alibaba/MNN

3.9

Adjusted Score

3.9

Raw Score

100%

Time Factor

2026-08-01

Last Push

15.8K

Stars

C++

Language

1.6M

Lines of Code

6.0K

Files

5.4K

Pattern Hits

2026-08-02

Scan Date

0.02

HC Hit Rate

What These Metrics Mean

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

Score History

This chart maps the temporal evolution of the adjusted synthetic code score across successive scan runs. An upward trajectory indicates ongoing incorporation of AI-generated code or expanding LLM-assisted scaffolding; a stable or declining trajectory may reflect active human refactoring, code removal, or the adoption of stricter authorship policies. The dashed secondary line (right axis) independently tracks total raw pattern hit count, which can diverge from the normalised score when codebase size changes significantly between scans.

Severity Breakdown

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

CRITICAL 30HIGH 99MEDIUM 211LOW 5015

Directory Score Breakdown

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

Pattern Findings

The scanner identified 5355 distinct pattern matches across 21 syntactic categories. Each entry below represents a discrete location in the source code where the engine recorded a statistically significant AI authorship indicator. Expand any category row to inspect the individual file paths, line numbers, code snippets, and the lexical context (CODE, COMMENT, or STRING) in which each match was detected.

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

Over-Commented Block3935 hits · 3575 pts

Severity	File	Line	Snippet	Context
LOW	test.sh	1	#!/usr/bin/env bash	COMMENT
LOW	test.sh	21	# Defaults to taobao-mnn/Qwen2.5-0.5B-Instruct-MNN.	COMMENT
LOW	test.sh	1141	# Unit-test matrix for the Android arm64 on-device driver.	COMMENT
LOW	test.sh	1161	opencl-buffer) [[ "${tag}" == "opencl-buffer" ]] ;;	COMMENT
LOW	.readthedocs.yaml	21	# builder: "dirhtml"	COMMENT
LOW	benchmark/benchmark.cpp	1	//	COMMENT
LOW	benchmark/benchmark.cpp	21	#undef min	COMMENT
LOW	benchmark/benchmark.h	1	//	COMMENT
LOW	benchmark/benchmarkExprModels.cpp	1	#include <iostream>	COMMENT
LOW	benchmark/exprModels/ShuffleNetExpr.hpp	1	//	COMMENT
LOW	benchmark/exprModels/ResNetExpr.cpp	1	//	COMMENT
LOW	benchmark/exprModels/SqueezeNetExpr.hpp	1	//	COMMENT
LOW	benchmark/exprModels/MobileNetExpr.cpp	1	//	COMMENT
LOW	benchmark/exprModels/GoogLeNetExpr.cpp	1	//	COMMENT
LOW	benchmark/exprModels/SqueezeNetExpr.cpp	1	//	COMMENT
LOW	benchmark/exprModels/MobileNetExpr.hpp	1	//	COMMENT
LOW	benchmark/exprModels/GoogLeNetExpr.hpp	1	//	COMMENT
LOW	benchmark/exprModels/ShuffleNetExpr.cpp	1	//	COMMENT
LOW	benchmark/exprModels/ExprModels.hpp	1	#ifndef ExprModels_hpp	COMMENT
LOW	benchmark/exprModels/ResNetExpr.hpp	1	//	COMMENT
LOW	demo/exec/transformerDemo.cpp	1	//	COMMENT
LOW	demo/exec/multiPose.cpp	1	//	COMMENT
LOW	demo/exec/multiPose.cpp	21		COMMENT
LOW	demo/exec/multiPose.cpp	361	const auto rgbaPtr = reinterpret_cast<uint8_t*>(inputImage);	COMMENT
LOW	demo/exec/expressDemo.cpp	1	//	COMMENT
LOW	demo/exec/pictureRecognition_batch.cpp	1	//	COMMENT
LOW	demo/exec/expressMakeModel.cpp	1	//	COMMENT
LOW	demo/exec/multithread_imgrecog.cpp	1	//	COMMENT
LOW	demo/exec/pictureRecognition.cpp	1	//	COMMENT
LOW	demo/exec/pictureRecognition_module.cpp	1	//	COMMENT
LOW	demo/exec/segment.cpp	1	//	COMMENT
LOW	demo/exec/pictureRotate.cpp	1	//	COMMENT
LOW	demo/exec/nluDemo.cpp	1	//	COMMENT
LOW	3rd_party/musa_compat/CMakeLists.txt	1	# MUSA Compatibility Layer CMake Configuration	COMMENT
LOW	3rd_party/half/half.hpp	1	// half - IEEE 754-based half-precision floating point library.	COMMENT
LOW	3rd_party/half/half.hpp	21		COMMENT
LOW	3rd_party/half/half.hpp	41	#if __has_feature(cxx_noexcept) && !defined(HALF_ENABLE_CPP11_NOEXCEPT)	COMMENT
LOW	3rd_party/half/half.hpp	61	#define HALF_ENABLE_CPP11_LONG_LONG 1	COMMENT
LOW	3rd_party/half/half.hpp	81	#elif defined(_MSC_VER) //Visual C++	COMMENT
LOW	3rd_party/half/half.hpp	101		COMMENT
LOW	3rd_party/half/half.hpp	121	#define HALF_NOTHROW noexcept	COMMENT
LOW	3rd_party/half/half.hpp	141	#include <functional>	COMMENT
LOW	3rd_party/half/half.hpp	161	#ifndef HALF_ROUND_STYLE	COMMENT
LOW	3rd_party/half/half.hpp	181	/// half-precision multiplication followed by an addition. Due to the internal single-precision implementation of all	COMMENT
LOW	3rd_party/half/half.hpp	201	#define FP_INFINITE 3	COMMENT
LOW	3rd_party/half/half.hpp	281	/// Unsigned integer of (at least) 32 bits width.	COMMENT
LOW	3rd_party/half/half.hpp	341	template<typename T,typename U> struct result : enable<expr,T,U> {};	COMMENT
LOW	3rd_party/half/half.hpp	361	}	COMMENT
LOW	3rd_party/half/half.hpp	381	/// \param arg value to query	COMMENT
LOW	3rd_party/half/half.hpp	621	return hbits;	COMMENT
LOW	3rd_party/half/half.hpp	941	/// \param value binary representation of half-precision value	COMMENT
LOW	3rd_party/half/half.hpp	1001	}	COMMENT
LOW	3rd_party/half/half.hpp	1061	/// \tparam T type of concrete half expression	COMMENT
LOW	3rd_party/half/half.hpp	1081	/// \return reference to this half	COMMENT
LOW	3rd_party/half/half.hpp	1141	/// While this returns an actual half-precision value, half literals can unfortunately not be constant expressions due	COMMENT
LOW	3rd_party/half/half.hpp	1161	/// \param x first operand	COMMENT
LOW	3rd_party/half/half.hpp	1281	/// \param y second operand	COMMENT
LOW	3rd_party/half/half.hpp	1301	#else	COMMENT
LOW	3rd_party/half/half.hpp	1361	/// Binary logarithm implementation.	COMMENT
LOW	3rd_party/half/half.hpp	1401	#else	COMMENT
3875 more matches not shown…

