NVlabs/LongLive

19.6

Adjusted Score

19.6

Raw Score

100%

Time Factor

2026-07-09

Last Push

2.5K

Stars

Python

Language

40.0K

Lines of Code

203

Files

551

Pattern Hits

2026-07-14

Scan Date

0.21

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 1HIGH 42MEDIUM 23LOW 485

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 551 distinct pattern matches across 16 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.

Cross-File Repetition40 hits · 200 pts

Severity	File	Snippet	Context
HIGH	fouroversix/scripts/ptq/coordinators/local.py	run any tasks that should be used to calibrate models for a given ptq method and set of parameters before running evalua	STRING
HIGH	fouroversix/scripts/ptq/coordinators/modal.py	run any tasks that should be used to calibrate models for a given ptq method and set of parameters before running evalua	STRING
HIGH	fouroversix/scripts/ptq/coordinators/base.py	run any tasks that should be used to calibrate models for a given ptq method and set of parameters before running evalua	STRING
HIGH	fouroversix/src/fouroversix/quantize/backend.py	return true if the cutlass backend is available on the current machine.	STRING
HIGH	fouroversix/src/fouroversix/matmul/backend.py	return true if the cutlass backend is available on the current machine.	STRING
HIGH	fouroversix/src/fouroversix/quantize/cuda/backend.py	return true if the cutlass backend is available on the current machine.	STRING
HIGH	fouroversix/src/fouroversix/quantize/triton/backend.py	return true if the cutlass backend is available on the current machine.	STRING
HIGH	fouroversix/src/fouroversix/quantize/pytorch/backend.py	return true if the cutlass backend is available on the current machine.	STRING
HIGH	fouroversix/src/fouroversix/matmul/pytorch.py	return true if the cutlass backend is available on the current machine.	STRING
HIGH	fouroversix/src/fouroversix/matmul/cutlass/backend.py	return true if the cutlass backend is available on the current machine.	STRING
HIGH	fouroversix/src/fouroversix/quantize/backend.py	return true if the pytorch backend supports the given input and quantization configuration.	STRING
HIGH	fouroversix/src/fouroversix/quantize/cuda/backend.py	return true if the pytorch backend supports the given input and quantization configuration.	STRING
HIGH	fouroversix/src/fouroversix/quantize/triton/backend.py	return true if the pytorch backend supports the given input and quantization configuration.	STRING
HIGH	fouroversix/src/fouroversix/quantize/pytorch/backend.py	return true if the pytorch backend supports the given input and quantization configuration.	STRING
HIGH	fouroversix/src/fouroversix/quantize/backend.py	quantize a tensor to fp4 using the pytorch backend. args: x (torch.tensor): the input tensor to quantize. config (quanti	STRING
HIGH	…oversix/src/fouroversix/quantize/transformer_engine.py	quantize a tensor to fp4 using the pytorch backend. args: x (torch.tensor): the input tensor to quantize. config (quanti	STRING
HIGH	fouroversix/src/fouroversix/quantize/cuda/backend.py	quantize a tensor to fp4 using the pytorch backend. args: x (torch.tensor): the input tensor to quantize. config (quanti	STRING
HIGH	fouroversix/src/fouroversix/quantize/triton/backend.py	quantize a tensor to fp4 using the pytorch backend. args: x (torch.tensor): the input tensor to quantize. config (quanti	STRING
HIGH	fouroversix/src/fouroversix/quantize/pytorch/backend.py	quantize a tensor to fp4 using the pytorch backend. args: x (torch.tensor): the input tensor to quantize. config (quanti	STRING
HIGH	fouroversix/src/fouroversix/model/modules/linear.py	return high precision parameters to be quantized and deleted.	STRING
HIGH	fouroversix/src/fouroversix/model/modules/qwen.py	return high precision parameters to be quantized and deleted.	STRING
HIGH	fouroversix/src/fouroversix/model/modules/gpt_oss.py	return high precision parameters to be quantized and deleted.	STRING
HIGH	fouroversix/src/fouroversix/model/modules/linear_ori.py	return high precision parameters to be quantized and deleted.	STRING
HIGH	fouroversix/src/fouroversix/model/modules/linear.py	get the size of a single element, in bytes, for a parameter.	STRING
HIGH	fouroversix/src/fouroversix/model/modules/qwen.py	get the size of a single element, in bytes, for a parameter.	STRING
HIGH	fouroversix/src/fouroversix/model/modules/gpt_oss.py	get the size of a single element, in bytes, for a parameter.	STRING
HIGH	fouroversix/src/fouroversix/model/modules/linear_ori.py	get the size of a single element, in bytes, for a parameter.	STRING
HIGH	fouroversix/src/fouroversix/model/modules/linear.py	prepare this layer for post-training quantization by quantizing the weight, storing the quantized weight, and deleting t	STRING
HIGH	fouroversix/src/fouroversix/model/modules/qwen.py	prepare this layer for post-training quantization by quantizing the weight, storing the quantized weight, and deleting t	STRING
HIGH	fouroversix/src/fouroversix/model/modules/gpt_oss.py	prepare this layer for post-training quantization by quantizing the weight, storing the quantized weight, and deleting t	STRING
HIGH	fouroversix/src/fouroversix/model/modules/linear_ori.py	prepare this layer for post-training quantization by quantizing the weight, storing the quantized weight, and deleting t	STRING
HIGH	wan_5b/textimage2video.py	initializes the image-to-video generation model components. args: config (easydict): object containing model parameters	STRING
HIGH	wan_5b/text2video.py	initializes the image-to-video generation model components. args: config (easydict): object containing model parameters	STRING
HIGH	wan_5b/image2video.py	initializes the image-to-video generation model components. args: config (easydict): object containing model parameters	STRING
HIGH	wan_5b/textimage2video.py	configures a model object. this includes setting evaluation modes, applying distributed parallel strategy, and handling	STRING
HIGH	wan_5b/text2video.py	configures a model object. this includes setting evaluation modes, applying distributed parallel strategy, and handling	STRING
HIGH	wan_5b/image2video.py	configures a model object. this includes setting evaluation modes, applying distributed parallel strategy, and handling	STRING
HIGH	wan_5b/distributed/sequence_parallel.py	forward pass through the diffusion model args: x (list[tensor]): list of input video tensors, each with shape [c_in, f,	STRING
HIGH	wan_5b/modules/causal_model.py	forward pass through the diffusion model args: x (list[tensor]): list of input video tensors, each with shape [c_in, f,	STRING
HIGH	wan_5b/modules/model.py	forward pass through the diffusion model args: x (list[tensor]): list of input video tensors, each with shape [c_in, f,	STRING

