ggml speech-to-text inference for 16+ model families
This report summarizes scanner evidence for handy-computer/transcribe.cpp, a C++ project with 2,001 GitHub stars. SynthScan examined 689,692 lines of code across 1977 source files, recording 2878 pattern matches distributed across 21 syntactic categories. The legacy adjusted pattern signal is 5.9 (Moderate pattern signal); it is not an authorship estimate.
Use the findings to ask whether maintainers can explain intent, verify dependencies, simplify unnecessary scaffolding, and focus review effort. Risk Payload v1 adds ownership, supply-chain, over-engineering, and review-effort vectors when available. These are prompts for human verification, not probabilities, security findings, or measured maintenance costs.
Use each vector to guide manual verification of the underlying findings.
Can maintainers explain intent, assumptions, and trade-offs? 2082 finding(s).
Are imports declared and package/API names verified? 210 finding(s).
Are wrappers, guards, and repeated structures necessary? 835 finding(s).
Which changes need focused human review? 2878 finding(s).
This chart tracks the legacy adjusted pattern score across scan runs. Changes can reflect source changes, scan coverage, or detector/scoring versions; they do not measure AI adoption or prove that a team refactored code. Compare runs only when their scoring versions are compatible. The dashed line tracks the raw pattern-hit count.
Groups findings by configured severity weight. Severity is not a confidence estimate, security rating, or proof that a package is missing. Verify imports against the project's manifests and inspect each finding in its local design context.
This horizontal bar chart shows where configured pattern scores accumulate across directories. Treat the bars as locations to inspect, not as evidence of authorship or code quality. Use the file-level findings to decide whether a directory needs a focused ownership, dependency, complexity, or review check.
The scanner identified 2878 distinct pattern matches across 21 syntactic categories. Each entry below represents a discrete location in the source code where the engine recorded a detector match to inspect in context. Expand any category row to inspect the individual file paths, line numbers, code snippets, and the lexical context (CODE, COMMENT, or STRING) in which each match was detected.
Reading the findings table: The Severity column indicates the configured finding weight (CRITICAL / HIGH / MEDIUM / LOW). The Context column identifies whether the match occurred inside executable code, an inline comment, or a string literal — legacy comment-context matches receive a ×1.5 score weight. The ⚡ bolt icon marks clustered matches: three or more patterns within a 10-line window, each receiving an additional ×1.5 density multiplier under the legacy scoring model.
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| LOW | CMakeLists.txt | 61 | endif() | COMMENT |
| LOW | CMakeLists.txt | 121 | # silently ignoring it. | COMMENT |
| LOW | CMakeLists.txt | 141 | # is a teardown/coexistence liability for an embeddable library, and the native | COMMENT |
| LOW | CMakeLists.txt | 161 | # single switch behind every official-wheel CPU posture (the per-flag lists | COMMENT |
| LOW | CMakeLists.txt | 201 | # ----------------------------------------------------------------------------- | COMMENT |
| LOW | CMakeLists.txt | 221 | # initializer zero by C aggregate-init rules. The | COMMENT |
| LOW | CMakeLists.txt | 241 | -fsanitize=address,undefined | COMMENT |
| LOW | CMakeLists.txt | 261 | # ----------------------------------------------------------------------------- | COMMENT |
| LOW | CMakeLists.txt | 281 | set(GGML_HIP ${TRANSCRIBE_HIP} CACHE BOOL "" FORCE) | COMMENT |
| LOW | CMakeLists.txt | 321 | if(TRANSCRIBE_IS_APPLE_SILICON AND NOT DEFINED CACHE{GGML_NATIVE}) | COMMENT |
| LOW | CMakeLists.txt | 341 | # This used to be a per-consumer SPLIT because ggml's native barrier deadlocked | COMMENT |
| LOW | CMakeLists.txt | 361 | # but pin them anyway so the contract is explicit. | COMMENT |
| LOW | CMakeLists.txt | 381 | "(backend modules are shared libraries loaded at runtime)") | COMMENT |
| LOW | pyproject.toml | 1 | # transcribe-cpp-native: the default native provider package for the | COMMENT |
| LOW | pyproject.toml | 121 | "**/__pycache__/", | COMMENT |
| LOW | pyproject.toml | 161 | cmake.args = ["-DTRANSCRIBE_METAL=ON", "-DGGML_NATIVE=OFF"] | COMMENT |
| LOW | pyproject.toml | 201 | "VK_TOOLCHAIN_CACHE", | COMMENT |
| LOW | pyproject.toml | 241 | # Keep `uv run <script>` at the repo root from syncing (= compiling) this | COMMENT |
| LOW | tools/CMakeLists.txt | 1 | # tools/ - developer tools that are NOT part of the public ABI. | COMMENT |
| LOW | tools/transcribe-quantize/policy.h | 1 | // transcribe-quantize policy: per-tensor bucket classification + per-preset | COMMENT |
| LOW | tools/transcribe-quantize/policy.h | 21 | COMMENT | |
| LOW | tools/transcribe-quantize/policy.h | 41 | // granite 4.x keeps the PyTorch Conv1d layout [1, in, out] (ne0 == 1), | COMMENT |
| LOW | tools/transcribe-quantize/policy.h | 61 | // Fallback when the inner dim doesn't divide linear_main's block | COMMENT |
| LOW | tools/transcribe-quantize/main.cpp | 1 | // transcribe-quantize - re-quantize a transcribe.cpp GGUF. | COMMENT |
| LOW | tools/transcribe-quantize/main.cpp | 21 | // pass without any external file handling. | COMMENT |
| LOW | tools/transcribe-quantize/main.cpp | 101 | COMMENT | |
| LOW | tools/transcribe-quantize/policy.cpp | 41 | } | COMMENT |
| LOW | tools/transcribe-quantize/policy.cpp | 121 | return Bucket::Norm; | COMMENT |
| LOW | tools/transcribe-quantize/policy.cpp | 161 | } | COMMENT |
| LOW | tools/transcribe-quantize/policy.cpp | 181 | } | COMMENT |
| LOW | tools/transcribe-quantize/policy.cpp | 201 | // preset: the head is 0.2M params, and a quantized head could move | COMMENT |
| LOW | tools/transcribe-quantize/policy.cpp | 221 | // conv (GET_CONV → F32/F16 only). | COMMENT |
| LOW | tools/transcribe-quantize/policy.cpp | 241 | // embeddings to Q6_K. | COMMENT |
| LOW | tools/transcribe-bench/main.cpp | 1 | // transcribe-bench - one-shot perf measurement for a (model, sample) pair. | COMMENT |
| LOW | include/transcribe/extensions.h | 21 | #include "transcribe/sortformer.h" | COMMENT |
| LOW | tests/ecapa_tdnn_smoke.cpp | 1 | // ecapa_tdnn_smoke.cpp - end-to-end smoke of the ecapa_tdnn family (LANGID | COMMENT |
| LOW | tests/granite5_ctc_smoke.cpp | 1 | // granite5_ctc_smoke.cpp - hermetic Granite Speech 5.0 loader smoke. | COMMENT |
| LOW | tests/qwen3_asr_bpe_parity.cpp | 1 | // qwen3_asr_bpe_parity.cpp - byte-level BPE encoder parity test. | COMMENT |
| LOW | tests/qwen3_asr_bpe_parity.cpp | 21 | // | COMMENT |
| LOW | tests/whisper_bin_tokenize_parity.cpp | 1 | // whisper_bin_tokenize_parity.cpp - tokenizer parity between the | COMMENT |
| LOW | tests/whisper_bin_tokenize_parity.cpp | 21 | // Gated by both: | COMMENT |
| LOW | tests/backend_probe_order_unit.cpp | 1 | // backend_probe_order_unit.cpp - unit tests for GPU probe ordering. | COMMENT |
| LOW | tests/ecapa_tdnn_real_smoke.cpp | 1 | // ecapa_tdnn_real_smoke.cpp - the real VoxLingua107 ECAPA-TDNN GGUF through | COMMENT |
| LOW | tests/diarize_dispatch_unit.cpp | 1 | // diarize_dispatch_unit.cpp - DIARIZE role dispatcher (transcribe-diarize.cpp) | COMMENT |
| LOW | tests/run_dispatch_unit.cpp | 1 | // run_dispatch_unit.cpp - dispatcher-level transcribe_run behavior tests. | COMMENT |
