transcribe-cpp-native 0.0.11__tar.gz → 0.1.0__tar.gz

{transcribe_cpp_native-0.0.11 → transcribe_cpp_native-0.1.0}/Cargo.lock

@@ -315,7 +315,7 @@ dependencies = [
 
 [[package]]
 name = "transcribe-cpp"
-version = "0.0.11"
+version = "0.1.0"
 dependencies = [
  "hound",
  "log",
@@ -325,7 +325,7 @@ dependencies = [
 
 [[package]]
 name = "transcribe-cpp-sys"
-version = "0.0.11"
+version = "0.1.0"
 dependencies = [
  "cmake",
  "serde_json",

{transcribe_cpp_native-0.0.11 → transcribe_cpp_native-0.1.0}/Cargo.toml

@@ -11,7 +11,7 @@
 
 [package]
 name = "transcribe-cpp-sys"
-version = "0.0.11"
+version = "0.1.0"
 edition = "2021"
 rust-version = "1.74"
 description = "Native FFI bindings for transcribe.cpp, a C/C++ speech-to-text library built on ggml"

{transcribe_cpp_native-0.0.11 → transcribe_cpp_native-0.1.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: transcribe-cpp-native
-Version: 0.0.11
+Version: 0.1.0
 Summary: Prebuilt native library for transcribe-cpp (speech-to-text, ggml)
 Keywords: transcription,speech-to-text,asr,ggml,native
 Author: The transcribe.cpp authors

{transcribe_cpp_native-0.0.11 → transcribe_cpp_native-0.1.0}/bindings/python/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "hatchling.build"
 
 [project]
 name = "transcribe-cpp"
-version = "0.0.11"
+version = "0.1.0"
 description = "Python bindings for transcribe.cpp"
 readme = "README.md"
 # 3.8 is EOL (2024-10); 3.9 is the floor. The binding is ctypes-only, so there
@@ -34,13 +34,13 @@ classifiers = [
 # packaging fix still resolves); the import-time version/header-hash check in
 # _library.py is the runtime backstop. check_version_sync.py gates this pin
 # against include/transcribe.h.
-dependencies = ["transcribe-cpp-native==0.0.11.*"]
+dependencies = ["transcribe-cpp-native==0.1.0.*"]
 
 [project.optional-dependencies]
 # Opt-in accelerator providers — ADDITIVE: they install alongside the default
 # provider and the best one wins at runtime. Same base-version pin contract
 # as the hard dependency (gated by check_version_sync.py).
-cu12 = ["transcribe-cpp-native-cu12==0.0.11.*"]
+cu12 = ["transcribe-cpp-native-cu12==0.1.0.*"]
 # Test-only deps. Run with: uv run --extra test pytest (from bindings/python).
 # numpy is here so the numpy PCM-input tests run in every lane instead of
 # silently skipping wherever numpy happens to be absent.

{transcribe_cpp_native-0.0.11 → transcribe_cpp_native-0.1.0}/bindings/python/src/transcribe_cpp/__init__.py

@@ -46,7 +46,7 @@ from .errors import (
     raise_for_status,
 )
 
-__version__ = "0.0.11"
+__version__ = "0.1.0"
 
 # String-enum types, exported so callers (and type checkers) can name them.
 Backend = Literal["auto", "cpu", "metal", "vulkan", "cpu_accel", "cuda"]

{transcribe_cpp_native-0.0.11 → transcribe_cpp_native-0.1.0}/bindings/rust/transcribe-cpp/Cargo.toml

@@ -1,6 +1,6 @@
 [package]
 name = "transcribe-cpp"
-version = "0.0.11"
+version = "0.1.0"
 edition = "2021"
 rust-version = "1.74"
 description = "Rust bindings for transcribe.cpp, a C/C++ speech-to-text library built on ggml"
@@ -40,7 +40,7 @@ dynamic-backends = ["shared", "transcribe-cpp-sys/dynamic-backends"]
 # `transcribe-cpp --no-default-features`, silently re-enabling Metal. The backend
 # features below forward explicitly, so the default (`metal`) still reaches sys
 # via `metal = ["transcribe-cpp-sys/metal"]`.
-transcribe-cpp-sys = { version = "0.0.11", path = "../../..", default-features = false }
+transcribe-cpp-sys = { version = "0.1.0", path = "../../..", default-features = false }
 thiserror = "2"
 log = "0.4"
 