Cross-File Repetition95 hits · 475 pts

Severity	File	Line	Snippet	Context
HIGH	…ls/mnncompress/mnncompress/tensorflow/LSQ_quantizer.py	0	only invoke this when you need to save the model for inference	STRING
HIGH	…ls/mnncompress/mnncompress/tensorflow/EMA_quantizer.py	0	only invoke this when you need to save the model for inference	STRING
HIGH	…ls/mnncompress/mnncompress/tensorflow/OAQ_quantizer.py	0	only invoke this when you need to save the model for inference	STRING
HIGH	source/backend/opencl/schema/CLCache/Shader.py	0	this method is deprecated. please switch to getrootas.	STRING
HIGH	source/backend/opencl/schema/CLCache/BackendInfo.py	0	this method is deprecated. please switch to getrootas.	STRING
HIGH	source/backend/opencl/schema/CLCache/Autotuning.py	0	this method is deprecated. please switch to getrootas.	STRING
HIGH	source/backend/opencl/schema/CLCache/Cache.py	0	this method is deprecated. please switch to getrootas.	STRING
HIGH	source/backend/opencl/schema/CLCache/OpInfo.py	0	this method is deprecated. please switch to getrootas.	STRING
HIGH	source/backend/opencl/schema/CLCache/GemmInfo.py	0	this method is deprecated. please switch to getrootas.	STRING
HIGH	source/backend/opencl/schema/CLCache/TensorInfo.py	0	this method is deprecated. please switch to getrootas.	STRING
HIGH	…mnn/sherpa-mnn/python/tests/test_offline_recognizer.py	0	args: wave_filename: path to a wave file. it should be single channel and each sample should be 16-bit. its sample rate	STRING
HIGH	…pa-mnn/sherpa-mnn/python/tests/test_keyword_spotter.py	0	args: wave_filename: path to a wave file. it should be single channel and each sample should be 16-bit. its sample rate	STRING
HIGH	…nn/sherpa-mnn/python/tests/test_speaker_recognition.py	0	args: wave_filename: path to a wave file. it should be single channel and each sample should be 16-bit. its sample rate	STRING
HIGH	…-mnn/sherpa-mnn/python/tests/test_online_recognizer.py	0	args: wave_filename: path to a wave file. it should be single channel and each sample should be 16-bit. its sample rate	STRING
HIGH	…/python-api-examples/spoken-language-identification.py	0	args: wave_filename: path to a wave file. it should be single channel and each sample should be 16-bit. its sample rate	STRING
HIGH	…ples/offline-websocket-client-decode-files-paralell.py	0	args: wave_filename: path to a wave file. it should be single channel and each sample should be 16-bit. its sample rate	STRING
HIGH	…sherpa-mnn/python-api-examples/offline-decode-files.py	0	args: wave_filename: path to a wave file. it should be single channel and each sample should be 16-bit. its sample rate	STRING
HIGH	…/sherpa-mnn/python-api-examples/online-decode-files.py	0	args: wave_filename: path to a wave file. it should be single channel and each sample should be 16-bit. its sample rate	STRING
HIGH	…orks/sherpa-mnn/python-api-examples/keyword-spotter.py	0	args: wave_filename: path to a wave file. it should be single channel and each sample should be 16-bit. its sample rate	STRING
HIGH	…n-api-examples/online-zipformer-ctc-hlg-decode-file.py	0	args: wave_filename: path to a wave file. it should be single channel and each sample should be 16-bit. its sample rate	STRING
HIGH	…es/offline-websocket-client-decode-files-sequential.py	0	args: wave_filename: path to a wave file. it should be single channel and each sample should be 16-bit. its sample rate	STRING
HIGH	…on-api-examples/vad-remove-non-speech-segments-alsa.py	0	the device name specifies which microphone to use in case there are several on your system. you can use arecord -l to fi	STRING
HIGH	…aker-identification-with-vad-non-streaming-asr-alsa.py	0	the device name specifies which microphone to use in case there are several on your system. you can use arecord -l to fi	STRING
HIGH	…nition-from-microphone-with-endpoint-detection-alsa.py	0	the device name specifies which microphone to use in case there are several on your system. you can use arecord -l to fi	STRING
HIGH	…/frameworks/sherpa-mnn/python-api-examples/vad-alsa.py	0	the device name specifies which microphone to use in case there are several on your system. you can use arecord -l to fi	STRING
HIGH	…n-api-examples/offline-sense-voice-ctc-decode-files.py	0	please download model files from https://github.com/k2-fsa/sherpa-onnx/releases/tag/asr-models	STRING
HIGH	…-api-examples/inverse-text-normalization-online-asr.py	0	please download model files from https://github.com/k2-fsa/sherpa-onnx/releases/tag/asr-models	STRING
HIGH	…n/python-api-examples/offline-nemo-ctc-decode-files.py	0	please download model files from https://github.com/k2-fsa/sherpa-onnx/releases/tag/asr-models	STRING
HIGH	…n-api-examples/offline-nemo-transducer-decode-files.py	0	please download model files from https://github.com/k2-fsa/sherpa-onnx/releases/tag/asr-models	STRING
HIGH	…nn/python-api-examples/offline-whisper-decode-files.py	0	please download model files from https://github.com/k2-fsa/sherpa-onnx/releases/tag/asr-models	STRING
HIGH	…api-examples/inverse-text-normalization-offline-asr.py	0	please download model files from https://github.com/k2-fsa/sherpa-onnx/releases/tag/asr-models	STRING
HIGH	…/python-api-examples/offline-moonshine-decode-files.py	0	please download model files from https://github.com/k2-fsa/sherpa-onnx/releases/tag/asr-models	STRING
HIGH	…on-api-examples/offline-telespeech-ctc-decode-files.py	0	please download model files from https://github.com/k2-fsa/sherpa-onnx/releases/tag/asr-models	STRING
HIGH	…thon-api-examples/offline-fire-red-asr-decode-files.py	0	please download model files from https://github.com/k2-fsa/sherpa-onnx/releases/tag/asr-models	STRING
HIGH	…nn/python-api-examples/online-nemo-ctc-decode-files.py	0	please download model files from https://github.com/k2-fsa/sherpa-onnx/releases/tag/asr-models	STRING
HIGH	…sherpa-mnn/python-api-examples/non_streaming_server.py	0	it specifies the spoken language in the input file. example values: en, fr, de, zh, jp. available languages for multilin	STRING
HIGH	…xamples/two-pass-speech-recognition-from-microphone.py	0	it specifies the spoken language in the input file. example values: en, fr, de, zh, jp. available languages for multilin	STRING
HIGH	…sherpa-mnn/python-api-examples/offline-decode-files.py	0	it specifies the spoken language in the input file. example values: en, fr, de, zh, jp. available languages for multilin	STRING
HIGH	…s/sherpa-mnn/python-api-examples/generate-subtitles.py	0	it specifies the spoken language in the input file. example values: en, fr, de, zh, jp. available languages for multilin	STRING
HIGH	…aker-identification-with-vad-non-streaming-asr-alsa.py	0	it specifies the spoken language in the input file. example values: en, fr, de, zh, jp. available languages for multilin	STRING
HIGH	…-mnn/python-api-examples/vad-with-non-streaming-asr.py	0	it specifies the spoken language in the input file. example values: en, fr, de, zh, jp. available languages for multilin	STRING
HIGH	…s/speaker-identification-with-vad-non-streaming-asr.py	0	it specifies the spoken language in the input file. example values: en, fr, de, zh, jp. available languages for multilin	STRING
HIGH	…sherpa-mnn/python-api-examples/non_streaming_server.py	0	for multilingual models, if you specify translate, the output will be in english.	STRING
HIGH	…s/sherpa-mnn/python-api-examples/generate-subtitles.py	0	for multilingual models, if you specify translate, the output will be in english.	STRING
HIGH	…aker-identification-with-vad-non-streaming-asr-alsa.py	0	for multilingual models, if you specify translate, the output will be in english.	STRING
HIGH	…xamples/two-pass-speech-recognition-from-microphone.py	0	for multilingual models, if you specify translate, the output will be in english.	STRING
HIGH	…-mnn/python-api-examples/vad-with-non-streaming-asr.py	0	for multilingual models, if you specify translate, the output will be in english.	STRING
HIGH	…sherpa-mnn/python-api-examples/offline-decode-files.py	0	for multilingual models, if you specify translate, the output will be in english.	STRING
HIGH	…s/speaker-identification-with-vad-non-streaming-asr.py	0	for multilingual models, if you specify translate, the output will be in english.	STRING
HIGH	…sherpa-mnn/python-api-examples/non_streaming_server.py	0	number of tail padding frames. we have removed the 30-second constraint from whisper, so you need to choose the amount o	STRING
HIGH	…s/sherpa-mnn/python-api-examples/generate-subtitles.py	0	number of tail padding frames. we have removed the 30-second constraint from whisper, so you need to choose the amount o	STRING
HIGH	…aker-identification-with-vad-non-streaming-asr-alsa.py	0	number of tail padding frames. we have removed the 30-second constraint from whisper, so you need to choose the amount o	STRING
HIGH	…xamples/two-pass-speech-recognition-from-microphone.py	0	number of tail padding frames. we have removed the 30-second constraint from whisper, so you need to choose the amount o	STRING
HIGH	…-mnn/python-api-examples/vad-with-non-streaming-asr.py	0	number of tail padding frames. we have removed the 30-second constraint from whisper, so you need to choose the amount o	STRING
HIGH	…sherpa-mnn/python-api-examples/offline-decode-files.py	0	number of tail padding frames. we have removed the 30-second constraint from whisper, so you need to choose the amount o	STRING
HIGH	…s/speaker-identification-with-vad-non-streaming-asr.py	0	number of tail padding frames. we have removed the 30-second constraint from whisper, so you need to choose the amount o	STRING
HIGH	…sherpa-mnn/python-api-examples/non_streaming_server.py	0	used only when --first-decoding-method is modified_beam_search. it specifies number of active paths to keep during decod	STRING
HIGH	…hon-api-examples/speech-recognition-from-microphone.py	0	used only when --first-decoding-method is modified_beam_search. it specifies number of active paths to keep during decod	STRING
HIGH	…/sherpa-mnn/python-api-examples/online-decode-files.py	0	used only when --first-decoding-method is modified_beam_search. it specifies number of active paths to keep during decod	STRING
HIGH	…rks/sherpa-mnn/python-api-examples/streaming_server.py	0	used only when --first-decoding-method is modified_beam_search. it specifies number of active paths to keep during decod	STRING
35 more matches not shown…

Decorative Section Separators131 hits · 408 pts

Severity	File	Line	Snippet	Context
MEDIUM	build_lib.sh	296	# ============================================================================	COMMENT
MEDIUM	build_lib.sh	298	# ============================================================================	COMMENT
MEDIUM	build_lib.sh	410	# ============================================================================	COMMENT
MEDIUM	build_lib.sh	412	# ============================================================================	COMMENT
MEDIUM	build_lib.sh	489	# ============================================================================	COMMENT
MEDIUM	build_lib.sh	491	# ============================================================================	COMMENT
MEDIUM	build_lib.sh	648	# ============================================================================	COMMENT
MEDIUM	build_lib.sh	650	# ============================================================================	COMMENT
MEDIUM	build_lib.sh	727	# ============================================================================	COMMENT
MEDIUM	build_lib.sh	729	# ============================================================================	COMMENT
MEDIUM	build_lib.sh	784	# ============================================================================	COMMENT
MEDIUM	build_lib.sh	786	# ============================================================================	COMMENT
MEDIUM	test.sh	43	# ─────────────────────────────────────────────────────────────────────────────	COMMENT
MEDIUM	test.sh	45	# ─────────────────────────────────────────────────────────────────────────────	COMMENT
MEDIUM	test.sh	136	# ─────────────────────────────────────────────────────────────────────────────	COMMENT
MEDIUM	test.sh	138	# ─────────────────────────────────────────────────────────────────────────────	COMMENT
MEDIUM	test.sh	155	# ─────────────────────────────────────────────────────────────────────────────	COMMENT
MEDIUM	test.sh	157	# ─────────────────────────────────────────────────────────────────────────────	COMMENT
MEDIUM	test.sh	199	# ─────────────────────────────────────────────────────────────────────────────	COMMENT
MEDIUM	test.sh	201	# ─────────────────────────────────────────────────────────────────────────────	COMMENT
MEDIUM	test.sh	275	# ─────────────────────────────────────────────────────────────────────────────	COMMENT
MEDIUM	test.sh	277	# ─────────────────────────────────────────────────────────────────────────────	COMMENT
MEDIUM	test.sh	331	# ─────────────────────────────────────────────────────────────────────────────	COMMENT
MEDIUM	test.sh	333	# ─────────────────────────────────────────────────────────────────────────────	COMMENT
MEDIUM	test.sh	366	# ─────────────────────────────────────────────────────────────────────────────	COMMENT
MEDIUM	test.sh	368	# ─────────────────────────────────────────────────────────────────────────────	COMMENT
MEDIUM	test.sh	478	# ─────────────────────────────────────────────────────────────────────────────	COMMENT
MEDIUM	test.sh	480	# ─────────────────────────────────────────────────────────────────────────────	COMMENT
MEDIUM	test.sh	536	# ─────────────────────────────────────────────────────────────────────────────	COMMENT
MEDIUM	test.sh	543	# ─────────────────────────────────────────────────────────────────────────────	COMMENT
MEDIUM	test.sh	614	# ─────────────────────────────────────────────────────────────────────────────	COMMENT
MEDIUM	test.sh	616	# ─────────────────────────────────────────────────────────────────────────────	COMMENT
MEDIUM	test.sh	632	# ─────────────────────────────────────────────────────────────────────────────	COMMENT
MEDIUM	test.sh	634	# ─────────────────────────────────────────────────────────────────────────────	COMMENT
MEDIUM	test.sh	748	# ─────────────────────────────────────────────────────────────────────────────	COMMENT
MEDIUM	test.sh	751	# ─────────────────────────────────────────────────────────────────────────────	COMMENT
MEDIUM	test.sh	791	# ─────────────────────────────────────────────────────────────────────────────	COMMENT
MEDIUM	test.sh	793	# ─────────────────────────────────────────────────────────────────────────────	COMMENT
MEDIUM	test.sh	827	# ─────────────────────────────────────────────────────────────────────────────	COMMENT
MEDIUM	test.sh	829	# ─────────────────────────────────────────────────────────────────────────────	COMMENT
MEDIUM	test.sh	962	# ─────────────────────────────────────────────────────────────────────────────	COMMENT
MEDIUM	test.sh	964	# ─────────────────────────────────────────────────────────────────────────────	COMMENT
MEDIUM	test.sh	996	# ─────────────────────────────────────────────────────────────────────────────	COMMENT
MEDIUM	test.sh	999	# ─────────────────────────────────────────────────────────────────────────────	COMMENT
MEDIUM	test.sh	1137	# ─────────────────────────────────────────────────────────────────────────────	COMMENT
MEDIUM	test.sh	1151	# ─────────────────────────────────────────────────────────────────────────────	COMMENT
MEDIUM	test.sh	1258	# ─────────────────────────────────────────────────────────────────────────────	COMMENT
MEDIUM	test.sh	1260	# ─────────────────────────────────────────────────────────────────────────────	COMMENT
MEDIUM	test.sh	1318	# ─────────────────────────────────────────────────────────────────────────────	COMMENT
MEDIUM	test.sh	1320	# ─────────────────────────────────────────────────────────────────────────────	COMMENT
MEDIUM⚡	tools/script/generate_podspec.sh	2	# =============================================================================	COMMENT
MEDIUM⚡	tools/script/generate_podspec.sh	15	# =============================================================================	COMMENT
MEDIUM⚡	tools/script/generate_podspec.sh	215	# =============================================================================	COMMENT
MEDIUM⚡	tools/script/generate_podspec.sh	217	# =============================================================================	COMMENT
MEDIUM⚡	tools/script/generate_podspec.sh	318	# =============================================================================	COMMENT
MEDIUM⚡	tools/script/generate_podspec.sh	320	# =============================================================================	COMMENT
MEDIUM	tools/script/generate_podspec.sh	93	# =============================================================================	COMMENT
MEDIUM	tools/script/generate_podspec.sh	95	# =============================================================================	COMMENT
MEDIUM	tools/script/generate_podspec.sh	123	# =============================================================================	COMMENT
MEDIUM	tools/script/generate_podspec.sh	125	# =============================================================================	COMMENT
71 more matches not shown…