Unused Imports105 hits · 103 pts

Severity	File	Line	Context
LOW	pipeline/__init__.py	1	CODE
LOW	pipeline/__init__.py	2	CODE
LOW	pipeline/self_forcing_training.py	12	CODE
LOW	fouroversix/scripts/generate_kernels.py	1	CODE
LOW	fouroversix/scripts/resources.py	1	CODE
LOW	fouroversix/scripts/test_on_modal.py	1	CODE
LOW	fouroversix/scripts/speedtest/matmul.py	1	CODE
LOW	fouroversix/scripts/speedtest/quantize.py	1	CODE
LOW	fouroversix/scripts/ptq/utils.py	1	CODE
LOW	fouroversix/scripts/ptq/coordinators/__init__.py	1	CODE
LOW	fouroversix/scripts/ptq/coordinators/__init__.py	2	CODE
LOW	fouroversix/scripts/ptq/evaluators/rtn.py	1	CODE
LOW	fouroversix/scripts/ptq/evaluators/spinquant.py	1	CODE
LOW	fouroversix/scripts/ptq/evaluators/__init__.py	1	CODE
LOW	fouroversix/scripts/ptq/evaluators/awq.py	1	CODE
LOW	fouroversix/scripts/ptq/evaluators/smoothquant.py	1	CODE
LOW	fouroversix/scripts/ptq/evaluators/evaluator.py	1	CODE
LOW	fouroversix/scripts/train/__main__.py	1	CODE
LOW	fouroversix/src/fouroversix/__init__.py	3	CODE
LOW	fouroversix/src/fouroversix/__init__.py	4	CODE
LOW	fouroversix/src/fouroversix/__init__.py	4	CODE
LOW	fouroversix/src/fouroversix/__init__.py	4	CODE
LOW	fouroversix/src/fouroversix/__init__.py	4	CODE
LOW	fouroversix/src/fouroversix/__init__.py	4	CODE
LOW	fouroversix/src/fouroversix/__init__.py	11	CODE
LOW	fouroversix/src/fouroversix/__init__.py	11	CODE
LOW	fouroversix/src/fouroversix/__init__.py	11	CODE
LOW	fouroversix/src/fouroversix/__init__.py	12	CODE
LOW	fouroversix/src/fouroversix/__init__.py	12	CODE
LOW	fouroversix/src/fouroversix/__init__.py	12	CODE
LOW	fouroversix/src/fouroversix/__init__.py	12	CODE
LOW	fouroversix/src/fouroversix/__init__.py	12	CODE
LOW	fouroversix/src/fouroversix/__init__.py	13	CODE
LOW	fouroversix/src/fouroversix/quantize/__init__.py	1	CODE
LOW	fouroversix/src/fouroversix/quantize/__init__.py	2	CODE
LOW	fouroversix/src/fouroversix/quantize/__init__.py	3	CODE
LOW	fouroversix/src/fouroversix/quantize/__init__.py	4	CODE
LOW	…oversix/src/fouroversix/quantize/transformer_engine.py	33	CODE
LOW	fouroversix/src/fouroversix/quantize/cuda/backend.py	29	CODE
LOW	fouroversix/src/fouroversix/quantize/cuda/__init__.py	1	CODE
LOW	fouroversix/src/fouroversix/quantize/triton/kernel.py	1	CODE
LOW	fouroversix/src/fouroversix/quantize/triton/__init__.py	1	CODE
LOW	…uroversix/src/fouroversix/quantize/pytorch/__init__.py	1	CODE
LOW	…roversix/src/fouroversix/quantize/pytorch/reference.py	1	CODE
LOW	fouroversix/src/fouroversix/model/quantize.py	1	CODE
LOW	fouroversix/src/fouroversix/model/__init__.py	1	CODE
LOW	fouroversix/src/fouroversix/model/__init__.py	1	CODE
LOW	fouroversix/src/fouroversix/model/__init__.py	2	CODE
LOW	fouroversix/src/fouroversix/model/__init__.py	3	CODE
LOW	fouroversix/src/fouroversix/model/__init__.py	3	CODE
LOW	fouroversix/src/fouroversix/model/modules/__init__.py	1	CODE
LOW	fouroversix/src/fouroversix/model/modules/__init__.py	2	CODE
LOW	fouroversix/src/fouroversix/matmul/__init__.py	1	CODE
LOW	fouroversix/src/fouroversix/matmul/cutlass/backend.py	27	CODE
LOW	fouroversix/src/fouroversix/matmul/cutlass/__init__.py	1	CODE
LOW	utils/dataset.py	9	CODE
LOW	utils/dataset.py	10	CODE
LOW	utils/dataset.py	15	CODE
LOW	utils/rope_triton.py	17	CODE
LOW	utils/nvfp4_checkpoint.py	10	CODE
45 more matches not shown…

