd_mx_gemm.py	387	# ----------------------------------------------------------------------------	COMMENT
MEDIUM	…ernels/benchmarks/misc/comparison/bench_amd_mx_gemm.py	627	# ----------------------------------------------------------------------------	COMMENT
MEDIUM	…ernels/benchmarks/misc/comparison/bench_amd_mx_gemm.py	629	# ----------------------------------------------------------------------------	COMMENT
MEDIUM	…ernels/benchmarks/misc/comparison/bench_amd_mx_gemm.py	727	# ----------------------------------------------------------------------------	COMMENT
MEDIUM	…ernels/benchmarks/misc/comparison/bench_amd_mx_gemm.py	729	# ----------------------------------------------------------------------------	COMMENT
MEDIUM	…nels/benchmarks/misc/comparison/bench_amd_mx_gemm.yaml	51	# ---------------------------------------------------------------------------	COMMENT
MEDIUM	…nels/benchmarks/misc/comparison/bench_amd_mx_gemm.yaml	53	# ---------------------------------------------------------------------------	COMMENT
MEDIUM	…nels/benchmarks/misc/comparison/bench_amd_mx_gemm.yaml	74	# ---------------------------------------------------------------------------	COMMENT
MEDIUM	…nels/benchmarks/misc/comparison/bench_amd_mx_gemm.yaml	76	# ---------------------------------------------------------------------------	COMMENT
MEDIUM	…nels/benchmarks/misc/comparison/bench_amd_mx_gemm.yaml	97	# ---------------------------------------------------------------------------	COMMENT
MEDIUM	…nels/benchmarks/misc/comparison/bench_amd_mx_gemm.yaml	99	# ---------------------------------------------------------------------------	COMMENT
MEDIUM	…nels/benchmarks/misc/comparison/bench_amd_mx_gemm.yaml	130	# ---------------------------------------------------------------------------	COMMENT
MEDIUM	…nels/benchmarks/misc/comparison/bench_amd_mx_gemm.yaml	135	# ---------------------------------------------------------------------------	COMMENT
MEDIUM	…nels/benchmarks/misc/comparison/bench_amd_mx_gemm.yaml	161	# ---------------------------------------------------------------------------	COMMENT
MEDIUM	…nels/benchmarks/misc/comparison/bench_amd_mx_gemm.yaml	164	# ---------------------------------------------------------------------------	COMMENT
MEDIUM	…nels/benchmarks/misc/comparison/bench_amd_mx_gemm.yaml	190	# ---------------------------------------------------------------------------	COMMENT
MEDIUM	…nels/benchmarks/misc/comparison/bench_amd_mx_gemm.yaml	198	# ---------------------------------------------------------------------------	COMMENT
MEDIUM	…/kernels/benchmarks/misc/comparison/setup_bench_env.py	106	# ---------------------------------------------------------------------------	COMMENT
MEDIUM	…/kernels/benchmarks/misc/comparison/setup_bench_env.py	108	# ---------------------------------------------------------------------------	COMMENT
MEDIUM	…/kernels/benchmarks/misc/comparison/setup_bench_env.py	159	# ---------------------------------------------------------------------------	COMMENT
MEDIUM	…/kernels/benchmarks/misc/comparison/setup_bench_env.py	161	# ---------------------------------------------------------------------------	COMMENT
MEDIUM	max/kernels/benchmarks/misc/comparison/bench_amd_mla.py	255	# ----------------------------------------------------------------------------	COMMENT
MEDIUM	max/kernels/benchmarks/misc/comparison/bench_amd_mla.py	257	# ----------------------------------------------------------------------------	COMMENT
MEDIUM⚡	max/kernels/benchmarks/misc/comparison/bench_amd_mla.py	844	# ----------------------------------------------------------------------------	COMMENT
MEDIUM⚡	max/kernels/benchmarks/misc/comparison/bench_amd_mla.py	846	# ----------------------------------------------------------------------------	COMMENT
MEDIUM	max/kernels/benchmarks/misc/comparison/bench_amd_mla.py	1519	# ----------------------------------------------------------------------------	COMMENT
MEDIUM	max/kernels/benchmarks/misc/comparison/bench_amd_mla.py	1521	# ----------------------------------------------------------------------------	COMMENT
MEDIUM	…kernels/benchmarks/gpu/bench_matmul_reducescatter.yaml	37	# -----------------------------------------------------------------------------	COMMENT
MEDIUM	…kernels/benchmarks/gpu/bench_matmul_reducescatter.yaml	39	# -----------------------------------------------------------------------------	COMMENT
MEDIUM	…kernels/benchmarks/gpu/bench_matmul_reducescatter.yaml	62	# -----------------------------------------------------------------------------	COMMENT
MEDIUM	…kernels/benchmarks/gpu/bench_matmul_reducescatter.yaml	64	# -----------------------------------------------------------------------------	COMMENT
MEDIUM	…kernels/benchmarks/gpu/bench_matmul_reducescatter.yaml	87	# =============================================================================	COMMENT
MEDIUM	…kernels/benchmarks/gpu/bench_matmul_reducescatter.yaml	89	# =============================================================================	COMMENT
MEDIUM	…kernels/benchmarks/gpu/bench_matmul_reducescatter.yaml	140	# -----------------------------------------------------------------------------	COMMENT
MEDIUM	…kernels/benchmarks/gpu/bench_matmul_reducescatter.yaml	143	# -----------------------------------------------------------------------------	COMMENT
MEDIUM	max/python/docs/check_api_coverage.py	65	# -----------------------------------------------------------------------------	COMMENT
MEDIUM	max/python/docs/check_api_coverage.py	67	# -----------------------------------------------------------------------------	COMMENT
MEDIUM	max/python/docs/check_api_coverage.py	246	# -----------------------------------------------------------------------------	COMMENT
MEDIUM	max/python/docs/check_api_coverage.py	248	# -----------------------------------------------------------------------------	COMMENT
MEDIUM	max/python/docs/check_api_coverage.py	454	# -----------------------------------------------------------------------------	COMMENT
MEDIUM	max/python/docs/check_api_coverage.py	456	# -----------------------------------------------------------------------------	COMMENT
MEDIUM	max/python/docs/check_api_coverage.py	584	# -----------------------------------------------------------------------------	COMMENT
MEDIUM	max/python/docs/check_api_coverage.py	586	# -----------------------------------------------------------------------------	COMMENT
MEDIUM	max/python/docs/check_api_coverage.py	711	# -----------------------------------------------------------------------------	COMMENT
MEDIUM	max/python/docs/check_api_coverage.py	713	# -----------------------------------------------------------------------------	COMMENT
MEDIUM	max/python/docs/check_api_coverage.py	783	# -----------------------------------------------------------------------------	COMMENT
MEDIUM	max/python/docs/check_api_coverage.py	785	# -----------------------------------------------------------------------------	COMMENT
MEDIUM	max/python/max/tree.py	822	# ─── reads ──────────────────────────────────────────────────────────────────	COMMENT
MEDIUM	max/python/max/tree.py	937	# ─── transform ──────────────────────────────────────────────────────────────	COMMENT
MEDIUM	max/python/max/benchmark/benchmark_shared/metrics.py	1351	# ---------------------------------------------------------------------------	COMMENT
MEDIUM	max/python/max/benchmark/benchmark_shared/metrics.py	1353	# ---------------------------------------------------------------------------	COMMENT
MEDIUM	max/python/max/benchmark/benchmark_shared/model_csv.py	120	# ---------------------------------------------------------------------------	COMMENT
MEDIUM	max/python/max/benchmark/benchmark_shared/model_csv.py	122	# ---------------------------------------------------------------------------	COMMENT
MEDIUM	max/python/max/benchmark/benchmark_shared/model_csv.py	262	# ---------------------------------------------------------------------------	COMMENT
MEDIUM	max/python/max/benchmark/benchmark_shared/model_csv.py	266	# ---------------------------------------------------------------------------	COMMENT
MEDIUM	max/python/max/benchmark/benchmark_shared/model_csv.py	330	# ---------------------------------------------------------------------------	COMMENT
MEDIUM	max/python/max/benchmark/benchmark_shared/model_csv.py	332	# ---------------------------------------------------------------------------	COMMENT
1736 more matches not shown…

Unused Imports5066 hits · 4643 pts

Severity	File	Line	Context
LOW	…er-connector-pyo3/python/kv_tier_connector/__init__.py	21	CODE
LOW	…er-connector-pyo3/python/kv_tier_connector/__init__.py	23	CODE
LOW	…er-connector-pyo3/python/kv_tier_connector/__init__.py	23	CODE
LOW	max/kernels/test/gpu/fuzz/fuzz.py	39	CODE
LOW	…rnels/test/gpu/compute_sanitizer/classify_synccheck.py	46	CODE
LOW	…ernels/benchmarks/misc/comparison/bench_amd_mx_gemm.py	52	CODE
LOW	…ernels/benchmarks/misc/comparison/bench_ep_baseline.py	30	CODE
LOW	…s/benchmarks/misc/comparison/bench_amd_m3_moe_chain.py	52	CODE
LOW	…s/benchmarks/misc/comparison/bench_blackwell_decode.py	21	CODE
LOW	max/kernels/benchmarks/misc/comparison/moe_routing.py	20	CODE
LOW	…nchmarks/misc/comparison/bench_blackwell_mla_decode.py	19	CODE
LOW	…hmarks/misc/comparison/bench_allreduce_subgraph_max.py	21	CODE
LOW	max/kernels/benchmarks/misc/comparison/sweep_decode.py	33	CODE
LOW	…nchmarks/misc/comparison/bench_mega_ffn_nvfp4_graph.py	29	CODE
LOW	…/benchmarks/misc/comparison/bench_blackwell_prefill.py	17	CODE
LOW	…benchmarks/misc/comparison/bench_allreduce_subgraph.py	31	CODE
LOW	…s/misc/comparison/bench_flashinfer_trtllm_moe_nvfp4.py	36	CODE
LOW	max/kernels/benchmarks/misc/comparison/bench_softmax.py	17	CODE
LOW	max/kernels/benchmarks/misc/comparison/gpu_telemetry.py	25	CODE
LOW	max/kernels/benchmarks/misc/comparison/bench_amd_mha.py	18	CODE
LOW	…s/benchmarks/misc/comparison/bench_mla_sparse_graph.py	30	CODE
LOW	max/kernels/benchmarks/misc/comparison/bench_amd_mla.py	22	CODE
LOW	…rnels/benchmarks/misc/comparison/sparse_mla_indices.py	21	CODE
LOW	…/misc/comparison/bench_flashinfer_trtllm_mla_sparse.py	27	CODE
LOW	max/kernels/benchmarks/sweep/sweeptime.py	25	CODE
LOW	max/kernels/benchmarks/autotune/terminal_viz.py	15	CODE
LOW	max/kernels/benchmarks/autotune/kbench.py	13	CODE
LOW	max/kernels/benchmarks/autotune/kbench_model.py	13	CODE
LOW	max/kernels/benchmarks/autotune/tuning_codegen.py	16	CODE
LOW	max/python/docs/check_api_coverage.py	40	CODE
LOW	max/python/max/tree.py	50	CODE
LOW	max/python/max/_validation_hooks.py	22	CODE
LOW	max/python/max/_interpreter.py	31	CODE
LOW	max/python/max/_mlir_context.py	14	CODE
LOW	…/max/benchmark/sweep_benchmark_serving_result_utils.py	23	CODE
LOW	max/python/max/benchmark/sweep_benchmark_serving.py	16	CODE
LOW	max/python/max/benchmark/serve_replay.py	16	CODE
LOW	max/python/max/benchmark/benchmark_serving.py	16	CODE
LOW	max/python/max/benchmark/results_to_csv.py	60	CODE
LOW	max/python/max/benchmark/benchmark_shared/iteration.py	21	CODE
LOW	max/python/max/benchmark/benchmark_shared/multi_turn.py	16	CODE
LOW	max/python/max/benchmark/benchmark_shared/metrics.py	26	CODE
LOW	max/python/max/benchmark/benchmark_shared/metrics.py	40	CODE
LOW	max/python/max/benchmark/benchmark_shared/metrics.py	40	CODE
LOW	…/python/max/benchmark/benchmark_shared/steady_state.py	27	CODE
LOW	max/python/max/benchmark/benchmark_shared/config.py	16	CODE
LOW	…hon/max/benchmark/benchmark_shared/prometheus_fetch.py	21	CODE
LOW	…max/benchmark/benchmark_shared/backend_metric_names.py	29	CODE
LOW	…ax/benchmark/benchmark_shared/serving_result_output.py	19	CODE
LOW	…python/max/benchmark/benchmark_shared/backend_names.py	34	CODE
LOW	max/python/max/benchmark/benchmark_shared/request.py	16	CODE
LOW	…ython/max/benchmark/benchmark_shared/server_metrics.py	16	CODE
LOW	…thon/max/benchmark/benchmark_shared/serving_metrics.py	16	CODE
LOW	max/python/max/benchmark/benchmark_shared/workload.py	16	CODE
LOW	max/python/max/benchmark/benchmark_shared/sse.py	19	CODE
LOW	…x/benchmark/benchmark_shared/gpu_metrics_prometheus.py	34	CODE
LOW	…/max/benchmark/benchmark_shared/gpu_metrics_scraper.py	33	CODE
LOW	max/python/max/benchmark/benchmark_shared/warmup.py	36	CODE
LOW	max/python/max/benchmark/benchmark_shared/utils.py	16	CODE
LOW	…x/python/max/benchmark/benchmark_shared/single_turn.py	16	CODE
5006 more matches not shown…