Hallucination Indicators30 hits · 320 pts

Severity	File	Line	Snippet	Context
CRITICAL	transformers/llm/export/llmexport.py	122	self.args.tie_word_embeddings = not self.args.seperate_embed and self.model.lm.lm.weight.equal(self.model.embed.	CODE
CRITICAL	…java/com/alibaba/mnnllm/android/utils/MmapUtilsTest.kt	36	every { com.alibaba.mls.api.ApplicationProvider.get() } returns application	CODE
CRITICAL	…/alibaba/mnnllm/android/benchmark/BenchmarkFragment.kt	613	return com.alibaba.mnnllm.android.modelmarket.TagMapper.getDisplayTagList(modelItem.tags).take(3)	CODE
CRITICAL	…main/java/com/alibaba/mnnllm/android/llm/LlmSession.kt	447	com.alibaba.mnnllm.android.benchmark.BenchmarkResult(	CODE
CRITICAL	…libaba/mnnllm/android/chat/ModelSelectionViewHolder.kt	51	return com.alibaba.mnnllm.android.modelmarket.TagMapper.getDisplayTagList(modelItem.tags).take(3)	CODE
CRITICAL	…libaba/mnnllm/android/chat/chatlist/ChatViewHolders.kt	268	com.alibaba.mnnllm.android.utils.ImageUtils.showImageMenu(it.context, imageUri)	CODE
CRITICAL	…libaba/mnnllm/android/chat/chatlist/ChatViewHolders.kt	422	com.alibaba.mnnllm.android.utils.ImageUtils.showImageMenu(v.context, imageUri)	CODE
CRITICAL	…ba/mnnllm/android/chat/input/AttachmentPickerModule.kt	328	val thumbnail = com.alibaba.mnnllm.android.utils.VideoThumbnailUtils.generateVideoThumbnail(	CODE
CRITICAL	…baba/mnnllm/android/chat/input/VoiceRecordingModule.kt	73	val isAudioModel = com.alibaba.mnnllm.android.model.ModelTypeUtils.isAudioModel(activity.modelId!!)	CODE
CRITICAL	…libaba/mnnllm/android/chat/voice/VoiceChatPresenter.kt	212	val userData = com.alibaba.mnnllm.android.chat.model.ChatDataItem(com.alibaba.mnnllm.android.chat.ch	CODE
CRITICAL	…alibaba/mnnllm/android/chat/voice/VoiceChatFragment.kt	360	com.alibaba.mnnllm.android.utils.ImageUtils.compressImageFile(photoFile)	CODE
CRITICAL	…alibaba/mnnllm/android/chat/voice/VoiceChatFragment.kt	561	if (com.alibaba.mnnllm.android.model.ModelTypeUtils.isVisualModel(modelId)) {	CODE
CRITICAL	…mnnllm/android/chat/voice/VoiceModelMarketViewModel.kt	232	com.alibaba.mnnllm.android.mainsettings.MainSettings.setDefaultTtsModel(context, modelId)	CODE
CRITICAL	…mnnllm/android/chat/voice/VoiceModelMarketViewModel.kt	238	com.alibaba.mnnllm.android.mainsettings.MainSettings.setDefaultAsrModel(context, modelId)	CODE
CRITICAL	…ba/mnnllm/android/mainsettings/MainSettingsFragment.kt	143	val sheet = com.alibaba.mnnllm.android.chat.voice.VoiceModelMarketBottomSheet.newInstance()	CODE
CRITICAL	…/alibaba/mnnllm/android/modelmarket/ModelRepository.kt	23	com.alibaba.mnnllm.android.MnnLlmApplication.getAppContext()	CODE
CRITICAL⚡	…/alibaba/mnnllm/android/modelmarket/ModelRepository.kt	72	return com.alibaba.mnnllm.android.utils.PreferenceUtils.getBoolean(context, KEY_ALLOW_NETWORK_MARKET_DATA, true)	CODE
CRITICAL⚡	…/alibaba/mnnllm/android/modelmarket/ModelRepository.kt	76	return com.alibaba.mnnllm.android.utils.PreferenceUtils.getBoolean(context, "debug_enable_network_delay", false)	CODE
CRITICAL⚡	…/alibaba/mnnllm/android/modelmarket/ModelRepository.kt	80	val env = com.alibaba.mnnllm.android.utils.PreferenceUtils.getString(context, KEY_MARKET_ENV, "prod")	CODE
CRITICAL⚡	…/alibaba/mnnllm/android/modelmarket/ModelRepository.kt	87	com.alibaba.mnnllm.android.utils.PreferenceUtils.setString(context, KEY_MARKET_ENV, normalized)	CODE
CRITICAL	…n/java/com/alibaba/mnnllm/android/main/MainActivity.kt	554	com.google.android.material.dialog.MaterialAlertDialogBuilder(this)	CODE
CRITICAL	…ava/com/alibaba/mnnllm/android/widgets/BottomTabBar.kt	20	) : com.google.android.material.card.MaterialCardView(context, attrs, defStyle) {	CODE
CRITICAL	…ibaba/mnnllm/api/openai/network/application/Routing.kt	71	val context = com.alibaba.mnnllm.android.MnnLlmApplication.getAppContext()	CODE
CRITICAL	…aba/mnnllm/api/openai/service/ApiServiceCoordinator.kt	84	val serverEventManager = com.alibaba.mnnllm.api.openai.manager.ServerEventManager.getInstance()	CODE
CRITICAL	…aba/mnnllm/api/openai/service/ApiServiceCoordinator.kt	159	com.alibaba.mnnllm.api.openai.manager.ServerEventManager.getInstance().resetRuntimeState()	CODE
CRITICAL	…aba/mnnllm/api/openai/service/ApiServiceCoordinator.kt	214	com.alibaba.mnnllm.api.openai.manager.ServerEventManager.getInstance().resetRuntimeState()	CODE
CRITICAL	…nnllm/android/debug/DefaultBenchmarkDebugController.kt	80	context = com.alibaba.mnnllm.android.MnnLlmApplication.getAppContext(),	CODE
CRITICAL	…a/com/alibaba/mls/api/download/ms/MsModelDownloader.kt	141	val marketSize = com.alibaba.mls.api.download.DownloadPersistentData.getMarketSizeTotal(ApplicationP	CODE
CRITICAL	…a/com/alibaba/mls/api/download/ml/MLModelDownloader.kt	156	val marketSize = com.alibaba.mls.api.download.DownloadPersistentData.getMarketSizeTotal(ApplicationP	CODE
CRITICAL	…a/com/alibaba/mls/api/download/hf/HfModelDownloader.kt	195	val marketSize = com.alibaba.mls.api.download.DownloadPersistentData.getMarketSizeTotal(ApplicationP	CODE

Unused Imports276 hits · 260 pts

Severity	File	Line	Context
LOW	benchmark/scripts/bench_pc.py	10	CODE
LOW	benchmark/scripts/bench_pc.py	11	CODE
LOW	benchmark/scripts/bench_pc.py	14	CODE
LOW	benchmark/scripts/bench_pc.py	30	CODE
LOW	benchmark/scripts/tvm/ios_bert.py	6	CODE
LOW	benchmark/scripts/tvm/ios_bert.py	8	CODE
LOW	benchmark/scripts/tvm/ios_bert.py	8	CODE
LOW	benchmark/scripts/tvm/ios_bert.py	9	CODE
LOW	benchmark/scripts/tvm/ios_bert.py	10	CODE
LOW	benchmark/scripts/tvm/ios_bert.py	11	CODE
LOW	benchmark/scripts/tvm/ios_tuning.py	4	CODE
LOW	benchmark/scripts/tvm/ios_tuning.py	9	CODE
LOW	pymnn/test/unit_test.py	1162	CODE
LOW	pymnn/test/playground/main.py	7	CODE
LOW	pymnn/test/playground/cases/model_test.py	7	CODE
LOW	pymnn/test/playground/cases/__init__.py	1	CODE
LOW	pymnn/test/playground/cases/__init__.py	1	CODE
LOW	pymnn/pip_package/setup.py	4	CODE
LOW	pymnn/pip_package/build_wheel.py	4	CODE
LOW	pymnn/pip_package/MNN/__init__.py	4	CODE
LOW	pymnn/pip_package/MNN/__init__.py	6	CODE
LOW	pymnn/pip_package/MNN/__init__.py	7	CODE
LOW	pymnn/pip_package/MNN/__init__.py	8	CODE
LOW	pymnn/pip_package/MNN/__init__.py	9	CODE
LOW	pymnn/pip_package/MNN/__init__.py	10	CODE
LOW	pymnn/pip_package/MNN/__init__.py	11	CODE
LOW	pymnn/pip_package/MNN/__init__.py	12	CODE
LOW	pymnn/pip_package/MNN/tools/mnnconvert.py	4	CODE
LOW	pymnn/pip_package/MNN/tools/mnnconvert.py	7	CODE
LOW	pymnn/pip_package/MNN/tools/mnnconvert.py	9	CODE
LOW	pymnn/pip_package/MNN/tools/__init__.py	1	CODE
LOW	pymnn/pip_package/MNN/tools/__init__.py	1	CODE
LOW	pymnn/pip_package/MNN/tools/__init__.py	1	CODE
LOW	pymnn/pip_package/MNN/tools/__init__.py	2	CODE
LOW	pymnn/pip_package/MNN/tools/mnnquant.py	4	CODE
LOW	pymnn/pip_package/MNN/tools/mnnquant.py	6	CODE
LOW	pymnn/pip_package/MNN/tools/mnn.py	4	CODE
LOW	pymnn/pip_package/MNN/tools/mnn.py	5	CODE
LOW	pymnn/pip_package/MNN/tools/mnn.py	6	CODE
LOW	pymnn/pip_package/MNN/tools/utils/log.py	11	CODE
LOW	pymnn/pip_package/MNN/tools/utils/__init__.py	1	CODE
LOW	pymnn/pip_package/MNN/nn/__init__.py	2	CODE
LOW	pymnn/pip_package/MNN/optim/__init__.py	1	CODE
LOW	pymnn/pip_package/MNN/expr/__init__.py	6	CODE
LOW	pymnn/pip_package/MNN/numpy/__init__.py	1372	CODE
LOW	pymnn/pip_package/MNN/numpy/__init__.py	1373	CODE
LOW	pymnn/pip_package/MNN/audio/__init__.py	1	CODE
LOW	pymnn/pip_package/MNN/audio/__init__.py	3	CODE
LOW	pymnn/pip_package/MNN/audio/__init__.py	4	CODE
LOW	pymnn/pip_package/MNN/audio/__init__.py	5	CODE
LOW	pymnn/pip_package/MNN/cv/__init__.py	1	CODE
LOW	pymnn/pip_package/MNN/data/__init__.py	1	CODE
LOW	pymnn/examples/MNNExpr/mobilenet_demo.py	4	CODE
LOW	pymnn/examples/MNNExpr/gpu_io_interface.py	5	CODE
LOW	pymnn/examples/MNNExpr/gpu_express_demo.py	3	CODE
LOW	pymnn/examples/MNNExpr/mnn_numpy_cv_demo.py	4	CODE
LOW	pymnn/examples/MNNTrain/mnist/train_mnist.py	1	CODE
LOW	pymnn/examples/MNNTrain/mnist/dataset.py	6	CODE
LOW	…ples/MNNTrain/mobilenet_finetune/mobilenet_transfer.py	1	CODE
LOW	pymnn/examples/MNNTrain/simple/solve_equation.py	1	CODE
216 more matches not shown…