Hyper-Verbose Identifiers113 hits · 103 pts

Severity	File	Line	Snippet	Context
LOW	inference_sp.py	99	def _expected_inference_samples(config):	CODE
LOW	inference_sp.py	166	def materialize_quantized_generator(model, device, materialize_fn, stage_desc, is_main_process):	CODE
LOW	inference_sp.py	176	def configure_generator_torch_compile(pipeline, config, is_main_process):	CODE
LOW	inference.py	145	def _expected_inference_samples(config):	CODE
LOW	inference.py	223	def materialize_quantized_generator(model, device, materialize_fn, stage_desc):	CODE
LOW	inference.py	237	def configure_generator_torch_compile(pipeline, config):	CODE
LOW⚡	pipeline/causal_diffusion_inference.py	954	def _set_all_modules_sink_size(self, sink_size_value: int):	CODE
LOW⚡	pipeline/causal_diffusion_inference.py	968	def _set_all_modules_global_sink_size(self, value: int):	CODE
LOW	pipeline/causal_diffusion_inference.py	861	def _initialize_crossattn_cache(self, batch_size, dtype, device):	CODE
LOW	pipeline/causal_diffusion_inference.py	895	def _initialize_sample_scheduler(self, noise):	CODE
LOW	pipeline/causal_diffusion_inference.py	918	def _set_all_modules_max_attention_size(self, local_attn_size_value: int):	CODE
LOW	pipeline/causal_diffusion_inference.py	1001	def _update_sink_for_scene_cut(self, kv_cache, current_num_frames):	CODE
LOW⚡	pipeline/self_forcing_training.py	372	def _set_all_modules_sink_size(self, sink_size_value: int):	CODE
LOW⚡	pipeline/self_forcing_training.py	384	def _set_all_modules_global_sink_size(self, value: int):	CODE
LOW	pipeline/self_forcing_training.py	82	def _build_default_denoising_step_list(self, sampling_steps):	CODE
LOW	pipeline/self_forcing_training.py	111	def generate_chunk_with_cache(	CODE
LOW	pipeline/self_forcing_training.py	288	def inference_with_trajectory(	CODE
LOW	pipeline/self_forcing_training.py	409	def _inference_with_trajectory_inner(	CODE
LOW	pipeline/self_forcing_training.py	698	def _initialize_crossattn_cache(self, batch_size, dtype, device):	CODE
LOW	pipeline/self_forcing_training.py	744	def _set_all_modules_max_attention_size(self, local_attn_size_value: int):	CODE
LOW	fouroversix/setup.py	32	def get_cuda_bare_metal_version() -> Version \| None:	CODE
LOW	fouroversix/tests/test_correctness.py	52	def test_backend_outputs_are_consistent( # noqa: C901, PLR0915	CODE
LOW	fouroversix/scripts/resources.py	52	def has_untracked_or_unstaged_changes(self) -> bool:	CODE
LOW	fouroversix/scripts/resources.py	150	def install_fouroversix_non_editable() -> None:	CODE
LOW	fouroversix/scripts/ptq/evaluators/smoothquant.py	196	def get_calibration_experiments(	CODE
LOW	fouroversix/src/fouroversix/quantize/utils.py	14	def get_no_random_sign_vector(device: int) -> torch.Tensor:	CODE
LOW	fouroversix/src/fouroversix/model/quantize.py	35	def should_replace_existing_modules_in_model(	CODE
LOW	fouroversix/src/fouroversix/model/modules/linear.py	271	def _serialize_quantized_weight(	CODE
LOW	fouroversix/src/fouroversix/model/modules/linear.py	322	def quantized_weight_transposed(self) -> QuantizedTensor \| None:	CODE
LOW	tests/test_i2v_teacher_forcing_context.py	96	def _load_causal_diffusion_with_stubs():	CODE
LOW	tests/test_fp8_inference.py	19	def test_strict_compile_wrapper_does_not_shadow_torch(monkeypatch):	CODE
LOW	utils/memory.py	61	def fake_diffusers_current_device(model: torch.nn.Module, target_device: torch.device):	CODE
LOW	utils/memory.py	96	def move_model_to_device_with_memory_preservation(model, target_device, preserved_memory_gb=0):	CODE
LOW	utils/memory.py	112	def offload_model_from_device_for_memory_preservation(model, target_device, preserved_memory_gb=0):	CODE
LOW	utils/torch_compile_utils.py	86	def configure_module_call_torch_compile(	CODE
LOW	utils/dataset.py	200	def _load_captions_from_folder(self, folder: Path):	CODE
LOW	utils/dataset.py	422	def _get_video_files_in_folder(self, folder_path):	CODE
LOW	utils/dataset.py	451	def _check_folder_has_enough_frames(self, folder_path):	CODE
LOW	utils/dataset.py	533	def _can_sample_from_position(self, total_frames, original_fps, num_frames, start_frame):	CODE
LOW	utils/dataset.py	552	def _can_complete_all_segments_without_wrap(self, video_files, start_video_idx, start_frame):	CODE
LOW	utils/dataset.py	629	def _sample_frames_from_video(self, video_path, num_frames, start_frame=0):	CODE
LOW	utils/dataset.py	696	def _sample_frames_from_video_ffmpeg(self, video_path, num_frames, start_frame=0):	CODE
LOW	utils/dataset.py	750	def _try_get_item_from_folder(self, folder_idx, deterministic: bool = False):	CODE
LOW	utils/lightvae_5b_wrapper.py	18	def _extract_checkpoint_state_dict(raw):	CODE
LOW	utils/lightvae_5b_wrapper.py	31	def _map_lightvae_key_to_wanvae(key):	CODE
LOW	utils/lightvae_5b_wrapper.py	106	def _normalize_vae_state_dict(raw_state):	CODE
LOW	utils/lightvae_5b_wrapper.py	117	def infer_lightvae_pruning_rate_from_ckpt(vae_path, full_decoder_conv1_out=1024):	CODE
LOW	utils/position_embedding_utils.py	15	def select_temporal_offset_for_sample(	CODE
LOW	utils/rope_triton.py	84	def _split_complex_to_cos_sin(freqs_complex: torch.Tensor) -> tuple[torch.Tensor, torch.Tensor]:	CODE
LOW	utils/nvfp4_checkpoint.py	38	def unwrap_generator_state_dict(checkpoint: object, use_ema: bool = False) -> object:	CODE
LOW	utils/nvfp4_checkpoint.py	50	def clean_fsdp_state_dict_keys(state_dict: Mapping[str, torch.Tensor]) -> dict[str, torch.Tensor]:	CODE
LOW	utils/nvfp4_checkpoint.py	55	def build_model_quantization_config(config, keep_master_weights: bool = False):	CODE
LOW	utils/nvfp4_checkpoint.py	134	def drop_fouroversix_master_weights(model: nn.Module) -> list[str]:	CODE
LOW	utils/wan_5b_wrapper.py	316	def enable_gradient_checkpointing(self) -> None:	CODE
LOW	utils/quant.py	90	def _resolve_modules_to_not_convert(	CODE
LOW	utils/quant.py	119	def _get_default_filtered_modules(filter_profile: str \| None) -> list[str]:	CODE
LOW	utils/quant.py	137	def _warn_for_te_config_mismatch(model_quant_config: ModelQuantizationConfig) -> None:	CODE
LOW	utils/quant.py	273	def _copy_tensor_into_parameter(	CODE
LOW	utils/quant.py	350	def quantize_model_with_optional_te(	CODE
LOW	utils/quant.py	425	def _materialize_transformer_engine_weights_for_inference(	CODE
53 more matches not shown…