Cross-File Repetition705 hits · 3525 pts

Severity	File	Line	Snippet	Context
HIGH	max/python/max/nn/legacy/linear.py	0	create sharded views of this moegate module across multiple devices. args: devices: iterable of devices to place the sha	STRING
HIGH	max/python/max/nn/legacy/conv.py	0	create sharded views of this moegate module across multiple devices. args: devices: iterable of devices to place the sha	STRING
HIGH	…thon/max/nn/legacy/attention/multi_latent_attention.py	0	create sharded views of this moegate module across multiple devices. args: devices: iterable of devices to place the sha	STRING
HIGH	…/max/nn/legacy/attention/multi_latent_attention_fp8.py	0	create sharded views of this moegate module across multiple devices. args: devices: iterable of devices to place the sha	STRING
HIGH	max/python/max/nn/legacy/norm/rms_norm.py	0	create sharded views of this moegate module across multiple devices. args: devices: iterable of devices to place the sha	STRING
HIGH	…/max/pipelines/architectures/gemma3/layers/rms_norm.py	0	create sharded views of this moegate module across multiple devices. args: devices: iterable of devices to place the sha	STRING
HIGH	max/python/max/nn/legacy/norm/layer_norm.py	0	create sharded views of this moegate module across multiple devices. args: devices: iterable of devices to place the sha	STRING
HIGH	max/python/max/nn/legacy/moe/moe.py	0	create sharded views of this moegate module across multiple devices. args: devices: iterable of devices to place the sha	STRING
HIGH	…/pipelines/architectures/deepseekV3/layers/moe_gate.py	0	create sharded views of this moegate module across multiple devices. args: devices: iterable of devices to place the sha	STRING
HIGH	…/max/pipelines/architectures/gemma4/layers/rms_norm.py	0	create sharded views of this moegate module across multiple devices. args: devices: iterable of devices to place the sha	STRING
HIGH	…on/max/pipelines/architectures/qwen2_5vl/nn/decoder.py	0	create sharded views of this moegate module across multiple devices. args: devices: iterable of devices to place the sha	STRING
HIGH	…pelines/architectures/qwen2_5vl/nn/vision_attention.py	0	create sharded views of this moegate module across multiple devices. args: devices: iterable of devices to place the sha	STRING
HIGH	…ython/max/pipelines/architectures/internvl/internvl.py	0	create sharded views of this moegate module across multiple devices. args: devices: iterable of devices to place the sha	STRING
HIGH	…/pipelines/architectures/kimik2_5/layers/vision/mlp.py	0	create sharded views of this moegate module across multiple devices. args: devices: iterable of devices to place the sha	STRING
HIGH	max/python/max/nn/legacy/linear.py	0	applies the mlp2 transformation to the input. args: x: input tensor to transform. returns: the transformed tensor after	STRING
HIGH	max/python/max/nn/common_layers/mlp.py	0	applies the mlp2 transformation to the input. args: x: input tensor to transform. returns: the transformed tensor after	STRING
HIGH	…/pipelines/architectures/kimik2_5/layers/vision/mlp.py	0	applies the mlp2 transformation to the input. args: x: input tensor to transform. returns: the transformed tensor after	STRING
HIGH	…/python/max/nn/legacy/attention/attention_with_rope.py	0	an opaque kv cache optimized attention mechanism with rope.	STRING
HIGH	…thon/max/nn/legacy/attention/multi_latent_attention.py	0	an opaque kv cache optimized attention mechanism with rope.	STRING
HIGH	…/max/nn/legacy/attention/multi_latent_attention_fp8.py	0	an opaque kv cache optimized attention mechanism with rope.	STRING
HIGH	…/python/max/nn/legacy/attention/attention_with_rope.py	0	the concatenation of q, k, and v bias weight vectors.	STRING
HIGH	…/max/pipelines/architectures/olmo3/layers/attention.py	0	the concatenation of q, k, and v bias weight vectors.	STRING
HIGH	…nes/architectures/gpt_oss_modulev3/layers/attention.py	0	the concatenation of q, k, and v bias weight vectors.	STRING
HIGH	…ines/architectures/gemma3_modulev3/layers/attention.py	0	the concatenation of q, k, and v bias weight vectors.	STRING
HIGH	…on/max/pipelines/architectures/qwen2_5vl/nn/decoder.py	0	the concatenation of q, k, and v bias weight vectors.	STRING
HIGH	…lines/architectures/olmo2_modulev3/layers/attention.py	0	the concatenation of q, k, and v bias weight vectors.	STRING
HIGH	…/python/max/nn/legacy/attention/attention_with_rope.py	0	applies data-parallel fp8 latent attention and returns the output.	STRING
HIGH	…thon/max/nn/legacy/attention/multi_latent_attention.py	0	applies data-parallel fp8 latent attention and returns the output.	STRING
HIGH	…/max/nn/legacy/attention/multi_latent_attention_fp8.py	0	applies data-parallel fp8 latent attention and returns the output.	STRING
HIGH	…thon/max/nn/legacy/attention/multi_latent_attention.py	0	creates the prefill planning metadata required by fp8 mla prefill kernels. args: input_row_offsets: ragged row offsets t	STRING
HIGH	…/python/max/nn/common_layers/multi_latent_attention.py	0	creates the prefill planning metadata required by fp8 mla prefill kernels. args: input_row_offsets: ragged row offsets t	STRING
HIGH	max/python/max/nn/common_layers/quant_mla.py	0	creates the prefill planning metadata required by fp8 mla prefill kernels. args: input_row_offsets: ragged row offsets t	STRING
HIGH	…/max/nn/legacy/attention/multi_latent_attention_fp8.py	0	creates the prefill planning metadata required by fp8 mla prefill kernels. args: input_row_offsets: ragged row offsets t	STRING
HIGH	…n/max/nn/legacy/transformer/distributed_transformer.py	0	stack of attention, feedforward, and rmsnorm layers.	STRING
HIGH	max/python/max/nn/legacy/transformer/transformer.py	0	stack of attention, feedforward, and rmsnorm layers.	STRING
HIGH	…es/architectures/pixtral/vision_encoder/transformer.py	0	stack of attention, feedforward, and rmsnorm layers.	STRING
HIGH	…ectures/pixtral_modulev3/vision_encoder/transformer.py	0	stack of attention, feedforward, and rmsnorm layers.	STRING
HIGH	max/python/max/nn/legacy/moe/sigmoid_router.py	0	replicates the correction bias alongside the base gate weights.	STRING
HIGH	…/pipelines/architectures/deepseekV3/layers/moe_gate.py	0	replicates the correction bias alongside the base gate weights.	STRING
HIGH	…/max/pipelines/architectures/laguna/layers/moe_gate.py	0	replicates the correction bias alongside the base gate weights.	STRING
HIGH	max/python/max/nn/common_layers/rotary_embedding.py	0	hyperparameter used to control the frequency scaling of the sinusoidal components of the embeddings.	STRING
HIGH	…tectures/pixtral/vision_encoder/rotary_embedding_2d.py	0	hyperparameter used to control the frequency scaling of the sinusoidal components of the embeddings.	STRING
HIGH	…pixtral_modulev3/vision_encoder/rotary_embedding_2d.py	0	hyperparameter used to control the frequency scaling of the sinusoidal components of the embeddings.	STRING
HIGH	…x/python/max/_interpreter_ops/elementwise_binary_gc.py	0	the input-dtype set an op is swept over (see ``_supported_dtypes``).	STRING
HIGH	max/python/max/_interpreter_ops/reduce_axis_gc.py	0	the input-dtype set an op is swept over (see ``_supported_dtypes``).	STRING
HIGH	max/python/max/_interpreter_ops/unary_elementwise_gc.py	0	the input-dtype set an op is swept over (see ``_supported_dtypes``).	STRING
HIGH	…x/python/max/_interpreter_ops/elementwise_binary_gc.py	0	conservative swept dtype set for a (dtype_class, device).	STRING
HIGH	max/python/max/_interpreter_ops/reduce_axis_gc.py	0	conservative swept dtype set for a (dtype_class, device).	STRING
HIGH	max/python/max/_interpreter_ops/unary_elementwise_gc.py	0	conservative swept dtype set for a (dtype_class, device).	STRING
HIGH	…x/python/max/_interpreter_ops/elementwise_binary_gc.py	0	graph ``sym_name`` and cache key for one (op, device, dtype).	STRING
HIGH	max/python/max/_interpreter_ops/unary_elementwise_gc.py	0	graph ``sym_name`` and cache key for one (op, device, dtype).	STRING
HIGH	max/python/max/_interpreter_ops/topk_gc.py	0	graph ``sym_name`` and cache key for one (op, device, dtype).	STRING
HIGH	…x/python/max/_interpreter_ops/elementwise_binary_gc.py	0	per-slot counterpart of :meth:`build_module`: one device only.	STRING
HIGH	max/python/max/_interpreter_ops/reduce_axis_gc.py	0	per-slot counterpart of :meth:`build_module`: one device only.	STRING
HIGH	max/python/max/_interpreter_ops/gather_gc.py	0	per-slot counterpart of :meth:`build_module`: one device only.	STRING
HIGH	max/python/max/_interpreter_ops/scatter_gc.py	0	per-slot counterpart of :meth:`build_module`: one device only.	STRING
HIGH	max/python/max/_interpreter_ops/scatter_nd_gc.py	0	the same-bit-width unsigned int a dtype is bit-cast to for copying. raises: notimplementederror: for sub-byte dtypes (su	STRING
HIGH	max/python/max/_interpreter_ops/gather_gc.py	0	the same-bit-width unsigned int a dtype is bit-cast to for copying. raises: notimplementederror: for sub-byte dtypes (su	STRING
HIGH	max/python/max/_interpreter_ops/scatter_gc.py	0	the same-bit-width unsigned int a dtype is bit-cast to for copying. raises: notimplementederror: for sub-byte dtypes (su	STRING
HIGH	max/python/max/pipelines/sampling/sampling_config.py	0	the section name to use when loading this config from a maxconfig file. this is used to differentiate between different	STRING
645 more matches not shown…

Docstring Block Structure626 hits · 2912 pts

Severity	File	Line	Snippet	Context
HIGH	…/kernels/benchmarks/misc/comparison/setup_bench_env.py	112	Ensure a source repository exists, cloning if necessary. Args: name: Repository key in REPOS dict (e.g., "d	STRING
HIGH	max/python/max/tree.py	261	Names a node class, so a serialized tree structure can refer to it. .. code-block:: python @tree.kind("pag	STRING
HIGH	max/python/max/tree.py	377	Takes one interior node apart, one level deep. Args: value: An interior node, as :func:`is_node` judges it.	STRING
HIGH	max/python/max/tree.py	490	Flattens ``tree`` using this structure to locate the leaves. The walk descends into ``tree`` wherever this stru	STRING
HIGH	max/python/max/tree.py	624	Splits ``tree`` into its leaves and the structure around them. Args: tree: The value to take apart.	STRING
HIGH	max/python/max/tree.py	688	Rebuilds a tree from a structure and its leaves. The inverse of :func:`flatten`, always building a fresh tree.	STRING
HIGH	max/python/max/tree.py	859	Returns ``tree``'s leaves keyed by their dotted path. The result is what :func:`update` consumes. Args:	STRING
HIGH	max/python/max/tree.py	947	Builds a new tree with each leaf replaced by what ``f`` returns. With extra trees in ``rest``, ``f`` receives one l	STRING
HIGH	max/python/max/tree.py	988	Writes path-keyed values into ``tree``, in place. The counterpart of :func:`paths`, taking the same knobs. Only lea	STRING
HIGH	max/python/max/_interpreter.py	234	Execute an MO graph and return output buffers. Args: graph: The finalized MO graph to execute. inpu	STRING
HIGH	max/python/max/benchmark/sweep_benchmark_serving.py	152	Persists and reports each iteration of a sweep as it is produced. For each iteration with a result, writes ``config	STRING
HIGH	max/python/max/benchmark/benchmark_serving.py	201	Choose the GPU-stats recorder for this run, or None to disable it. The remote sources are source selectors, not add	STRING
HIGH	max/python/max/benchmark/benchmark_serving.py	1462	Runs a serving benchmark sweep. Yields one :class:`BenchmarkRunResult` per (concurrency, request rate) combinat	STRING
HIGH	max/python/max/benchmark/benchmark_shared/config.py	293	Resolve one argument into parsed JSON, from any of its spellings. A string is inline JSON when it starts with ``inl	STRING
HIGH	…hon/max/benchmark/benchmark_shared/prometheus_fetch.py	27	Fetch raw metrics text from a Prometheus endpoint. Args: url: Prometheus metrics endpoint URL. time	STRING
HIGH	…ython/max/benchmark/benchmark_shared/server_metrics.py	131	Get a histogram by name and optional labels. Args: metric_name: Base metric name (such as "maxserve	STRING
HIGH	…ython/max/benchmark/benchmark_shared/server_metrics.py	192	Get the metrics URL for a backend. Args: backend: Backend name (Backend enum) base_url: Base API UR	STRING
HIGH	…ython/max/benchmark/benchmark_shared/server_metrics.py	329	Fetch and parse metrics for a backend. Convenience function that combines get_metrics_url, fetch_metrics, and parse	STRING
HIGH	max/python/max/benchmark/benchmark_shared/workload.py	44	Loads a workload YAML file's options. A workload file is either flat, with options at the root, or selects the	STRING
HIGH	…x/benchmark/benchmark_shared/lora_benchmark_manager.py	112	Get LoRA name for a given request index. Args: request_idx: The request index Returns:	STRING
HIGH	max/python/max/benchmark/benchmark_shared/model_csv.py	406	Loads a result JSON file and returns one flattened row per result. Supports the shapes the benchmarks emit: -	STRING
HIGH	max/python/max/benchmark/benchmark_shared/model_csv.py	461	Chooses the ordered CSV columns from those available in the input. The default (no options) is ``profile.summary``,	STRING
HIGH	max/python/max/benchmark/benchmark_shared/model_csv.py	615	Reads result JSON ``inputs`` and writes a CSV with selected columns. Args: inputs: The result J	STRING
HIGH	max/python/max/benchmark/benchmark_shared/model_csv.py	878	Applies a function to every key in a mapping and returns a new dict. Args: f: The function to a	STRING
HIGH	…n/max/benchmark/benchmark_shared/datasets/interface.py	79	Factory method to create the appropriate dataset subclass based on dataset name. This factory method automatica	STRING
HIGH	…n/max/benchmark/benchmark_shared/datasets/interface.py	142	Get the appropriate dataset class for the given dataset name. Args: dataset_name: Name of the datas	STRING
HIGH	…n/max/benchmark/benchmark_shared/datasets/interface.py	223	Sample requests from the dataset. This is the standardized interface that all dataset implementations must foll	STRING
HIGH	…x/benchmark/benchmark_shared/datasets/agentic_tools.py	96	Build the per-tool load profiles from a ``tools`` mapping. Args: profiles: A mapping holding a ``tools`` li	STRING
HIGH	…ython/max/benchmark/benchmark_shared/datasets/types.py	252	Read an image file as raw bytes and encode in base64. Preserves the exact original file data, with no transformatio	STRING
HIGH	…/python/max/benchmark/benchmark_shared/datasets/all.py	96	Resolve ``--agentic-tool-profiles``, rejecting incoherent combinations. Args: args: The parsed benchmark co	STRING
HIGH	…ax/benchmark/benchmark_shared/datasets/distribution.py	105	Parse a distribution parameter that the caller must supply. Args: param: An int or float, or a dist	STRING
HIGH	…ax/benchmark/benchmark_shared/datasets/distribution.py	127	Parse a distribution parameter into a concrete distribution instance. Args: param: An int or float,	STRING
HIGH	…ax/benchmark/benchmark_shared/datasets/distribution.py	275	Parse a string like "Cat(v1:w1, v2:w2, ...)" into a CategoricalDistribution. Args: schema: A string	STRING
HIGH	…ax/benchmark/benchmark_shared/datasets/distribution.py	336	Parse a string like "N(mean, std)" into a NormalDistribution. Args: schema: A string in the format	STRING
HIGH	…ax/benchmark/benchmark_shared/datasets/distribution.py	383	Parse a string like "U(lower, upper)" into a UniformDistribution. Args: schema: A string in the for	STRING
HIGH	…ax/benchmark/benchmark_shared/datasets/distribution.py	429	Parse a string like "G(shape, scale)" into a GammaDistribution. Args: schema: A string in the forma	STRING
HIGH	…ax/benchmark/benchmark_shared/datasets/distribution.py	482	Parse a string like "LN(mean, std)" into a LogNormalDistribution. Args: schema: A string in the for	STRING
HIGH	…ax/benchmark/benchmark_shared/datasets/distribution.py	548	Parse a string like "Burr12(c, d, scale)" into a Burr12Distribution. Args: schema: A string in the	STRING
HIGH	…ax/benchmark/benchmark_shared/datasets/distribution.py	603	Parse a string like "DU(lower, upper)" into a DiscreteUniformDistribution. Args: schema: A string i	STRING
HIGH	…ax/benchmark/benchmark_shared/datasets/distribution.py	662	Parse a string like "NB(n, p)" into a NegativeBinomialDistribution. Args: schema: A string in the f	STRING
HIGH	…chmark/benchmark_shared/datasets/image_augmentation.py	94	Sample pixel dimensions from long-side and aspect-ratio distributions. Args: long_side_dist: Distribution f	STRING
HIGH	max/python/max/experimental/torch/torch.py	82	Converts a MAX DType to the corresponding torch dtype. Args: dtype: The MAX DType to convert. Returns:	STRING
HIGH	max/python/max/experimental/torch/torch.py	99	Converts a torch dtype to the corresponding MAX DType. Args: dtype: The torch dtype to convert. Return	STRING
HIGH	…python/max/experimental/cascade/grpc_runtime/client.py	197	Fetch and decode a single result from the remote server by id. Returns raw JSON-like data (:py:data:`CascadeVal	STRING
HIGH	…python/max/experimental/cascade/grpc_runtime/client.py	236	Stream items from a bound server-side iterator, decoded one by one. Yields raw JSON-like data (:py:data:`Cascad	STRING
HIGH	max/python/max/dtype/dtype.py	88	Converts a NumPy dtype to the corresponding DType. Args: dtype: The NumPy dtype to convert. Returns:	STRING
HIGH	max/python/max/nn/_weight_utils.py	380	Builds the weight registry for a model. Each entry of ``arrays`` is checked against the weight it loads into, c	STRING
HIGH	max/python/max/nn/hyper_connection.py	118	Computes the mHC gates and collapses the streams for the sublayer. Args: hidden_streams: The parall	STRING
HIGH	max/python/max/nn/module.py	290	Defines the computation performed by the module. Users must override this method in their subclass to define th	STRING
HIGH	max/python/max/nn/module.py	513	Transfers all module parameters to a single device. Records the target as :attr:`device`. Build the input types	STRING
HIGH	max/python/max/nn/module.py	641	Traces ``forward`` and compiles it into a callable. The graph refers to each parameter by name, and ``weights``	STRING
HIGH	max/python/max/nn/module.py	745	Traces ``forward`` and builds the weight registry it binds. This is the tracing half of :meth:`compile`. Use it	STRING
HIGH	max/python/max/nn/legacy/linear.py	522	Applies a linear transformation to the input data. Args: x: Input :class:`~max.graph.TensorValue` o	STRING
HIGH	max/python/max/nn/legacy/linear.py	744	Applies a linear transformation to the input data. Args: x: Input sequence of :class:`~max.graph.Te	STRING
HIGH	max/python/max/nn/legacy/hyper_connection.py	139	Computes the mHC gates and collapses the streams for the sublayer. Args: hidden_streams: The parall	STRING
HIGH	max/python/max/nn/legacy/kernels.py	215	Computes fused query, key, and value projections with padded input. This is for non-ragged (padded batch) inputs wh	STRING
HIGH	max/python/max/nn/legacy/kernels.py	351	Prepares chunked QKV projections and appends to a paged KV cache. Applies partial NeoX RoPE to Q/K, scales V in FP3	STRING
HIGH	max/python/max/nn/legacy/kernels.py	1341	Computes MiniMax-M3's fused QKV + index-QK projections in one MXFP8 GEMM. A 5-way fusion: ``input`` and ``wqkv`` ca	STRING
HIGH	max/python/max/nn/legacy/kernels.py	1489	Computes MiniMax-M3's fused QKV + index-QK projections in one MXFP6 GEMM. A 5-way fusion: ``input`` and ``wqkv`` ca	STRING
HIGH	max/python/max/nn/legacy/kernels.py	1631	Computes MiniMax-M3's fused QKV + index-QK projections in one BF16 GEMM. Non-scaled BF16 analog of ``_fused_qkv_ind	STRING
566 more matches not shown…