{transcribe_cpp_native-0.0.11 → transcribe_cpp_native-0.1.0}/bindings/swift/Sources/TranscribeCpp/TranscribeCpp.swift

@@ -17,7 +17,7 @@ public enum Transcribe {
     /// Version this binding was built against. Pinned here for the pre-1.0
     /// base-version load gate; the version-sync milestone will generate it from
     /// `include/transcribe.h` (the single source of truth) rather than hardcode.
-    public static let compiledVersion = "0.0.11"
+    public static let compiledVersion = "0.1.0"
 
     /// `MAJOR.MINOR.PATCH` of the linked native library.
     public static func version() -> String { String(cString: transcribe_version()) }

{transcribe_cpp_native-0.0.11 → transcribe_cpp_native-0.1.0}/bindings/typescript/package-lock.json

@@ -1,12 +1,12 @@
 {
   "name": "transcribe-cpp",
-  "version": "0.0.11",
+  "version": "0.1.0",
   "lockfileVersion": 3,
   "requires": true,
   "packages": {
     "": {
       "name": "transcribe-cpp",
-      "version": "0.0.11",
+      "version": "0.1.0",
       "license": "MIT",
       "dependencies": {
         "koffi": "^3.0.2"
@@ -19,11 +19,11 @@
         "node": ">=22"
       },
       "optionalDependencies": {
-        "@transcribe-cpp/darwin-arm64-metal": "0.0.11",
-        "@transcribe-cpp/darwin-x64-cpu": "0.0.11",
-        "@transcribe-cpp/linux-arm64-cpu-vulkan": "0.0.11",
-        "@transcribe-cpp/linux-x64-cpu-vulkan": "0.0.11",
-        "@transcribe-cpp/win32-x64-cpu-vulkan": "0.0.11"
+        "@transcribe-cpp/darwin-arm64-metal": "0.1.0",
+        "@transcribe-cpp/darwin-x64-cpu": "0.1.0",
+        "@transcribe-cpp/linux-arm64-cpu-vulkan": "0.1.0",
+        "@transcribe-cpp/linux-x64-cpu-vulkan": "0.1.0",
+        "@transcribe-cpp/win32-x64-cpu-vulkan": "0.1.0"
       }
     },
     "node_modules/@koromix/koffi-darwin-arm64": {

{transcribe_cpp_native-0.0.11 → transcribe_cpp_native-0.1.0}/bindings/typescript/package.json

@@ -1,6 +1,6 @@
 {
   "name": "transcribe-cpp",
-  "version": "0.0.11",
+  "version": "0.1.0",
   "description": "TypeScript/Node.js bindings for transcribe.cpp — a C/C++ speech-to-text library built on ggml",
   "type": "module",
   "exports": {
@@ -38,11 +38,11 @@
     "koffi": "^3.0.2"
   },
   "optionalDependencies": {
-    "@transcribe-cpp/darwin-arm64-metal": "0.0.11",
-    "@transcribe-cpp/darwin-x64-cpu": "0.0.11",
-    "@transcribe-cpp/linux-x64-cpu-vulkan": "0.0.11",
-    "@transcribe-cpp/linux-arm64-cpu-vulkan": "0.0.11",
-    "@transcribe-cpp/win32-x64-cpu-vulkan": "0.0.11"
+    "@transcribe-cpp/darwin-arm64-metal": "0.1.0",
+    "@transcribe-cpp/darwin-x64-cpu": "0.1.0",
+    "@transcribe-cpp/linux-x64-cpu-vulkan": "0.1.0",
+    "@transcribe-cpp/linux-arm64-cpu-vulkan": "0.1.0",
+    "@transcribe-cpp/win32-x64-cpu-vulkan": "0.1.0"
   },
   "devDependencies": {
     "@types/node": "^22.0.0",