Verbosity Indicators43 hits · 64 pts

Severity	File	Line	Snippet	Context
LOW⚡	pipeline/causal_diffusion_inference.py	67	# Step 2: Initialize scheduler	COMMENT
LOW	pipeline/causal_diffusion_inference.py	39	# Step 1: Initialize all models	COMMENT
LOW	pipeline/causal_diffusion_inference.py	232	# Step 1: Initialize KV cache to all zeros	COMMENT
LOW	pipeline/causal_diffusion_inference.py	266	# Step 2: Cache context feature	COMMENT
LOW	pipeline/causal_diffusion_inference.py	437	# Step 3: Temporal denoising loop	COMMENT
LOW	pipeline/causal_diffusion_inference.py	741	# Step 4: Decode the output	COMMENT
LOW	pipeline/self_forcing_training.py	446	# Step 1: Initialize KV cache to all zeros	COMMENT
LOW	pipeline/self_forcing_training.py	468	# Step 2: Cache context feature	COMMENT
LOW	pipeline/self_forcing_training.py	486	# Step 3: Temporal denoising loop	COMMENT
LOW	utils/dataset.py	545	# We need to check if we can get num_frames frames	COMMENT
LOW	model/diffusion.py	30	# Step 2: Initialize all hyperparameters	COMMENT
LOW	model/diffusion.py	171	# Step 2: Randomly sample a timestep and add noise to denoiser inputs	COMMENT
LOW	model/diffusion.py	258	# Step 3: Noise augmentation, also add small noise to clean context latents	COMMENT
LOW	model/diffusion.py	331	# Step 5: Store prediction errors into error buffer.	COMMENT
LOW⚡	model/dmd.py	487	# Step 5: Debugging Log	COMMENT
LOW	model/dmd.py	47	# Step 2: Initialize all dmd hyperparameters	COMMENT
LOW	model/dmd.py	98	# Step 1: Compute the fake score	COMMENT
LOW	model/dmd.py	119	# Step 2: Compute the real score	COMMENT
LOW	model/dmd.py	138	# Step 3: Compute the DMD gradient (DMD paper eq. 7).	COMMENT
LOW	model/dmd.py	213	# Step 1: Randomly sample timestep based on the given schedule and corresponding noise	COMMENT
LOW	model/dmd.py	247	# Step 2: Compute the KL grad	COMMENT
LOW	model/dmd.py	291	# Step 1: Unroll generator to obtain fake videos	COMMENT
LOW	model/dmd.py	307	# Step 2: Compute the DMD loss	COMMENT
LOW	model/dmd.py	368	# Step 1: Run generator on backward simulated noisy input	COMMENT
LOW	model/dmd.py	410	# Step 2: Compute the fake prediction	COMMENT
LOW	model/dmd.py	450	# Step 3: Compute the denoising loss for the fake critic	COMMENT
LOW⚡	trainer/diffusion.py	60	# Step 1: Initialize the distributed training environment (rank, seed, dtype, logging etc.)	COMMENT
LOW	trainer/diffusion.py	152	# Step 2: Initialize the model and optimizer	COMMENT
LOW	trainer/diffusion.py	333	# Step 3: Initialize the dataloader	COMMENT
LOW	trainer/diffusion.py	736	# Step 1: Get the next batch of text prompts	COMMENT
LOW	trainer/diffusion.py	759	# Step 2: Extract the conditional infos	COMMENT
LOW	trainer/diffusion.py	792	# Step 3: Train the generator	COMMENT
LOW	trainer/diffusion.py	821	# Step 4: Update EMA (if enabled and after start step)	COMMENT
LOW	trainer/diffusion.py	856	# Step 5: Logging	COMMENT
LOW	trainer/distillation.py	41	# Step 1: Initialize the distributed training environment (rank, seed, dtype, logging etc.)	COMMENT
LOW	trainer/distillation.py	79	# Step 2: Initialize the model	COMMENT
LOW	trainer/distillation.py	406	# Step 3: Set up EMA parameter containers	COMMENT
LOW	trainer/distillation.py	430	# Step 4: Initialize the optimizer	COMMENT
LOW	trainer/distillation.py	447	# Step 5: Initialize the dataloader	COMMENT
LOW	trainer/distillation.py	531	# Step 6: Initialize the validation dataloader for visualization (fixed prompts)	COMMENT
LOW	trainer/distillation.py	1000	# Step 1: Get the next batch of text prompts	COMMENT
LOW	trainer/distillation.py	1040	# Step 2: Extract the conditional infos	COMMENT
LOW	trainer/distillation.py	1071	# Step 3: Store gradients for the generator (if training the generator)	COMMENT