Empty Error Handlers153 hits · 765 pts

Severity	File	Line	Snippet	Context
HIGH	max/kernels/benchmarks/misc/utils/bench.py	52	except ImportError:	CODE
HIGH	max/kernels/benchmarks/misc/utils/bench.py	385	except (AssertionError, ZeroDivisionError, RuntimeError):	CODE
HIGH	max/kernels/benchmarks/misc/utils/bench.py	302	except (ValueError, ZeroDivisionError):	CODE
HIGH	max/kernels/benchmarks/autotune/kbench_model.py	1296	except Exception:	CODE
HIGH	max/kernels/benchmarks/autotune/kbench_model.py	259	except Exception:	CODE
HIGH	max/kernels/benchmarks/autotune/kprofile.py	637	except:	CODE
HIGH	max/python/docs/sphinx_ext/docstring_sources.py	141	except Exception:	CODE
HIGH	max/python/max/_interpreter.py	77	except Exception:	CODE
HIGH	max/python/max/benchmark/serve_replay.py	57	except Exception:	CODE
HIGH	max/python/max/benchmark/serve_replay.py	80	except Exception:	CODE
HIGH	max/python/max/benchmark/serve_replay.py	94	except Exception:	CODE
HIGH	max/python/max/benchmark/benchmark_shared/utils.py	152	except (urllib.error.URLError, ConnectionError, OSError):	CODE
HIGH	…/python/max/benchmark/benchmark_shared/datasets/all.py	158	except Exception:	CODE
HIGH	…ax/benchmark/benchmark_shared/datasets/distribution.py	153	except ValueError:	CODE
HIGH	…python/max/experimental/cascade/grpc_runtime/server.py	131	except asyncio.CancelledError:	CODE
HIGH	…ython/max/experimental/cascade/grpc_runtime/subproc.py	72	except FileNotFoundError:	CODE
HIGH	…python/max/experimental/cascade/grpc_runtime/client.py	82	except grpc.RpcError:	CODE
HIGH	…ython/max/experimental/cascade/http_runtime/subproc.py	128	except aiohttp.ClientError:	CODE
HIGH	max/python/max/nn/legacy/parallel/numpy_parallel_ops.py	100	except Exception:	CODE
HIGH	max/python/max/nn/legacy/parallel/numpy_parallel_ops.py	108	except Exception:	CODE
HIGH	max/python/max/nn/legacy/parallel/numpy_parallel_ops.py	126	except Exception:	CODE
HIGH	max/python/max/config/__init__.py	387	except ValueError:	CODE
HIGH	max/python/max/pipelines/context/video.py	241	except ValueError:	CODE
HIGH	max/python/max/pipelines/context/video.py	580	except av.error.FFmpegError:	CODE
HIGH	max/python/max/pipelines/context/video.py	603	except av.error.FFmpegError:	CODE
HIGH	max/python/max/pipelines/context/video.py	621	except av.error.FFmpegError:	CODE
HIGH	max/python/max/pipelines/architectures/__init__.py	473	except ModuleNotFoundError:	CODE
HIGH	max/python/max/pipelines/architectures/__init__.py	481	except ModuleNotFoundError:	CODE
HIGH	max/python/max/pipelines/architectures/__init__.py	488	except ModuleNotFoundError:	CODE
HIGH	max/python/max/pipelines/architectures/__init__.py	496	except ModuleNotFoundError:	CODE
HIGH	max/python/max/pipelines/architectures/__init__.py	503	except ModuleNotFoundError:	CODE
HIGH	max/python/max/pipelines/architectures/__init__.py	511	except ModuleNotFoundError:	CODE
HIGH	max/python/max/pipelines/architectures/__init__.py	521	except ModuleNotFoundError:	CODE
HIGH	max/python/max/pipelines/architectures/__init__.py	531	except ModuleNotFoundError:	CODE
HIGH	max/python/max/pipelines/architectures/__init__.py	543	except ModuleNotFoundError:	CODE
HIGH	…/python/max/pipelines/architectures/hf_config_shims.py	67	except ValueError:	CODE
HIGH	…/python/max/pipelines/architectures/hf_config_shims.py	79	except ValueError:	CODE
HIGH	…/python/max/pipelines/architectures/hf_config_shims.py	144	except ValueError:	CODE
HIGH	…/python/max/pipelines/architectures/hf_config_shims.py	206	except ValueError:	CODE
HIGH	…/python/max/pipelines/architectures/hf_config_shims.py	317	except ValueError:	CODE
HIGH	…/python/max/pipelines/architectures/hf_config_shims.py	356	except ValueError:	CODE
HIGH	…/python/max/pipelines/architectures/hf_config_shims.py	688	except ValueError:	CODE
HIGH	…/python/max/pipelines/architectures/hf_config_shims.py	51	except ValueError:	CODE
HIGH	…/python/max/pipelines/architectures/hf_config_shims.py	720	except ValueError:	CODE
HIGH	…/python/max/pipelines/architectures/hf_config_shims.py	750	except ValueError:	CODE
HIGH	…thon/max/pipelines/architectures/laguna/tool_parser.py	81	except json.JSONDecodeError:	CODE
HIGH	…hon/max/pipelines/architectures/laguna/model_config.py	578	except (AttributeError, TypeError):	CODE
HIGH	…hon/max/pipelines/architectures/laguna/model_config.py	595	except (AttributeError, TypeError):	CODE
HIGH	…pelines/architectures/gemma4_assistant/model_config.py	54	except ValueError:	CODE
HIGH	…ython/max/pipelines/architectures/hy_v4/tool_parser.py	267	except Exception:	CODE
HIGH	…thon/max/pipelines/architectures/gemma4/tool_parser.py	78	except ValueError:	CODE
HIGH	…elines/architectures/deepseekV41/engram_shared_host.py	118	except OSError:	CODE
HIGH	…elines/architectures/deepseekV41/engram_shared_host.py	153	except FileExistsError:	CODE
HIGH	…elines/architectures/deepseekV41/engram_shared_host.py	626	except OSError:	CODE
HIGH	…elines/architectures/deepseekV41/engram_shared_host.py	770	except OSError:	CODE
HIGH	…pipelines/architectures/deepseekV41/weight_adapters.py	190	except OSError:	CODE
HIGH	…pipelines/architectures/deepseekV41/weight_adapters.py	257	except OSError:	CODE
HIGH	…pipelines/architectures/deepseekV41/weight_adapters.py	277	except OSError:	CODE
HIGH	…pelines/architectures/kimik2_5_modulev3/tool_parser.py	84	except (ValueError, IndexError):	CODE
HIGH	…on/max/pipelines/architectures/kimik2_5/tool_parser.py	84	except (ValueError, IndexError):	CODE
93 more matches not shown…

Deep Nesting699 hits · 628 pts

Severity	File	Line	Context
LOW	max/kernels/test/gpu/fuzz/fuzz.py	1067	CODE
LOW	max/kernels/test/gpu/fuzz/fuzz.py	1333	CODE
LOW	max/kernels/benchmarks/misc/analyze_4wave_autotune.py	55	CODE
LOW	max/kernels/benchmarks/misc/utils/bench.py	61	CODE
LOW	max/kernels/benchmarks/misc/utils/bench.py	186	CODE
LOW	max/kernels/benchmarks/misc/utils/bench.py	273	CODE
LOW	…ernels/benchmarks/misc/comparison/bench_ep_baseline.py	457	CODE
LOW	…s/benchmarks/misc/comparison/bench_amd_m3_moe_chain.py	247	CODE
LOW	…s/benchmarks/misc/comparison/bench_mxfp8_quant_vllm.py	48	CODE
LOW	…s/benchmarks/misc/comparison/bench_mxfp8_quant_vllm.py	176	CODE
LOW	max/kernels/benchmarks/misc/comparison/sweep_decode.py	50	CODE
LOW	…nchmarks/misc/comparison/bench_mega_ffn_nvfp4_graph.py	322	CODE
LOW	…/benchmarks/misc/comparison/bench_blackwell_prefill.py	444	CODE
LOW	…s/misc/comparison/bench_flashinfer_trtllm_moe_nvfp4.py	268	CODE
LOW	max/kernels/benchmarks/autotune/kbench_model.py	1384	CODE
LOW	max/kernels/benchmarks/autotune/kbench_model.py	697	CODE
LOW	max/kernels/benchmarks/autotune/kbench_model.py	747	CODE
LOW	max/kernels/benchmarks/autotune/kbench_model.py	842	CODE
LOW	max/kernels/benchmarks/autotune/kbench_model.py	934	CODE
LOW	max/kernels/benchmarks/autotune/kbench_model.py	1648	CODE
LOW	max/kernels/benchmarks/autotune/kbench_model.py	1879	CODE
LOW	max/kernels/benchmarks/autotune/kbench_model.py	2040	CODE
LOW	max/kernels/benchmarks/autotune/kbench_model.py	2078	CODE
LOW	max/kernels/benchmarks/autotune/kbench_model.py	2136	CODE
LOW	max/kernels/benchmarks/autotune/tuning_codegen.py	95	CODE
LOW	max/kernels/benchmarks/autotune/kplot.py	111	CODE
LOW	max/kernels/benchmarks/autotune/kprofile.py	750	CODE
LOW	max/python/docs/check_api_coverage.py	91	CODE
LOW	max/python/docs/check_api_coverage.py	217	CODE
LOW	max/python/docs/check_api_coverage.py	469	CODE
LOW	max/python/docs/sphinx_ext/docstring_sources.py	70	CODE
LOW	max/python/docs/sphinx_ext/source_files.py	62	CODE
LOW	max/python/docs/sphinx_ext/source_files.py	115	CODE
LOW	max/python/max/tree.py	685	CODE
LOW	max/python/max/tree.py	981	CODE
LOW	max/python/max/tree.py	737	CODE
LOW	max/python/max/tree.py	1015	CODE
LOW	max/python/max/_interpreter.py	279	CODE
LOW	max/python/max/benchmark/sweep_benchmark_serving.py	145	CODE
LOW	max/python/max/benchmark/serve_replay.py	65	CODE
LOW	max/python/max/benchmark/benchmark_serving.py	295	CODE
LOW	max/python/max/benchmark/benchmark_serving.py	866	CODE
LOW	max/python/max/benchmark/benchmark_serving.py	1374	CODE
LOW	max/python/max/benchmark/benchmark_shared/multi_turn.py	86	CODE
LOW	max/python/max/benchmark/benchmark_shared/multi_turn.py	794	CODE
LOW	max/python/max/benchmark/benchmark_shared/metrics.py	1455	CODE
LOW	max/python/max/benchmark/benchmark_shared/metrics.py	387	CODE
LOW	…ax/benchmark/benchmark_shared/serving_result_output.py	248	CODE
LOW	…ax/benchmark/benchmark_shared/serving_result_output.py	459	CODE
LOW	max/python/max/benchmark/benchmark_shared/request.py	545	CODE
LOW	max/python/max/benchmark/benchmark_shared/request.py	1525	CODE
LOW	…ython/max/benchmark/benchmark_shared/server_metrics.py	262	CODE
LOW	…ython/max/benchmark/benchmark_shared/server_metrics.py	583	CODE
LOW	…thon/max/benchmark/benchmark_shared/serving_metrics.py	304	CODE
LOW	max/python/max/benchmark/benchmark_shared/sse.py	82	CODE
LOW	…/max/benchmark/benchmark_shared/gpu_metrics_scraper.py	171	CODE
LOW	max/python/max/benchmark/benchmark_shared/warmup.py	247	CODE
LOW	max/python/max/benchmark/benchmark_shared/utils.py	100	CODE
LOW	max/python/max/benchmark/benchmark_shared/model_csv.py	811	CODE
LOW	…nchmark/benchmark_shared/datasets/tool_augmentation.py	43	CODE
639 more matches not shown…