| LOW | tests/gigaam_workspace_release_smoke.cpp | 1 | // Real-model regression test: the dispatcher releases the per-run compute | COMMENT |
| LOW | tests/backend_init_throw_unit.cpp | 1 | // backend_init_throw_unit.cpp - device-init exception containment. | COMMENT |
| LOW | tests/moonshine_streaming_stream_parity.cpp | 1 | // moonshine_streaming_stream_parity.cpp - real-model gated test that | COMMENT |
| LOW | tests/moonshine_streaming_stream_parity.cpp | 21 | // | COMMENT |
| LOW | tests/whisper_e2e_smoke.cpp | 1 | // whisper_e2e_smoke.cpp - real-model gated public-ABI test for Whisper. | COMMENT |
| LOW | tests/whisper_e2e_smoke.cpp | 321 | // recording) so the default-threshold no-speech path exercises | COMMENT |
| LOW | tests/qwen3_asr_smoke.cpp | 1 | // qwen3_asr_smoke.cpp - default-ctest loader smoke for the Qwen3-ASR | COMMENT |
| LOW | tests/qwen3_asr_smoke.cpp | 21 | // 4. ggml_tensor::ne for representative tensors matches the | COMMENT |
| LOW | tests/qwen3_asr_smoke.cpp | 41 | // - qwen3_asr_real_smoke_{0_6b,1_7b}.cpp: env-gated structural tests | COMMENT |
| LOW | tests/moss_diarize_parser_unit.cpp | 1 | // moss_diarize_parser_unit.cpp - grammar tests for the MOSS emergent | COMMENT |
| LOW | tests/CMakeLists.txt | 1 | # transcribe.cpp test suite. | COMMENT |
| LOW | tests/CMakeLists.txt | 21 | # Fixture pipeline rationale: previous attempts ran the generator at | COMMENT |
| LOW | tests/CMakeLists.txt | 81 | NAME transcribe_extension_umbrella_check | COMMENT |
| LOW | tests/CMakeLists.txt | 601 | add_dependencies(transcribe_loader_smoke fixtures) | COMMENT |
| LOW | tests/CMakeLists.txt | 641 | # ----------------------------------------------------------------------------- | COMMENT |
| 1497 more matches not shown… | ||||
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| MEDIUM⚡ | CMakeLists.txt | 98 | # ----------------------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | CMakeLists.txt | 100 | # ----------------------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | CMakeLists.txt | 107 | # ----------------------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | CMakeLists.txt | 109 | # ----------------------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | CMakeLists.txt | 259 | # ----------------------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | CMakeLists.txt | 261 | # ----------------------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | CMakeLists.txt | 269 | # ----------------------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | CMakeLists.txt | 271 | # ----------------------------------------------------------------------------- | COMMENT |
| MEDIUM | CMakeLists.txt | 3 | # ----------------------------------------------------------------------------- | COMMENT |
| MEDIUM | CMakeLists.txt | 5 | # ----------------------------------------------------------------------------- | COMMENT |
| MEDIUM | CMakeLists.txt | 49 | # ----------------------------------------------------------------------------- | COMMENT |
| MEDIUM | CMakeLists.txt | 51 | # ----------------------------------------------------------------------------- | COMMENT |
| MEDIUM | CMakeLists.txt | 70 | # ----------------------------------------------------------------------------- | COMMENT |
| MEDIUM | CMakeLists.txt | 72 | # ----------------------------------------------------------------------------- | COMMENT |
| MEDIUM | CMakeLists.txt | 201 | # ----------------------------------------------------------------------------- | COMMENT |
| MEDIUM | CMakeLists.txt | 203 | # ----------------------------------------------------------------------------- | COMMENT |
| MEDIUM | CMakeLists.txt | 441 | # ----------------------------------------------------------------------------- | COMMENT |
| MEDIUM | CMakeLists.txt | 443 | # ----------------------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | tests/CMakeLists.txt | 369 | # ----------------------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | tests/CMakeLists.txt | 372 | # ----------------------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | tests/CMakeLists.txt | 382 | # ----------------------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | tests/CMakeLists.txt | 384 | # ----------------------------------------------------------------------------- | COMMENT |
| MEDIUM | tests/CMakeLists.txt | 34 | # ----------------------------------------------------------------------------- | COMMENT |
| MEDIUM | tests/CMakeLists.txt | 36 | # ----------------------------------------------------------------------------- | COMMENT |
| MEDIUM | tests/CMakeLists.txt | 86 | # ----------------------------------------------------------------------------- | COMMENT |
| MEDIUM | tests/CMakeLists.txt | 88 | # ----------------------------------------------------------------------------- | COMMENT |
| MEDIUM | tests/CMakeLists.txt | 106 | # ----------------------------------------------------------------------------- | COMMENT |
| MEDIUM | tests/CMakeLists.txt | 108 | # ----------------------------------------------------------------------------- | COMMENT |
| MEDIUM | tests/CMakeLists.txt | 127 | # ----------------------------------------------------------------------------- | COMMENT |
| MEDIUM | tests/CMakeLists.txt | 129 | # ----------------------------------------------------------------------------- | COMMENT |
| MEDIUM | tests/CMakeLists.txt | 148 | # ----------------------------------------------------------------------------- | COMMENT |
| MEDIUM | tests/CMakeLists.txt | 150 | # ----------------------------------------------------------------------------- | COMMENT |
| MEDIUM | tests/CMakeLists.txt | 172 | # ----------------------------------------------------------------------------- | COMMENT |
| MEDIUM | tests/CMakeLists.txt | 174 | # ----------------------------------------------------------------------------- | COMMENT |
| MEDIUM | tests/CMakeLists.txt | 221 | # ----------------------------------------------------------------------------- | COMMENT |
| MEDIUM | tests/CMakeLists.txt | 223 | # ----------------------------------------------------------------------------- | COMMENT |
| MEDIUM | tests/CMakeLists.txt | 237 | # ----------------------------------------------------------------------------- | COMMENT |
| MEDIUM | tests/CMakeLists.txt | 239 | # ----------------------------------------------------------------------------- | COMMENT |
| MEDIUM | tests/CMakeLists.txt | 253 | # ----------------------------------------------------------------------------- | COMMENT |
| MEDIUM | tests/CMakeLists.txt | 268 | # ----------------------------------------------------------------------------- | COMMENT |
| MEDIUM | tests/CMakeLists.txt | 281 | # ----------------------------------------------------------------------------- | COMMENT |
| MEDIUM | tests/CMakeLists.txt | 296 | # ----------------------------------------------------------------------------- | COMMENT |
| MEDIUM | tests/CMakeLists.txt | 298 | # ----------------------------------------------------------------------------- | COMMENT |
| MEDIUM | tests/CMakeLists.txt | 312 | # ----------------------------------------------------------------------------- | COMMENT |
| MEDIUM | tests/CMakeLists.txt | 314 | # ----------------------------------------------------------------------------- | COMMENT |
| MEDIUM | tests/CMakeLists.txt | 328 | # ----------------------------------------------------------------------------- | COMMENT |