{transcribe_cpp_native-0.0.11 → transcribe_cpp_native-0.1.0}/docs/models/granite-speech-4.1-2b-plus.md

@@ -12,9 +12,10 @@ final-layer hidden states (doubling the projector K/V input from 1024 to
 
 Offline multilingual speech-to-text with word-level timestamps. Covers
 English plus French, German, Spanish, and Portuguese (no Japanese on this
-variant). Takes a 16 kHz mono WAV and produces a transcript, optionally
-interleaved with `[SS:N]` word-timestamp markers (centiseconds since
-segment start).
+variant). Takes a 16 kHz mono WAV and produces a transcript; with
+`--timestamps word` it returns per-word start/end times. Internally the model
+emits `[T:N]` end-of-word centisecond markers; the runtime parses them into
+structured word timestamps and returns a clean transcript.
 
 This variant is transcription-only. Unlike the base
 [`granite-speech-4.1-2b`](granite-speech-4.1-2b.md), it does not perform
@@ -74,11 +75,20 @@ build/bin/transcribe-cli \
   samples/jfk.wav
 ```
 
-Output includes `[SS:N]` markers (centiseconds since segment start)
-interleaved with words:
+The runtime returns a clean transcript plus structured per-word start/end
+times. (Internally the model emits `[T:N]` markers giving each word's end time
+in centiseconds, modulo 1000 with a 10 s rollover that the runtime unwraps; the
+markers are stripped from the returned text.)
 