Hyper-Verbose Identifiers253 hits · 248 pts

Severity	File	Line	Snippet	Context
LOW⚡	3rd_party/flatbuffers/tests/py_test.py	1179	def test_default_monster_name(self):	CODE
LOW⚡	3rd_party/flatbuffers/tests/py_test.py	1182	def test_default_monster_inventory_item(self):	CODE
LOW⚡	3rd_party/flatbuffers/tests/py_test.py	1185	def test_default_monster_inventory_length(self):	CODE
LOW⚡	3rd_party/flatbuffers/tests/py_test.py	1188	def test_default_monster_color(self):	CODE
LOW⚡	3rd_party/flatbuffers/tests/py_test.py	1191	def test_nondefault_monster_color(self):	CODE
LOW⚡	3rd_party/flatbuffers/tests/py_test.py	1203	def test_default_monster_testtype(self):	CODE
LOW⚡	3rd_party/flatbuffers/tests/py_test.py	1206	def test_default_monster_test_field(self):	CODE
LOW⚡	3rd_party/flatbuffers/tests/py_test.py	1215	def test_default_monster_testarrayofstring(self):	CODE
LOW⚡	3rd_party/flatbuffers/tests/py_test.py	1218	def test_default_monster_testarrayofstring_length(self):	CODE
LOW⚡	3rd_party/flatbuffers/tests/py_test.py	1221	def test_default_monster_testarrayoftables(self):	CODE
LOW⚡	3rd_party/flatbuffers/tests/py_test.py	1224	def test_nondefault_monster_testarrayoftables(self):	CODE
LOW⚡	3rd_party/flatbuffers/tests/py_test.py	1271	def test_default_monster_testnestedflatbuffer(self):	CODE
LOW⚡	3rd_party/flatbuffers/tests/py_test.py	1274	def test_default_monster_testnestedflatbuffer_length(self):	CODE
LOW⚡	3rd_party/flatbuffers/tests/py_test.py	1277	def test_nondefault_monster_testnestedflatbuffer(self):	CODE
LOW⚡	3rd_party/flatbuffers/tests/py_test.py	1482	def test_object_is_nested_error(self):	CODE
LOW⚡	3rd_party/flatbuffers/tests/py_test.py	1488	def test_object_is_not_nested_error(self):	CODE
LOW⚡	3rd_party/flatbuffers/tests/py_test.py	1493	def test_struct_is_not_inline_error(self):	CODE
LOW⚡	3rd_party/flatbuffers/tests/py_test.py	1504	def test_create_string_is_nested_error(self):	CODE
LOW⚡	3rd_party/flatbuffers/tests/py_test.py	1511	def test_create_byte_vector_is_nested_error(self):	CODE
LOW⚡	3rd_party/flatbuffers/tests/py_test.py	1518	def test_finished_bytes_error(self):	CODE
LOW	3rd_party/flatbuffers/tests/py_test.py	458	def test_create_arbitrary_string(self):	CODE
LOW	3rd_party/flatbuffers/tests/py_test.py	662	def test_create_numpy_vector_bool(self):	CODE
LOW	3rd_party/flatbuffers/tests/py_test.py	695	def test_create_numpy_vector_reject_strings(self):	CODE
LOW	3rd_party/flatbuffers/tests/py_test.py	718	def test_create_numpy_vector_reject_object(self):	CODE
LOW	3rd_party/flatbuffers/tests/py_test.py	748	def test_vtable_with_one_true_bool(self):	CODE
LOW	3rd_party/flatbuffers/tests/py_test.py	764	def test_vtable_with_one_default_bool(self):	CODE
LOW	3rd_party/flatbuffers/tests/py_test.py	825	def test_vtable_with_empty_vector(self):	CODE
LOW	3rd_party/flatbuffers/tests/py_test.py	841	def test_vtable_with_empty_vector_of_byte_and_some_scalars(self):	CODE
LOW	3rd_party/flatbuffers/tests/py_test.py	934	def test_table_with_some_elements(self):	CODE
LOW	3rd_party/flatbuffers/tests/py_test.py	957	def test__one_unfinished_table_and_one_finished_table(self):	CODE
LOW	3rd_party/flatbuffers/tests/py_test.py	1081	def make_monster_from_generated_code(sizePrefix = False):	CODE
LOW	3rd_party/flatbuffers/tests/py_test.py	1165	def test_nondefault_monster_mana(self):	CODE
LOW	3rd_party/flatbuffers/tests/py_test.py	1248	def test_default_monster_testarrayoftables_length(self):	CODE
LOW	3rd_party/flatbuffers/tests/py_test.py	1251	def test_nondefault_monster_enemy(self):	CODE
LOW	3rd_party/flatbuffers/tests/py_test.py	1309	def test_nondefault_monster_testempty(self):	CODE
LOW	3rd_party/flatbuffers/tests/py_test.py	1329	def test_default_monster_testbool(self):	CODE
LOW	3rd_party/flatbuffers/tests/py_test.py	1332	def test_nondefault_monster_testbool(self):	CODE
LOW	3rd_party/flatbuffers/tests/py_test.py	1344	def test_default_monster_testhashes(self):	CODE
LOW	3rd_party/flatbuffers/tests/py_test.py	1354	def test_nondefault_monster_testhashes(self):	CODE
LOW	3rd_party/flatbuffers/tests/py_test.py	1380	def test_getrootas_for_nonroot_table(self):	CODE
LOW	3rd_party/flatbuffers/tests/py_test.py	1401	def test_vtable_deduplication(self):	CODE
LOW	3rd_party/flatbuffers/tests/py_test.py	1655	def backward_compatible_run_tests(**kwargs):	CODE
LOW	3rd_party/flatbuffers/tests/union_vector/Movie.php	153	public static function createCharactersTypeVector(FlatBufferBuilder $builder, array $data)	CODE
LOW	3rd_party/flatbuffers/tests/union_vector/Movie.php	167	public static function startCharactersTypeVector(FlatBufferBuilder $builder, $numElems)	CODE
LOW	3rd_party/flatbuffers/tests/union_vector/Attacker.php	26	public static function AttackerBufferHasIdentifier(ByteBuffer $buf)	CODE
LOW⚡	…d_party/flatbuffers/tests/MyGame/InParentNamespace.php	17	public static function getRootAsInParentNamespace(ByteBuffer $bb)	CODE
LOW⚡	…d_party/flatbuffers/tests/MyGame/InParentNamespace.php	23	public static function InParentNamespaceIdentifier()	CODE
LOW⚡	…d_party/flatbuffers/tests/MyGame/InParentNamespace.php	28	public static function InParentNamespaceBufferHasIdentifier(ByteBuffer $buf)	CODE
LOW⚡	…d_party/flatbuffers/tests/MyGame/InParentNamespace.php	33	public static function InParentNamespaceExtension()	CODE
LOW	3rd_party/flatbuffers/tests/MyGame/Example2/Monster.php	28	public static function MonsterBufferHasIdentifier(ByteBuffer $buf)	CODE
LOW	…party/flatbuffers/tests/MyGame/Example/TypeAliases.php	28	public static function TypeAliasesBufferHasIdentifier(ByteBuffer $buf)	CODE
LOW⚡	…ffers/tests/MyGame/Example/TestSimpleTableWithEnum.php	17	public static function getRootAsTestSimpleTableWithEnum(ByteBuffer $bb)	CODE
LOW⚡	…ffers/tests/MyGame/Example/TestSimpleTableWithEnum.php	23	public static function TestSimpleTableWithEnumIdentifier()	CODE
LOW⚡	…ffers/tests/MyGame/Example/TestSimpleTableWithEnum.php	28	public static function TestSimpleTableWithEnumBufferHasIdentifier(ByteBuffer $buf)	CODE
LOW⚡	…ffers/tests/MyGame/Example/TestSimpleTableWithEnum.php	33	public static function TestSimpleTableWithEnumExtension()	CODE
LOW	…ffers/tests/MyGame/Example/TestSimpleTableWithEnum.php	63	public static function startTestSimpleTableWithEnum(FlatBufferBuilder $builder)	CODE
LOW	…ffers/tests/MyGame/Example/TestSimpleTableWithEnum.php	72	public static function createTestSimpleTableWithEnum(FlatBufferBuilder $builder, $color)	CODE
LOW	…ffers/tests/MyGame/Example/TestSimpleTableWithEnum.php	94	public static function endTestSimpleTableWithEnum(FlatBufferBuilder $builder)	CODE
LOW	…_party/flatbuffers/tests/MyGame/Example/Referrable.php	28	public static function ReferrableBufferHasIdentifier(ByteBuffer $buf)	CODE
LOW⚡	3rd_party/flatbuffers/tests/MyGame/Example/Monster.php	524	public function getVectorOfWeakReferences($j)	CODE
193 more matches not shown…