Structural Annotation Overuse42 hits · 62 pts

Severity	File	Line	Snippet	Context
LOW⚡	pipeline/causal_diffusion_inference.py	67	# Step 2: Initialize scheduler	COMMENT
LOW	pipeline/causal_diffusion_inference.py	39	# Step 1: Initialize all models	COMMENT
LOW	pipeline/causal_diffusion_inference.py	232	# Step 1: Initialize KV cache to all zeros	COMMENT
LOW	pipeline/causal_diffusion_inference.py	266	# Step 2: Cache context feature	COMMENT
LOW	pipeline/causal_diffusion_inference.py	437	# Step 3: Temporal denoising loop	COMMENT
LOW	pipeline/causal_diffusion_inference.py	741	# Step 4: Decode the output	COMMENT
LOW	pipeline/self_forcing_training.py	446	# Step 1: Initialize KV cache to all zeros	COMMENT
LOW	pipeline/self_forcing_training.py	468	# Step 2: Cache context feature	COMMENT
LOW	pipeline/self_forcing_training.py	486	# Step 3: Temporal denoising loop	COMMENT
LOW	model/diffusion.py	30	# Step 2: Initialize all hyperparameters	COMMENT
LOW	model/diffusion.py	171	# Step 2: Randomly sample a timestep and add noise to denoiser inputs	COMMENT
LOW	model/diffusion.py	258	# Step 3: Noise augmentation, also add small noise to clean context latents	COMMENT
LOW	model/diffusion.py	331	# Step 5: Store prediction errors into error buffer.	COMMENT
LOW⚡	model/dmd.py	487	# Step 5: Debugging Log	COMMENT
LOW	model/dmd.py	47	# Step 2: Initialize all dmd hyperparameters	COMMENT
LOW	model/dmd.py	98	# Step 1: Compute the fake score	COMMENT
LOW	model/dmd.py	119	# Step 2: Compute the real score	COMMENT
LOW	model/dmd.py	138	# Step 3: Compute the DMD gradient (DMD paper eq. 7).	COMMENT
LOW	model/dmd.py	213	# Step 1: Randomly sample timestep based on the given schedule and corresponding noise	COMMENT
LOW	model/dmd.py	247	# Step 2: Compute the KL grad	COMMENT
LOW	model/dmd.py	291	# Step 1: Unroll generator to obtain fake videos	COMMENT
LOW	model/dmd.py	307	# Step 2: Compute the DMD loss	COMMENT
LOW	model/dmd.py	368	# Step 1: Run generator on backward simulated noisy input	COMMENT
LOW	model/dmd.py	410	# Step 2: Compute the fake prediction	COMMENT
LOW	model/dmd.py	450	# Step 3: Compute the denoising loss for the fake critic	COMMENT
LOW⚡	trainer/diffusion.py	60	# Step 1: Initialize the distributed training environment (rank, seed, dtype, logging etc.)	COMMENT
LOW	trainer/diffusion.py	152	# Step 2: Initialize the model and optimizer	COMMENT
LOW	trainer/diffusion.py	333	# Step 3: Initialize the dataloader	COMMENT
LOW	trainer/diffusion.py	736	# Step 1: Get the next batch of text prompts	COMMENT
LOW	trainer/diffusion.py	759	# Step 2: Extract the conditional infos	COMMENT
LOW	trainer/diffusion.py	792	# Step 3: Train the generator	COMMENT
LOW	trainer/diffusion.py	821	# Step 4: Update EMA (if enabled and after start step)	COMMENT
LOW	trainer/diffusion.py	856	# Step 5: Logging	COMMENT
LOW	trainer/distillation.py	41	# Step 1: Initialize the distributed training environment (rank, seed, dtype, logging etc.)	COMMENT
LOW	trainer/distillation.py	79	# Step 2: Initialize the model	COMMENT
LOW	trainer/distillation.py	406	# Step 3: Set up EMA parameter containers	COMMENT
LOW	trainer/distillation.py	430	# Step 4: Initialize the optimizer	COMMENT
LOW	trainer/distillation.py	447	# Step 5: Initialize the dataloader	COMMENT
LOW	trainer/distillation.py	531	# Step 6: Initialize the validation dataloader for visualization (fixed prompts)	COMMENT
LOW	trainer/distillation.py	1000	# Step 1: Get the next batch of text prompts	COMMENT
LOW	trainer/distillation.py	1040	# Step 2: Extract the conditional infos	COMMENT
LOW	trainer/distillation.py	1071	# Step 3: Store gradients for the generator (if training the generator)	COMMENT