 ```
-[SS:23] And [SS:52] so [SS:92] my [SS:121] fellow [SS:154] Americans ...
+text: and so my fellow americans ask not what your country can do for you ask what you can do for your country
+words: 22
+  [   0.30 ->    0.57] and
+  [   0.57 ->    0.95] so
+  [   0.95 ->    1.25] my
+  [   1.25 ->    1.60] fellow
+  [   1.60 ->    2.13] americans
+  ...
 ```
 
 ## Performance
@@ -143,7 +153,7 @@ Linux 6.18 (Fedora 43), transcribe.cpp `dbe5814`.
 | Transcribe (English)        | Yes    |
 | Transcribe (fr/de/es/pt)    | Yes    |
 | Translate                   | No (ASR-only variant; use the base [granite-speech-4.1-2b](granite-speech-4.1-2b.md) for translation) |
-| Word-level timestamps       | Yes (`--timestamps word`, `[SS:N]` markers) |
+| Word-level timestamps       | Yes (`--timestamps word`, structured per-word t0/t1 parsed from the model's `[T:N]` markers) |
 | Speaker diarization         | No (upstream supports via prompt; not exposed in v1 of transcribe.cpp) |
 
 ## Numerical Validation

{transcribe_cpp_native-0.0.11 → transcribe_cpp_native-0.1.0}/docs/models/granite-speech.md

@@ -22,7 +22,7 @@ perspective they are one family of GGUFs.
   LLM head, autoregressive; 1.31% BF16 / 1.32% Q8_0 WER on LibriSpeech
   test-clean. Improved punctuation and casing over 4.0-1b.
 - **Most features.** `granite-speech-4.1-2b-plus` — same 4.1 base plus
-  word-level timestamps (`[SS:N]` centisecond markers). Encoder
+  word-level timestamps (parsed from the model's `[T:N]` centisecond markers). Encoder
   concatenates a mid-layer and the final layer (`cat_hidden_layers=[3]`),
   doubling projector K/V width.
 - **Fastest, ASR-only.** `granite-speech-4.1-2b-nar` — non-autoregressive
@@ -101,8 +101,8 @@ Translation (`granite-4.0-1b-speech`, `granite-speech-4.1-2b`):
   from the audio. The `-plus` variant is ASR-only and does not translate.
 
 Plus only (`granite-speech-4.1-2b-plus`):
-- **Word-level timestamps** as `[SS:N]` centisecond markers interleaved
-  with words (`--timestamps word`).
+- **Word-level timestamps** via `--timestamps word` — structured per-word
+  start/end times, parsed from the model's `[T:N]` centisecond markers.
 
 NAR only (`granite-speech-4.1-2b-nar`):
 - **Single-pass non-autoregressive decode** — fastest of the four.

{transcribe_cpp_native-0.0.11 → transcribe_cpp_native-0.1.0}/docs/models/nemotron-3.5-asr-streaming-0.6b.md

@@ -74,15 +74,15 @@ context, NVIDIA's self-reported FLEURS en-US WER is **7.91%** (and an
 | --- | ---: | ---: |
 | Reference (NeMo) | 7.99 | 3.03 |
 | F32    | 7.97 | 3.04 |
-| F16    | 7.97 | 3.04 |
-| Q8_0   | 7.88 | 3.05 |
-| Q6_K   | 8.02 | 3.08 |
+| F16    | 7.97 | 3.03 |
+| Q8_0   | 7.88 | 3.06 |
+| Q6_K   | 8.02 | 3.07 |
 | Q5_K_M | 8.15 | 3.10 |
-| Q4_K_M | 8.49 | 3.30 |
+| Q4_K_M | 8.49 | 3.28 |
 
 The F32 reference dtype meets the measured-Oracle gate on both datasets.
 F16/Q8_0/Q6_K/Q5_K_M land inside the reference 95% CI; Q4_K_M carries the
-largest quantization loss (+0.50 on FLEURS, +0.27 on LibriSpeech) but is
+largest quantization loss (+0.50 on FLEURS, +0.25 on LibriSpeech) but is
 accepted for shipping.
 
 ## Quick Start

{transcribe_cpp_native-0.0.11 → transcribe_cpp_native-0.1.0}/docs/porting/families/granite.md

@@ -155,7 +155,7 @@ Allowed statuses: `PASS`, `SKIP - not exposed by runtime`, `ACCEPTED GAP - <reas
 | Transcribe | granite-speech-4.1-2b-plus | explicit language hint | `build/bin/transcribe-cli -m models/granite-speech-4.1-2b-plus/granite-speech-4.1-2b-plus-BF16.gguf --language en samples/jfk.wav` | non-empty plausible English transcript | PASS (granite-4 system-role template; transcript matches reference within one comma) |