| MEDIUM | tests/CMakeLists.txt | 330 | # ----------------------------------------------------------------------------- | COMMENT |
| MEDIUM | tests/CMakeLists.txt | 352 | # ----------------------------------------------------------------------------- | COMMENT |
| MEDIUM | tests/CMakeLists.txt | 354 | # ----------------------------------------------------------------------------- | COMMENT |
| MEDIUM | tests/CMakeLists.txt | 399 | # ----------------------------------------------------------------------------- | COMMENT |
| MEDIUM | tests/CMakeLists.txt | 401 | # ----------------------------------------------------------------------------- | COMMENT |
| MEDIUM | tests/CMakeLists.txt | 418 | # ----------------------------------------------------------------------------- | COMMENT |
| MEDIUM | tests/CMakeLists.txt | 420 | # ----------------------------------------------------------------------------- | COMMENT |
| MEDIUM | tests/CMakeLists.txt | 440 | # ----------------------------------------------------------------------------- | COMMENT |
| MEDIUM | tests/CMakeLists.txt | 442 | # ----------------------------------------------------------------------------- | COMMENT |
| MEDIUM | tests/CMakeLists.txt | 459 | # ----------------------------------------------------------------------------- | COMMENT |
| MEDIUM | tests/CMakeLists.txt | 476 | # ----------------------------------------------------------------------------- | COMMENT |
| MEDIUM | tests/CMakeLists.txt | 478 | # ----------------------------------------------------------------------------- | COMMENT |
| MEDIUM | tests/CMakeLists.txt | 501 | # ----------------------------------------------------------------------------- | COMMENT |
| MEDIUM | tests/CMakeLists.txt | 504 | # ----------------------------------------------------------------------------- | COMMENT |
| 395 more matches not shown… | ||||
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| LOW | scripts/intake.py | 159 | def _floating_dtype_distribution(dist: dict[str, int]) -> dict[str, int]: | CODE |
| LOW | scripts/intake.py | 190 | def _safetensors_dtype_distribution(repo: str, revision: str) -> dict[str, int]: | CODE |
| LOW | scripts/convert-funasr_nano.py | 140 | def _patch_fun_asr_nano_imports() -> None: | CODE |
| LOW⚡ | scripts/dump_reference_parakeet_nemo.py | 115 | def _patch_conformer_for_offline(force_regular_att_style: bool = False) -> None: | CODE |
| LOW | scripts/dump_reference_parakeet_nemo.py | 458 | def apply_att_context_right_override(model, att_context_right) -> None: | CODE |
| LOW | scripts/dump_reference_parakeet_nemo.py | 680 | def _dump_transducer_first_step(model, enc_t, dump): | CODE |
| LOW | scripts/convert-granite5_ctc.py | 358 | def emit_mel_filterbank_and_window(writer: GGUFWriter, hp: dict) -> None: | CODE |
| LOW | scripts/dump_reference_canary_nemo.py | 214 | def capture_encoder_intermediates(model, blocks: list[int]): | CODE |
| LOW | scripts/dump_reference_canary_nemo.py | 252 | def capture_decoder_intermediates(model, dec_blocks: list[int]): | CODE |
| LOW | scripts/convert-gigaam.py | 245 | def extract_charwise_tokenizer(vocab: list[str], blank_piece: str = "<blank>") -> dict: | CODE |
| LOW | scripts/dump_reference_funasr_nano_funasr.py | 72 | def _patch_fun_asr_nano_imports() -> None: | CODE |
| LOW | scripts/validate.py | 57 | def normalize_text_for_compare(text: str) -> str: | CODE |
| LOW | scripts/convert-parakeet.py | 505 | def _load_nemo_archive_directly(nemo_path: Path): | CODE |
| LOW | scripts/convert-parakeet.py | 794 | def load_hf_safetensors_model(model_spec: str, revision: str | None) -> _HfSafetensorsModel: | CODE |
| LOW | scripts/convert-parakeet.py | 936 | def _resolve_att_context_size(raw) -> tuple[int, int]: | CODE |
| LOW | scripts/convert-parakeet.py | 963 | def _resolve_att_context_size_choices(raw) -> list[tuple[int, int]]: | CODE |
| LOW | scripts/convert-parakeet.py | 991 | def _resolve_att_chunk_context_size(raw) -> tuple[list[int], list[int], list[int]]: | CODE |
| LOW | scripts/convert-parakeet.py | 1020 | def _resolve_conv_context_size(raw, kernel_size: int) -> tuple[int, int]: | CODE |
| LOW | scripts/convert-granite.py | 141 | def granite_translation_pairs(asr_langs: list[str]) -> list[str]: | CODE |
| LOW | scripts/validate_streaming.py | 199 | def expand_pattern_tolerances(ref_dir: Path, pattern_path: Path, | CODE |
| LOW | scripts/dump_reference_canary_qwen_nemo.py | 324 | def register_perception_hooks(perception_parts: dict, blocks: list[int]) -> dict[str, CaptureHook]: | CODE |
| LOW | scripts/preflight.py | 178 | def _expected_dtype_from_intake(data: dict[str, Any]) -> str | None: | CODE |
| LOW | scripts/preflight.py | 302 | def _nemo_preprocessor_to_pre(model_config: dict) -> dict | None: | CODE |
| LOW | scripts/preflight.py | 532 | def _frontend_from_preprocessor(pre: dict) -> dict[str, Any]: | CODE |
| LOW | scripts/preflight.py | 560 | def _canonical_normalize_for_compare(raw): | CODE |
| LOW | scripts/preflight.py | 680 | def check_architecture_sanity(declared, gguf_kvs, reference, gate: Gate) -> CheckResult: | CODE |
| LOW | scripts/convert-granite_nar.py | 248 | def emit_mel_filterbank_and_window(writer: GGUFWriter, hp: dict) -> None: | CODE |
| LOW | scripts/bench/run.py | 222 | def parse_quant_from_filename(path: Path) -> str | None: | CODE |
| LOW⚡ | scripts/catalog/test_db_mapping.py | 40 | def test_accuracy_rows_reach_the_database(): | CODE |
| LOW⚡ | scripts/catalog/test_db_mapping.py | 44 | def test_speed_rows_reach_the_database(): | CODE |
| LOW⚡ | scripts/catalog/test_db_mapping.py | 48 | def test_download_rows_reach_the_database(): | CODE |
| LOW | scripts/catalog/profiles.py | 97 | def canonical_fleurs_language(tag: str, record: dict) -> str | None: | CODE |
| LOW | scripts/catalog/profiles.py | 236 | def has_measurement_provenance(row: dict) -> bool: | CODE |
| LOW | scripts/lib/gguf_common.py | 344 | def declare_missing_capabilities(writer) -> list[str]: | CODE |
| LOW | scripts/lib/test_gguf_writer.py | 50 | def test_move_reorders_bulk_to_end(): | CODE |
| LOW | scripts/lib/test_gguf_writer.py | 65 | def test_move_is_noop_without_bulk_keys(): | CODE |
| LOW | scripts/lib/test_gguf_writer.py | 75 | def test_split_scalars_precede_trailer(): | CODE |
| LOW | scripts/lib/test_gguf_writer.py | 92 | def test_factory_writes_trailer_on_disk(): | CODE |
| LOW | scripts/lib/test_gguf_writer.py | 114 | def test_bulk_keys_are_tokenizer_only(): | CODE |
| LOW | scripts/lib/test_quant_policy_sync.py | 219 | def test_linear_keeps_reference_dtype() -> None: | CODE |
| LOW | scripts/lib/test_quant_policy_sync.py | 226 | def test_known_drift_is_pinned() -> None: | CODE |
| LOW | scripts/hf_cards/generate.py | 251 | def build_transcribe_cpp_block(ctx: dict) -> str: | CODE |
| LOW | scripts/wer/run_reference_funasr_nano.py | 36 | def _patch_fun_asr_nano_imports() -> None: | CODE |
| LOW | scripts/wer/test_der.py | 14 | def test_matches_pyannote_tutorial_fixture(self) -> None: | CODE |
| LOW⚡ | scripts/wer/test_der.py | 31 | def test_permuted_labels_are_free(self) -> None: | CODE |
| LOW⚡ | scripts/wer/test_der.py | 40 | def test_confusion_uses_optimal_one_to_one_mapping(self) -> None: | CODE |
| LOW⚡ | scripts/wer/test_der.py | 50 | def test_miss_and_false_alarm_include_nonspeech(self) -> None: | CODE |
| LOW⚡ | scripts/wer/test_der.py | 59 | def test_overlap_uses_speaker_time(self) -> None: | CODE |
| LOW⚡ | scripts/wer/test_der.py | 66 | def test_collar_excludes_both_sides_of_reference_boundaries(self) -> None: | CODE |