Structural Annotation Overuse107 hits · 187 pts

Severity	File	Line	Snippet	Context
LOW⚡	benchmark/scripts/tvm/tvmc_cuda_test.py	14	model = tvmc.load("/home/mnnteam/benchmark/models/onnx/" + model_name + ".onnx") #Step 1: Load	CODE
LOW⚡	benchmark/scripts/tvm/tvmc_cuda_test.py	24	tvmc.compile(model, target=device_name, tuning_records=log_file, package_path=compile_file) #Step 2: Compile	CODE
LOW⚡	benchmark/scripts/tvm/tvmc_cuda_test.py	31	result = tvmc.run(new_package, device=device_name, repeat=100) #Step 3: Run	CODE
LOW⚡	tools/script/generate_podspec.sh	329	# Step 1: Check file exists	COMMENT
LOW⚡	tools/script/generate_podspec.sh	334	# Step 2: Validate syntax (only if file exists)	COMMENT
LOW⚡	tools/script/generate_podspec.sh	341	# Step 3: Validate file references	COMMENT
LOW⚡	tools/script/generate_podspec.sh	346	# Step 4: Full validation (if requested)	COMMENT
LOW⚡	test/op/LinearAttentionTest.cpp	162	// Step 4: Scale Q	COMMENT
LOW⚡	test/op/LinearAttentionTest.cpp	168	// Step 5: Gated Delta Rule with persistent state	COMMENT
LOW⚡	test/op/LinearAttentionTest.cpp	1161	// Step 2: simulate the meta updates Llm performs around eraseHistory.	COMMENT
LOW⚡	test/op/LinearAttentionTest.cpp	1167	// Step 3: prefill Y. onResize sees remove>0 -> isExplicitRollback;	COMMENT
LOW	test/op/LinearAttentionTest.cpp	66	// Step 1: Build conv_input = cat(convState, qkv) along dim L	COMMENT
LOW	test/op/LinearAttentionTest.cpp	107	// Step 2: Split Q, K, V with GQA expansion	COMMENT
LOW	test/op/LinearAttentionTest.cpp	144	// Step 3: L2 Norm	COMMENT
LOW	test/op/LinearAttentionTest.cpp	1143	// Step 1: prefill X. metaA.previous=0 here makes onResize treat this as	COMMENT
LOW	source/backend/metal/MetalKVCacheManager.mm	359	// Step 3: Move the kvcache from temporary buffers in memory to disk	COMMENT
LOW	source/backend/metal/MetalLinearAttentionShader.hpp	601	// Step 1: Decay state + compute v_pred	COMMENT
LOW⚡	source/backend/metal/MetalLinearAttentionShader.hpp	613	// Step 2: Compute delta	COMMENT
LOW⚡	source/backend/metal/MetalLinearAttentionShader.hpp	616	// Step 3: Update state + compute output	COMMENT
LOW	source/backend/cpu/CPUKVCacheManager.cpp	197	// Step 1: Copy the old kvcache from files to temporary buffers in memory	COMMENT
LOW⚡	source/backend/cpu/CPUKVCacheManager.cpp	220	// Step 2: Resize the kvcache files and remap them	COMMENT
LOW⚡	source/backend/cpu/CPUKVCacheManager.cpp	224	// Step 3: Move the kvcache from temporary buffers in memory to disk	COMMENT
LOW	source/backend/cpu/CPUKVCacheManager.cpp	284	// Step 4: Release the temporary buffers	COMMENT
LOW	source/backend/cpu/CPULinearAttention.cpp	349	// Step 1: Depthwise Conv1D + SiLU	COMMENT
LOW	source/backend/cpu/CPULinearAttention.cpp	389	// Step 2: Split Q, K, V	COMMENT
LOW	source/backend/cpu/CPULinearAttention.cpp	426	// Step 3: Optional L2 Normalization	COMMENT
LOW⚡	source/backend/cpu/CPULinearAttention.cpp	448	// Step 4: Scale Q	COMMENT
LOW⚡	source/backend/cpu/CPULinearAttention.cpp	453	// Step 5: Gated Delta Rule	COMMENT
LOW	source/backend/cpu/CPULinearAttention.cpp	1171	// Step 2: y = C_ * conv_out, transpose to output [B, L, 1, H]	COMMENT
LOW⚡	source/backend/cpu/compute/TurboQuant.hpp	84	// Step 1: Apply sign flips	COMMENT
LOW⚡	source/backend/cpu/compute/TurboQuant.hpp	88	// Step 2: Butterfly stages (log2(32) = 5 stages)	COMMENT
LOW⚡	source/backend/cpu/compute/TurboQuant.hpp	99	// Step 3: Normalize	COMMENT
LOW⚡	source/backend/cpu/compute/TurboQuant.hpp	109	// Step 1: Copy input	COMMENT
LOW⚡	source/backend/cpu/compute/TurboQuant.hpp	113	// Step 2: Butterfly stages (WHT is self-inverse up to scaling)	COMMENT
LOW⚡	source/backend/cpu/compute/TurboQuant.hpp	178	// Step 2: Normalize	COMMENT
LOW⚡	source/backend/cpu/compute/TurboQuant.hpp	185	// Step 3: Apply WHT forward	COMMENT
LOW⚡	source/backend/cpu/compute/TurboQuant.hpp	189	// Step 4: Find nearest codebook index for each coordinate	COMMENT
LOW⚡	source/backend/cpu/compute/TurboQuant.hpp	195	// Step 5: Pack 3-bit indices (4 groups of 8 → 12 bytes)	COMMENT
LOW⚡	source/backend/cpu/compute/TurboQuant.hpp	203	// Step 1: Read scale	COMMENT
LOW⚡	source/backend/cpu/compute/TurboQuant.hpp	208	// Step 2: Unpack 3-bit indices and look up centroids	COMMENT
LOW⚡	source/backend/cpu/compute/TurboQuant.hpp	218	// Step 3: Apply inverse WHT	COMMENT
LOW⚡	source/backend/cpu/compute/TurboQuant.hpp	222	// Step 4: Scale back	COMMENT
LOW⚡	source/backend/cpu/compute/TurboQuant.hpp	233	// Step 1: Read scale	COMMENT
LOW⚡	source/backend/cpu/compute/TurboQuant.hpp	238	// Step 2: Unpack 3-bit indices and look up centroids	COMMENT
LOW⚡	source/backend/cpu/compute/TurboQuant.hpp	248	// Step 3: Apply inverse WHT	COMMENT
LOW⚡	source/backend/cpu/compute/TurboQuant.hpp	252	// Step 4: Scale and write to strided destination	COMMENT
LOW	source/backend/cpu/compute/TurboQuant.hpp	124	// Step 3: Normalize and undo sign flips	COMMENT
LOW	source/backend/cpu/compute/TurboQuant.hpp	164	// Step 1: Compute RMS scale	COMMENT
LOW⚡	source/backend/cpu/arm/CommonOptFunctionNeon.cpp	584	// Step 1: Transpose 2x2 blocks of 2-element vectors	COMMENT
LOW⚡	source/backend/cpu/arm/CommonOptFunctionNeon.cpp	588	// Step 2: Combine the results to get the full transpose	COMMENT
LOW	…rce/backend/hexagon/htp-ops-lib/include/dsp/hvx_math.h	8	// NOTE: This function is intended to use in safe softmax (x <= 0, overflow handing is omitted)	COMMENT
LOW	docs/tools/compress.md	203	#### Step 1: 先转为 float MNN	COMMENT
LOW	docs/tools/compress.md	209	#### Step 2: 准备数据集，编写量化json文件	COMMENT
LOW	docs/tools/compress.md	308	#### Step 3: 使用 `quantized.out` 或 `mnnquant` 进行离线量化	COMMENT
LOW	skills/general-debug/SKILL.md	87	#### Step 1: 复现并最小化	COMMENT
LOW	skills/general-debug/SKILL.md	92	#### Step 2: 用地址等式定位别名（关键手法）	COMMENT
LOW	skills/general-debug/SKILL.md	107	#### Step 3: 从 `onResize` 找根因	COMMENT
LOW⚡	skills/general-debug/SKILL.md	126	#### Step 4: 跨 op tensor 别名（fusion / 图重排场景）	COMMENT
LOW⚡	skills/general-debug/SKILL.md	134	#### Step 5: 越界写 / 未初始化	COMMENT
LOW⚡	skills/general-debug/SKILL.md	142	#### Step 6: 生命周期错位（栈上临时 buffer）	COMMENT
47 more matches not shown…

Deep Nesting185 hits · 184 pts

Severity	File	Line	Context
LOW	benchmark/scripts/tvm/ios_bert.py	119	CODE
LOW	benchmark/scripts/tvm/ios_tuning.py	72	CODE
LOW	benchmark/scripts/tvm/android_tuning.py	50	CODE
LOW	3rd_party/flatbuffers/tests/py_test.py	236	CODE
LOW	pymnn/test/model_test.py	10	CODE
LOW	pymnn/test/model_test.py	50	CODE
LOW	pymnn/test/model_test.py	71	CODE
LOW	pymnn/test/benchmark.py	47	CODE
LOW	pymnn/test/unit_test.py	384	CODE
LOW	pymnn/test/unit_test.py	666	CODE
LOW	pymnn/pip_package/build_deps.py	74	CODE
LOW	pymnn/pip_package/MNN/tools/mnnconvert.py	19	CODE
LOW	pymnn/pip_package/MNN/expr/__init__.py	81	CODE
LOW	pymnn/pip_package/MNN/numpy/linalg/__init__.py	4	CODE
LOW	tools/converter/tools/test4TF.py	56	CODE
LOW	tools/converter/tools/auto_quant.py	17	CODE
LOW	tools/script/analyzeQuantErrors.py	81	CODE
LOW	tools/script/arm_assembly.py	11	CODE
LOW	tools/script/register.py	3	CODE
LOW	tools/script/register.py	93	CODE
LOW	tools/script/register.py	155	CODE
LOW	tools/script/register.py	9	CODE
LOW	tools/script/register.py	100	CODE
LOW	tools/script/register.py	163	CODE
LOW	tools/script/prue_mnn_ops.py	34	CODE
LOW	tools/script/testMNNFromTf.py	46	CODE
LOW	tools/script/arm2binary.py	70	CODE
LOW	tools/script/arm2binary.py	82	CODE
LOW	tools/script/arm2binary.py	194	CODE
LOW	tools/script/turnMNN2Expr.py	66	CODE
LOW	tools/script/testMNNFromOnnx.py	129	CODE
LOW	tools/script/gen_unicode_data.py	72	CODE
LOW	tools/script/opencl_kernel_check.py	64	CODE
LOW	tools/script/opencl_kernel_check.py	135	CODE
LOW	tools/mnncompress/mnncompress/common/helper.py	4	CODE
LOW	…ompress/mnncompress/tensorflow/sensitivity_analyzer.py	43	CODE
LOW	…ls/mnncompress/mnncompress/tensorflow/LSQ_quantizer.py	200	CODE
LOW	…ls/mnncompress/mnncompress/tensorflow/LSQ_quantizer.py	498	CODE
LOW	…ls/mnncompress/mnncompress/tensorflow/LSQ_quantizer.py	634	CODE
LOW	…ls/mnncompress/mnncompress/tensorflow/EMA_quantizer.py	200	CODE
LOW	…ls/mnncompress/mnncompress/tensorflow/EMA_quantizer.py	499	CODE
LOW	…ls/mnncompress/mnncompress/tensorflow/EMA_quantizer.py	633	CODE
LOW	…ress/mnncompress/tensorflow/TaylorFO_channel_pruner.py	145	CODE
LOW	…mnncompress/mnncompress/tensorflow/weight_quantizer.py	112	CODE
LOW	…mnncompress/mnncompress/tensorflow/weight_quantizer.py	152	CODE
LOW	…ress/mnncompress/tensorflow/SA_prune_ratio_searcher.py	108	CODE
LOW	…ress/mnncompress/tensorflow/SA_prune_ratio_searcher.py	262	CODE
LOW	…ls/mnncompress/mnncompress/tensorflow/OAQ_quantizer.py	203	CODE
LOW	…ls/mnncompress/mnncompress/tensorflow/OAQ_quantizer.py	640	CODE
LOW	…ls/mnncompress/mnncompress/tensorflow/OAQ_quantizer.py	746	CODE
LOW	…ls/mnncompress/mnncompress/tensorflow/OAQ_quantizer.py	779	CODE
LOW	…s/mnncompress/mnncompress/tensorflow/SIMD_OC_pruner.py	145	CODE
LOW	…nncompress/mnncompress/tensorflow/SNIP_level_pruner.py	141	CODE
LOW	…ls/mnncompress/mnncompress/tensorflow/decomposition.py	118	CODE
LOW	…ls/mnncompress/mnncompress/tensorflow/decomposition.py	215	CODE
LOW	tools/mnncompress/mnncompress/tensorflow/admm_pruner.py	78	CODE
LOW	tools/mnncompress/mnncompress/tensorflow/admm_pruner.py	95	CODE
LOW	tools/mnncompress/mnncompress/tensorflow/admm_pruner.py	161	CODE
LOW	tools/mnncompress/mnncompress/tensorflow/helpers.py	367	CODE
LOW	tools/mnncompress/mnncompress/pytorch/LSQ_quantizer.py	190	CODE
125 more matches not shown…