 | Transcribe | granite-speech-4.1-2b-plus | auto / no language hint | `build/bin/transcribe-cli -m models/granite-speech-4.1-2b-plus/granite-speech-4.1-2b-plus-BF16.gguf samples/jfk.wav` | non-empty plausible transcript | PASS |
 | Translate | granite-speech-4.1-2b-plus | n/a | n/a | runtime rejects `--translate` (`supports_translate=false`) | SKIP - not exposed by runtime (upstream model card lists -plus as ASR-only; the base granite-speech-4.1-2b is the AST variant. The fused LLM can be prompt-coerced to translate — the smoke produced "Ich bitte meine Mitbürger, ..." — but the converter advertises `stt.capability.translate=false`, so the dispatcher rejects the task.) |
-| Word timestamps | granite-speech-4.1-2b-plus | text-protocol output | `build/bin/transcribe-cli -m models/granite-speech-4.1-2b-plus/granite-speech-4.1-2b-plus-BF16.gguf --timestamps word samples/jfk.wav` | `[T:N]` centisecond markers interleaved with the transcript text | PASS (model emits `[SS:N]` markers: e.g. "[SS:23] And [SS:52] so [SS:92] my [SS:121] fellow [SS:154] Americans ..."). Spec says `[T:N]`; granite uses `[SS:N]` literally. |
+| Word timestamps | granite-speech-4.1-2b-plus | structured per-word timestamps | `build/bin/transcribe-cli -m models/granite-speech-4.1-2b-plus/granite-speech-4.1-2b-plus-BF16.gguf --timestamps word samples/jfk.wav` | clean transcript + per-word t0/t1, parsed from the model's `[T:N]` end-of-word centisecond markers | PASS (model emits `[T:N]` markers, e.g. "_ [T:30] and [T:57] so [T:95] my [T:125] fellow [T:160] americans ..."; the runtime drops the `_` placeholders, unwraps the mod-1000 / 10 s rollover, and returns structured words. Verified against IBM's transformers reference with the model-card timestamps prompt; the earlier `[SS:N]` reading was an artifact of an incorrect prompt string.) |
 | Speaker diarization | granite-speech-4.1-2b-plus | multi-speaker audio | n/a | `[Speaker N]:` tags preceding speaker turns in the text | SKIP - not exposed by runtime (user-deferred to later) |
 | Incremental decoding | granite-speech-4.1-2b-plus | prefix_text continuation | n/a | transcript continues from supplied prefix | SKIP - not exposed by runtime |
 

{transcribe_cpp_native-0.0.11 → transcribe_cpp_native-0.1.0}/docs/tools/hf-metadata-schema.md

@@ -43,7 +43,11 @@ transcribe_cpp:
 
 The `<dataset>` suffix is the spec's `wer.metadata_key` (default
 `librispeech_test_clean`); the `<machine>` suffix is the `perf:` rig key with `-`
-mapped to `_`. Standard HF keys (`license`, `language`, `pipeline_tag`,
+mapped to `_`. A spec can also publish secondary benchmarks inline: any dataset-named key
+under `wer:` whose value is a `{quant: wer%}` map (e.g. `librispeech_test_clean:`)
+is emitted as its own `wer_<dataset>` block alongside the headline one — so a
+model whose per-quant column is e.g. FLEURS can still expose LibriSpeech
+machine-readably. Add a dataset by adding a key; no wrapper needed. Standard HF keys (`license`, `language`, `pipeline_tag`,
 `base_model`, `tags`, …) are emitted alongside and unchanged by this block.
 
 ## Reading it

{transcribe_cpp_native-0.0.11 → transcribe_cpp_native-0.1.0}/include/transcribe.h

@@ -146,8 +146,8 @@
  * exactly and refuses to load a native provider that does not match.
  */
 #define TRANSCRIBE_VERSION_MAJOR 0
-#define TRANSCRIBE_VERSION_MINOR 0
-#define TRANSCRIBE_VERSION_PATCH 11
+#define TRANSCRIBE_VERSION_MINOR 1
+#define TRANSCRIBE_VERSION_PATCH 0
 
 #define TRANSCRIBE_VERSION_STRINGIZE_(x) #x
 #define TRANSCRIBE_VERSION_STRINGIZE(x)  TRANSCRIBE_VERSION_STRINGIZE_(x)

{transcribe_cpp_native-0.0.11 → transcribe_cpp_native-0.1.0}/scripts/convert-whisper.py

@@ -170,6 +170,11 @@ def detect_reference_dtype(safetensors_path: Path) -> tuple[str, LlamaFileType,
         return ("F32", LlamaFileType.ALL_F32, GGMLQuantizationType.F32)