| LOW | scripts/wer/test_der.py | 78 | def test_untimed_sentinel_is_rejected(self) -> None: | CODE |
| LOW | scripts/wer/test_der.py | 92 | def test_sanitize_clamps_hypothesis_to_duration(self) -> None: | CODE |
| LOW | scripts/wer/test_der.py | 111 | def test_sanitize_drops_degenerate_hypothesis_rows(self) -> None: | CODE |
| LOW | scripts/wer/remote/subprocess_io.py | 9 | def run_subprocess_capturing_stderr( | CODE |
| LOW | scripts/wer/remote/modal_sweep.py | 317 | def _prepare_manifest_for_cell( | CODE |
| LOW | scripts/wer/remote/modal_sweep.py | 335 | def _subset_manifest_for_cell( | CODE |
| LOW | scripts/wer/remote/modal_sweep.py | 824 | def _register_reference_runner(gpu_id: str): | CODE |
| LOW | bindings/python/tests/test_compute_lock.py | 151 | def test_calls_on_two_sessions_never_overlap(fake, monkeypatch): | CODE |
| LOW | bindings/python/tests/test_compute_lock.py | 186 | def test_every_compute_site_holds_the_lock(fake, monkeypatch): | CODE |
| LOW | bindings/python/tests/test_compute_lock.py | 262 | def test_cancel_while_queued_is_kept(fake, monkeypatch): | CODE |
| LOW | bindings/python/tests/test_compute_lock.py | 294 | def test_session_close_during_in_flight_call_defers_free(fake, monkeypatch): | CODE |
| 176 more matches not shown… | ||||
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| LOW | scripts/intake.py | 24 | CODE | |
| LOW | scripts/dump_reference_moss_author.py | 42 | CODE | |
| LOW | scripts/convert-funasr_nano.py | 88 | CODE | |
| LOW | scripts/convert-funasr_nano.py | 143 | CODE | |
| LOW | scripts/convert-canary-qwen.py | 69 | CODE | |
| LOW | scripts/convert-silero_vad.py | 17 | CODE | |
| LOW | scripts/convert-moonshine_streaming.py | 52 | CODE | |
| LOW | scripts/convert-voxtral_realtime.py | 43 | CODE | |
| LOW | scripts/compose-multitalker-bundle.py | 44 | CODE | |
| LOW | scripts/dump_reference_whisper_transformers.py | 26 | CODE | |
| LOW | scripts/batch_parity.py | 42 | CODE | |
| LOW | scripts/compare_tensors.py | 61 | CODE | |
| LOW | scripts/dump_reference_voxtral_realtime_transformers.py | 68 | CODE | |
| LOW | scripts/convert-moss.py | 76 | CODE | |
| LOW | scripts/dump_reference_sortformer_nemo.py | 46 | CODE | |
| LOW | scripts/dump_reference_multitalker_nemo.py | 32 | CODE | |
| LOW | scripts/validate_buffered_streaming.py | 46 | CODE | |
| LOW | scripts/convert-moonshine.py | 142 | CODE | |
| LOW | scripts/convert-sortformer.py | 30 | CODE | |
| LOW | scripts/batch_tensor_parity.py | 28 | CODE | |
| LOW | scripts/audit_gguf_metadata.py | 28 | CODE | |
| LOW | scripts/audit_gguf_metadata.py | 32 | CODE | |
| LOW | scripts/dump_reference_voxtral_transformers.py | 45 | CODE | |
| LOW | scripts/quantize-all.py | 34 | CODE | |
| LOW | scripts/convert-cohere.py | 72 | CODE | |
| LOW | scripts/dump_reference_parakeet_nemo.py | 39 | CODE | |
| LOW | scripts/convert-granite5_ctc.py | 61 | CODE | |
| LOW | scripts/dump_reference_canary_nemo.py | 62 | CODE | |
| LOW | scripts/whisper_long_form_parity.py | 28 | CODE | |
| LOW | scripts/convert-gigaam.py | 81 | CODE | |
| LOW | scripts/convert-gigaam.py | 84 | CODE | |
| LOW | scripts/dump_reference_gigaam_author.py | 41 | CODE | |
| LOW | scripts/dump_reference_funasr_nano_funasr.py | 49 | CODE | |
| LOW | scripts/dump_reference_funasr_nano_funasr.py | 53 | CODE | |
| LOW | scripts/dump_reference_funasr_nano_funasr.py | 74 | CODE | |
| LOW | scripts/validate.py | 37 | CODE | |
| LOW | scripts/dump_reference_sensevoice_funasr.py | 57 | CODE | |
| LOW | scripts/dump_reference_sensevoice_funasr.py | 60 | CODE | |
| LOW | scripts/dump_reference_sensevoice_funasr.py | 63 | CODE | |
| LOW | scripts/convert-parakeet.py | 73 | CODE | |
| LOW | scripts/dump_reference_ecapa_tdnn_speechbrain.py | 18 | CODE | |
| LOW | scripts/convert-medasr.py | 59 | CODE | |
| LOW | scripts/dump_reference_medasr_transformers.py | 72 | CODE | |
| LOW | scripts/quant_accuracy.py | 61 | CODE | |
| LOW | scripts/convert-granite.py | 89 | CODE | |
| LOW | scripts/convert-sensevoice.py | 90 | CODE | |
| LOW | scripts/convert-sensevoice.py | 93 | CODE | |
| LOW | scripts/validate_streaming.py | 46 | CODE | |
| LOW | scripts/convert-ecapa_tdnn.py | 14 | CODE | |
| LOW | scripts/dump_reference_canary_qwen_nemo.py | 58 | CODE | |
| LOW | scripts/dump_reference_canary_qwen_nemo.py | 61 | CODE | |
| LOW | scripts/dump_reference_silero_vad_author.py | 23 | CODE | |
| LOW | scripts/dump_reference_qwen3_asr_author.py | 36 | CODE | |
| LOW | scripts/dump_reference_qwen3_asr_author.py | 40 | CODE | |
| LOW | scripts/dump_reference_granite_transformers.py | 40 | CODE | |
| LOW | scripts/dump_reference_granite_transformers.py | 43 | CODE | |
| LOW | scripts/dump_reference_granite_transformers.py | 44 | CODE | |
| LOW | scripts/dump_reference_granite5_ctc_transformers.py | 86 | CODE | |
| LOW | scripts/convert-canary.py | 57 | CODE | |
| LOW | scripts/convert-voxtral.py | 34 | CODE | |
| 142 more matches not shown… | ||||
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| LOW | tools/transcribe-quantize/main.cpp | 427 | // Step 2: quantize fp32 → dst_type into the destination | COMMENT |
| LOW | .claude/skills/porting-6-bench/SKILL.md | 85 | ### Step 0: Catalog record (execute, first port only) | COMMENT |
| LOW⚡ | .claude/skills/porting-6-bench/SKILL.md | 109 | ### Step 1: Matrix presence (execute) | COMMENT |
| LOW⚡ | .claude/skills/porting-6-bench/SKILL.md | 118 | ### Step 2: Rebuild transcribe-bench (execute) | COMMENT |
| LOW⚡ | .claude/skills/porting-6-bench/SKILL.md | 127 | ### Step 3: Bench scope (ask-point) | COMMENT |
| LOW | .claude/skills/porting-6-bench/SKILL.md | 145 | ### Step 4: Baseline capture (execute) | COMMENT |
| LOW | .claude/skills/porting-6-bench/SKILL.md | 179 | ### Step 5: Iteration loop (human-driven) | COMMENT |
| LOW | .claude/skills/porting-6-bench/SKILL.md | 219 | ### Step 6: Sign-off | COMMENT |
| LOW | .claude/skills/porting-2-oracle/SKILL.md | 59 | ### Step 1: Per-family env and dumper | COMMENT |
| LOW | .claude/skills/porting-2-oracle/SKILL.md | 74 | ### Step 2: Complete the golden manifest (execute) | COMMENT |
| LOW | .claude/skills/porting-2-oracle/SKILL.md | 93 | ### Step 3: Reference dumps (execute) | COMMENT |
| LOW | .claude/skills/porting-2-oracle/SKILL.md | 120 | ### Step 4: Reference transcripts (execute) | COMMENT |
| LOW | .claude/skills/porting-2-oracle/SKILL.md | 134 | ### Step 5: Generate dump_coverage.json (execute) | COMMENT |
| LOW | .claude/skills/porting-2-oracle/SKILL.md | 152 | ### Step 6: Provisional tolerances from per-tensor magnitude (execute) | COMMENT |
| LOW | .claude/skills/porting-2-oracle/SKILL.md | 177 | ### Step 7: Reference WER baseline (execute) | COMMENT |
| LOW | .claude/skills/porting-2-oracle/SKILL.md | 229 | ### Step 8: Sign-off | COMMENT |
| LOW⚡ | .claude/skills/porting-5-quants/SKILL.md | 38 | ### Step 1: Build the quantizer (execute) | COMMENT |
| LOW⚡ | .claude/skills/porting-5-quants/SKILL.md | 44 | ### Step 2: Run quantize-all (execute) | COMMENT |
| LOW⚡ | .claude/skills/porting-5-quants/SKILL.md | 54 | ### Step 3: CLI output-validity smoke (execute) | COMMENT |
| LOW | .claude/skills/porting-5-quants/SKILL.md | 73 | ### Step 4: Publish quants to a private HF repo (execute) | COMMENT |
| LOW | .claude/skills/porting-5-quants/SKILL.md | 84 | ### Step 5: Preliminary 512-utterance WER sweep (execute) | COMMENT |
| LOW | .claude/skills/porting-5-quants/SKILL.md | 111 | ### Step 6: Sign-off | COMMENT |