Self-Referential Comments233 hits · 579 pts

Severity	File	Line	Snippet	Context
MEDIUM	…local_transformers_setup/cleanup_local_transformers.sh	30	# Create a temp file to build the cleaned version	COMMENT
MEDIUM	…enchmarks/misc/utils/bench_vllm_flashinfer_decoding.py	88	# Create the block tables.	COMMENT
MEDIUM	…nchmarks/misc/utils/bench_vllm_flashinfer_attention.py	82	# Create the block tables.	COMMENT
MEDIUM	…ernels/benchmarks/misc/utils/warp_profiling/profile.py	237	# Create a dictionary of dataframes, one for each group	COMMENT
MEDIUM	…benchmarks/misc/comparison/bench_allreduce_subgraph.py	193	# Create a CPU group for coordination (CustomAllreduce needs this)	COMMENT
MEDIUM	…chmarks/gpu/linalg/bench_matmul_llama3_70B_4xH100.yaml	20	# This file is facilitate development before GTC and should be combined	COMMENT
MEDIUM	max/python/docs/build-rsts.sh	71	# Create the directory if it does not exist	COMMENT
MEDIUM	max/python/max/_interpreter.py	248	# Create a new interpreter slots dictionary for this execution.	COMMENT
MEDIUM	max/python/max/benchmark/benchmark_serving.py	385	# Create a request driver instance without pbar for test prompt	COMMENT
MEDIUM	…hmark/benchmark_shared/datasets/arxiv_summarization.py	75	# Create a summarization prompt	COMMENT
MEDIUM	…hmark/benchmark_shared/datasets/arxiv_summarization.py	112	# Create the full prompt	COMMENT
MEDIUM	…/python/max/nn/legacy/attention/multihead_attention.py	166	# Create a module instance for this device	COMMENT
MEDIUM	max/python/max/nn/legacy/moe/quant_strategy.py	372	# Create a dummy expert scales tensor with shape (num_experts,) for	COMMENT
MEDIUM	max/python/max/config/__init__.py	1052	# Create a wrapper class to override parse_args method	COMMENT
MEDIUM	max/python/max/graph/quantization.py	80	# Create a quantized weight using Q4_K encoding	STRING
MEDIUM	max/python/max/graph/graph.py	910	# Create an always-ready chain that is never advanced by the graph.	COMMENT
MEDIUM	max/python/max/graph/type.py	136	# Create a GPU device reference (default id=0)	STRING
MEDIUM	max/python/max/graph/type.py	139	# Create a CPU device with specific id	STRING
MEDIUM	max/python/max/graph/type.py	375	# Create a tensor type with float32 elements and static dimensions 2x3	STRING
MEDIUM	max/python/max/graph/type.py	543	# Create an opaque type with parameters	STRING
MEDIUM⚡	max/python/max/graph/value.py	378	# Create a sample matrix	STRING
MEDIUM⚡	max/python/max/graph/value.py	381	# Create a Graph context to work with tensors	STRING
MEDIUM⚡	max/python/max/graph/value.py	383	# Create a constant tensor from the matrix	STRING
MEDIUM⚡	max/python/max/graph/value.py	488	# Create a 2x2 matrix	STRING
MEDIUM⚡	max/python/max/graph/value.py	491	# Create a Graph context to work with tensors	STRING
MEDIUM⚡	max/python/max/graph/value.py	493	# Create a constant tensor from the matrix	STRING
MEDIUM⚡	max/python/max/graph/value.py	518	# Create a matrix with float32 values	STRING
MEDIUM⚡	max/python/max/graph/value.py	521	# Create a Graph context to work with tensors	STRING
MEDIUM⚡	max/python/max/graph/value.py	523	# Create a constant tensor from the matrix	STRING
MEDIUM⚡	max/python/max/graph/value.py	543	# Create a 2x2 matrix (2-dimensional array)	STRING
MEDIUM⚡	max/python/max/graph/value.py	546	# Create a Graph context to work with tensors	STRING
MEDIUM⚡	max/python/max/graph/value.py	548	# Create a constant tensor from the matrix	STRING
MEDIUM⚡	max/python/max/graph/value.py	575	# Create a 2x2 matrix	STRING
MEDIUM⚡	max/python/max/graph/value.py	578	# Create a Graph context to work with tensors	STRING
MEDIUM⚡	max/python/max/graph/value.py	580	# Create a constant tensor from the matrix	STRING
MEDIUM⚡	max/python/max/graph/value.py	611	# Create a 2x2 matrix	STRING
MEDIUM⚡	max/python/max/graph/value.py	614	# Create a Graph context to work with tensors	STRING
MEDIUM⚡	max/python/max/graph/value.py	616	# Create a constant tensor from the matrix	STRING
MEDIUM⚡	max/python/max/graph/value.py	645	# Create a 2x2 matrix	STRING
MEDIUM⚡	max/python/max/graph/value.py	648	# Create a Graph context to work with tensors	STRING
MEDIUM⚡	max/python/max/graph/value.py	650	# Create a constant tensor from the matrix	STRING
MEDIUM⚡	max/python/max/graph/value.py	678	# Create a matrix with float32 values	STRING
MEDIUM⚡	max/python/max/graph/value.py	681	# Create a Graph context to work with tensors	STRING
MEDIUM⚡	max/python/max/graph/value.py	683	# Create a constant tensor from the matrix	STRING
MEDIUM	max/python/max/graph/value.py	81	# Create a graph context	STRING
MEDIUM	max/python/max/graph/value.py	768	# Create a 2x3 matrix	STRING
MEDIUM	max/python/max/graph/value.py	812	# Create a 2x2 matrix	STRING
MEDIUM	max/python/max/graph/value.py	816	# Create a tensor on the default device	STRING
MEDIUM	max/python/max/graph/value.py	844	# Define a 2x3 float32 input tensor for the graph	STRING
MEDIUM	max/python/max/graph/value.py	873	# Define a 2x3 float32 input tensor for the graph	STRING
MEDIUM	max/python/max/graph/value.py	902	# Define a 2x3 float32 input tensor for the graph	STRING
MEDIUM	max/python/max/graph/value.py	932	# Define a 2x3 float32 input tensor for the graph	STRING
MEDIUM	max/python/max/graph/value.py	964	# Define a 2x3 float32 input tensor for the graph	STRING
MEDIUM	max/python/max/graph/value.py	996	# Define a 2x3 float32 input tensor for the graph	STRING
MEDIUM	max/python/max/graph/dim.py	39	# Create a TensorType with a symbolic "batch" dimension and a static dimension of size 10	STRING
MEDIUM	max/python/max/graph/ops/slice_tensor.py	212	# Create a slice(None, None, None) index, for each dim indexed by ellipsis.	COMMENT
MEDIUM	max/python/max/_interpreter_ops/handlers.py	1206	# Create a 1D tensor of si64 values	COMMENT
MEDIUM	max/python/max/pipelines/context/tokens.py	31	# Create a token buffer with initial prompt tokens	STRING
MEDIUM	…elines/architectures/deepseekV3_2/layers/transforms.py	63	# This function is a slightly modified version of the	COMMENT
MEDIUM	…thon/max/pipelines/architectures/gpt_oss/layers/moe.py	66	# Initialize parent class	COMMENT
173 more matches not shown…

Modern Structural Boilerplate573 hits · 568 pts

Severity	File	Line	Snippet	Context
LOW	…er-connector-pyo3/python/kv_tier_connector/__init__.py	28	__all__ = ["TierConnector", "TierTransfer"]	CODE
LOW	…x/kernels/benchmarks/misc/comparison/install_sglang.py	18	def update_pip() -> None:	CODE
LOW	max/python/docs/sphinx_ext/docstring_sources.py	46	logger = logging.getLogger(__name__)	CODE
LOW	max/python/max/tree.py	60	__all__ = [	CODE
LOW	max/python/max/_eager_policy.py	24	logger = logging.getLogger(__name__)	CODE
LOW	max/python/max/benchmark/benchmark_serving.py	172	logger = logging.getLogger(__name__)	CODE
LOW	max/python/max/benchmark/results_to_csv.py	78	logger = logging.getLogger(__name__)	STRING
LOW	max/python/max/benchmark/benchmark_shared/iteration.py	36	logger = logging.getLogger(__name__)	CODE
LOW	max/python/max/benchmark/benchmark_shared/multi_turn.py	62	logger = logging.getLogger(__name__)	CODE
LOW	max/python/max/benchmark/benchmark_shared/metrics.py	61	__all__ = [	CODE
LOW	…max/benchmark/benchmark_shared/backend_metric_names.py	37	__all__ = [	CODE
LOW	…ax/benchmark/benchmark_shared/serving_result_output.py	73	logger = logging.getLogger(__name__)	CODE
LOW	max/python/max/benchmark/benchmark_shared/request.py	68	logger = logging.getLogger(__name__)	CODE
LOW	…ython/max/benchmark/benchmark_shared/server_metrics.py	46	logger = logging.getLogger(__name__)	CODE
LOW	…thon/max/benchmark/benchmark_shared/serving_metrics.py	54	logger = logging.getLogger(__name__)	CODE
LOW	max/python/max/benchmark/benchmark_shared/workload.py	28	logger = logging.getLogger(__name__)	CODE
LOW	…x/benchmark/benchmark_shared/gpu_metrics_prometheus.py	52	logger = logging.getLogger(__name__)	CODE
LOW	…/max/benchmark/benchmark_shared/gpu_metrics_scraper.py	52	logger = logging.getLogger(__name__)	CODE
LOW	max/python/max/benchmark/benchmark_shared/warmup.py	46	logger = logging.getLogger(__name__)	CODE
LOW	max/python/max/benchmark/benchmark_shared/utils.py	39	logger = logging.getLogger(__name__)	CODE
LOW	max/python/max/benchmark/benchmark_shared/utils.py	348	def set_ulimit(target_soft_limit: int = 65535) -> None:	CODE
LOW	…x/benchmark/benchmark_shared/lora_benchmark_manager.py	30	logger = logging.getLogger(__name__)	CODE
LOW	…x/python/max/benchmark/benchmark_shared/single_turn.py	65	logger = logging.getLogger(__name__)	CODE
LOW	max/python/max/benchmark/benchmark_shared/model_csv.py	74	logger = logging.getLogger(__name__)	CODE
LOW	…python/max/benchmark/benchmark_shared/workload_file.py	51	logger = logging.getLogger(__name__)	CODE
LOW	…nchmark/benchmark_shared/datasets/tool_augmentation.py	40	logger = logging.getLogger(__name__)	CODE
LOW	…chmark_shared/datasets/response_format_augmentation.py	44	logger = logging.getLogger(__name__)	CODE
LOW	…/benchmark/benchmark_shared/datasets/instruct_coder.py	40	logger = logging.getLogger(__name__)	CODE
LOW	…enchmark_shared/datasets/multiturn_distribution_fit.py	88	logger = logging.getLogger(__name__)	CODE
LOW	…nchmark/benchmark_shared/datasets/nemotron_opencode.py	43	logger = logging.getLogger(__name__)	CODE
LOW	…on/max/benchmark/benchmark_shared/datasets/__init__.py	56	__all__ = [	CODE
LOW	…/python/max/benchmark/benchmark_shared/datasets/all.py	90	logger = logging.getLogger(__name__)	CODE
LOW	…thon/max/benchmark/benchmark_shared/datasets/random.py	45	logger = logging.getLogger(__name__)	CODE
LOW	…ax/benchmark/benchmark_shared/datasets/agentic_code.py	42	logger = logging.getLogger(__name__)	CODE
LOW	…chmark/benchmark_shared/datasets/image_augmentation.py	49	logger = logging.getLogger(__name__)	CODE
LOW	…hmark/benchmark_shared/datasets/artificial_analysis.py	45	logger = logging.getLogger(__name__)	CODE
LOW	…n/max/benchmark/benchmark_shared/datasets/batch_job.py	42	logger = logging.getLogger(__name__)	CODE
LOW	…/max/benchmark/benchmark_shared/datasets/chat_judge.py	37	logger = logging.getLogger(__name__)	CODE
LOW	max/python/max/experimental/torch/__init__.py	27	__all__ = [	CODE
LOW	max/python/max/experimental/cascade/__init__.py	45	__all__ = [	CODE
LOW	…python/max/experimental/cascade/grpc_runtime/server.py	48	logger = logging.getLogger(__name__)	CODE
LOW	…python/max/experimental/cascade/grpc_runtime/client.py	71	logger = logging.getLogger(__name__)	CODE
LOW	…thon/max/experimental/cascade/grpc_runtime/__init__.py	20	__all__ = ["GrpcRuntimeClient", "SubprocGrpcRuntimeClient"]	CODE
LOW	…ython/max/experimental/cascade/grpc_runtime/runtime.py	20	__all__ = ["GrpcRuntimeClient", "SubprocGrpcRuntimeClient"]	CODE
LOW	max/python/max/experimental/cascade/core/__init__.py	29	__all__ = [	CODE
LOW	max/python/max/experimental/cascade/core/fd_limits.py	22	logger = logging.getLogger(__name__)	CODE
LOW	…python/max/experimental/cascade/http_runtime/server.py	65	logger = logging.getLogger(__name__)	CODE
LOW	…thon/max/experimental/cascade/http_runtime/__init__.py	28	__all__ = ["HttpRuntimeProxy", "SubprocHttpRuntime"]	CODE
LOW	…thon/max/experimental/cascade/workers/max_tokenizer.py	36	logger = logging.getLogger(__name__)	CODE
LOW	…n/max/experimental/cascade/workers/max_model_worker.py	50	logger = logging.getLogger(__name__)	CODE
LOW	…python/max/experimental/cascade/telemetry/_provider.py	48	logger = logging.getLogger(__name__)	CODE
LOW	…/python/max/experimental/cascade/telemetry/__init__.py	38	__all__ = [	CODE
LOW	max/python/max/experimental/cascade/serve/__init__.py	18	__all__ = ["CascadeFastAPI", "build_router"]	CODE
LOW	max/python/max/experimental/cascade/serve/main.py	64	logger = logging.getLogger(__name__)	CODE
LOW	max/python/max/_xgrammar/structural_tag.py	630	__all__ = [	CODE
LOW	max/python/max/_xgrammar/__init__.py	329	__all__ = [	CODE
LOW	max/python/max/nn/__init__.py	74	__all__ = [	CODE
LOW	max/python/max/nn/_compilation_timer.py	34	__all__ = [	CODE
LOW	max/python/max/nn/module.py	49	__all__ = [	CODE
LOW	max/python/max/nn/rope/__init__.py	17	__all__ = [	CODE
513 more matches not shown…