Excessive Try-Catch Wrapping115 hits · 176 pts

Severity	File	Line	Snippet	Context
LOW	3rd_party/flatbuffers/tests/py_test.py	51	except Exception as e:	CODE
LOW	pymnn/pip_package/MNN/expr/__init__.py	13	except Exception:	CODE
LOW	tools/script/arm_assembly.py	223	except Exception as e:	CODE
LOW	tools/script/arm_assembly.py	266	except Exception as e:	CODE
LOW	tools/script/arm_assembly.py	303	except Exception as e:	CODE
MEDIUM	tools/script/testMNNFromOnnx.py	260	print('Error is between ', last_right_node, ' and ', first_error_node)	CODE
MEDIUM	tools/script/genQNNModelsFromMNN.py	27	print(f"Error generating for socId: {socId}, hexagonArch: {hexagonArch}")	CODE
MEDIUM	tools/script/genQNNModelsFromMNN.py	30	print(f"Error output:\n{e.stderr}")	CODE
LOW	source/backend/qnn/npu_convert.py	40	except Exception as e:	CODE
LOW	source/backend/qnn/npu_convert.py	158	except Exception as e:	CODE
LOW	source/backend/qnn/npu_convert.py	212	except Exception as e:	CODE
MEDIUM	source/backend/qnn/npu_convert.py	29	def raise_stack_limit():	CODE
MEDIUM	transformers/llm/export/gguf2mnn.py	386	print("Error: Can't find weight for " + name)	CODE
MEDIUM	transformers/llm/export/gguf2mnn.py	419	print('Error: Bias is quant: ', tensormap[bias_name].tensor_type)	CODE
LOW	transformers/llm/export/llmexport.py	706	except Exception as e:	CODE
LOW	transformers/llm/export/llmexport.py	832	except Exception as e:	CODE
LOW	transformers/llm/export/npu/generate_llm_qnn.py	305	except Exception as e:	CODE
MEDIUM	transformers/llm/export/npu/generate_llm_qnn.py	306	print(f"Error parsing --image_sizes: {e}")	CODE
MEDIUM	transformers/llm/export/npu/generate_llm_qnn.py	331	print(f"Error: 'hidden_size' key not found in {config_file_path}")	CODE
MEDIUM	transformers/llm/export/utils/smooth_quantizer.py	520	print("Error: CPU only supports 8 bit feature map quantized")	CODE
LOW	transformers/llm/export/utils/model.py	135	except Exception:	CODE
LOW	transformers/llm/export/utils/tokenizer.py	47	except Exception:	CODE
LOW	transformers/llm/export/utils/tokenizer.py	618	except Exception:	CODE
LOW	transformers/llm/export/utils/tokenizer.py	632	except Exception:	CODE
LOW	transformers/llm/export/utils/tokenizer.py	768	except Exception as e:	CODE
LOW⚡	transformers/llm/export/utils/tokenizer.py	796	except Exception as e:	CODE
LOW⚡	transformers/llm/export/utils/tokenizer.py	799	except Exception as e2:	CODE
LOW⚡	transformers/llm/export/utils/tokenizer.py	805	except Exception as e2:	CODE
MEDIUM	transformers/llm/export/utils/tokenizer.py	6	def __init__(self, tokenizer_path, model_type, **kwargs):	CODE
LOW	transformers/llm/export/utils/mnn_converter.py	43	except Exception:	CODE
LOW	transformers/llm/export/utils/spinner.py	20	except Exception as e:	CODE
LOW	transformers/llm/export/utils/torch_utils.py	108	except Exception:	CODE
MEDIUM	transformers/llm/engine/tools/bench_ttfa.py	152	print(f"Error: {path} not found", file=sys.stderr)	CODE
LOW	transformers/diffusion/export/convert_mnn_sd35.py	30	except Exception as e:	CODE
LOW	transformers/diffusion/export/convert_mnn_sd35.py	73	except Exception as e:	CODE
LOW	transformers/diffusion/export/convert_mnn_sd35.py	83	except Exception:	CODE
LOW	transformers/diffusion/export/convert_mnn.py	51	except Exception:	CODE
LOW	transformers/diffusion/export/convert_mnn.py	67	except Exception:	CODE
LOW	transformers/diffusion/export/convert_mnn.py	102	except Exception:	CODE
MEDIUM	transformers/diffusion/export/wan/wan_onnx_export.py	121	def load_diffusers_pipeline(model_path, dtype, device, errors):	CODE
MEDIUM	transformers/diffusion/export/wan/wan_onnx_export.py	341	def load_official_wan(model_path, dtype, device, errors):	CODE
LOW⚡	transformers/diffusion/export/wan/wan_onnx_export.py	346	except Exception:	CODE
LOW⚡	transformers/diffusion/export/wan/wan_onnx_export.py	350	except Exception as e:	CODE
LOW⚡	transformers/diffusion/export/wan/wan_onnx_export.py	357	except Exception as e:	CODE
LOW⚡	transformers/diffusion/export/wan/wan_onnx_export.py	366	except Exception:	CODE
LOW	transformers/diffusion/export/wan/wan_onnx_export.py	124	except Exception as e:	CODE
LOW	transformers/diffusion/export/wan/wan_onnx_export.py	159	except Exception as e:	CODE
LOW	transformers/diffusion/export/wan/wan_onnx_export.py	251	except Exception as e:	CODE
LOW	transformers/diffusion/export/wan/wan_onnx_export.py	412	except Exception as e:	CODE
LOW	transformers/diffusion/export/wan/wan_onnx_export.py	459	except Exception:	CODE
LOW	transformers/diffusion/export/wan/wan_onnx_export.py	476	except Exception as e:	CODE
LOW	transformers/diffusion/export/wan/wan_onnx_export.py	512	except Exception:	CODE
LOW	transformers/diffusion/export/wan/wan_onnx_export.py	514	except Exception:	CODE
LOW	transformers/diffusion/export/wan/wan_onnx_export.py	576	except Exception as e:	CODE
LOW	transformers/diffusion/export/wan/wan_onnx_export.py	667	except Exception as e:	CODE
LOW	transformers/diffusion/export/wan/wan_onnx_export.py	721	except Exception as e:	CODE
LOW	transformers/diffusion/export/wan/wan_convert_mnn.py	177	except Exception:	CODE
LOW	transformers/diffusion/export/wan/wan_convert_mnn.py	200	except Exception as e:	CODE
LOW	transformers/diffusion/export/wan/wan_alignment.py	383	except Exception as exc:	CODE
LOW	transformers/diffusion/export/wan/wan_alignment.py	448	except Exception as exc:	CODE
55 more matches not shown…