     if dtypes == {"F16"}:
         return ("F16", LlamaFileType.MOSTLY_F16, GGMLQuantizationType.F16)
+    if dtypes == {"BF16"}:
+        # Fine-tunes (e.g. MediaTek's Breeze-ASR-25) ship bfloat16. Preserve
+        # it as the reference dtype; reference_dtype_for() downgrades the
+        # conv stem to F16 (loader has no BF16 conv kernel).
+        return ("BF16", LlamaFileType.MOSTLY_BF16, GGMLQuantizationType.BF16)
     if dtypes <= {"F32", "F16"}:
         # Mixed F32 + F16 — promote to F32 reference. Upcast on load.
         return ("F32", LlamaFileType.ALL_F32, GGMLQuantizationType.F32)
@@ -460,6 +465,7 @@ VARIANT_DISPLAY_NAMES: dict[str, str] = {
     "whisper-large-v2":       "Whisper Large v2",
     "whisper-large-v3":       "Whisper Large v3",
     "whisper-large-v3-turbo": "Whisper Large v3 Turbo",
+    "breeze-asr-25":          "Breeze-ASR-25",
 }
 
 
@@ -523,6 +529,14 @@ def convert(model_dir: Path, out_path: Path, variant: str, repo_id: str | None =
                 f"unknown whisper variant slug: {variant!r}; "
                 f"add it to VARIANT_DISPLAY_NAMES"
             )
+        # Stock Whisper variants are OpenAI's; community fine-tunes (e.g.
+        # MediaTek's Breeze-ASR-25) carry their own author/org, derived from
+        # the source repo so the metadata names the actual publisher.
+        org = repo_id.split("/")[0] if repo_id else "openai"
+        if org.lower() == "openai":
+            author, organization = "OpenAI", "openai"
+        else:
+            author, organization = org, org
         add_general_identity(
             writer,
             name=VARIANT_DISPLAY_NAMES[variant],
@@ -530,8 +544,8 @@ def convert(model_dir: Path, out_path: Path, variant: str, repo_id: str | None =
             size_label=size_label,
             file_type=REFERENCE_FILE_TYPE,
             languages=hp["languages"],
-            author="OpenAI",
-            organization="openai",
+            author=author,
+            organization=organization,
             license="apache-2.0",
             license_name="Apache License 2.0",
             license_link="https://www.apache.org/licenses/LICENSE-2.0",
@@ -687,11 +701,12 @@ def convert(model_dir: Path, out_path: Path, variant: str, repo_id: str | None =
             if src_name not in st_keys:
                 raise KeyError(f"safetensors missing tensor: {src_name!r}")
             t = st.get_tensor(src_name)
-            if t.dtype not in (torch.float32, torch.float16):
+            if t.dtype not in (torch.float32, torch.float16, torch.bfloat16):
                 raise ValueError(
-                    f"{src_name}: expected float32 or float16, got {t.dtype}"
+                    f"{src_name}: expected float32, float16, or bfloat16, "
+                    f"got {t.dtype}"
                 )
-            # encode_for_gguf wants fp32 input. F16 → F32 is exact, so we
+            # encode_for_gguf wants fp32 input. F16/BF16 → F32 is exact, so we
             # upcast on read and let encode_for_gguf re-pack to the
             # per-tensor target dtype (which respects REFERENCE_GGML_TYPE).
             arr = transform(t.float().numpy())

transcribe_cpp_native-0.1.0/scripts/hf_cards/breeze-asr-25.yaml

@@ -0,0 +1,141 @@
+# Spec for the HF README of handy-computer/Breeze-ASR-25-gguf.
+# Consumed by scripts/hf_cards/generate.py.
+
+hf_repo: MediaTek-Research/Breeze-ASR-25
+target_repo: handy-computer/Breeze-ASR-25-gguf
+# Pure fine-tune of whisper-large-v2 (identical architecture), so the model
+# page points at the whisper-large-v2 doc rather than a Breeze-specific one.
+transcribe_docs_url: https://github.com/handy-computer/transcribe.cpp/blob/main/docs/models/whisper-large-v2.md
+