Cross-Language Confusion114 hits · 538 pts

Severity	File	Line	Snippet	Context
HIGH	max/kernels/benchmarks/autotune/kdiff.py	50	export KUBECONFIG=/dev/null ; # recommended	STRING
HIGH	…max/benchmark/benchmark_shared/backend_metric_names.py	66	corresponding ``BenchmarkResult`` field stays null) rather than guessing	STRING
HIGH	max/python/max/nn/common_layers/quant_mla.py	259	"""Fused ``q_a_proj \|\| kv_a_proj_with_mqa`` (bf16 or FP8)."""	STRING
HIGH	…thon/max/pipelines/architectures/glm5_1/tool_parser.py	110	# case; decode only non-string JSON (numbers, booleans, null, arrays,	COMMENT
HIGH	…pipelines/architectures/deepseekV41/image_processor.py	101	"DeepSeek-V4.1 vision_config.max_wh_ratio must be null, got"	CODE
HIGH	max/python/max/pipelines/kv_cache/config.py	143	default=KVConnectorType.null,	CODE
HIGH	…lines/kv_cache/paged_kv_cache/recurrent_coordinator.py	39	Every block behind it is null, a published checkpoint, or one a	STRING
HIGH	…elines/kv_cache/paged_kv_cache/kv_group_coordinator.py	98	every block below it is null, published, or permanently incomplete.	STRING
HIGH	max/python/max/pipelines/lib/embeddings_pipeline.py	147	tracer.push("prepare_response")	CODE
HIGH	max/python/max/profiler/tracing.py	137	tracer.push("child_operation")	STRING
HIGH	max/python/max/profiler/tracing.py	144	tracer.push("child_operation")	STRING
HIGH	max/python/max/profiler/tracing.py	164	self.push(message, color)	CODE
HIGH	max/python/max/profiler/tracing.py	219	self.push(message, color)	CODE
HIGH	max/python/max/serve/parser/json_utils.py	25	(?<boolean> true \| false \| null )	CODE
HIGH	max/python/max/serve/schemas/_decisions.py	59	# scalars (numbers, booleans, null) are not accepted.	COMMENT
HIGH	max/python/max/serve/worker_interface/zmq_interface.py	332	response_queue=self.response_queue_config.push(),	CODE
HIGH	max/python/max/serve/worker_interface/_zmq_queue.py	273	return self.push(), self.pull()	CODE
HIGH⚡	…ests/integration/unorganized/test_memory_estimation.py	569	([DeviceSpec.cpu()], KVConnectorType.null, 0),	CODE
HIGH⚡	…ests/integration/unorganized/test_memory_estimation.py	570	([DeviceSpec.accelerator(id=0)], KVConnectorType.null, 0),	CODE
HIGH⚡	…ests/integration/unorganized/test_memory_estimation.py	574	KVConnectorType.null,	CODE
HIGH⚡	…ests/integration/unorganized/test_memory_estimation.py	579	KVConnectorType.null,	CODE
HIGH⚡	…ests/integration/unorganized/test_memory_estimation.py	584	KVConnectorType.null,	CODE
HIGH⚡	…ests/integration/unorganized/test_memory_estimation.py	636	(1, KVConnectorType.null, 1),	CODE
HIGH⚡	…ests/integration/unorganized/test_memory_estimation.py	638	(2, KVConnectorType.null, 1),	CODE
HIGH	max/tests/integration/graph/test_horizontal_fusion.py	176	accumulation buffer from `dtypeChanges && containsMatmulOrConvOrConcat`,	CODE
HIGH	…tectures/mimo_v2_modulev3/test_serving_numerics_gpu.py	59	"""Bounds on KL(HF \|\| MAX) over the whole vocabulary, and on top-1	STRING
HIGH	…tectures/mimo_v2_modulev3/test_serving_numerics_gpu.py	86	"""Each row's KL(gold \|\| test)."""	STRING
HIGH	…rchitectures/mimo_v2_modulev3/test_v2_v3_parity_gpu.py	60	"""V3's mean KL(HF \|\| V3) over every row of both prompts may be at most	STRING
HIGH	…rchitectures/mimo_v2_modulev3/test_v2_v3_parity_gpu.py	83	"""Each row's KL(gold \|\| test)."""	STRING
HIGH⚡	…chitectures/dspark_gemma4/test_dspark_gemma4_module.py	700	"""Mean per-position KL(ref \|\| max) over fp32 softmax distributions."""	STRING
HIGH⚡	…ectures/dspark_draft/test_dspark_speculators_module.py	399	"""Mean per-position KL(ref \|\| max) over fp32 softmax distributions."""	STRING
HIGH	…ion/architectures/mimo_v2/test_serving_numerics_gpu.py	40	"""Bounds on KL(HF \|\| MAX) over the whole vocabulary, and on top-1	STRING
HIGH	…ion/architectures/mimo_v2/test_serving_numerics_gpu.py	59	"""Each row's KL(gold \|\| test)."""	STRING
HIGH	…tion/architectures/mimo_v2/check_reference_fixtures.py	62	"""The fixtures' ``kl64``: KL(gold \|\| test) over gold's top 64 + tail."""	STRING
HIGH⚡	…/integration/architectures/qwen3_5/test_tool_parser.py	494	# string \| null: a number-looking value is not null, so it stays text.	COMMENT
HIGH⚡	…tegration/pipelines/tokenization/test_xgrammar_shim.py	4019	({"type": "object"}, '{"a": ["x", {"b": null}]}', '{"a": 1 zzz}'),	CODE
HIGH⚡	…tegration/pipelines/tokenization/test_xgrammar_shim.py	4022	'{"a": ["x", {"b": null}]}',	CODE
HIGH⚡	…tegration/pipelines/tokenization/test_xgrammar_shim.py	4025	({"type": "array"}, '[1, ["x", {"b": null}]]', "[1 zzz]"),	CODE
HIGH	…/pipelines/tokenization/test_tool_grammar_fill_cost.py	67	_COMMAND = "cd /tmp && ls -la \| grep -v total"	CODE
HIGH	…okenization/test_kimi_reasoning_toolarg_enforcement.py	161	# anyOf(string-enum \| null): once the value opens with a quote, only the	COMMENT
HIGH⚡	…/accuracy/model_evals/test_exacto_gpqa_answer_audit.py	68	"scores":{"mcq_scorer":{"answer":null,"value":"I"}},	CODE
HIGH⚡	…/accuracy/model_evals/test_exacto_gpqa_answer_audit.py	79	"target":"A","scores":{"mcq_scorer":{"answer":null,"value":"I"}},	CODE
HIGH	…ation/accuracy/llm_fuzz/scenarios/structured_output.py	207	# (non-dict, null, recursive $ref) must be rejected.	COMMENT
HIGH	…tion/accuracy/llm_fuzz/scenarios/models/kimi_battle.py	103	f"Expected str\|null, got {type(data['value']).__name__}"	CODE
HIGH	…ation/accuracy/llm_fuzz/scenarios/models/glm_battle.py	222	detail=f"nickname should be string or null, got {type(data['nickname']).__name__}",	CODE
HIGH	…ation/accuracy/llm_fuzz/scenarios/models/glm_battle.py	297	detail=f"value should be string or null, got {type(data['value']).__name__}",	CODE
HIGH⚡	…llm_fuzz/scenarios/validation/tc_schema_enforcement.py	1263	# 15. anyOf with nullable string: anyOf: [{type:string},{type:null}]	COMMENT
HIGH⚡	…llm_fuzz/scenarios/validation/tc_schema_enforcement.py	1257	"and reason is null (must be null, not a string)."	CODE
HIGH⚡	…llm_fuzz/scenarios/validation/tc_schema_enforcement.py	1426	f"status={status!r} not in enum ['active','inactive',null]"	CODE
HIGH⚡	…llm_fuzz/scenarios/validation/tc_schema_enforcement.py	1461	f"not in [true, false, null]"	CODE
HIGH⚡	…llm_fuzz/scenarios/validation/tc_schema_enforcement.py	2812	"string 'none' instead of null. Do not use null, "	CODE
HIGH⚡	…llm_fuzz/scenarios/validation/tc_schema_enforcement.py	4085	f"name: expected string or null, got {type(name).__name__}"	CODE
HIGH⚡	…llm_fuzz/scenarios/validation/tc_schema_enforcement.py	4091	f"age: expected integer or null, got "	CODE
HIGH⚡	…llm_fuzz/scenarios/validation/tc_schema_enforcement.py	4098	f"tags: expected array or null, got "	CODE
HIGH	…llm_fuzz/scenarios/validation/tc_schema_enforcement.py	139	"integer vs number, null, nested objects, fixed property order"	CODE
HIGH	…llm_fuzz/scenarios/validation/tc_schema_enforcement.py	918	f"nickname: expected string or null, got {type(nickname).__name__}"	CODE
HIGH	…llm_fuzz/scenarios/validation/tc_schema_enforcement.py	1180	f"nickname: expected string or null, got {type(nn).__name__}"	CODE
HIGH	…llm_fuzz/scenarios/validation/tc_schema_enforcement.py	1231	f"reason: expected null, got "	CODE
HIGH	…llm_fuzz/scenarios/validation/tc_schema_enforcement.py	1279	f"middle_name: expected string or null, "	CODE
HIGH	…llm_fuzz/scenarios/validation/tc_schema_enforcement.py	1874	f"score: expected number or null, "	CODE
54 more matches not shown…

Excessive Try-Catch Wrapping510 hits · 499 pts

Severity	File	Line	Snippet	Context
LOW	…egration-test/python/test_signal_handler_python_gil.py	86	except Exception as e:	CODE
LOW	…ernels/benchmarks/misc/comparison/bench_amd_mx_gemm.py	593	except Exception as e:	CODE
LOW	…ernels/benchmarks/misc/comparison/bench_amd_mx_gemm.py	780	except Exception as e:	CODE
LOW	…s/benchmarks/misc/comparison/bench_amd_m3_moe_chain.py	737	except Exception as e:	CODE
MEDIUM	…s/benchmarks/misc/comparison/bench_blackwell_decode.py	59	print(f"Error: flashinfer not available: {e}")	CODE
LOW	…s/benchmarks/misc/comparison/bench_blackwell_decode.py	507	except Exception as e:	CODE
LOW	…s/benchmarks/misc/comparison/bench_blackwell_decode.py	525	except Exception as e:	CODE
LOW	…s/benchmarks/misc/comparison/bench_mxfp8_quant_vllm.py	93	except Exception:	CODE
LOW	…s/benchmarks/misc/comparison/bench_mxfp8_quant_vllm.py	196	except Exception:	CODE
MEDIUM	…nchmarks/misc/comparison/bench_blackwell_mla_decode.py	126	print(f"Error: flashinfer not available: {e}")	CODE
LOW	…nchmarks/misc/comparison/bench_blackwell_mla_decode.py	1049	except Exception as e:	CODE
LOW	…nchmarks/misc/comparison/bench_blackwell_mla_decode.py	1070	except Exception as e:	CODE
LOW	max/kernels/benchmarks/misc/comparison/sweep_decode.py	169	except Exception as e:	CODE
LOW	max/kernels/benchmarks/misc/comparison/sweep_decode.py	193	except Exception as e:	CODE
LOW	…/kernels/benchmarks/misc/comparison/setup_bench_env.py	380	except Exception as e:	CODE
LOW	…/kernels/benchmarks/misc/comparison/setup_bench_env.py	567	except Exception as e:	CODE
LOW	…/kernels/benchmarks/misc/comparison/setup_bench_env.py	592	except Exception as e:	CODE
LOW	…/kernels/benchmarks/misc/comparison/setup_bench_env.py	608	except Exception as e:	CODE
MEDIUM	…/benchmarks/misc/comparison/bench_blackwell_prefill.py	52	print(f"Error: flashinfer not available: {e}")	CODE
MEDIUM	…/benchmarks/misc/comparison/bench_blackwell_prefill.py	65	print(f"Error: flash-attn-4 not available: {e}")	CODE
LOW	…/benchmarks/misc/comparison/bench_blackwell_prefill.py	504	except Exception as e:	CODE
LOW	…/benchmarks/misc/comparison/bench_blackwell_prefill.py	524	except Exception as e:	CODE
LOW	…/benchmarks/misc/comparison/bench_blackwell_prefill.py	543	except Exception as e:	CODE
LOW	…benchmarks/misc/comparison/bench_allreduce_subgraph.py	148	except Exception as e:	CODE
LOW	…benchmarks/misc/comparison/bench_allreduce_subgraph.py	196	except Exception:	CODE
LOW	…benchmarks/misc/comparison/bench_allreduce_subgraph.py	226	except Exception as e:	CODE
LOW	…benchmarks/misc/comparison/bench_allreduce_subgraph.py	267	except Exception as e:	CODE
MEDIUM	…benchmarks/misc/comparison/bench_allreduce_subgraph.py	355	print("Error: This benchmark requires mpirun for multi-GPU execution.")	CODE
LOW	max/kernels/benchmarks/misc/comparison/bench_softmax.py	136	except Exception as e:	CODE
LOW	max/kernels/benchmarks/misc/comparison/bench_amd_mha.py	501	except Exception as e:	CODE
LOW⚡	max/kernels/benchmarks/misc/comparison/bench_amd_mla.py	838	except Exception as e:	CODE
LOW	max/kernels/benchmarks/misc/comparison/bench_amd_mla.py	1287	except Exception as e:	CODE
LOW	max/kernels/benchmarks/misc/comparison/bench_amd_mla.py	1488	except Exception as e:	CODE
LOW	max/kernels/benchmarks/misc/comparison/bench_amd_mla.py	1558	except Exception as e:	CODE
LOW	max/kernels/benchmarks/autotune/kbench.py	720	except Exception:	CODE
MEDIUM	max/kernels/benchmarks/autotune/kbench.py	717	def main() -> None:	CODE
LOW	max/kernels/benchmarks/autotune/kbench_model.py	160	except Exception as e:	CODE
LOW	max/kernels/benchmarks/autotune/kbench_model.py	169	except Exception as exc:	CODE
LOW	max/kernels/benchmarks/autotune/kbench_model.py	259	except Exception:	CODE
LOW	max/kernels/benchmarks/autotune/kbench_model.py	638	except Exception as e:	CODE
LOW	max/kernels/benchmarks/autotune/kbench_model.py	1262	except Exception as e:	CODE
LOW	max/kernels/benchmarks/autotune/kbench_model.py	1296	except Exception:	CODE
LOW	max/kernels/benchmarks/autotune/tests/test_kbench.py	104	except Exception:	CODE
LOW	max/python/docs/sphinx_ext/docstring_sources.py	104	except Exception:	CODE
LOW	max/python/docs/sphinx_ext/docstring_sources.py	141	except Exception:	CODE
LOW	max/python/docs/sphinx_ext/typing_xrefs.py	63	except Exception:	CODE
LOW	max/python/docs/sphinx_ext/source_files.py	137	except Exception:	CODE
LOW	max/python/docs/sphinx_ext/source_files.py	159	except Exception:	CODE
LOW	max/python/docs/sphinx_ext/source_files.py	174	except Exception:	CODE
LOW	max/python/docs/sphinx_ext/source_files.py	245	except Exception:	CODE
LOW	max/python/max/tree.py	157	except Exception:	CODE
LOW	max/python/max/_interpreter.py	50	except Exception as _e:	CODE
LOW	max/python/max/_interpreter.py	77	except Exception:	CODE
LOW	max/python/max/benchmark/serve_replay.py	57	except Exception:	CODE
LOW	max/python/max/benchmark/serve_replay.py	80	except Exception:	CODE
LOW	max/python/max/benchmark/serve_replay.py	94	except Exception:	CODE
LOW	max/python/max/benchmark/benchmark_serving.py	503	except Exception as e:	CODE
LOW	max/python/max/benchmark/benchmark_serving.py	544	except Exception:	CODE
LOW	max/python/max/benchmark/benchmark_serving.py	786	except Exception as e:	CODE
LOW	max/python/max/benchmark/benchmark_shared/request.py	608	except Exception:	CODE
450 more matches not shown…