Verbosity Indicators87 hits · 156 pts

Severity	File	Line	Snippet	Context
LOW⚡	benchmark/scripts/tvm/tvmc_cuda_test.py	14	model = tvmc.load("/home/mnnteam/benchmark/models/onnx/" + model_name + ".onnx") #Step 1: Load	CODE
LOW⚡	benchmark/scripts/tvm/tvmc_cuda_test.py	24	tvmc.compile(model, target=device_name, tuning_records=log_file, package_path=compile_file) #Step 2: Compile	CODE
LOW⚡	benchmark/scripts/tvm/tvmc_cuda_test.py	31	result = tvmc.run(new_package, device=device_name, repeat=100) #Step 3: Run	CODE
LOW⚡	tools/script/generate_podspec.sh	329	# Step 1: Check file exists	COMMENT
LOW⚡	tools/script/generate_podspec.sh	334	# Step 2: Validate syntax (only if file exists)	COMMENT
LOW⚡	tools/script/generate_podspec.sh	341	# Step 3: Validate file references	COMMENT
LOW⚡	tools/script/generate_podspec.sh	346	# Step 4: Full validation (if requested)	COMMENT
LOW⚡	test/op/LinearAttentionTest.cpp	162	// Step 4: Scale Q	COMMENT
LOW⚡	test/op/LinearAttentionTest.cpp	168	// Step 5: Gated Delta Rule with persistent state	COMMENT
LOW⚡	test/op/LinearAttentionTest.cpp	1161	// Step 2: simulate the meta updates Llm performs around eraseHistory.	COMMENT
LOW⚡	test/op/LinearAttentionTest.cpp	1167	// Step 3: prefill Y. onResize sees remove>0 -> isExplicitRollback;	COMMENT
LOW	test/op/LinearAttentionTest.cpp	66	// Step 1: Build conv_input = cat(convState, qkv) along dim L	COMMENT
LOW	test/op/LinearAttentionTest.cpp	107	// Step 2: Split Q, K, V with GQA expansion	COMMENT
LOW	test/op/LinearAttentionTest.cpp	144	// Step 3: L2 Norm	COMMENT
LOW	test/op/LinearAttentionTest.cpp	1143	// Step 1: prefill X. metaA.previous=0 here makes onResize treat this as	COMMENT
LOW	source/backend/metal/MetalKVCacheManager.mm	359	// Step 3: Move the kvcache from temporary buffers in memory to disk	COMMENT
LOW	source/backend/metal/MetalLinearAttentionShader.hpp	601	// Step 1: Decay state + compute v_pred	COMMENT
LOW⚡	source/backend/metal/MetalLinearAttentionShader.hpp	613	// Step 2: Compute delta	COMMENT
LOW⚡	source/backend/metal/MetalLinearAttentionShader.hpp	616	// Step 3: Update state + compute output	COMMENT
LOW	source/backend/cpu/CPUKVCacheManager.cpp	197	// Step 1: Copy the old kvcache from files to temporary buffers in memory	COMMENT
LOW⚡	source/backend/cpu/CPUKVCacheManager.cpp	220	// Step 2: Resize the kvcache files and remap them	COMMENT
LOW⚡	source/backend/cpu/CPUKVCacheManager.cpp	224	// Step 3: Move the kvcache from temporary buffers in memory to disk	COMMENT
LOW	source/backend/cpu/CPUKVCacheManager.cpp	284	// Step 4: Release the temporary buffers	COMMENT
LOW	source/backend/cpu/CPULinearAttention.cpp	349	// Step 1: Depthwise Conv1D + SiLU	COMMENT
LOW	source/backend/cpu/CPULinearAttention.cpp	389	// Step 2: Split Q, K, V	COMMENT
LOW	source/backend/cpu/CPULinearAttention.cpp	426	// Step 3: Optional L2 Normalization	COMMENT
LOW⚡	source/backend/cpu/CPULinearAttention.cpp	448	// Step 4: Scale Q	COMMENT
LOW⚡	source/backend/cpu/CPULinearAttention.cpp	453	// Step 5: Gated Delta Rule	COMMENT
LOW	source/backend/cpu/CPULinearAttention.cpp	1171	// Step 2: y = C_ * conv_out, transpose to output [B, L, 1, H]	COMMENT
LOW	source/backend/cpu/compute/ConvInt8TiledExecutor.cpp	288	if (calOcMain <= ocMain) { // The purpose of this function is to increase the value of ocMain.	CODE
LOW⚡	source/backend/cpu/compute/TurboQuant.hpp	84	// Step 1: Apply sign flips	COMMENT
LOW⚡	source/backend/cpu/compute/TurboQuant.hpp	88	// Step 2: Butterfly stages (log2(32) = 5 stages)	COMMENT
LOW⚡	source/backend/cpu/compute/TurboQuant.hpp	99	// Step 3: Normalize	COMMENT
LOW⚡	source/backend/cpu/compute/TurboQuant.hpp	109	// Step 1: Copy input	COMMENT
LOW⚡	source/backend/cpu/compute/TurboQuant.hpp	113	// Step 2: Butterfly stages (WHT is self-inverse up to scaling)	COMMENT
LOW⚡	source/backend/cpu/compute/TurboQuant.hpp	178	// Step 2: Normalize	COMMENT
LOW⚡	source/backend/cpu/compute/TurboQuant.hpp	185	// Step 3: Apply WHT forward	COMMENT
LOW⚡	source/backend/cpu/compute/TurboQuant.hpp	189	// Step 4: Find nearest codebook index for each coordinate	COMMENT
LOW⚡	source/backend/cpu/compute/TurboQuant.hpp	195	// Step 5: Pack 3-bit indices (4 groups of 8 → 12 bytes)	COMMENT
LOW⚡	source/backend/cpu/compute/TurboQuant.hpp	203	// Step 1: Read scale	COMMENT
LOW⚡	source/backend/cpu/compute/TurboQuant.hpp	208	// Step 2: Unpack 3-bit indices and look up centroids	COMMENT
LOW⚡	source/backend/cpu/compute/TurboQuant.hpp	218	// Step 3: Apply inverse WHT	COMMENT
LOW⚡	source/backend/cpu/compute/TurboQuant.hpp	222	// Step 4: Scale back	COMMENT
LOW⚡	source/backend/cpu/compute/TurboQuant.hpp	233	// Step 1: Read scale	COMMENT
LOW⚡	source/backend/cpu/compute/TurboQuant.hpp	238	// Step 2: Unpack 3-bit indices and look up centroids	COMMENT
LOW⚡	source/backend/cpu/compute/TurboQuant.hpp	248	// Step 3: Apply inverse WHT	COMMENT
LOW⚡	source/backend/cpu/compute/TurboQuant.hpp	252	// Step 4: Scale and write to strided destination	COMMENT
LOW	source/backend/cpu/compute/TurboQuant.hpp	124	// Step 3: Normalize and undo sign flips	COMMENT
LOW	source/backend/cpu/compute/TurboQuant.hpp	164	// Step 1: Compute RMS scale	COMMENT
LOW⚡	source/backend/cpu/arm/CommonOptFunctionNeon.cpp	584	// Step 1: Transpose 2x2 blocks of 2-element vectors	COMMENT
LOW⚡	source/backend/cpu/arm/CommonOptFunctionNeon.cpp	588	// Step 2: Combine the results to get the full transpose	COMMENT
LOW	…Chat/MNNLLMiOS/InferenceEngine/SanaDiffusionSession.mm	75	// Step 1: Load LLM on background thread	COMMENT
LOW	…Chat/MNNLLMiOS/InferenceEngine/SanaDiffusionSession.mm	87	// Step 2: Load Diffusion on background thread	COMMENT
LOW	…alibaba/mnnllm/android/benchmark/LeaderboardService.kt	203	// Step 1: Initiate rank query	COMMENT
LOW	…alibaba/mnnllm/android/benchmark/LeaderboardService.kt	262	// Step 2: Poll for final result	COMMENT
LOW	…in/java/com/alibaba/mnnllm/android/utils/ImageUtils.kt	137	// Step 1: Decode only dimensions to calculate scaling factor without loading full bitmap into memory	COMMENT
LOW⚡	…in/java/com/alibaba/mnnllm/android/utils/ImageUtils.kt	153	// Step 2: Decode the bitmap with calculated scale	COMMENT
LOW⚡	…in/java/com/alibaba/mnnllm/android/utils/ImageUtils.kt	158	// Step 3: Handle orientation using EXIF data	COMMENT
LOW	…in/java/com/alibaba/mnnllm/android/utils/ImageUtils.kt	177	// Step 4: Optimization - Combine precise scaling and rotation into a single operation	COMMENT
LOW⚡	…in/java/com/alibaba/mnnllm/android/utils/ImageUtils.kt	203	// Step 5: Save the compressed bitmap back to the original file	COMMENT
27 more matches not shown…

Redundant / Tautological Comments55 hits · 89 pts

Severity	File	Line	Snippet	Context
LOW	pymnn/pip_package/MNN/llm/__init__.py	39	# Set back to C++ object	COMMENT
LOW⚡	tools/script/generate_podspec.sh	10	# ./tools/script/generate_podspec.sh -c # Check if podspec is in sync	COMMENT
LOW⚡	tools/script/generate_podspec.sh	219	# Check if podspec file exists	COMMENT
LOW	tools/script/generate_podspec.sh	268	# Check if pattern contains glob	COMMENT
LOW	tools/script/generate_podspec.sh	276	# Check if file exists	COMMENT
LOW	tools/script/turnMNN2Expr.py	165	# Return output	COMMENT
LOW	transformers/llm/export/gguf/gguf_reader.py	184	# Check if it's a simple scalar type.	COMMENT
LOW	transformers/llm/export/utils/model.py	105	# Check if skip_weight mode is enabled (load structure only, no weights)	COMMENT
LOW	transformers/llm/export/utils/omni_quantizer.py	479	# Check if we already have prepared data	COMMENT
LOW	apps/Android/MnnLlmChat/scripts/test_build.sh	32	# Check if we're in the right directory	COMMENT
LOW	apps/Android/MnnLlmChat/scripts/setup_release.sh	32	# Check if we're in the right directory	COMMENT
LOW	apps/Android/MnnLlmChat/scripts/setup_release.sh	40	# Check if release.config already exists	COMMENT
LOW	apps/Android/MnnLlmChat/scripts/qnn.sh	66	# Check if directory exists	COMMENT
LOW	apps/Android/MnnLlmChat/scripts/release.sh	139	# Check if we're in the right directory	COMMENT
LOW	apps/Android/MnnLlmChat/scripts/release.sh	152	# Check if Gradle wrapper exists	COMMENT
LOW	apps/Android/MnnLlmChat/scripts/release.sh	281	# Check if fastlane is available	COMMENT
LOW⚡	apps/Android/MnnLlmChat/scripts/push_model_test.sh	56	# Check if we're in the right directory	COMMENT
LOW⚡	apps/Android/MnnLlmChat/scripts/push_model_test.sh	62	# Check if ADB is available	COMMENT
LOW⚡	apps/Android/MnnLlmChat/scripts/push_model_test.sh	68	# Check if modelscope cache directory exists	COMMENT
LOW⚡	apps/Android/MnnLlmChat/scripts/push_model_test.sh	74	# Check if models directory exists (either MNN subdirectory or main models directory)	COMMENT
LOW⚡	apps/Android/MnnLlmChat/scripts/push_model_test.sh	82	# Check if build_64 directory exists	COMMENT
LOW	apps/Android/MnnLlmChat/scripts/push_model_test.sh	94	# Check if any device is connected	COMMENT
LOW	apps/Android/MnnLlmChat/scripts/push_model_test.sh	104	# Check if device is rooted (needed for /data/local/tmp access)	COMMENT
LOW	apps/Android/MnnLlmChat/scripts/push_model_test.sh	275	# Set LD_LIBRARY_PATH to include current directory for .so files	COMMENT
LOW	apps/Android/MnnLlmChat/scripts/push_model_test.sh	297	# Check if libMNN.so is present	COMMENT
LOW	apps/Android/MnnLlmChat/scripts/push_model_test.sh	305	# Check if llm_demo is executable	COMMENT
LOW	apps/Android/MnnLlmChat/scripts/get_repo_info.py	55	# Check if the model has a HuggingFace source	COMMENT
LOW	apps/mnncli/android_debug.sh	92	# Check if adb is available	COMMENT
LOW⚡	apps/mnncli/android_debug.sh	202	# Check if video path is provided	COMMENT
LOW⚡	apps/mnncli/android_debug.sh	209	# Check if specified video file exists on device	COMMENT
LOW⚡	apps/mnncli/android_debug.sh	219	# Check if model path is provided	COMMENT
LOW	apps/mnncli/build.sh	118	# Check if build was successful	COMMENT
LOW	apps/mnncli/android_build.sh	100	# Check if ANDROID_NDK is set	COMMENT
LOW	apps/mnncli/android_build.sh	108	# Check if ANDROID_NDK exists	COMMENT
LOW	apps/mnncli/android_build.sh	142	# Check if download was successful	COMMENT
LOW	apps/mnncli/android_build.sh	252	# Check if build was successful	COMMENT
LOW	apps/mnncli/android_build.sh	260	# Check if libMNN.so was created	COMMENT
LOW	apps/mnncli/build_integrated.sh	62	# Check if build was successful	COMMENT
LOW	apps/mnncli/test.sh	152	# Check if executable exists	COMMENT
LOW	apps/mnncli/test/test_hf_download.sh	12	# Check if we're in the right directory	COMMENT
LOW	apps/mnncli/test/test_hf_download.sh	100	# Check if test executable exists	COMMENT
LOW	apps/mnncli/test/test_download_integration.sh	160	# Check if output contains expected pattern	COMMENT
LOW	apps/mnncli/test/test_download_integration.sh	185	# Check if output contains error message	COMMENT
LOW	apps/mnncli/test/test_download_integration.sh	209	# Check if output contains expected pattern	COMMENT
LOW	apps/mnncli/test/test_download_integration.sh	233	# Check if output contains help information	COMMENT
LOW	apps/mnncli/test/test_download_integration.sh	257	# Check if output contains version information	COMMENT
LOW	apps/mnncli/test/test_download_integration.sh	276	# Check if mnncli binary exists	COMMENT
LOW⚡	apps/mnncli/test/create_test_video.sh	10	# Check if ffmpeg is available	COMMENT
LOW	apps/mnncli/test/test_model_repository.sh	12	# Check if we're in the right directory	COMMENT
LOW	apps/mnncli/test/test_model_repository.sh	38	# Check if test executable exists	COMMENT
LOW	apps/mnncli/test/test_hf_download_standalone.sh	12	# Check if we're in the right directory	COMMENT
LOW	apps/mnncli/test/test_hf_download_standalone.sh	48	# Check if mnncli executable exists	COMMENT
LOW	apps/mnncli/test/test_download_simple.sh	151	# Check if output contains expected pattern	COMMENT
LOW	apps/mnncli/test/test_download_simple.sh	177	# Check if output contains error message	COMMENT
LOW	apps/mnncli/test/test_download_simple.sh	196	# Check if mnncli binary exists	COMMENT