| LOW | .claude/skills/porting-1-intake/SKILL.md | 34 | ### Step 1: Mechanical draft (execute) | COMMENT |
| LOW | .claude/skills/porting-1-intake/SKILL.md | 48 | ### Step 2: Human-judgment fields | COMMENT |
| LOW | .claude/skills/porting-1-intake/SKILL.md | 104 | ### Step 3: Schema validation (execute) | COMMENT |
| LOW⚡ | .claude/skills/porting-1-intake/SKILL.md | 115 | ### Step 4: Preflight Gate A (execute) | COMMENT |
| LOW⚡ | .claude/skills/porting-1-intake/SKILL.md | 123 | ### Step 5: Family doc draft (execute) | COMMENT |
| LOW⚡ | .claude/skills/porting-1-intake/SKILL.md | 132 | ### Step 6: Golden manifest skeleton (execute) | COMMENT |
| LOW | .claude/skills/porting-1-intake/SKILL.md | 154 | ### Step 7: Sign-off | COMMENT |
| LOW | .claude/skills/porting-8-ship/SKILL.md | 46 | ### Step 1: Verify inputs and GGUF metadata (execute) | COMMENT |
| LOW | .claude/skills/porting-8-ship/SKILL.md | 117 | ### Step 2: Ingest measurements (execute) | COMMENT |
| LOW | .claude/skills/porting-8-ship/SKILL.md | 131 | ### Step 3: Validate the publication profile (execute) | COMMENT |
| LOW | .claude/skills/porting-8-ship/SKILL.md | 141 | ### Step 4: Author prose (execute + ask-point) | COMMENT |
| LOW | .claude/skills/porting-8-ship/SKILL.md | 254 | ### Step 5: Render docs and card (execute) | COMMENT |
| LOW | .claude/skills/porting-8-ship/SKILL.md | 268 | ### Step 6: Verify no drift (execute) | COMMENT |
| LOW | .claude/skills/porting-8-ship/SKILL.md | 279 | ### Step 7: Human review (ask-point) | COMMENT |
| LOW | .claude/skills/porting-8-ship/SKILL.md | 290 | ### Step 8: Upload (execute) | COMMENT |
| LOW | .claude/skills/porting-4-cpp/SKILL.md | 62 | ### Step 0: Family-level forward-map.md (execute) | COMMENT |
| LOW | .claude/skills/porting-4-cpp/SKILL.md | 84 | ### Step 1: Sibling-variant shortcut (execute) | COMMENT |
| LOW | .claude/skills/porting-4-cpp/SKILL.md | 101 | ### Step 2: Implement src/arch/<family>/ — drive from the forward map | COMMENT |
| LOW | .claude/skills/porting-4-cpp/SKILL.md | 165 | ### Step 3: First validate.py run (execute) | COMMENT |
| LOW | .claude/skills/porting-4-cpp/SKILL.md | 189 | ### Step 4: Finalize tolerances (execute) | COMMENT |
| LOW | .claude/skills/porting-4-cpp/SKILL.md | 243 | ### Step 5: Human review of tolerances | COMMENT |
| LOW | .claude/skills/porting-4-cpp/SKILL.md | 254 | ### Step 6: Ref-dtype validation green (execute) | COMMENT |
| LOW | .claude/skills/porting-4-cpp/SKILL.md | 264 | ### Step 7: Frontend parity (execute) | COMMENT |
| LOW | .claude/skills/porting-4-cpp/SKILL.md | 283 | ### Step 8: Run the family-doc Capability Validation table (execute) | COMMENT |
| LOW | .claude/skills/porting-4-cpp/SKILL.md | 320 | ### Step 9: Batch parity (execute) | COMMENT |
| LOW | .claude/skills/porting-4-cpp/SKILL.md | 361 | ### Step 10: Publish ref-dtype to private HF repo (execute) | COMMENT |
| LOW | .claude/skills/porting-4-cpp/SKILL.md | 372 | ### Step 11: Full ref-dtype WER gate (execute) | COMMENT |
| LOW | .claude/skills/porting-4-cpp/SKILL.md | 407 | ### Step 12: Sign-off | COMMENT |
| LOW | .claude/skills/porting-3-convert/SKILL.md | 32 | ### Step 1: Per-family converter | COMMENT |
| LOW | .claude/skills/porting-3-convert/SKILL.md | 72 | ### Step 2: Identify reference dtype (read intake) | COMMENT |
| LOW | .claude/skills/porting-3-convert/SKILL.md | 86 | ### Step 3: Run the converter (execute) | COMMENT |
| LOW | .claude/skills/porting-3-convert/SKILL.md | 108 | ### Step 4: Converter manifest (execute) | COMMENT |
| LOW | .claude/skills/porting-3-convert/SKILL.md | 140 | ### Step 5: Structural checks (execute) | COMMENT |
| LOW | .claude/skills/porting-3-convert/SKILL.md | 164 | ### Step 6: Sign-off | COMMENT |
| LOW | .claude/skills/porting-7-wer/SKILL.md | 43 | ### Step 1: Acceptance manifest (execute or ask-point) | COMMENT |
| LOW | .claude/skills/porting-7-wer/SKILL.md | 73 | ### Step 2: Run the publication profile (execute) | COMMENT |
| LOW | .claude/skills/porting-7-wer/SKILL.md | 96 | ### Step 3: Ref-dtype WER limit (execute) | COMMENT |
| LOW | .claude/skills/porting-7-wer/SKILL.md | 153 | ### Step 4: Acceptance cells outside the profile (execute) | COMMENT |
| 49 more matches not shown… | ||||
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| LOW | scripts/intake.py | 196 | except Exception: | CODE |
| LOW⚡ | scripts/convert-funasr_nano.py | 151 | except Exception as e: | CODE |
| LOW | scripts/compare_tensors.py | 154 | except Exception as e: | CODE |
| LOW | scripts/compare_tensors.py | 158 | except Exception as e: | CODE |
| LOW | scripts/dump_reference_voxtral_realtime_transformers.py | 325 | except Exception: # noqa: BLE001 - fall through to next decode form | CODE |
| LOW | scripts/audit_gguf_metadata.py | 223 | except Exception as e: # noqa: BLE001 - report and continue the sweep | CODE |
| LOW | scripts/audit_gguf_metadata.py | 230 | except Exception as e: # noqa: BLE001 | CODE |
| LOW | scripts/dump_reference_voxtral_transformers.py | 272 | except Exception: # noqa: BLE001 - fall through to next decode form | CODE |
| LOW | scripts/quantize-all.py | 72 | except Exception: | CODE |
| LOW | scripts/dump_reference_parakeet_nemo.py | 224 | except Exception as e2: | CODE |
| LOW | scripts/dump_reference_parakeet_nemo.py | 242 | except Exception as e2: | CODE |
| LOW | scripts/dump_reference_parakeet_nemo.py | 335 | except Exception: | CODE |
| LOW | scripts/dump_reference_parakeet_nemo.py | 505 | except Exception as e: | CODE |
| LOW | scripts/dump_reference_parakeet_nemo.py | 792 | except Exception as e: | CODE |
| LOW | scripts/dump_reference_canary_nemo.py | 551 | except Exception as exc: | CODE |
| LOW | scripts/convert-parakeet.py | 593 | except Exception as e: | CODE |
| LOW | scripts/convert-parakeet.py | 603 | except Exception as e: | CODE |
| LOW | scripts/validate_streaming.py | 186 | except Exception: | CODE |
| LOW⚡ | scripts/dump_reference_canary_qwen_nemo.py | 477 | except Exception as exc: | CODE |
| LOW⚡ | scripts/dump_reference_canary_qwen_nemo.py | 483 | except Exception as exc: | CODE |
| LOW | scripts/dump_reference_canary_qwen_nemo.py | 535 | except Exception: | CODE |
| LOW | scripts/convert-canary.py | 204 | except Exception: | CODE |
| LOW | scripts/diar/ingest_ami_seglst.py | 72 | except Exception as exc: | CODE |
| LOW | scripts/catalog/sync_capabilities.py | 156 | except Exception as exc: # noqa: BLE001 - any transport failure is just "unavailable" | CODE |
| LOW | scripts/wer/run_reference_funasr_nano.py | 160 | except Exception: | CODE |
| LOW | scripts/wer/run_reference_funasr_nano.py | 212 | except Exception as e: | CODE |
| LOW | scripts/wer/run.py | 367 | except Exception: | CODE |
| LOW | scripts/wer/run.py | 446 | except Exception as e: | CODE |
| LOW | scripts/wer/run_reference_medasr_transformers.py | 145 | except Exception as e: | CODE |
| LOW | scripts/wer/run_reference_granite_transformers.py | 110 | except Exception: | CODE |
| LOW⚡ | scripts/wer/run_reference_granite_transformers.py | 293 | except Exception as e: | CODE |
| LOW⚡ | scripts/wer/run_reference_granite_transformers.py | 299 | except Exception: | CODE |
| LOW⚡ | scripts/wer/run_reference_granite_transformers.py | 304 | except Exception as e2: | CODE |
| LOW | scripts/wer/run_reference_granite5_ctc_transformers.py | 159 | except Exception as e: | CODE |
| LOW | scripts/wer/run_reference_parakeet_streaming_nemo.py | 103 | except Exception as e2: | CODE |
| LOW | scripts/wer/run_reference_parakeet_streaming_nemo.py | 130 | except Exception as e: | CODE |