Excessive Try-Catch Wrapping52 hits · 57 pts

Severity	File	Line	Snippet	Context
LOW	inference_sp.py	49	except Exception as exc:	CODE
LOW	inference.py	84	except Exception as e:	CODE
MEDIUM	pipeline/causal_diffusion_inference.py	468	def _vae_thread_fn():	CODE
LOW⚡	pipeline/causal_diffusion_inference.py	64	except Exception as exc:	CODE
LOW⚡	pipeline/causal_diffusion_inference.py	356	except Exception:	CODE
LOW⚡	pipeline/causal_diffusion_inference.py	361	except Exception:	CODE
LOW⚡	pipeline/causal_diffusion_inference.py	366	except Exception:	CODE
LOW⚡	pipeline/causal_diffusion_inference.py	936	except Exception:	CODE
LOW⚡	pipeline/causal_diffusion_inference.py	946	except Exception:	CODE
LOW⚡	pipeline/causal_diffusion_inference.py	951	except Exception:	CODE
LOW⚡	pipeline/causal_diffusion_inference.py	965	except Exception:	CODE
LOW⚡	pipeline/causal_diffusion_inference.py	974	except Exception:	CODE
LOW	pipeline/causal_diffusion_inference.py	495	except Exception as exc:	CODE
LOW⚡	pipeline/self_forcing_training.py	348	except Exception:	CODE
LOW⚡	pipeline/self_forcing_training.py	353	except Exception:	CODE
LOW⚡	pipeline/self_forcing_training.py	358	except Exception:	CODE
LOW⚡	pipeline/self_forcing_training.py	381	except Exception:	CODE
LOW⚡	pipeline/self_forcing_training.py	390	except Exception:	CODE
LOW	pipeline/self_forcing_training.py	764	except Exception:	CODE
LOW	pipeline/self_forcing_training.py	773	except Exception:	CODE
LOW	utils/torch_compile_utils.py	49	except Exception as exc:	CODE
LOW	utils/torch_compile_utils.py	75	except Exception as exc:	CODE
LOW	utils/torch_compile_utils.py	112	except Exception as exc:	CODE
MEDIUM	utils/dataset.py	433	def get_numeric_key(path):	CODE
LOW⚡	utils/dataset.py	211	except Exception:	CODE
LOW	utils/dataset.py	230	except Exception:	CODE
LOW	utils/dataset.py	418	except Exception:	CODE
LOW	utils/dataset.py	469	except Exception:	CODE
LOW	utils/dataset.py	953	except Exception as exc:	CODE
LOW	utils/lightvae_5b_wrapper.py	123	except Exception as exc:	CODE
LOW	utils/lightvae_5b_wrapper.py	138	except Exception:	CODE
LOW	utils/quant.py	282	except Exception:	CODE
LOW	utils/quant.py	294	except Exception as copy_exc:	CODE
LOW	utils/quant.py	719	except Exception as exc: # pragma: no cover - exercised only when extension is stale/missing	CODE
LOW	wan_5b/utils/prompt_extend.py	185	except Exception as e:	CODE
LOW	wan_5b/utils/prompt_extend.py	248	except Exception as e:	CODE
LOW	wan_5b/utils/utils.py	53	except Exception as e:	CODE
LOW	wan_5b/utils/utils.py	75	except Exception as e:	CODE
LOW	wan_5b/modules/causal_model_sp_ulysses.py	46	except Exception:	CODE
LOW	wan_5b/modules/attention.py	11	except Exception:	CODE
LOW	wan_5b/modules/attention.py	36	except Exception:	CODE
LOW	scripts/decode_vae_latents.py	148	except Exception as e:	CODE
LOW	scripts/decode_lightvae_latents.py	324	except Exception as exc:	CODE
LOW	scripts/decode_lightvae_latents.py	334	except Exception as exc:	CODE
LOW⚡	model/dmd.py	484	except Exception:	CODE
LOW	model/dmd.py	334	except Exception:	CODE
LOW⚡	trainer/diffusion.py	50	except Exception as e:	CODE
LOW	trainer/diffusion.py	629	except Exception as e:	CODE
LOW	trainer/diffusion.py	646	except Exception as e:	CODE
LOW	trainer/diffusion.py	885	except Exception:	CODE
LOW⚡	trainer/distillation.py	1265	except Exception as e:	CODE
LOW	trainer/distillation.py	908	except Exception as e:	CODE

Decorative Section Separators20 hits · 49 pts

Severity	File	Line	Snippet	Context
MEDIUM⚡	utils/dataset.py	164	# ------------------------------------------------------------------	COMMENT
MEDIUM⚡	utils/dataset.py	166	# ------------------------------------------------------------------	COMMENT
MEDIUM⚡	utils/dataset.py	175	# ------------------------------------------------------------------	COMMENT
MEDIUM⚡	utils/dataset.py	177	# ------------------------------------------------------------------	COMMENT
MEDIUM⚡	utils/dataset.py	215	# ------------------------------------------------------------------	COMMENT
MEDIUM⚡	utils/dataset.py	217	# ------------------------------------------------------------------	COMMENT
MEDIUM	wan_5b/modules/causal_model.py	246	# -------------------------------	STRING
MEDIUM	wan_5b/modules/causal_model.py	249	# -------------------------------	STRING
MEDIUM	wan_5b/modules/causal_model.py	264	# -------------------------------	STRING
MEDIUM	wan_5b/modules/causal_model.py	269	# -------------------------------	STRING
MEDIUM	wan_5b/modules/causal_model.py	1217	# ==========================================	STRING
MEDIUM	wan_5b/modules/causal_model.py	1219	# ==========================================	STRING
MEDIUM	wan_5b/modules/causal_model.py	1231	# ==========================================	STRING
MEDIUM	wan_5b/modules/causal_model.py	1233	# ==========================================	STRING
MEDIUM	wan_5b/modules/causal_model.py	1250	# ==========================================	STRING
MEDIUM	wan_5b/modules/causal_model.py	1252	# ==========================================	STRING
MEDIUM	trainer/distillation.py	601	# ----------------------------------------------------------------------------------------------------------	COMMENT
MEDIUM	trainer/distillation.py	603	# ----------------------------------------------------------------------------------------------------------	COMMENT
MEDIUM	trainer/distillation.py	1332	# --------------------------------------------------------------------------------------------------------------	COMMENT
MEDIUM	trainer/distillation.py	1334	# --------------------------------------------------------------------------------------------------------------	COMMENT