AI Structural Patterns40 hits · 40 pts

Severity	File	Line	Context
LOW	pymnn/test/playground/cases/model_test.py	27	CODE
LOW	pymnn/pip_package/MNN/numpy/__init__.py	652	CODE
LOW	pymnn/pip_package/MNN/audio/__init__.py	74	CODE
LOW	pymnn/pip_package/MNN/audio/__init__.py	81	CODE
LOW	tools/script/testMNNFromTf.py	114	CODE
LOW	tools/script/testMNNFromOnnx.py	274	CODE
LOW	tools/script/testMNNFromTflite.py	144	CODE
LOW	tools/script/testMNNFromTorch.py	87	CODE
LOW	tools/mnncompress/mnncompress/common/log.py	128	CODE
LOW	tools/mnncompress/mnncompress/tensorflow/hooks.py	13	CODE
LOW	tools/mnncompress/mnncompress/tensorflow/hooks.py	22	CODE
LOW	tools/mnncompress/mnncompress/tensorflow/hooks.py	31	CODE
LOW	tools/mnncompress/mnncompress/tensorflow/hooks.py	54	CODE
LOW	…ress/mnncompress/tensorflow/TaylorFO_channel_pruner.py	18	CODE
LOW	…ress/mnncompress/tensorflow/SA_prune_ratio_searcher.py	13	CODE
LOW	…ls/mnncompress/mnncompress/tensorflow/OAQ_quantizer.py	87	CODE
LOW	…s/mnncompress/mnncompress/tensorflow/SIMD_OC_pruner.py	18	CODE
LOW	…nncompress/mnncompress/tensorflow/SNIP_level_pruner.py	18	CODE
LOW	…compress/mnncompress/tensorflow/overflow_custom_ops.py	231	CODE
LOW	transformers/llm/eval/evaluate_chat_ceval.py	121	CODE
LOW	transformers/llm/export/utils/transformers.py	1140	CODE
LOW	transformers/llm/export/utils/transformers.py	1155	CODE
LOW	transformers/llm/export/utils/token2wav.py	67	CODE
LOW	transformers/llm/export/utils/token2wav.py	297	CODE
LOW	transformers/llm/export/utils/audio.py	482	CODE
LOW	transformers/llm/export/utils/omni_quantizer.py	14	CODE
LOW	…-mnn/sherpa-mnn/python/sherpa_mnn/online_recognizer.py	41	CODE
LOW	…-mnn/sherpa-mnn/python/sherpa_mnn/online_recognizer.py	315	CODE
LOW	…-mnn/sherpa-mnn/python/sherpa_mnn/online_recognizer.py	438	CODE
LOW	…-mnn/sherpa-mnn/python/sherpa_mnn/online_recognizer.py	567	CODE
LOW	…-mnn/sherpa-mnn/python/sherpa_mnn/online_recognizer.py	685	CODE
LOW	…pa-mnn/sherpa-mnn/python/sherpa_mnn/keyword_spotter.py	30	CODE
LOW	…mnn/sherpa-mnn/python/sherpa_mnn/offline_recognizer.py	42	CODE
LOW	…mnn/sherpa-mnn/python/sherpa_mnn/offline_recognizer.py	182	CODE
LOW	…mnn/sherpa-mnn/python/sherpa_mnn/offline_recognizer.py	264	CODE
LOW	…mnn/sherpa-mnn/python/sherpa_mnn/offline_recognizer.py	337	CODE
LOW	…mnn/sherpa-mnn/python/sherpa_mnn/offline_recognizer.py	412	CODE
LOW	…mnn/sherpa-mnn/python/sherpa_mnn/offline_recognizer.py	486	CODE
LOW	…mnn/sherpa-mnn/python/sherpa_mnn/offline_recognizer.py	729	CODE
LOW	…mnn/sherpa-mnn/python/sherpa_mnn/offline_recognizer.py	802	CODE

AI Slop Vocabulary12 hits · 36 pts

Severity	File	Line	Snippet	Context
MEDIUM	3rd_party/flatbuffers/include/flatbuffers/flatbuffers.h	611	// Essentially, this supports 2 std::vectors in a single buffer.	COMMENT
MEDIUM	3rd_party/flatbuffers/samples/sample_binary.py	19	# to facilitate executing this script in the `samples` folder, and to root	COMMENT
MEDIUM	3rd_party/flatbuffers/src/idl_gen_rust.cpp	279	// conflicts with these keywords. However, that's a fairly nuanced	COMMENT
MEDIUM	source/backend/metal/MetalBackend.hpp	34	// encode-reuse / ICB leverage ceiling. Not compiled in production builds.	COMMENT
MEDIUM	…/opencl/execution/buffer/ConvBufLowMemoryExecution.cpp	847	// Create image1d_buffer_t for input to leverage texture cache (int4 only)	COMMENT
MEDIUM	transformers/llm/engine/src/httplib.h	8949	// Cross-host/scheme redirect - create new client with robust setup	COMMENT
MEDIUM	transformers/llm/engine/src/httplib.h	8954	// New method for robust redirect client creation	COMMENT
MEDIUM	transformers/llm/engine/src/httplib.h	9022	// Setup client with robust configuration	COMMENT
MEDIUM	transformers/llm/engine/src/httplib.h	9030	// New method for robust client setup (based on basic_manual_redirect.cpp logic)	COMMENT
MEDIUM	…LlmChat/app/src/main/cpp/video/byte_buffer_decoder.cpp	235	// Non-blocking poll with limited retries to be robust around startup/drain	COMMENT
MEDIUM	…LlmChat/app/src/main/cpp/video/byte_buffer_decoder.cpp	357	// or observe OUTPUT EOS from the codec. This makes EOS handling robust.	COMMENT
MEDIUM	apps/frameworks/model_downloader/cpp/src/file_utils.cpp	154	// robust lookup with support for provider prefixes and multiple storage layouts.	COMMENT

Self-Referential Comments12 hits · 22 pts

Severity	File	Line	Snippet	Context
MEDIUM	source/backend/opencl/schema/CLCache/Shader.py	21	"""This method is deprecated. Please switch to GetRootAs."""	STRING
MEDIUM	source/backend/opencl/schema/CLCache/BackendInfo.py	21	"""This method is deprecated. Please switch to GetRootAs."""	STRING
MEDIUM	source/backend/opencl/schema/CLCache/Autotuning.py	21	"""This method is deprecated. Please switch to GetRootAs."""	STRING
MEDIUM	source/backend/opencl/schema/CLCache/Cache.py	21	"""This method is deprecated. Please switch to GetRootAs."""	STRING
MEDIUM	source/backend/opencl/schema/CLCache/OpInfo.py	21	"""This method is deprecated. Please switch to GetRootAs."""	STRING
MEDIUM	source/backend/opencl/schema/CLCache/GemmInfo.py	21	"""This method is deprecated. Please switch to GetRootAs."""	STRING
MEDIUM	source/backend/opencl/schema/CLCache/TensorInfo.py	21	"""This method is deprecated. Please switch to GetRootAs."""	STRING
MEDIUM	transformers/llm/export/utils/transformers.py	406	"""This function is intended to align with the l2norm implementation in the FLA library."""	STRING
MEDIUM⚡	apps/Android/MnnLlmChat/scripts/get_repo_info.py	105	# Define the keys that contain lists of models to process	COMMENT
MEDIUM⚡	apps/mnncli/test/create_test_video.sh	3	# Create a simple test video for Android testing	COMMENT
MEDIUM⚡	apps/mnncli/test/create_test_video.sh	18	# Create a simple test video (1 second, 320x240, black frame with text)	COMMENT
MEDIUM	…rks/sherpa-mnn/python-api-examples/offline-tts-play.py	381	"""This function is called whenever max_num_sentences sentences	STRING

Example Usage Blocks10 hits · 16 pts

Severity	File	Line	Snippet	Context
LOW	pymnn/pip_package/build_manylinux2014.sh	3	# Usage:	COMMENT
LOW⚡	tools/script/generate_podspec.sh	8	# Usage:	COMMENT
LOW	package_scripts/harmony/build.sh	10	# Usage:	COMMENT
LOW	transformers/llm/engine/ios/ios_llm_bench.sh	6	# Usage:	COMMENT
LOW	apps/Android/MnnLlmChat/tools/stetho_open.py	7	## Usage:	COMMENT
LOW	…/MnnLlmChat/tests/smoke/scripts/16_regress_voice_ui.sh	17	# Usage:	COMMENT
LOW	…t/tests/smoke/scripts/noui/15_regress_voice_dumpapp.sh	15	# Usage:	COMMENT
LOW	apps/Android/MnnLlmChat/scripts/push_model_test.sh	6	# Usage:	COMMENT
LOW	apps/frameworks/sherpa-mnn/test_tts_batch.sh	8	# Usage:	COMMENT
LOW	apps/frameworks/sherpa-mnn/sherpa-mnn/c-api/c-api.h	271	/// Usage example:	COMMENT

Cross-Language Confusion3 hits · 15 pts

Severity	File	Line	Snippet	Context
HIGH	tools/script/llm_upload.py	127	mkdir build && cd build	CODE
HIGH	tools/script/llm_upload.py	185	mkdir build && cd build	CODE
HIGH	tools/script/llm_upload.py	239	mkdir build && cd build	CODE

Synthetic Comment Markers1 hit · 5 pts

Severity	File	Line	Snippet	Context
HIGH	CONTRIBUTING.md	59	We welcome contributions generated by AI assistants (Cursor, Claude Code, GitHub Copilot, etc.).	CODE

TODO Padding4 hits · 4 pts

Severity	File	Line	Snippet	Context
LOW	3rd_party/flatbuffers/include/flatbuffers/flexbuffers.h	123	// TODO: implement proper support for 8/16bit floats, or decide not to	COMMENT
LOW	3rd_party/flatbuffers/src/reflection.cpp	82	s += "(struct)"; // TODO: implement this as well.	CODE
LOW	3rd_party/flatbuffers/src/reflection.cpp	110	return "[(elements)]"; // TODO: implement this as well.	CODE
LOW	3rd_party/flatbuffers/src/reflection.cpp	111	case reflection::Union: return "(union)"; // TODO: implement this as well.	CODE

Modern AI Meta-Vocabulary1 hit · 3 pts

Severity	File	Line	Snippet	Context
MEDIUM	transformers/llm/export/llmexport.py	231	# embedding model's embed	COMMENT

Modern Structural Boilerplate2 hits · 2 pts

Severity	File	Line	Snippet	Context
LOW	transformers/llm/finetune/mnn_qlora.py	82	def set_random_seed(seed: int) -> None:	CODE
LOW	transformers/llm/export/gguf/gguf_reader.py	37	logger = logging.getLogger(__name__)	CODE

Slop Phrases1 hit · 2 pts

Severity	File	Line	Snippet	Context
LOW	transformers/llm/export/utils/transformers.py	1303	# the current expert. We need to make sure to multiply the output hidden	COMMENT

Analysis Overview

What These Metrics Mean

Score History

Severity Breakdown

Directory Score Breakdown

Pattern Findings