| LOW | scripts/wer/run_reference_parakeet_streaming_nemo.py | 247 | except Exception as e: | CODE |
| LOW | scripts/wer/run_reference_cohere_transformers.py | 188 | except Exception: | CODE |
| LOW | scripts/wer/run_reference_cohere_transformers.py | 198 | except Exception as e2: | CODE |
| LOW | scripts/wer/run_reference_sensevoice.py | 139 | except Exception as e: | CODE |
| LOW | scripts/wer/run_reference_voxtral_transformers.py | 116 | except Exception: # noqa: BLE001 | CODE |
| LOW⚡ | scripts/wer/run_reference_voxtral_transformers.py | 190 | except Exception as e: # noqa: BLE001 | CODE |
| LOW⚡ | scripts/wer/run_reference_voxtral_transformers.py | 196 | except Exception: | CODE |
| LOW⚡ | scripts/wer/run_reference_voxtral_transformers.py | 200 | except Exception as e2: # noqa: BLE001 | CODE |
| LOW | …pts/wer/run_reference_voxtral_realtime_transformers.py | 144 | except Exception: # noqa: BLE001 | CODE |
| LOW⚡ | …pts/wer/run_reference_voxtral_realtime_transformers.py | 250 | except Exception as e: # noqa: BLE001 | CODE |
| LOW⚡ | …pts/wer/run_reference_voxtral_realtime_transformers.py | 256 | except Exception: | CODE |
| LOW⚡ | …pts/wer/run_reference_voxtral_realtime_transformers.py | 260 | except Exception as e2: # noqa: BLE001 | CODE |
| LOW | scripts/wer/run_reference_moss_author.py | 169 | except Exception as e: | CODE |
| LOW | scripts/wer/run_reference_parakeet_nemo.py | 155 | except Exception as e2: | CODE |
| LOW | scripts/wer/run_reference_parakeet_nemo.py | 274 | except Exception: | CODE |
| LOW | scripts/wer/run_reference_parakeet_nemo.py | 283 | except Exception as e2: | CODE |
| LOW | scripts/wer/run_reference_gigaam_author.py | 116 | except Exception as e: | CODE |
| LOW | scripts/wer/run_reference_canary_qwen_nemo.py | 112 | except Exception as exc: | CODE |
| LOW | scripts/wer/run_reference_canary_qwen_nemo.py | 118 | except Exception as exc: | CODE |
| LOW | scripts/wer/run_reference_canary_qwen_nemo.py | 168 | except Exception as e: | CODE |
| LOW | scripts/wer/run_reference_qwen3_asr_author.py | 183 | except Exception as e: | CODE |
| LOW | scripts/wer/run_reference_canary_nemo.py | 77 | except Exception as exc: | CODE |
| LOW | scripts/wer/run_reference_canary_nemo.py | 156 | except Exception as e: | CODE |
| LOW | scripts/wer/run_reference_granite_nar_transformers.py | 142 | except Exception as e: | CODE |
| 32 more matches not shown… | ||||
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| HIGH | scripts/dump_reference_parakeet_nemo.py | 299 | except (ImportError, AttributeError): | CODE |
| HIGH | scripts/dump_reference_parakeet_nemo.py | 1048 | except AttributeError: | CODE |
| HIGH | scripts/dump_reference_canary_nemo.py | 189 | except (ImportError, AttributeError): | CODE |
| HIGH | scripts/validate.py | 300 | except OSError: | CODE |
| HIGH | scripts/dump_reference_ecapa_tdnn_speechbrain.py | 50 | except RuntimeError: | CODE |
| HIGH | scripts/dump_reference_canary_qwen_nemo.py | 198 | except (ImportError, AttributeError): | CODE |
| HIGH | scripts/dump_reference_canary_qwen_nemo.py | 535 | except Exception: | CODE |
| HIGH | scripts/preflight.py | 112 | except (UnicodeDecodeError, TypeError, IndexError): | CODE |
| HIGH | …/wer/run_reference_parakeet_buffered_streaming_nemo.py | 112 | except AttributeError: | CODE |
| HIGH | scripts/wer/remote/modal_sweep.py | 753 | except Exception: | CODE |
| HIGH | bindings/python/tests/test_device_select.py | 19 | except transcribe_cpp.BackendError: | CODE |
| HIGH | bindings/python/_generate/generate.py | 101 | except Exception: | CODE |
| HIGH | bindings/python/src/transcribe_cpp/__init__.py | 409 | except Exception: | CODE |
| HIGH | bindings/python/src/transcribe_cpp/__init__.py | 1395 | except Exception: | CODE |
| HIGH | bindings/python/src/transcribe_cpp/__init__.py | 1480 | except Exception: | CODE |
| HIGH | bindings/python/src/transcribe_cpp/__init__.py | 1963 | except Exception: | CODE |
| HIGH | bindings/python/src/transcribe_cpp/__init__.py | 1157 | except Exception: | CODE |
| HIGH | …hon-native-cu12/transcribe_cpp_native_cu12/__init__.py | 74 | except OSError: | CODE |
| HIGH | …hon-native-cu12/transcribe_cpp_native_cu12/__init__.py | 70 | except OSError: | CODE |
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| HIGH | scripts/dump_reference_moss_author.py | 0 | record the first output tensor a module emits during forward. | STRING |
| HIGH | scripts/dump_reference_voxtral_transformers.py | 0 | record the first output tensor a module emits during forward. | STRING |
| HIGH | scripts/dump_reference_gigaam_author.py | 0 | record the first output tensor a module emits during forward. | STRING |
| HIGH | scripts/dump_reference_funasr_nano_funasr.py | 0 | record the first output tensor a module emits during forward. | STRING |
| HIGH | scripts/dump_reference_sensevoice_funasr.py | 0 | record the first output tensor a module emits during forward. | STRING |
| HIGH | scripts/dump_reference_canary_qwen_nemo.py | 0 | record the first output tensor a module emits during forward. | STRING |
| HIGH | scripts/dump_reference_qwen3_asr_author.py | 0 | record the first output tensor a module emits during forward. | STRING |
| HIGH | scripts/dump_reference_granite_transformers.py | 0 | record the first output tensor a module emits during forward. | STRING |
| HIGH | scripts/dump_reference_voxtral_realtime_transformers.py | 0 | resolve a model argument to (model_id_or_path, local_files_only). | STRING |
| HIGH | scripts/dump_reference_voxtral_transformers.py | 0 | resolve a model argument to (model_id_or_path, local_files_only). | STRING |
| HIGH | scripts/dump_reference_granite_transformers.py | 0 | resolve a model argument to (model_id_or_path, local_files_only). | STRING |
| HIGH | scripts/dump_reference_medasr_transformers.py | 0 | attach hooks at every dump point. returns a name -> hook dict. | STRING |
| HIGH | scripts/dump_reference_granite5_ctc_transformers.py | 0 | attach hooks at every dump point. returns a name -> hook dict. | STRING |
| HIGH | scripts/dump_reference_granite_nar_transformers.py | 0 | attach hooks at every dump point. returns a name -> hook dict. | STRING |
| HIGH | ggml/examples/gpt-2/convert-ckpt-to-ggml.py | 0 | returns list of utf-8 byte and a corresponding list of unicode strings. the reversible bpe codes work on unicode strings | STRING |
| HIGH | ggml/examples/gpt-2/convert-h5-to-ggml.py | 0 | returns list of utf-8 byte and a corresponding list of unicode strings. the reversible bpe codes work on unicode strings | STRING |
| HIGH | ggml/examples/gpt-2/convert-cerebras-to-ggml.py | 0 | returns list of utf-8 byte and a corresponding list of unicode strings. the reversible bpe codes work on unicode strings | STRING |
| HIGH | ggml/examples/gpt-j/convert-h5-to-ggml.py | 0 | returns list of utf-8 byte and a corresponding list of unicode strings. the reversible bpe codes work on unicode strings | STRING |
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| LOW | scripts/convert-canary-qwen.py | 602 | CODE | |
| LOW | scripts/tokenizer-parity-fixture.py | 66 | CODE | |
| LOW | scripts/convert-moonshine_streaming.py | 122 | CODE | |
| LOW | scripts/convert-moonshine_streaming.py | 679 | CODE | |
| LOW | scripts/dump_reference_whisper_transformers.py | 402 | CODE | |
| LOW | scripts/compare_tensors.py | 233 | CODE | |
| LOW | scripts/dump_reference_multitalker_nemo.py | 110 | CODE | |
| LOW | scripts/validate_buffered_streaming.py | 159 | CODE | |
| LOW | scripts/validate_buffered_streaming.py | 218 | CODE | |