Deep Nesting47 hits · 36 pts

Severity	File	Line	Context
LOW	inference_sp.py	33	CODE
LOW	inference.py	61	CODE
LOW	pipeline/causal_diffusion_inference.py	167	CODE
LOW	pipeline/causal_diffusion_inference.py	369	CODE
LOW	pipeline/causal_diffusion_inference.py	468	CODE
LOW	pipeline/self_forcing_training.py	111	CODE
LOW	pipeline/self_forcing_training.py	409	CODE
LOW	fouroversix/setup.py	52	CODE
LOW	fouroversix/tests/test_correctness.py	52	CODE
LOW	fouroversix/scripts/ptq/evaluators/evaluator.py	59	CODE
LOW	utils/dataset.py	750	CODE
LOW	utils/position_embedding_utils.py	56	CODE
LOW	utils/error_buffer.py	239	CODE
LOW	utils/nvfp4_checkpoint.py	134	CODE
LOW	utils/quant.py	477	CODE
LOW	utils/lora_utils.py	19	CODE
LOW	wan_5b/utils/fm_solvers.py	412	CODE
LOW	wan_5b/utils/fm_solvers.py	483	CODE
LOW	wan_5b/utils/utils.py	106	CODE
LOW	wan_5b/utils/qwen_vl_utils.py	85	CODE
LOW	wan_5b/utils/qwen_vl_utils.py	328	CODE
LOW	wan_5b/modules/attention.py	84	CODE
LOW	wan_5b/modules/vae2_1.py	68	CODE
LOW	wan_5b/modules/vae2_1.py	101	CODE
LOW	wan_5b/modules/t5.py	27	CODE
LOW	wan_5b/modules/causal_model.py	90	CODE
LOW	wan_5b/modules/causal_model.py	327	CODE
LOW	wan_5b/modules/causal_model.py	1371	CODE
LOW	wan_5b/modules/vae2_2.py	73	CODE
LOW	wan_5b/modules/vae2_2.py	111	CODE
LOW	scripts/decode_vae_latents.py	72	CODE
LOW	scripts/decode_lightvae_latents.py	104	CODE
LOW	model/diffusion.py	15	CODE
LOW	trainer/diffusion.py	25	CODE
LOW	trainer/diffusion.py	56	CODE
LOW	trainer/diffusion.py	552	CODE
LOW	trainer/diffusion.py	578	CODE
LOW	trainer/diffusion.py	637	CODE
LOW	trainer/diffusion.py	911	CODE
LOW	trainer/diffusion.py	1016	CODE
LOW	trainer/diffusion.py	1080	CODE
LOW	trainer/distillation.py	37	CODE
LOW	trainer/distillation.py	831	CODE
LOW	trainer/distillation.py	857	CODE
LOW	trainer/distillation.py	994	CODE
LOW	trainer/distillation.py	1152	CODE
LOW	trainer/distillation.py	1271	CODE

AI Structural Patterns36 hits · 34 pts

Severity	File	Line	Context
LOW	inference.py	562	CODE
LOW	pipeline/self_forcing_training.py	17	CODE
LOW	pipeline/causal_diffusion_inference_sp.py	32	CODE
LOW	fouroversix/scripts/resources.py	188	CODE
LOW	fouroversix/scripts/ptq/evaluators/evaluator.py	59	CODE
LOW	utils/dataset.py	280	CODE
LOW	utils/lightvae_5b_wrapper.py	162	CODE
LOW	utils/error_buffer.py	41	CODE
LOW	utils/wan_5b_wrapper.py	184	CODE
LOW	utils/wan_5b_wrapper.py	209	CODE
LOW	utils/wan_5b_wrapper.py	274	CODE
LOW	utils/wan_5b_wrapper.py	278	CODE
LOW	utils/wan_5b_wrapper.py	453	CODE
LOW	utils/scheduler.py	108	CODE
LOW	utils/quant.py	585	CODE
LOW	wan_5b/textimage2video.py	36	CODE
LOW	wan_5b/textimage2video.py	162	CODE
LOW	wan_5b/textimage2video.py	239	CODE
LOW	wan_5b/textimage2video.py	413	CODE
LOW	wan_5b/text2video.py	33	CODE
LOW	wan_5b/text2video.py	203	CODE
LOW	wan_5b/image2video.py	35	CODE
LOW	wan_5b/image2video.py	206	CODE
LOW	wan_5b/distributed/sequence_parallel.py	380	CODE
LOW	wan_5b/utils/fm_solvers.py	131	CODE
LOW	wan_5b/utils/fm_solvers_unipc.py	79	CODE
LOW	wan_5b/modules/causal_model_sp_ulysses.py	113	CODE
LOW	wan_5b/modules/causal_model_sp_ulysses.py	342	CODE
LOW	wan_5b/modules/causal_model_sp_ulysses.py	441	CODE
LOW	wan_5b/modules/attention.py	84	CODE
LOW	wan_5b/modules/attention.py	260	CODE
LOW	wan_5b/modules/model.py	325	CODE
LOW	wan_5b/modules/causal_model.py	327	CODE
LOW	wan_5b/modules/causal_model.py	829	CODE
LOW	wan_5b/modules/causal_model.py	974	CODE
LOW	wan_5b/modules/vae2_2.py	735	CODE