+# Pinned at download (the snapshot the GGUFs were converted from).
+upstream_commit: cffe7ccb404d025296a00758d0a33468bec3a9d0
+pin_date: 2026-06-29
+
+validation:
+  reference: transformers
+  commit: 3848875
+  date: 2026-06-29
+
+license: apache-2.0
+license_display: Apache-2.0
+pipeline_tag: automatic-speech-recognition
+
+# Breeze-ASR-25 is a Whisper-large-v2 fine-tune. It inherits Whisper's
+# 99-language tokenizer, but MediaTek optimized and validated it for
+# Taiwanese Mandarin (Traditional Chinese) and English only, so we advertise
+# just those two — listing all 99 would overclaim coverage the fine-tune
+# does not target.
+languages:
+  - zh
+  - en
+tags:
+  - gguf
+  - transcribe.cpp
+  - asr
+  - speech-to-text
+  - whisper
+  - mediatek
+  - mandarin
+  - traditional-chinese
+  - code-switching
+
+summary: |
+  MediaTek Research's Breeze-ASR-25 — a fine-tune of OpenAI Whisper large-v2,
+  converted to GGUF for transcribe.cpp. Optimized for Taiwanese Mandarin
+  (it emits **Traditional Chinese**) and English, with explicit support for
+  Mandarin-English code-switching (intra- and inter-sentential). Trained on
+  ~11,749 hours: 10,000 h synthetic Mandarin (ODC Synth), 1,738 h English
+  (CommonVoice17) and 11 h real code-switch (NTUML2021). Architecturally
+  identical to Whisper large-v2 (encoder-decoder transformer, 30-second
+  windows with chunked long-form decoding); it retains Whisper's 99-language
+  tokenizer, but only Mandarin and English are optimized and validated —
+  other languages remain technically accessible but out of scope.
+
+default_quant_index: 2  # Q8_0
+
+# Capability flags for the transcribe_cpp metadata block. Same Whisper
+# code path as whisper-large-v2 (the library exposes translation, language
+# detection and segment timestamps for any Whisper GGUF), though this
+# fine-tune is specialized for zh/en transcription.
+capabilities:
+  streaming: false
+  translate: true
+  lang_detect: true
+  timestamps: segment  # none | segment | word | token
+
+# Speedup-over-realtime (×RT) per rig/backend, published raw as rtf_<machine>
+# in the metadata block. Breeze-ASR-25 is byte-for-byte the same architecture
+# and tensor shapes as Whisper large-v2 (1.5B params, 32+32 layers,
+# d_model 1280); RTF is a function of shapes/quant/backend, not weight values,
+# so these are inherited verbatim from whisper-large-v2 rather than re-benched.
+perf:
+  m4-max:
+    metal: 23.7
+    cpu: 1.5
+  ryzen-4750u:
+    vulkan: 2.1
+    cpu: 0.6
+
+wer:
+  source: LibriSpeech test-clean
+  metadata_key: librispeech_test_clean
+  source2: FLEURS zh
+  notes: |
+    Two benchmarks, both full test splits, decoded on a Modal L40S with the
+    transcribe.cpp default recipe (greedy + temperature fallback,
+    suppress_tokens, segment timestamps), batch 1.
+
+    **English — WER, LibriSpeech test-clean (2620 utterances).** Standard
+    Whisper/`EnglishTextNormalizer` scoring.
+
+    **Chinese — CER, FLEURS `cmn_hans_cn` test (945 utterances).** Breeze
+    emits **Traditional** Chinese while FLEURS references are **Simplified**,
+    so the CER reported here is computed after folding *both* hypothesis and
+    reference to Simplified with OpenCC (`t2s`) — otherwise the raw,
+    script-mismatched CER is ~35% and meaningless. FLEURS ships no Traditional
+    Mandarin split, which is why a script-normalized Simplified set is used.
+
+    Quantization is effectively free on both languages: every quant down to
+    Q4_K_M (1.0 GB) sits within run-to-run noise of the BF16 reference
+    (English 2.25-2.29%, Chinese 8.08-8.12%).
+
+  # Secondary per-quant metric → wer_fleurs_zh in the metadata block.
+  fleurs_zh:
+    bf16: 8.12
+    f16: 8.11
+    q8_0: 8.10
+    q6_k: 8.12