| LOW | scripts/convert-moonshine.py | 229 | CODE | |
| LOW | scripts/convert-moonshine.py | 789 | CODE | |
| LOW | scripts/audit_gguf_metadata.py | 158 | CODE | |
| LOW | scripts/convert-cohere.py | 119 | CODE | |
| LOW | scripts/convert-cohere.py | 411 | CODE | |
| LOW | scripts/dump_reference_parakeet_nemo.py | 201 | CODE | |
| LOW | scripts/dump_reference_parakeet_nemo.py | 339 | CODE | |
| LOW | scripts/dump_reference_canary_nemo.py | 424 | CODE | |
| LOW | scripts/convert-gigaam.py | 196 | CODE | |
| LOW | scripts/validate.py | 627 | CODE | |
| LOW | scripts/validate.py | 900 | CODE | |
| LOW | scripts/convert-parakeet.py | 794 | CODE | |
| LOW | scripts/convert-parakeet.py | 861 | CODE | |
| LOW | scripts/convert-parakeet.py | 1579 | CODE | |
| LOW | scripts/convert-medasr.py | 205 | CODE | |
| LOW | scripts/convert-granite.py | 979 | CODE | |
| LOW | scripts/convert-sensevoice.py | 225 | CODE | |
| LOW | scripts/validate_streaming.py | 231 | CODE | |
| LOW | scripts/validate_streaming.py | 323 | CODE | |
| LOW | scripts/dump_reference_canary_qwen_nemo.py | 324 | CODE | |
| LOW | scripts/convert-qwen3_asr.py | 769 | CODE | |
| LOW | scripts/convert-whisper.py | 794 | CODE | |
| LOW | scripts/preflight.py | 219 | CODE | |
| LOW | scripts/preflight.py | 277 | CODE | |
| LOW | scripts/bench/run.py | 177 | CODE | |
| LOW | scripts/bench/run.py | 428 | CODE | |
| LOW | scripts/diar/run_cpp_multitalker.py | 34 | CODE | |
| LOW | scripts/ci/vulkan_degradation_check.py | 60 | CODE | |
| LOW | scripts/ci/extract_native_bundle.py | 49 | CODE | |
| LOW | scripts/ci/check_lane_mirror.py | 118 | CODE | |
| LOW | scripts/ci/build_wheel_index.py | 72 | CODE | |
| LOW | scripts/catalog/ingest_accuracy.py | 135 | CODE | |
| LOW | scripts/catalog/check.py | 118 | CODE | |
| LOW | scripts/catalog/check.py | 243 | CODE | |
| LOW | scripts/catalog/render.py | 107 | CODE | |
| LOW | scripts/catalog/render.py | 446 | CODE | |
| LOW | scripts/catalog/ingest_perf.py | 145 | CODE | |
| LOW | scripts/catalog/ingest_perf.py | 200 | CODE | |
| LOW | scripts/catalog/profiles.py | 145 | CODE | |
| LOW | scripts/catalog/profiles.py | 216 | CODE | |
| LOW | scripts/langid/run.py | 134 | CODE | |
| LOW | scripts/langid/bench.py | 47 | CODE | |
| LOW | scripts/lib/gguf_common.py | 306 | CODE | |
| LOW | scripts/wer/run_reference_funasr_nano.py | 65 | CODE | |
| LOW | scripts/wer/run.py | 199 | CODE | |
| LOW | scripts/wer/ingest.py | 112 | CODE | |
| LOW | scripts/wer/run_reference_medasr_transformers.py | 29 | CODE | |
| LOW | scripts/wer/run_reference_granite_transformers.py | 37 | CODE | |
| LOW | scripts/wer/run_reference_granite5_ctc_transformers.py | 36 | CODE | |
| LOW | scripts/wer/run_reference_cohere_transformers.py | 35 | CODE | |
| LOW | scripts/wer/run_reference_sensevoice.py | 36 | CODE | |
| 25 more matches not shown… | ||||
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| LOW | tools/transcribe-quantize/main.cpp | 427 | // Step 2: quantize fp32 → dst_type into the destination | COMMENT |
| LOW | scripts/dump_reference_whisper_transformers.py | 445 | # Step 0: the prompt-pass already ran; redo its last-row logits | COMMENT |
| LOW⚡ | ggml/tests/test-backend-ops.cpp | 2041 | // Step 1: create input tensors that don't depend on any other tensors: | COMMENT |
| LOW⚡ | ggml/tests/test-backend-ops.cpp | 2048 | // Step 2: use the op that you want to test in the GGML compute graph. | COMMENT |
| LOW⚡ | ggml/tests/test-backend-ops.cpp | 2052 | // Step 3: return the output tensor. | COMMENT |
| LOW | ggml/src/ggml-et/et-kernels/src/norm_f32.c | 84 | // Step 1: sum for mean | COMMENT |
| LOW | ggml/src/ggml-et/et-kernels/src/norm_f32.c | 110 | // Step 2: compute (x - mean) → dst, accumulate variance | COMMENT |
| LOW | ggml/src/ggml-et/et-kernels/src/norm_f32.c | 143 | // Step 3: apply scale to centered values in dst | COMMENT |
| LOW | ggml/src/ggml-et/et-kernels/src/rms_norm_f32.c | 100 | // Step 1: Compute sum of squares using 8-wide vectors | COMMENT |
| LOW⚡ | ggml/src/ggml-et/et-kernels/src/rms_norm_f32.c | 124 | // Step 2: scale = rsqrt(mean + eps) | COMMENT |
| LOW⚡ | ggml/src/ggml-et/et-kernels/src/rms_norm_f32.c | 132 | // Step 3: Apply scaling: broadcast scale once, reuse across loop | COMMENT |
| LOW | ggml/src/ggml-et/et-kernels/src/rwkv_wkv7_f32.c | 177 | // Step 1: sa = dot(a, state_prev[i]) | COMMENT |
| LOW | ggml/src/ggml-et/et-kernels/src/rwkv_wkv7_f32.c | 196 | // Step 2: state update + result accumulation | COMMENT |
| LOW | ggml/src/ggml-webgpu/ggml-webgpu.cpp | 1446 | // For set rows specifically, we need to check if src and idx are empty | COMMENT |
| LOW⚡ | ggml/src/ggml-openvino/openvino/op/gated_delta_net.cpp | 318 | // Step 1: Apply decay gate to state | COMMENT |
| LOW⚡ | ggml/src/ggml-openvino/openvino/op/gated_delta_net.cpp | 323 | // Step 2: delta = (v - S @ k) * beta | COMMENT |
| LOW⚡ | ggml/src/ggml-openvino/openvino/op/gated_delta_net.cpp | 330 | // Step 3: state += outer(delta, k) | COMMENT |
| LOW⚡ | ggml/src/ggml-openvino/openvino/op/gated_delta_net.cpp | 336 | // Step 4: attn_out = S @ q * scale | COMMENT |
| LOW | ggml/src/ggml-cann/aclnn_ops.cpp | 2853 | // Step 5: multiply by attn_factor | COMMENT |
| LOW | ggml/src/ggml-cann/aclnn_ops.cpp | 2870 | // Step 6: repeat | COMMENT |
| LOW | ggml/src/ggml-cann/aclnn_ops.cpp | 3134 | // Step 1: Prepare trans tensors for F16 type conversion to F32 if needed | COMMENT |
| LOW | ggml/src/ggml-cann/aclnn_ops.cpp | 3159 | // Step 2: Prepare head tensors for tail splitting if needed | COMMENT |
| LOW | ggml/src/ggml-cann/aclnn_ops.cpp | 3188 | // Step 3: Execute RotaryPositionEmbedding | COMMENT |
| LOW | ggml/src/ggml-cann/aclnn_ops.cpp | 3235 | // Step 4: Copy unrotated tail portion from source to destination | COMMENT |
| LOW | ggml/src/ggml-cann/aclnn_ops.cpp | 3264 | // Step 5: Cast back to F16 if needed | COMMENT |
| LOW | ggml/src/ggml-cann/aclnn_ops.cpp | 3950 | // Step 1: cast the src0 (Query) to fp16 if needed | COMMENT |
| LOW | ggml/src/ggml-cann/aclnn_ops.cpp | 3973 | // Step 2: create the acl tensors for src1 (Key), src2 (Value), | COMMENT |
| LOW | ggml/src/ggml-cann/aclnn_ops.cpp | 4017 | // Step 3: create the PSEShift tensor if needed | COMMENT |
| LOW | ggml/src/ggml-cann/aclnn_ops.cpp | 4086 | // Step 4: set the inputs for FusedInferAttention. | COMMENT |
| LOW | ggml/src/ggml-cann/aclnn_ops.cpp | 4147 | // Step 6: post-processing — slice padded output and/or cast to f32 | COMMENT |
| LOW⚡ | src/transcribe-kaldi-fbank.cpp | 187 | // Step 1: read frame, upscale. | COMMENT |
| LOW⚡ | src/transcribe-kaldi-fbank.cpp | 193 | // Step 2: remove DC offset (mean subtraction). | COMMENT |
| LOW⚡ | src/transcribe-kaldi-fbank.cpp | 203 | // Step 3: kaldi preemphasis with special x[0] handling. The | COMMENT |
| LOW⚡ | src/transcribe-kaldi-fbank.cpp | 212 | // Step 4: apply hamming window. | COMMENT |
| LOW⚡ | src/transcribe-kaldi-fbank.cpp | 217 | // Step 5: pad to padded_n_fft and FFT. | COMMENT |
| LOW⚡ | src/transcribe-kaldi-fbank.cpp | 225 | // Step 6: power spectrum. | COMMENT |
| LOW | src/transcribe-kaldi-fbank.cpp | 249 | // Step 9: LFR stacking. Mirrors funasr.frontends.wav_frontend.apply_lfr. | COMMENT |
| LOW | src/transcribe-kaldi-fbank.cpp | 278 | // Step 10: optional per-feature CMVN: (x + shift) * scale. | COMMENT |
| LOW⚡ | src/conformer/conformer.cpp | 88 | // Step 1: prepend a [1, T_q, H, B] column of zeros along axis 0. | COMMENT |