Modern Structural Boilerplate30 hits · 30 pts

Severity	File	Line	Snippet	Context
LOW	pipeline/__init__.py	4	__all__ = [	CODE
LOW	fouroversix/scripts/ptq/coordinators/__init__.py	4	__all__ = ["LocalEvaluationCoordinator", "ModalEvaluationCoordinator"]	CODE
LOW	fouroversix/scripts/ptq/evaluators/utils.py	10	def set_random_seeds(seed: int \| None = None) -> None:	CODE
LOW	fouroversix/src/fouroversix/__init__.py	17	__all__ = [	CODE
LOW	fouroversix/src/fouroversix/quantize/__init__.py	6	__all__ = ["QuantizationConfig", "QuantizedTensor", "get_rht_matrix", "quantize_to_fp4"]	CODE
LOW	fouroversix/src/fouroversix/quantize/cuda/__init__.py	3	__all__ = ["CUDAQuantizeBackend"]	CODE
LOW	fouroversix/src/fouroversix/quantize/triton/__init__.py	3	__all__ = ["TritonQuantizeBackend"]	CODE
LOW	…uroversix/src/fouroversix/quantize/pytorch/__init__.py	3	__all__ = ["PyTorchQuantizeBackend"]	CODE
LOW	…oversix/src/fouroversix/weight_conversions/__init__.py	13	__all__ = ["FourOverSixGptOssDeserialize", "GptOssWeightConverter", "WeightConversions"]	CODE
LOW	fouroversix/src/fouroversix/model/__init__.py	5	__all__ = [	CODE
LOW	fouroversix/src/fouroversix/model/modules/__init__.py	6	__all__ = ["FourOverSixGptOssMLP", "FourOverSixLinear"]	CODE
LOW	fouroversix/src/fouroversix/matmul/__init__.py	3	__all__ = ["fp4_matmul"]	CODE
LOW	fouroversix/src/fouroversix/matmul/cutlass/__init__.py	3	__all__ = ["CUTLASSMatmulBackend"]	CODE
LOW	utils/nvfp4_kernel.py	22	__all__ = ["fp4_dequantize", "static_blockwise_fp4_fake_quant"]	CODE
LOW	wan_5b/__init__.py	3	__all__ = ["WanI2V", "WanT2V", "WanTI2V"]	CODE
LOW	wan_5b/distributed/util.py	40	__all__ = [	CODE
LOW	wan_5b/distributed/ulysses.py	47	__all__ = ["distributed_attention", "distributed_flex_attention"]	CODE
LOW	wan_5b/utils/__init__.py	9	__all__ = [	CODE
LOW	wan_5b/utils/utils.py	12	__all__ = ['save_video', 'save_image', 'str2bool']	CODE
LOW	wan_5b/utils/qwen_vl_utils.py	23	logger = logging.getLogger(__name__)	CODE
LOW	wan_5b/modules/attention.py	78	__all__ = [	CODE
LOW	wan_5b/modules/vae2_1.py	10	__all__ = [	CODE
LOW	wan_5b/modules/t5.py	12	__all__ = [	CODE
LOW	wan_5b/modules/__init__.py	9	__all__ = [	CODE
LOW	wan_5b/modules/model.py	12	__all__ = ['WanModel']	CODE
LOW	wan_5b/modules/vae2_2.py	10	__all__ = [	CODE
LOW	wan_5b/modules/tokenizers.py	9	__all__ = ['HuggingfaceTokenizer']	CODE
LOW	model/__init__.py	3	__all__ = [	CODE
LOW	trainer/__init__.py	4	__all__ = [	CODE
LOW	trainer/sp_helper.py	18	__all__ = ["DEFAULT_SP_VAE_HALO_LATENTS", "SequenceParallelHelper"]	CODE

Over-Commented Block16 hits · 16 pts

Severity	File	Line	Snippet	Context
LOW	fouroversix/scripts/hadamard_code_gen.py	1	from pathlib import Path	COMMENT
LOW	…uroversix/src/fouroversix/quantize/quantized_tensor.py	161	f"Expected {expected_scale_factors} scale factors, got "	COMMENT
LOW	…uroversix/src/fouroversix/csrc/include/kernel_traits.h	81	// static constexpr int kSwizzleS = Is_nvfp4 ? 2 : 1; // 4 or 2 elements	COMMENT
LOW	…c/fouroversix/csrc/include/fp4_quant_launch_template.h	21	#define KERNEL_PARAM_MODIFIER __grid_constant__	COMMENT
LOW	…uroversix/src/fouroversix/csrc/include/static_switch.h	1	// Inspired by	COMMENT
LOW	utils/fp8.py	1	# Copyright (c) 2026 NVIDIA CORPORATION & AFFILIATES	COMMENT
LOW	utils/__init__.py	1	# Marker file: turn `utils/` from a namespace package into a regular package.	COMMENT
LOW	utils/rope_triton.py	21	import triton.language as tl	COMMENT
LOW	utils/kernel/__init__.py	1	# Copyright (c) 2026 NVIDIA CORPORATION & AFFILIATES	COMMENT
LOW	wan_5b/modules/attention.py	161	# we're trying to replace).	COMMENT
LOW	wan_5b/modules/causal_model.py	41	# concatenated freqs tensor each time. LRU keeps memory bounded.	COMMENT
LOW	wan_5b/modules/causal_model.py	61	# when disabled, stores (freqs_i_complex, None, None).	COMMENT
LOW	wan_5b/modules/causal_model.py	101	# iter-47 (grad-safety fix): the Triton RoPE kernel (rope_apply_triton) is a	COMMENT
LOW	wan_5b/modules/causal_model.py	121	# kernel will be used (it consumes bf16 directly).	COMMENT
LOW	wan_5b/modules/causal_model.py	501	# * global_sink_tokens: first N frames are permanently anchored	COMMENT
LOW	model/diffusion.py	41	self.device = device	COMMENT

Hallucination Indicators1 hit · 10 pts

Severity	File	Line	Snippet	Context
CRITICAL	fouroversix/scripts/ptq/evaluators/spinquant.py	230	model.lm_head.weight.data = model.model.embed_tokens.weight.data.clone()	CODE

Docstring Block Structure2 hits · 10 pts

Severity	File	Line	Snippet	Context
HIGH	wan_5b/utils/utils.py	80	Convert a string to a boolean. Supported true values: 'yes', 'true', 't', 'y', '1' Supported false values:	STRING
HIGH	wan_5b/utils/qwen_vl_utils.py	138	calculate the number of frames for video used for model inputs. Args: ele (dict): a dict contains the confi	STRING

Self-Referential Comments1 hit · 4 pts

Severity	File	Line	Snippet	Context
MEDIUM⚡	trainer/distillation.py	1276	# Define the specific modules we want to apply LoRA to	COMMENT

Redundant / Tautological Comments2 hits · 2 pts

Severity	File	Line	Snippet	Context
LOW	utils/dataset.py	473	# Check if we have enough frames for all segments	COMMENT
LOW	trainer/distillation.py	1156	# Check if we should train generator on this optimization step	COMMENT

Fake / Example Data1 hit · 1 pts

Severity	File	Line	Snippet	Context
LOW	model/base.py	51	self.fake_model_name = getattr(args, "fake_name", "Wan2.2-TI2V-5B")	CODE

Analysis Overview

What These Metrics Mean

Score History

Severity Breakdown

Directory Score Breakdown

Pattern Findings