AI Slop Vocabulary152 hits · 492 pts

Severity	File	Line	Snippet	Context
MEDIUM⚡	max/kernels/test/gpu/fuzz/fuzz.py	115	# target, and `corpus/mega_ffn/` would sit under the harness directory,	COMMENT
MEDIUM⚡	max/kernels/test/gpu/fuzz/fuzz.py	118	# `-I max/kernels/test/gpu/fuzz` for `_fuzz`, which makes the harness	COMMENT
MEDIUM⚡	max/kernels/test/gpu/fuzz/fuzz.py	124	# corpus out of the harness directory.	COMMENT
MEDIUM	…/kernels/test/gpu/compute_sanitizer/stress_schedule.sh	15	# Schedule-amplification harness for inter-block global-memory races.	COMMENT
MEDIUM	…/kernels/test/gpu/compute_sanitizer/stress_schedule.sh	26	# This harness amplifies the schedule with three knobs that need no kernel	COMMENT
MEDIUM	…ernels/benchmarks/misc/utils/warp_profiling/profile.py	119	# Create comprehensive visualization showing all SMs	COMMENT
MEDIUM	…ernels/benchmarks/misc/comparison/bench_amd_mx_gemm.py	43	# in that range, so they measure partly L2-warm. What this harness does	COMMENT
MEDIUM	…ernels/benchmarks/misc/comparison/bench_amd_mx_gemm.py	47	# the narrower ones. Absolute cold-HBM MXFP6 numbers need a harness that	COMMENT
MEDIUM	…chmarks/gpu/linalg/bench_matmul_llama3_70B_4xH100.yaml	20	# This file is facilitate development before GTC and should be combined	COMMENT
LOW	max/kernels/benchmarks/autotune/kbench_model.py	544	# just use compile-time parameters and ignore runtime variables.	COMMENT
MEDIUM	max/python/max/benchmark/benchmark_shared/metrics.py	916	# the harness didn't launch the server (e.g. benchmarking an external	COMMENT
LOW	max/python/max/nn/legacy/quant_ops.py	521	# activation scales are [1xm] which but in the dynamic scaled matmul we expect just use x_scale[0,0]	COMMENT
MEDIUM	max/python/max/nn/legacy/parallel/numpy_parallel_ops.py	268	# to better utilize available parallelism and memory bandwidth.	COMMENT
LOW	…/max_license/mlir_nanobind/tblgen/backends/nanobind.py	110	# - if there's multiple setters, just use the first one we find	COMMENT
MEDIUM	max/python/max/config/config_file_model.py	109	"""Optional section name for comprehensive/multi-section config files.	STRING
MEDIUM	max/python/max/driver/buffer.py	433	# Hardcoded but we should move to a robust versioning system if this	COMMENT
MEDIUM	max/python/max/graph/graph.py	1072	# This may over-declare but is deterministic and comprehensive.	COMMENT
MEDIUM	max/python/max/pipelines/logging_utils.py	43	"""Logs comprehensive pipeline and KVCache configuration information.	STRING
MEDIUM	max/python/max/pipelines/context/sampling_params.py	341	"""Logs comprehensive sampling parameters information.	STRING
MEDIUM	max/python/max/pipelines/lib/eplb_stats.py	407	# (wherever the harness injects them), so no separate host counter is	COMMENT
LOW	max/python/max/_entrypoints/pipelines.py	1405	# Help was requested and printed, just return	COMMENT
MEDIUM	max/python/max/serve/config.py	519	"""Log comprehensive server configuration information.	STRING
MEDIUM	…/max/serve/scheduler/batch_constructor/token_budget.py	101	"""Whether this budget may shrink the context via ``context.chunk`` in order to fit within the remaining capacit	STRING
MEDIUM	max/python/layer_benchmarks/precompile_layer.py	34	# A session only because the harness signature takes one; graph building	COMMENT
MEDIUM	max/python/layer_benchmarks/testbed/runner.py	102	# weights registry (and the reference weights a harness stashes	COMMENT
MEDIUM	max/python/layer_benchmarks/testbed/runner.py	231	# Compile first so harness can store weights for the torch reference.	COMMENT
MEDIUM⚡	…x/python/layer_benchmarks/testbed/kbench_entrypoint.py	44	# harness ABC so each harness declares its own shape keys.	COMMENT
MEDIUM⚡	…x/python/layer_benchmarks/testbed/kbench_entrypoint.py	48	# to the harness as static params.	COMMENT
MEDIUM	…x/python/layer_benchmarks/testbed/kbench_entrypoint.py	122	# Extract harness name from the config bundle.	COMMENT
MEDIUM	…x/python/layer_benchmarks/testbed/kbench_entrypoint.py	136	# Filter out control keys before passing to the harness.	COMMENT
MEDIUM	…ayer_benchmarks/testbed/harnesses/wan_vae_attention.py	165	# harness catches that regression), small enough that	COMMENT
MEDIUM	…sts/integration/graph/test_custom_kernel_validation.py	387	# Test multiple operations in sequence to ensure comprehensive validation	COMMENT
MEDIUM	…sts/integration/graph/test_custom_kernel_validation.py	341	"""Tests for comprehensive kernel signature validation."""	STRING
LOW	…integration/architectures/deepseekV3_2/test_indexer.py	329	# ``slice_freqs_cis`` and just pass ``rope.freqs_cis``.	COMMENT
MEDIUM	…architectures/mimo_v2_modulev3/test_base_graph_tree.py	110	# A fresh lazy context for the stage, as the CPU harness does: sharing	COMMENT
MEDIUM	…ration/architectures/gemma4_modulev3/test_attention.py	68	# ModuleV3 harness	COMMENT
MEDIUM	…ctures/dflash_mimo_v2_modulev3/test_fused_graph_cpu.py	288	# the ModuleV3 MiMo CPU harness does.	COMMENT
MEDIUM	max/tests/integration/accuracy/llm_fuzz_matrix.py	32	# exercises the harness end to end without taking an accelerator.	COMMENT
MEDIUM	max/tests/integration/accuracy/llm_fuzz_matrix.py	133	# Tiny model on CPU: a cheap end-to-end check of the fuzz harness and the	COMMENT
MEDIUM	max/tests/integration/accuracy/run_mmmu_video_eval.py	757	# Fail on a broken harness (no successful inferences / mostly errors) so	COMMENT
LOW	max/tests/integration/accuracy/verify_pipelines.py	1305	# Let generate_llm_logits.py validate the `--devices` CLI arg and just pass	COMMENT
MEDIUM⚡	…gration/accuracy/model_evals/run_omnidocbench_local.sh	19	# OmniDocBench scores with a third-party harness (pdf_validation.py) that needs	COMMENT
MEDIUM⚡	…gration/accuracy/model_evals/run_omnidocbench_local.sh	36	# Passed to the harness by environment, never written to	COMMENT
MEDIUM	…gration/accuracy/model_evals/run_omnidocbench_local.sh	59	# --refresh Re-clone the harness and rebuild its venv.	COMMENT
MEDIUM⚡	…gration/accuracy/model_evals/run_omnidocbench_local.sh	145	# while the harness appends its own; normalize so either form works.	COMMENT
MEDIUM⚡	…gration/accuracy/model_evals/run_omnidocbench_local.sh	150	# The harness takes a directory, so the pages to run are staged as symlinks:	COMMENT
MEDIUM⚡	…gration/accuracy/model_evals/run_omnidocbench_local.sh	156	# build_save_name() in the harness derives its output name from the prediction	COMMENT
MEDIUM	…gration/accuracy/model_evals/run_omnidocbench_local.sh	183	# into the harness venv would shadow that venv's own packages (the same hazard	COMMENT
MEDIUM	…gration/accuracy/model_evals/run_omnidocbench_local.sh	221	# The harness requires Python < 3.12.	COMMENT
MEDIUM	…gration/accuracy/model_evals/run_omnidocbench_local.sh	230	# The display-formula CDM metric renders LaTeX; without it the harness leaves	COMMENT
MEDIUM	…gration/accuracy/model_evals/run_omnidocbench_local.sh	256	# ImageMagick 6 on Ubuntu ships `convert`; the harness calls IM7's `magick`.	COMMENT
MEDIUM	…gration/accuracy/model_evals/run_omnidocbench_local.sh	395	# ./result is written relative to the harness's working directory.	COMMENT
MEDIUM	…ests/integration/accuracy/model_evals/exacto_report.py	51	#: The openbench harness normalizes the OpenAI finish_reason "length" onto	COMMENT
MEDIUM	…ests/integration/accuracy/model_evals/exacto_report.py	53	#: (verified against inspect_ai 0.3.125); "length" is kept for a harness that	COMMENT
MEDIUM	…ests/integration/accuracy/model_evals/exacto_report.py	237	# The task names the harness it reproduces; openbench's own tasks do not.	COMMENT
MEDIUM	…ests/integration/accuracy/model_evals/exacto_report.py	263	# Set by tasks that, like the harness, send media from a manifest: two	COMMENT
MEDIUM	…ests/integration/accuracy/model_evals/exacto_report.py	266	# The harness scores a request the server rejected (a 400 for an	COMMENT
MEDIUM	…ests/integration/accuracy/model_evals/exacto_report.py	290	#: chance — an API or harness failure, not the agent losing. ``max_steps`` is	COMMENT
MEDIUM	…ests/integration/accuracy/model_evals/exacto_report.py	466	# This module is executed by the pinned third-party harness venv (openbench or	COMMENT
MEDIUM⚡	…tegration/accuracy/model_evals/run_exacto_vgi_local.sh	20	# (benchmark_type "vgi_bench") with its own TypeScript harness,	COMMENT
92 more matches not shown…

AI Structural Patterns192 hits · 178 pts

Severity	File	Line	Context
LOW	…s/benchmarks/misc/comparison/bench_blackwell_decode.py	431	CODE
LOW	…/benchmarks/misc/comparison/bench_blackwell_prefill.py	339	CODE
LOW	max/kernels/benchmarks/misc/comparison/bench_amd_mha.py	146	CODE
LOW	max/kernels/benchmarks/misc/comparison/bench_amd_mha.py	349	CODE
LOW	max/kernels/benchmarks/misc/comparison/bench_amd_mha.py	390	CODE
LOW	max/kernels/benchmarks/misc/comparison/bench_amd_mha.py	431	CODE
LOW	max/kernels/benchmarks/autotune/kbench.py	105	CODE
LOW	max/kernels/benchmarks/autotune/kbench_model.py	523	CODE
LOW	max/python/max/benchmark/benchmark_shared/multi_turn.py	86	CODE
LOW	max/python/max/benchmark/benchmark_shared/multi_turn.py	436	CODE
LOW	max/python/max/benchmark/benchmark_shared/multi_turn.py	610	CODE
LOW	…/benchmark/benchmark_shared/datasets/instruct_coder.py	202	CODE
LOW	…enchmark_shared/datasets/multiturn_distribution_fit.py	551	CODE
LOW	…nchmark/benchmark_shared/datasets/nemotron_opencode.py	461	CODE
LOW	…ax/benchmark/benchmark_shared/datasets/agentic_code.py	295	CODE
LOW	max/python/max/nn/module.py	629	CODE
LOW	max/python/max/nn/conv.py	80	CODE
LOW	max/python/max/nn/legacy/kernels.py	502	CODE
LOW	max/python/max/nn/legacy/kernels.py	6138	CODE
LOW	max/python/max/nn/legacy/kernels.py	6364	CODE
LOW	max/python/max/nn/legacy/kernels.py	8369	CODE
LOW	max/python/max/nn/legacy/kernels.py	9291	CODE
LOW	max/python/max/nn/legacy/kernels.py	9716	CODE
LOW	max/python/max/nn/legacy/kernels.py	10138	CODE
LOW	max/python/max/nn/legacy/kernels.py	10404	CODE
LOW	max/python/max/nn/legacy/kernels.py	10908	CODE
LOW	max/python/max/nn/legacy/kernels.py	11059	CODE
LOW	max/python/max/nn/legacy/kernels.py	11209	CODE
LOW	max/python/max/nn/legacy/kernels.py	11257	CODE
LOW	max/python/max/nn/legacy/kernels.py	11343	CODE
LOW	max/python/max/nn/legacy/kernels.py	11430	CODE
LOW	max/python/max/nn/legacy/kernels.py	12896	CODE
LOW	max/python/max/nn/legacy/kernels.py	14357	CODE
LOW	max/python/max/nn/legacy/kernels.py	14432	CODE
LOW	max/python/max/nn/legacy/conv_transpose.py	88	CODE
LOW	max/python/max/nn/legacy/conv_transpose.py	297	CODE
LOW	max/python/max/nn/legacy/conv.py	96	CODE
LOW	max/python/max/nn/legacy/conv.py	382	CODE
LOW	max/python/max/nn/legacy/conv.py	586	CODE
LOW	max/python/max/nn/legacy/quant_ops.py	166	CODE
LOW	max/python/max/nn/legacy/quant_ops.py	199	CODE
LOW	…/python/max/nn/legacy/attention/attention_with_rope.py	60	CODE
LOW	…/python/max/nn/legacy/attention/attention_with_rope.py	986	CODE
LOW	…/python/max/nn/legacy/attention/attention_with_rope.py	1147	CODE
LOW	…thon/max/nn/legacy/attention/multi_latent_attention.py	91	CODE
LOW	…thon/max/nn/legacy/attention/multi_latent_attention.py	580	CODE
LOW	…/max/nn/legacy/attention/multi_latent_attention_fp8.py	59	CODE
LOW	max/python/max/nn/legacy/sampling/rejection_sampler.py	1129	CODE
LOW	max/python/max/nn/legacy/sampling/rejection_sampler.py	732	CODE
LOW	max/python/max/nn/legacy/moe/stacked_moe.py	302	CODE
LOW	max/python/max/nn/legacy/moe/moe.py	292	CODE
LOW	…/python/max/nn/common_layers/multi_latent_attention.py	71	CODE
LOW	max/python/max/nn/common_layers/quant_mla.py	79	CODE
LOW	max/python/max/graph/graph.py	847	CODE
LOW	max/python/max/graph/ops/nonzero.py	68	CODE
LOW	max/python/max/graph/ops/concat.py	97	CODE
LOW	max/python/max/pipelines/context/video.py	458	CODE
LOW	max/python/max/pipelines/context/video.py	739	CODE
LOW	…max/pipelines/architectures/laguna/layers/attention.py	75	CODE
LOW	…/max/pipelines/architectures/olmo3/layers/attention.py	216	CODE
132 more matches not shown…

Structural Annotation Overuse78 hits · 128 pts

Severity	File	Line	Snippet	Context
LOW	CONTRIBUTING.md	46	### Step 1: Evaluate and get buy-in on the change	COMMENT
LOW	CONTRIBUTING.md	95	### Step 2: Create a pull request	COMMENT
LOW	CONTRIBUTING.md	244	### Step 3: PR triage and review	COMMENT
LOW	CONTRIBUTING.md	257	### Step 4: Internal review and syncing	COMMENT
LOW	CONTRIBUTING.md	276	### Step 5: Review feedback and iteration	COMMENT
LOW⚡	…/max/pipelines/architectures/olmo3/layers/attention.py	170	# Step 1: QKV matmul.	COMMENT
LOW⚡	…/max/pipelines/architectures/olmo3/layers/attention.py	177	# Step 2: Apply full-dimension QK norm before rope.	COMMENT
LOW⚡	…/max/pipelines/architectures/olmo3/layers/attention.py	187	# Step 3: Fused rope + split + KV store.	COMMENT
LOW	…/max/pipelines/architectures/olmo3/layers/attention.py	202	# Step 5: Compute flash attention (reads K/V from cache)	COMMENT
LOW	…n/max/pipelines/architectures/qwen3vl_moe/tokenizer.py	800	# Step 1: Extract prompt from request and apply chat template if needed	COMMENT
LOW	…n/max/pipelines/architectures/qwen3vl_moe/tokenizer.py	837	# Step 2: Load and process images	COMMENT
LOW	…n/max/pipelines/architectures/qwen3vl_moe/tokenizer.py	855	# Step 3: Process images with custom image processor (if any) and expand <\|image_pad\|> placeholders in text	COMMENT
LOW	…n/max/pipelines/architectures/qwen3vl_moe/tokenizer.py	951	# Step 4: Tokenize the expanded text	COMMENT
LOW	…n/max/pipelines/architectures/qwen3vl_moe/tokenizer.py	1019	# Step 5: Assemble the vision encoder's inputs, in PROMPT order.	COMMENT
LOW⚡	…lines/architectures/olmo2_modulev3/layers/attention.py	144	# Step 1: QKV matmul.	COMMENT
LOW⚡	…lines/architectures/olmo2_modulev3/layers/attention.py	151	# Step 2: Apply full-dimension QK norm before rope.	COMMENT
LOW⚡	…lines/architectures/olmo2_modulev3/layers/attention.py	160	# Step 3: Fused rope + split + KV store.	COMMENT
LOW	…lines/architectures/olmo2_modulev3/layers/attention.py	175	# Step 5: Compute flash attention	COMMENT
LOW	max/python/max/pipelines/lib/eplb_rebalance.py	151	# Step 1: pack groups to nodes.	COMMENT
LOW⚡	max/python/max/pipelines/lib/eplb_rebalance.py	167	# Step 2: construct redundant experts within nodes.	COMMENT
LOW⚡	max/python/max/pipelines/lib/eplb_rebalance.py	175	# Step 3: pack physical experts to GPUs within each node.	COMMENT
LOW	max/tests/integration/accuracy/llm_fuzz/model_config.py	223	# Step 1: Try HuggingFace fetch using --model as repo ID	COMMENT
LOW	max/tests/integration/accuracy/llm_fuzz/model_config.py	236	# Step 2: Apply manual overrides (highest priority)	COMMENT
LOW⚡	…/accuracy/llm_fuzz/scenarios/tool_schema_validation.py	1371	# Step 1: initial tool call	COMMENT
LOW	…/accuracy/llm_fuzz/scenarios/tool_schema_validation.py	1390	# Step 2: provide tool result, ask follow-up that triggers another call	COMMENT
LOW	…/accuracy/llm_fuzz/scenarios/tool_schema_validation.py	1440	# Step 3: multi-tool conversation	COMMENT
LOW⚡	…/accuracy/llm_fuzz/scenarios/models/kimi_production.py	537	# Step 1: Structured output request	COMMENT
LOW⚡	…/accuracy/llm_fuzz/scenarios/models/kimi_production.py	561	# Step 2: Tool calling request	COMMENT
LOW⚡	…/accuracy/llm_fuzz/scenarios/models/kimi_production.py	576	# Step 3: Plain chat request	COMMENT
LOW	max/docs/accuracy-debugging.md	52	## Step 1: Dump the intermediate tensors with `debug_model`	COMMENT
LOW	max/docs/accuracy-debugging.md	79	## Step 2: Compare to find the area of divergence	COMMENT
LOW	max/docs/accuracy-debugging.md	106	## Step 3: Add fine-grained print logging	COMMENT
LOW	max/docs/accuracy-debugging.md	183	## Step 4: Export full tensors with `debug_model`	COMMENT
LOW	max/docs/accuracy-debugging.md	212	## Step 5: Compare the tensors with `compare_tensors`	COMMENT
LOW⚡	max/docs/site/code/develop/index/graph_simple.py	21	# Step 1: Define the graph structure	COMMENT
LOW⚡	max/docs/site/code/develop/index/graph_simple.py	30	# Step 2: Compile the graph	COMMENT
LOW⚡	max/docs/site/code/develop/index/graph_simple.py	35	# Step 3: Execute with data	COMMENT
LOW⚡	max/examples/diffusion/simple_offline_generation.py	707	# Step 6: Prepare inputs for the pipeline	COMMENT
LOW⚡	max/examples/diffusion/simple_offline_generation.py	767	# Step 8: Get the output for our request	COMMENT
LOW⚡	max/examples/diffusion/simple_offline_generation.py	778	# Step 9: Extract and save images from OutputImageContent	COMMENT
LOW	max/examples/diffusion/simple_offline_generation.py	413	# Step 1: Initialize pipeline configuration	COMMENT
LOW	max/examples/diffusion/simple_offline_generation.py	502	# Step 2: Initialize the tokenizer	COMMENT
LOW	max/examples/diffusion/simple_offline_generation.py	551	# Step 3: Initialize the pipeline	COMMENT
LOW	max/examples/diffusion/simple_offline_generation.py	575	# Step 4: Create an OpenResponsesRequest	COMMENT
LOW	max/examples/diffusion/simple_offline_generation.py	671	# Step 5: Create a PixelContext object from the request	COMMENT
LOW	max/examples/diffusion/simple_offline_generation.py	751	# Step 7: Execute the pipeline	COMMENT
LOW	bazel/internal/pytest-shard.py	72	# Step 1: Pull out anything marked with either	COMMENT
LOW	…RT/unittests/Driver/NIXL/UCX/MetadataExchangeTests.cpp	157	// Step 1: Get and load connection info only	COMMENT
LOW	…RT/unittests/Driver/NIXL/UCX/MetadataExchangeTests.cpp	169	// Step 2: Get partial metadata for first 3 buffers only	COMMENT
LOW	MLRT/lib/Driver/DeviceContext/DeviceContext.cpp	1021	// Step 1: Validate graphs and begin tracing on all streams	COMMENT
LOW⚡	MLRT/lib/Driver/DeviceContext/DeviceContext.cpp	1035	// Step 2: Execute once for all devices	COMMENT
LOW⚡	MLRT/lib/Driver/DeviceContext/DeviceContext.cpp	1041	// Step 3: Extract traces and compare to baseline for each graph	COMMENT
LOW⚡	MLRT/lib/Driver/DeviceContext/DeviceContext.cpp	1046	// Step 4: Compare with baseline (as sorted permutations to handle	COMMENT
LOW	Mojo/test/mojo-integration/extern-c-abi/README.md	158	### Step 1: Build the C Library	COMMENT
LOW	Mojo/test/mojo-integration/extern-c-abi/README.md	183	### Step 2: Build and Run a Test	COMMENT
LOW	Mojo/test/mojo-integration/extern-c-abi/README.md	195	### Step 3: Validate with FileCheck (Optional)	COMMENT
LOW	Mojo/docs/contributing/stdlib/adding-gpu-targets.md	350	### Step 1: Gather GPU information	COMMENT
LOW	Mojo/docs/contributing/stdlib/adding-gpu-targets.md	379	### Step 2: Create the target configuration	COMMENT
LOW	Mojo/docs/contributing/stdlib/adding-gpu-targets.md	404	### Step 3: Create the `GPUInfo` alias	COMMENT
LOW	Mojo/docs/contributing/stdlib/adding-gpu-targets.md	422	### Step 4: Update `_get_info_from_target`	COMMENT
18 more matches not shown…