| LOW⚡ | src/conformer/conformer.cpp | 98 | // Step 2: reshape (swap inner two axes; total elements unchanged). | COMMENT |
| LOW⚡ | src/conformer/conformer.cpp | 101 | // Step 3: drop the first row of axis 1. | COMMENT |
| LOW⚡ | src/conformer/conformer.cpp | 105 | // Step 4: materialize as contiguous (the view above is sparse: | COMMENT |
| LOW⚡ | src/conformer/conformer.cpp | 110 | // Step 5: reshape back to [pos_len, T_q, H, B]. | COMMENT |
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| MEDIUM | tests/whisper_bin_e2e_smoke.cpp | 170 | // robust against minor decode drift. | COMMENT |
| MEDIUM⚡ | scripts/dump_reference_moss_author.py | 55 | # GitHub inference harness (moss_transcribe_diarize/inference_utils.py | COMMENT |
| MEDIUM | scripts/dump_reference_moss_author.py | 369 | # Mirror the reference harness (inference_utils.prepare_inputs): | COMMENT |
| MEDIUM | scripts/convert-moss.py | 141 | # Canonical instruction from the OpenMOSS inference harness. The model has no | COMMENT |
| MEDIUM | scripts/dump_reference_sortformer_nemo.py | 273 | # Accepted for the validate.py harness contract; a diarizer takes no | COMMENT |
| MEDIUM | scripts/dump_reference_parakeet_nemo.py | 827 | # parity is the validate harness's job. | COMMENT |
| MEDIUM | scripts/dump_reference_parakeet_nemo.py | 947 | # for transducer; tensor for CTC). For the parity harness we | COMMENT |
| MEDIUM | scripts/dump_reference_parakeet_nemo.py | 968 | # sidecar so the harness doesn't have to re-derive them. | COMMENT |
| MEDIUM | scripts/dump_reference_parakeet_nemo.py | 1042 | # parity harness needs byte-stable mel features. pad_to=0 is also | COMMENT |
| LOW | scripts/dump_reference_funasr_nano_funasr.py | 267 | # fire reliably through FunASR's extract_fbank wrapper, so just call it | COMMENT |
| MEDIUM | scripts/convert-canary.py | 109 | # AST excludes Latvian (lv): NVIDIA's card notes seamless-m4t, used | COMMENT |
| MEDIUM | scripts/bench/run.py | 718 | # and compared (scripts/bench/compare.py). A gap here is a bench-harness | COMMENT |
| MEDIUM⚡ | scripts/catalog/common.py | 62 | # spec for the WER harness, a slug for a report filename, a key for the | COMMENT |
| MEDIUM⚡ | scripts/catalog/common.py | 70 | # harness agrees: `fleurs:zh` names a language and carries the split as a | COMMENT |
| MEDIUM | bindings/typescript/src/index.ts | 1046 | // whisper resolves it to "segment" (its robust path), no-timestamp | COMMENT |
| MEDIUM | bindings/swift/Sources/TranscribeCpp/Options.swift | 166 | // (its robust path), no-timestamp families resolve to .none. | COMMENT |
| MEDIUM | ggml/examples/mnist/mnist-common.h | 20 | // The physical batch size determines how many datapoints are processed in parallel, larger values utilize compute bette | COMMENT |
| MEDIUM | ggml/src/ggml-cpu/amx/mmq.cpp | 603 | // pack again with 128 to fully utilize vector length | COMMENT |
| MEDIUM | ggml/src/ggml-cpu/amx/mmq.cpp | 645 | // pack again with 128 to fully utilize vector length | COMMENT |
| MEDIUM | ggml/src/ggml-cpu/amx/mmq.cpp | 743 | // pack again with 128 to fully utilize vector length | COMMENT |
| MEDIUM | src/arch/voxtral_realtime/model.cpp | 1166 | // K=1 is the robust setting across backends (K>=2 costs long-form decode; | COMMENT |
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| HIGH | scripts/convert-gigaam.py | 350 | # absence of slaney area normalization (mel_norm=null). The C++ | COMMENT |
| HIGH | scripts/ci/modal_cuda_build.py | 109 | " && chmod +x /usr/local/bin/sccache && rm /tmp/sccache.tgz", | CODE |
| HIGH | scripts/catalog/render.py | 96 | raise RenderError("published_repo is null, so downloads have no URL") | CODE |
| HIGH | scripts/catalog/common.py | 21 | # The recipe half of the pointer may be null, meaning "any"; the identity | COMMENT |
| HIGH | scripts/lib/gguf_common.py | 257 | # null). canonicalize_normalize maps to our schema so the GGUF and the | COMMENT |
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| CRITICAL | docs/porting/families/canary_qwen.md | 18 | - Canonical reference: **NeMo** (`nvidia/NeMo`, ≥ v2.5.0) via `nemo.collections.speechlm2.models.SALM.from_pretrained('n | CODE |
| CRITICAL | docs/porting/families/canary.md | 25 | - Canonical reference: **NeMo** (`nvidia/NeMo`) via `nemo.collections.asr.models.EncDecMultiTaskModel.from_pretrained(<h | CODE |
| CRITICAL | scripts/langid/run.py | 66 | self.clf.mods.classifier.out.register_forward_hook( | CODE |
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| LOW | scripts/sync-ggml.sh | 20 | # Usage: | COMMENT |
| LOW | scripts/ci/build_xcframework.sh | 32 | # Usage: | COMMENT |
| LOW | scripts/ci/clang-format.sh | 9 | # Usage: | COMMENT |
| LOW | ggml/scripts/sync-whisper-am.sh | 5 | # Usage: | COMMENT |
| LOW | ggml/scripts/release.sh | 11 | # Usage: | COMMENT |
| LOW | ggml/scripts/sync-llama-am.sh | 5 | # Usage: | COMMENT |
| LOW | ggml/src/ggml-sycl/dpct/helper.hpp | 3020 | /// Example usage: | COMMENT |
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| MEDIUM⚡ | CMakeLists.txt | 117 | # reads, ref-mel loaders, and dump scaffolding all live behind this guard, so | COMMENT |
| MEDIUM | …g/parakeet/nemotron-3.5-asr-streaming-0.6b/intake.json | 191 | "NEW vs predecessor: `aux_ctc` head present (ctc_loss_weight=0.1 in cfg, ConvASRDecoder). This makes the model a hyb | CODE |
| MEDIUM | src/CMakeLists.txt | 4 | # shared scaffolding (loader, mel frontend, tokenizer, load-common, | COMMENT |
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| LOW | scripts/lib/gguf_common.py | 42 | CODE | |
| LOW | scripts/wer/remote/cache_paths.py | 8 | CODE | |
| LOW | scripts/wer/remote/modal_sweep.py | 418 | CODE | |
| LOW | scripts/wer/remote/modal_sweep.py | 952 | CODE | |
| LOW | scripts/wer/remote/modal_sweep.py | 1100 | CODE | |
| LOW | scripts/wer/remote/modal_sweep.py | 1258 | CODE | |
| LOW | scripts/wer/remote/modal_sweep.py | 607 | CODE | |
| LOW | bindings/python/src/transcribe_cpp/__init__.py | 2257 | CODE | |
| LOW | bindings/python/src/transcribe_cpp/__init__.py | 864 | CODE | |
| LOW | bindings/python/src/transcribe_cpp/__init__.py | 1511 | CODE | |
| LOW | bindings/python/src/transcribe_cpp/__init__.py | 1570 | CODE | |
| LOW | bindings/python/src/transcribe_cpp/__init__.py | 1672 | CODE | |
| LOW | ggml/examples/mnist/mnist-train-fc.py | 31 | CODE |
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| LOW | ggml/src/ggml.c | 3972 | // TODO: implement non F32 return | COMMENT |
| LOW | ggml/src/ggml.c | 3996 | // TODO: implement non F32 return | COMMENT |
| LOW | ggml/src/ggml.c | 5174 | // TODO: implement antialias for modes other than bilinear | COMMENT |
| LOW | ggml/src/ggml-zdnn/ggml-zdnn.cpp | 22 | // TODO: implement support for quantized types | COMMENT |
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| LOW | ggml/scripts/release.sh | 77 | # Check if we have the latest from master (skip in dry-run) | COMMENT |
| LOW⚡ | ggml/src/ggml-virtgpu/regenerate_remoting.py | 46 | # Check if clang-format is available | COMMENT |
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| MEDIUM | ggml/ci/run.sh | 367 | # Create a fresh python3 venv and enter it | COMMENT |
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| HIGH | ggml/src/ggml-virtgpu/regenerate_remoting.py | 3 | # Generated by Claude AI | STRING |
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| LOW | bindings/python/src/transcribe_cpp/__init__.py | 76 | __all__ = [ | CODE |
| LOW | bindings/python/src/transcribe_cpp/__init__.py | 413 | def set_log_callback(handler) -> None: | CODE |