Redundant / Tautological Comments76 hits · 112 pts

Severity	File	Line	Snippet	Context
LOW	max/kernels/benchmarks/autotune/kplot.py	93	y_text: bool = True # Print value on top of the bar	CODE
LOW	…ython/max/benchmark/benchmark_shared/datasets/types.py	275	# Check if extension is supported	COMMENT
LOW	max/python/max/nn/legacy/linear.py	331	# Check if head boundaries align with block boundaries	COMMENT
LOW	max/python/max/nn/legacy/kernels.py	13952	# Set min_top_p to the scalar value if provided, otherwise use top_p	COMMENT
LOW	max/python/max/nn/legacy/parallel/numpy_parallel_ops.py	229	# Check if all arrays are contiguous	COMMENT
LOW	max/python/max/nn/legacy/layer/layer.py	881	# Check if weight has symbolic dimensions	COMMENT
LOW	max/python/max/nn/legacy/kv_cache/cache_params.py	3767	# Check if we are allocating sufficient blocks.	COMMENT
LOW	max/python/max/nn/legacy/hooks/base_print_hook.py	140	# Print outputs.	COMMENT
LOW	max/python/max/config/__init__.py	75	# Check if the class has a custom enum mapping	COMMENT
LOW	max/python/max/config/__init__.py	133	# Check if this is Optional[T] (Union[T, None])	COMMENT
LOW	max/python/max/config/__init__.py	218	# Check if this is Optional[T] (Union[T, None])	COMMENT
LOW	max/python/max/config/__init__.py	534	# Check if this looks like a full MAXConfig file	COMMENT
LOW	max/python/max/config/__init__.py	543	# Check if the top-level keys are mostly class fields (individual config)	COMMENT
LOW	max/python/max/config/__init__.py	653	# Check if the base class has a _get_enum_mapping method	COMMENT
LOW	max/python/max/graph/weight.py	465	# Check if this is a partial function wrapping head_aware_col_sharding_strategy.	COMMENT
LOW	max/python/max/graph/weight.py	478	# Check if this is a partial function wrapping stacked_qkv_sharding_strategy.	COMMENT
LOW	max/python/max/graph/weight.py	496	# Check if this is a partial function wrapping gate_up_sharding_strategy.	COMMENT
LOW	max/python/max/graph/weights/load_gguf.py	90	# Check if a weight exists	STRING
LOW	max/python/max/graph/ops/stack.py	79	# Check if axis is within bounds	COMMENT
LOW	max/python/max/_interpreter_ops/handlers.py	1216	# Check if it's an index/integer type by looking at the attribute	COMMENT
LOW	…ax/pipelines/architectures/deepseekV4/encoding_dsv4.py	379	# Check if previous message has a task - if so, this is a task output (no thinking)	COMMENT
LOW	…python/max/pipelines/architectures/deepseekV3/model.py	196	# Check if EP should be configured	COMMENT
LOW	…hon/max/pipelines/architectures/qwen2_5vl/tokenizer.py	114	# Check if spatial merging is possible early	COMMENT
LOW	…rchitectures/autoencoders/autoencoder_kl_qwen_image.py	758	# Check if this weight needs numpy transformation	COMMENT
LOW	…/pipelines/architectures/deepseekV41/encoding_dsv41.py	715	# Check if previous message has a task - if so, this is a task output (no thinking)	COMMENT
LOW	…x/python/max/pipelines/architectures/kimik2_5/model.py	457	# Check if EP should be configured	COMMENT
LOW	…lines/modeling/dataprocessing/causal_attention_mask.py	70	# Set diagonal to k + 1 so that tokens attend to themselves.	COMMENT
LOW	…x/pipelines/kv_cache/paged_kv_cache/transfer_engine.py	167	# Set SO_REUSEADDR to avoid TIME_WAIT issues	COMMENT
LOW	…x/pipelines/kv_cache/paged_kv_cache/transfer_engine.py	1240	# Check if the relevant transport env vars are set. You can get away	COMMENT
LOW	…on/max/pipelines/kv_cache/paged_kv_cache/block_pool.py	101	# Check if a block with the same hash is already committed.	COMMENT
LOW	…pelines/kv_cache/paged_kv_cache/jenga_block_manager.py	959	# Check if we have enough blocks available to satisfy the demand.	COMMENT
LOW	max/python/max/pipelines/lib/registry.py	826	# Check if there's a secondary tokenizer (tokenizer_2) in the manifest	COMMENT
LOW	max/python/max/pipelines/lib/config/config.py	1373	# Check if the architecture is Llama3 (LlamaForCausalLM)	COMMENT
LOW	max/python/max/pipelines/lib/config/model_config.py	1465	# Check if file exists locally (direct, local repo, or cache).	COMMENT
LOW	max/python/max/pipelines/lib/pipeline_variants/utils.py	1186	# Check if any context has structured output	COMMENT
LOW	max/python/max/serve/scheduler/lora_scheduler_utils.py	59	# Check if (protected LoRAs + new LoRA) fits within capacity.	COMMENT
LOW	max/python/max/serve/scheduler/decode_scheduler.py	564	# Set dst_idx to -1 to denote pages which the decode already has due	COMMENT
LOW	…/scheduler/batch_constructor/text_batch_constructor.py	1749	# Check if the request fits in memory	COMMENT
LOW	…/scheduler/batch_constructor/text_batch_constructor.py	1880	# Check if it fits within the token budget. Under DP, a decode row	COMMENT
LOW	…/scheduler/batch_constructor/text_batch_constructor.py	1950	# Check if it fits within the token budget	COMMENT
LOW	…/scheduler/batch_constructor/text_batch_constructor.py	2699	# Set priority to the majority case, defaulting to TG if tied	COMMENT
LOW	max/python/max/serve/pipelines/llm.py	429	# Check if reasoning was disabled in the prompt. Use the	COMMENT
LOW	max/python/max/serve/pipelines/echo_gen.py	194	# Check if we have more tokens to echo and haven't reached max length	COMMENT
LOW⚡	max/python/max/serve/telemetry/common.py	428	# Check if the logger name starts with any of the allowed component prefixes	COMMENT
LOW	max/python/max/serve/router/openai_routes.py	3971	# Check if LoRA is enabled	COMMENT
LOW	max/python/max/serve/router/_image_resolution.py	997	# Check if path is within allowed roots before touching the filesystem.	COMMENT
LOW	max/python/max/serve/router/_image_resolution.py	1030	# Read the file with its own local-file size limit.	COMMENT
LOW	max/tests/integration/tools/debugging_utils.py	486	# Check if we found the tensors	COMMENT
LOW	max/tests/integration/graph/test_engine.py	342	# Check if the outputs really alias.	COMMENT
LOW	…/architectures/deepseekV3_2/test_hadamard_transform.py	35	# Check if outputs are close with appropriate tolerance	COMMENT
LOW	…ration/architectures/qwen2_5vl/test_image_processor.py	211	# Check if the difference is systematic	COMMENT
LOW	…ration/architectures/qwen2_5vl/test_image_processor.py	219	# Check if it's a scaling issue	COMMENT
LOW	…ion/architectures/qwen2_5vl/tests/test_patch_merger.py	206	# Check if outputs are close with appropriate tolerance	COMMENT
LOW	…on/architectures/qwen2_5vl/tests/test_qwen_vl_utils.py	379	# Check if both are numpy arrays or lists of PIL Images	COMMENT
LOW	…s/integration/architectures/idefics3/test_connector.py	291	# Check if outputs are close with appropriate tolerance	COMMENT
LOW	…on/architectures/idefics3/test_vision_encoder_layer.py	296	# Check if outputs are close with appropriate tolerance	COMMENT
LOW	…/integration/architectures/idefics3/test_vision_mlp.py	268	# Check if outputs are close with appropriate tolerance	COMMENT
LOW	…ration/architectures/idefics3/test_vision_attention.py	241	# Check if outputs are close with appropriate tolerance	COMMENT
LOW	…ests/integration/architectures/idefics3/torch_utils.py	51	# Check if prompt already contains chat template tokens	COMMENT
LOW	…egration/architectures/qwen3vl/test_image_processor.py	198	# Check if the difference is systematic	COMMENT
16 more matches not shown…

Verbosity Indicators54 hits · 92 pts

Severity	File	Line	Snippet	Context
LOW	max/python/max/_core/max_license/lib/Driver/Buffer.cpp	486	// If a buffer is not a contiguous slice, we need to check if the shape of	COMMENT
LOW⚡	…/max/pipelines/architectures/olmo3/layers/attention.py	170	# Step 1: QKV matmul.	COMMENT
LOW⚡	…/max/pipelines/architectures/olmo3/layers/attention.py	177	# Step 2: Apply full-dimension QK norm before rope.	COMMENT
LOW⚡	…/max/pipelines/architectures/olmo3/layers/attention.py	187	# Step 3: Fused rope + split + KV store.	COMMENT
LOW	…/max/pipelines/architectures/olmo3/layers/attention.py	202	# Step 5: Compute flash attention (reads K/V from cache)	COMMENT
LOW	…n/max/pipelines/architectures/qwen3vl_moe/tokenizer.py	800	# Step 1: Extract prompt from request and apply chat template if needed	COMMENT
LOW	…n/max/pipelines/architectures/qwen3vl_moe/tokenizer.py	837	# Step 2: Load and process images	COMMENT
LOW	…n/max/pipelines/architectures/qwen3vl_moe/tokenizer.py	855	# Step 3: Process images with custom image processor (if any) and expand <\|image_pad\|> placeholders in text	COMMENT
LOW	…n/max/pipelines/architectures/qwen3vl_moe/tokenizer.py	951	# Step 4: Tokenize the expanded text	COMMENT
LOW	…n/max/pipelines/architectures/qwen3vl_moe/tokenizer.py	1019	# Step 5: Assemble the vision encoder's inputs, in PROMPT order.	COMMENT
LOW⚡	…lines/architectures/olmo2_modulev3/layers/attention.py	144	# Step 1: QKV matmul.	COMMENT
LOW⚡	…lines/architectures/olmo2_modulev3/layers/attention.py	151	# Step 2: Apply full-dimension QK norm before rope.	COMMENT
LOW⚡	…lines/architectures/olmo2_modulev3/layers/attention.py	160	# Step 3: Fused rope + split + KV store.	COMMENT
LOW	…lines/architectures/olmo2_modulev3/layers/attention.py	175	# Step 5: Compute flash attention	COMMENT
LOW	max/python/max/pipelines/lib/eplb_rebalance.py	151	# Step 1: pack groups to nodes.	COMMENT
LOW⚡	max/python/max/pipelines/lib/eplb_rebalance.py	167	# Step 2: construct redundant experts within nodes.	COMMENT
LOW⚡	max/python/max/pipelines/lib/eplb_rebalance.py	175	# Step 3: pack physical experts to GPUs within each node.	COMMENT
LOW	max/tests/integration/accuracy/llm_fuzz/model_config.py	223	# Step 1: Try HuggingFace fetch using --model as repo ID	COMMENT
LOW	max/tests/integration/accuracy/llm_fuzz/model_config.py	236	# Step 2: Apply manual overrides (highest priority)	COMMENT
LOW⚡	…/accuracy/llm_fuzz/scenarios/tool_schema_validation.py	1371	# Step 1: initial tool call	COMMENT
LOW	…/accuracy/llm_fuzz/scenarios/tool_schema_validation.py	1390	# Step 2: provide tool result, ask follow-up that triggers another call	COMMENT
LOW	…/accuracy/llm_fuzz/scenarios/tool_schema_validation.py	1440	# Step 3: multi-tool conversation	COMMENT
LOW⚡	…/accuracy/llm_fuzz/scenarios/models/kimi_production.py	537	# Step 1: Structured output request	COMMENT
LOW⚡	…/accuracy/llm_fuzz/scenarios/models/kimi_production.py	561	# Step 2: Tool calling request	COMMENT
LOW⚡	…/accuracy/llm_fuzz/scenarios/models/kimi_production.py	576	# Step 3: Plain chat request	COMMENT
LOW⚡	max/docs/site/code/develop/index/graph_simple.py	21	# Step 1: Define the graph structure	COMMENT
LOW⚡	max/docs/site/code/develop/index/graph_simple.py	30	# Step 2: Compile the graph	COMMENT
LOW⚡	max/docs/site/code/develop/index/graph_simple.py	35	# Step 3: Execute with data	COMMENT
LOW⚡	max/examples/diffusion/simple_offline_generation.py	707	# Step 6: Prepare inputs for the pipeline	COMMENT
LOW⚡	max/examples/diffusion/simple_offline_generation.py	767	# Step 8: Get the output for our request	COMMENT
LOW⚡	max/examples/diffusion/simple_offline_generation.py	778	# Step 9: Extract and save images from OutputImageContent	COMMENT
LOW	max/examples/diffusion/simple_offline_generation.py	413	# Step 1: Initialize pipeline configuration	COMMENT
LOW	max/examples/diffusion/simple_offline_generation.py	502	# Step 2: Initialize the tokenizer	COMMENT
LOW	max/examples/diffusion/simple_offline_generation.py	551	# Step 3: Initialize the pipeline	COMMENT
LOW	max/examples/diffusion/simple_offline_generation.py	575	# Step 4: Create an OpenResponsesRequest	COMMENT
LOW	max/examples/diffusion/simple_offline_generation.py	671	# Step 5: Create a PixelContext object from the request	COMMENT
LOW	max/examples/diffusion/simple_offline_generation.py	751	# Step 7: Execute the pipeline	COMMENT
LOW	bazel/internal/pytest-shard.py	72	# Step 1: Pull out anything marked with either	COMMENT
LOW	…RT/unittests/Driver/NIXL/UCX/MetadataExchangeTests.cpp	157	// Step 1: Get and load connection info only	COMMENT
LOW	…RT/unittests/Driver/NIXL/UCX/MetadataExchangeTests.cpp	169	// Step 2: Get partial metadata for first 3 buffers only	COMMENT
LOW	MLRT/lib/Driver/DeviceContext/DeviceContext.cpp	1021	// Step 1: Validate graphs and begin tracing on all streams	COMMENT
LOW⚡	MLRT/lib/Driver/DeviceContext/DeviceContext.cpp	1035	// Step 2: Execute once for all devices	COMMENT
LOW⚡	MLRT/lib/Driver/DeviceContext/DeviceContext.cpp	1041	// Step 3: Extract traces and compare to baseline for each graph	COMMENT
LOW⚡	MLRT/lib/Driver/DeviceContext/DeviceContext.cpp	1046	// Step 4: Compare with baseline (as sorted permutations to handle	COMMENT
LOW	Mojo/lib/KGENToLLVM/LowerPOPToLLVMExternalCalls.cpp	226	// Step 4: Compute the final LLVM-level attributes we expect on the	COMMENT
LOW	Mojo/lib/KGENToLLVM/LowerPOPToLLVMExternalCalls.cpp	253	// Step 5: Lookup existing function (unified path - no early return!)	COMMENT
LOW	Mojo/lib/KGENToLLVM/LowerPOPToLLVMExternalCalls.cpp	273	// Step 6: Create function if needed (only branch on creation)	COMMENT
LOW	Mojo/lib/KGENToLLVM/LowerPOPToLLVMExternalCalls.cpp	301	// Step 7: Create call and handle the return value	COMMENT
LOW	Mojo/lib/KGENToLLVM/LowerKGENToLLVMRewriteCABIFns.cpp	189	// Step 1: Coerce arguments — replace original block args with C ABI args	COMMENT
LOW	Mojo/lib/MojoParser/ParserStmts.cpp	3223	// If we see a 'try' block in a context that cannot raise, we need to check if	COMMENT
LOW	Mojo/lib/MojoParser/ClosureEmitter.cpp	944	// Step 1: prepend a new origin slot for self and collect origin decls.	COMMENT
LOW⚡	Mojo/lib/MojoParser/ClosureEmitter.cpp	956	// Step 2: prepend explicit params for each captured parameter.	COMMENT
LOW⚡	Mojo/lib/MojoParser/ClosureEmitter.cpp	966	// Step 3: prepend the self runtime argument so it becomes arg[0].	COMMENT
LOW	Mojo/lib/MojoParser/ClosureEmitter.cpp	979	// Step 4: resolve depth-local index refs to named param/origin references.	COMMENT

Modern AI Meta-Vocabulary24 hits · 76 pts

Severity	File	Line	Snippet	Context
MEDIUM	…hmarks/misc/comparison/bench_blackwell_grouped_gemm.py	15	# Mimics bench.py scaffolding; uses naive float64 reference for correctness.	COMMENT
MEDIUM	…ython/max/benchmark/benchmark_shared/datasets/types.py	93	# Usually list[TextContentBlock], but not guaranteed: agentic-code passes	COMMENT
MEDIUM	max/python/max/graph/graph.py	835	# ``device_chains[DeviceRef.CPU()]`` is the host orchestration chain (the	COMMENT
MEDIUM	max/python/max/pipelines/architectures/laguna/arch.py	77	# scaffolding in ``model.py`` is in place but not yet validated.	COMMENT
MEDIUM	…on/max/pipelines/architectures/flux2/flux2_executor.py	734	# -- Denoising loop orchestration -----------------------------------------	COMMENT
MEDIUM	…elines/architectures/qwen3_embedding_modulev3/model.py	128	# Remove lm_head weights — embedding model doesn't use them	COMMENT
MEDIUM	max/python/max/serve/router/openresponses_routes.py	81	# Request is already parsed and validated via dependency injection	COMMENT
MEDIUM	max/python/max/serve/router/openai_routes.py	1869	# so replayed chain-of-thought is not silently dropped before the	COMMENT
MEDIUM	…tion_pipeline/test_overlap_text_generation_pipeline.py	417	# Worst-case boundary request: committed tokens fill the context window and	COMMENT
MEDIUM	max/tests/integration/accuracy/run_mmmu_video_eval.py	276	# not here: when a reasoning parser is active, chain-of-thought is split out	COMMENT
MEDIUM	max/tests/integration/accuracy/run_mmmu_video_eval.py	443	# The server's reasoning parser (if any) already removed chain-of-thought	COMMENT
MEDIUM	max/tests/integration/accuracy/run_longbench_v2.py	121	# Truncate prompt to fit context window	COMMENT
MEDIUM	…sts/integration/accuracy/model_evals/run_text_evals.sh	14	# Runs all non-agentic text evals in sequence against a running server. Each	COMMENT
MEDIUM⚡	…tion/accuracy/model_evals/run_exacto_taubench_local.sh	15	# tau2-bench airline — the agentic tool-calling half of OpenRouter's AutoExacto	COMMENT
MEDIUM⚡	…tion/accuracy/model_evals/run_exacto_taubench_local.sh	18	# This is the half of the gate we had no coverage for: the suite's other agentic	COMMENT
MEDIUM	max/tests/integration/accuracy/llm_fuzz/client.py	74	# derives it from the model's context window).	COMMENT
MEDIUM	max/tests/integration/accuracy/llm_fuzz/README.md	170	# Default: 8 short conversations plus 2 at half the context window,	COMMENT
MEDIUM	…ation/accuracy/llm_fuzz/scenarios/kv_cache_pressure.py	119	# Probe near the model's context window boundary.	COMMENT
MEDIUM	…ntegration/accuracy/llm_fuzz/scenarios/tool_calling.py	936	# multi-step agentic prompt encourages several tool calls; the	COMMENT
MEDIUM	…gration/accuracy/llm_fuzz/scenarios/models/kimi_3am.py	152	# 8K char tool result in history -- common in RAG pipelines	COMMENT
MEDIUM⚡	max/tests/tests/benchmark/test_benchmark_config.py	709	# --agentic-tool-profiles validation	COMMENT
MEDIUM	max/tests/tests/benchmark/test_benchmark_datasets.py	1541	# at run time, like agentic-code / instruct-coder).	COMMENT
MEDIUM	max/examples/embedding-knowledge-base/README.md	1	# Knowledge base with a text embedding model	COMMENT
MEDIUM	…/stdlib/scripts/unicode-data/GraphemeBreakProperty.txt	1200	B77D..B797 ; LVT # Lo [27] HANGUL SYLLABLE RAG..HANGUL SYLLABLE RAH	CODE

Cross-Language Confusion (JS/TS)13 hits · 62 pts

Severity	File	Line	Snippet	Context
HIGH	Mojo/test/mojo-lsp-server-node/src/hover.spec.ts	26	print(other)	CODE
HIGH	Mojo/test/mojo-lsp-server-node/src/hover.spec.ts	613	print("string")	CODE
HIGH	Mojo/test/mojo-lsp-server-node/src/hover.spec.ts	614	print(arg)	CODE
HIGH	Mojo/test/mojo-lsp-server-node/src/completion.spec.ts	349	elif arg.value:`	CODE
HIGH	Mojo/test/mojo-lsp-server-node/src/codeAction.spec.ts	22	print(repr(b))	CODE
HIGH	Mojo/test/mojo-lsp-server-node/src/definitions.spec.ts	33	print(y)	CODE
HIGH	Mojo/test/mojo-lsp-server-node/src/definitions.spec.ts	69	print(arg)	CODE
HIGH	Mojo/test/mojo-lsp-server-node/src/references.spec.ts	24	print(foo)	CODE
HIGH	Mojo/test/mojo-lsp-server-node/src/references.spec.ts	25	print("foo")	CODE
HIGH	Mojo/test/mojo-lsp-server-node/src/regression.spec.ts	20	print("hi")	STRING
HIGH	Mojo/test/mojo-lsp-server-node/src/rename.spec.ts	21	print(someVar)	CODE
HIGH	Mojo/test/mojo-lsp-server-node/src/rename.spec.ts	120	print(1 + 2)	CODE
HIGH	Mojo/test/mojo-lsp-server-node/src/rename.spec.ts	153	print(1 + 2)	CODE

Example Usage Blocks26 hits · 40 pts

Severity	File	Line	Snippet	Context
LOW	Kernels/lib/max_license/Tracing/Tracing.h	63	/// Example usage:	COMMENT
LOW	…rnels/benchmarks/misc/utils/cutlass_profiler_parser.py	18	# Usage:	COMMENT
LOW	…enchmarks/misc/utils/bench_vllm_flashinfer_decoding.py	17	# Usage:	COMMENT
LOW	…nchmarks/misc/utils/bench_vllm_flashinfer_attention.py	17	# Usage:	COMMENT
LOW	…s/misc/comparison/bench_mla_decode_sm100_kimi_fp8.yaml	21	# Usage:	COMMENT
LOW	…hmarks/misc/comparison/bench_blackwell_grouped_gemm.py	18	# Usage example:	COMMENT
LOW	…c/comparison/bench_mla_decode_sm100_deepseek_bf16.yaml	21	# Usage:	COMMENT
LOW	…/misc/comparison/bench_mla_decode_sm100_kimi_bf16.yaml	21	# Usage:	COMMENT
LOW	…sc/comparison/bench_mla_decode_sm100_deepseek_fp8.yaml	21	# Usage:	COMMENT
LOW	…kernels/benchmarks/gpu/nn/bench_fp8_index_prefill.yaml	18	# Usage:	COMMENT
LOW	…s/benchmarks/gpu/nn/bench_mla_sparse_prefill_tune.yaml	18	# Usage:	COMMENT
LOW	max/kernels/benchmarks/sweep/bench_dsa_block.yaml	32	# Usage:	COMMENT
LOW	…sweep/bench_kv_cache_ragged_flash_attention_paged.yaml	33	# Usage:	COMMENT
LOW	…ernels/benchmarks/sweep/bench_mla_index_fp8_paged.yaml	33	# Usage:	COMMENT
LOW	…/kernels/benchmarks/sweep/bench_mla_decode_sparse.yaml	34	# Usage:	COMMENT
LOW	…kernels/benchmarks/sweep/bench_mla_prefill_sparse.yaml	30	# Usage:	COMMENT
LOW	…benchmark/configs/serving_flux2_t2i_random_config.yaml	15	# Usage:	COMMENT
LOW⚡	…gration/accuracy/model_evals/run_omnidocbench_local.sh	28	# Usage:	COMMENT
LOW	…tegration/accuracy/model_evals/run_exacto_vgi_local.sh	75	# Usage:	COMMENT
LOW	…tion/accuracy/model_evals/run_exacto_taubench_local.sh	53	# Usage:	COMMENT
LOW⚡	…egration/accuracy/model_evals/run_exacto_gpqa_local.sh	59	# Usage:	COMMENT
LOW	…s/embedding-knowledge-base/src/embeddings/kb_system.py	134	# Example usage	COMMENT
LOW	Support/include/Support/ADT/ConcurrentAppendingVector.h	44	/// Example usage:	COMMENT
LOW	…/lib/Driver/DeviceContext/CUDA/CUDACompilationDevice.h	38	/// Example usage:	COMMENT
LOW	…RT/lib/Driver/DeviceContext/HIP/HIPCompilationDevice.h	41	/// Example usage:	COMMENT
LOW	GraphCompiler/utils/warm_cache_pipeline_models.sh	19	# Usage:	COMMENT

Fake / Example Data18 hits · 20 pts

Severity	File	Line	Snippet	Context
LOW	…ration/accuracy/llm_fuzz/scenarios/openrouter_tests.py	1772	lorem = "Lorem ipsum dolor sit amet, consectetur adipiscing elit. " * (	CODE
LOW	…ration/accuracy/llm_fuzz/scenarios/openrouter_tests.py	1772	lorem = "Lorem ipsum dolor sit amet, consectetur adipiscing elit. " * (	CODE
LOW	…scenarios/validation/basic_reasoning_and_tool_usage.py	150	"lorem ipsum dolor sit amet, consectetur adipiscing elit, sed do "	CODE
LOW	…scenarios/validation/basic_reasoning_and_tool_usage.py	150	"lorem ipsum dolor sit amet, consectetur adipiscing elit, sed do "	CODE
LOW	max/tests/tests/graph/ops/test_call.py	241	"placeholder",	CODE
LOW	…thon-extension-modules/feature-overview/test_module.py	78	person.change_name("John Doe")	CODE
LOW	…thon-extension-modules/feature-overview/test_module.py	79	assert person.name() == "John Doe"	CODE
LOW	…est/python-extension-modules/def-method/test_module.py	41	same_person = person._with("Jane Doe", 25)	CODE
LOW	…est/python-extension-modules/def-method/test_module.py	42	assert person.get_name() == "Jane Doe"	CODE
LOW	…est/python-extension-modules/def-method/test_module.py	59	assert person.get_name() == "Jane Doe"	CODE
LOW⚡	…est/python-extension-modules/def-method/test_module.py	85	person.set_name_and_age("John Doe", 25)	CODE
LOW⚡	…est/python-extension-modules/def-method/test_module.py	86	assert person.get_name() == "John Doe"	CODE
LOW⚡	…est/python-extension-modules/def-method/test_module.py	96	person.set_name_and_age("John Doe", 25)	CODE
LOW⚡	…est/python-extension-modules/def-method/test_module.py	105	person.set_name("John Doe")	CODE
LOW⚡	…est/python-extension-modules/def-method/test_module.py	106	assert person.get_name() == "John Doe"	CODE
LOW	Mojo/docs/site/releases/2022-11.md	40	a new "placeholder" attribute. This allows us to expose the power of the KGEN	CODE
LOW	Mojo/stdlib/test/builtin/test_file_dummy_input.txt	1	Lorem ipsum dolor sit amet, consectetur adipiscing elit.	CODE
LOW	Mojo/stdlib/test/builtin/test_file_dummy_input.txt	1	Lorem ipsum dolor sit amet, consectetur adipiscing elit.	CODE

Synthetic Comment Markers2 hits · 15 pts

Severity	File	Line	Snippet	Context
HIGH	MLRT/lib/Driver/DeviceContext/HAL/HALDeviceContext.cpp	574	// Keyed on the api as requested, so a repeat call avoids reloading the plugin	COMMENT
HIGH	Mojo/include/Mojo/Compiler/Target/TargetBackend.h	237	/// host policy (AddressSanitizer and ThreadSanitizer as requested in	COMMENT

Hallucination Indicators1 hit · 10 pts

Severity	File	Line	Snippet	Context
CRITICAL	max/docs/site/releases/v26.7.md	785	- Removed `max.experimental.nn.common_layers.functional_kernels.local_map()`.	CODE

Dead Code4 hits · 8 pts

Severity	File	Line	Context
MEDIUM	max/tests/tests/serve/test_start_stream.py	76	CODE
MEDIUM	max/tests/tests/serve/test_openresponses_routes.py	518	CODE
MEDIUM	max/tests/tests/serve/test_completions_stream_errors.py	116	CODE
MEDIUM	max/tests/tests/serve/test_completions_stream_errors.py	133	CODE

Slop Phrases2 hits · 3 pts

Severity	File	Line	Snippet	Context
LOW	max/tests/integration/accuracy/verify_pipelines.py	290	# Even if verdicts is empty, we want to make sure to call write. When we	COMMENT
LOW	Mojo/lib/MojoJupyter/enable_matplotlib.py	81	# end of our display data. We also make sure to flush the output stream	COMMENT

Overly Generic Function Names3 hits · 2 pts

Severity	File	Line	Snippet	Context
LOW	max/python/max/serve/process_control.py	278	async def run_task() -> None:	CODE
LOW	max/tests/integration/test_common/registry.py	34	def test_function(manage_production_models):	STRING
LOW	Mojo/stdlib/test/python/my_module.py	31	def my_function(name) -> str: # noqa: ANN001	CODE

TODO Padding1 hit · 2 pts

Severity	File	Line	Snippet	Context
LOW	GraphCompiler/lib/MODialect/Canonicalize.cpp	655	// TODO: implement this	COMMENT

modular/modular

Open-Source Code Quality, Security and Technical Debt Review