+    q5_k_m: 8.12
+    q4_k_m: 8.08
+
+quants:
+  - name: BF16
+    filename: Breeze-ASR-25-BF16.gguf
+    size: 3.10 GB
+    wer: 2.29%
+    wer2: 8.12%
+  - name: F16
+    filename: Breeze-ASR-25-F16.gguf
+    size: 3.11 GB
+    wer: 2.29%
+    wer2: 8.11%
+  - name: Q8_0
+    filename: Breeze-ASR-25-Q8_0.gguf
+    size: 1.67 GB
+    wer: 2.27%
+    wer2: 8.10%
+  - name: Q6_K
+    filename: Breeze-ASR-25-Q6_K.gguf
+    size: 1.30 GB
+    wer: 2.29%
+    wer2: 8.12%
+  - name: Q5_K_M
+    filename: Breeze-ASR-25-Q5_K_M.gguf
+    size: 1.16 GB
+    wer: 2.25%
+    wer2: 8.12%
+  - name: Q4_K_M
+    filename: Breeze-ASR-25-Q4_K_M.gguf
+    size: 1.00 GB
+    wer: 2.26%
+    wer2: 8.08%

{transcribe_cpp_native-0.0.11 → transcribe_cpp_native-0.1.0}/scripts/hf_cards/generate.py

@@ -50,20 +50,28 @@ def build_transcribe_cpp_block(spec: dict) -> str:
         return ""
 
     caps = spec.get("capabilities", {})
-    dataset_key = spec["wer"].get("metadata_key", "librispeech_test_clean")
-    block: dict = {
-        f"wer_{dataset_key}": {
-            q["name"].lower(): float(str(q["wer"]).rstrip("%")) for q in spec["quants"]
-        },
-        **{
-            f"rtf_{machine.replace('-', '_')}": backends
-            for machine, backends in spec["perf"].items()
-        },
-        "streaming": bool(caps.get("streaming", False)),
-        "translate": bool(caps.get("translate", False)),
-        "lang_detect": bool(caps.get("lang_detect", False)),
-        "timestamps": caps.get("timestamps", "none"),
+    wer = spec["wer"]
+    dataset_key = wer.get("metadata_key", "librispeech_test_clean")
+    block: dict = {}
+    # Headline dataset: per-quant WER taken from the `quants:` column.
+    block[f"wer_{dataset_key}"] = {
+        q["name"].lower(): float(str(q["wer"]).rstrip("%")) for q in spec["quants"]
     }
+    # Any additional per-quant WER maps listed inline under `wer:` (keyed by
+    # dataset name, e.g. `librispeech_test_clean:`) are emitted as their own
+    # `wer_<dataset>` blocks. Only dict values count as datasets; scalar keys
+    # (metadata_key, source, notes) are skipped.
+    for key, per_quant in wer.items():
+        if isinstance(per_quant, dict):
+            block[f"wer_{key}"] = {
+                str(q).lower(): float(str(v).rstrip("%")) for q, v in per_quant.items()
+            }
+    for machine, backends in spec["perf"].items():
+        block[f"rtf_{machine.replace('-', '_')}"] = backends
+    block["streaming"] = bool(caps.get("streaming", False))
+    block["translate"] = bool(caps.get("translate", False))
+    block["lang_detect"] = bool(caps.get("lang_detect", False))
+    block["timestamps"] = caps.get("timestamps", "none")
     dumped = yaml.safe_dump(
         {"transcribe_cpp": block}, sort_keys=False, default_flow_style=False
     )

{transcribe_cpp_native-0.0.11 → transcribe_cpp_native-0.1.0}/scripts/hf_cards/granite-4.0-1b-speech.yaml

@@ -50,7 +50,7 @@ capabilities:
   streaming: false
   translate: true
   lang_detect: false
-  timestamps: word  # none | segment | word | token
+  timestamps: none  # none | segment | word | token
 
 # Speedup-over-realtime (×RT) per rig/backend, from docs/models; published
 # raw as rtf_<machine> in the metadata block.
@@ -62,6 +62,7 @@ perf:
     metal: 11
     cpu: 5
   ryzen-4750u:
+    vulkan: 2.95
     cpu: 1.6
 
 wer:

{transcribe_cpp_native-0.0.11 → transcribe_cpp_native-0.1.0}/scripts/hf_cards/granite-speech-4.1-2b-nar.yaml

@@ -63,6 +63,7 @@ perf:
     metal: 18
     cpu: 4
   ryzen-4750u:
+    vulkan: 3.65
     cpu: 1.5
 
 wer:

{transcribe_cpp_native-0.0.11 → transcribe_cpp_native-0.1.0}/scripts/hf_cards/granite-speech-4.1-2b-plus.yaml

@@ -41,8 +41,9 @@ summary: |
   decoder) with two changes: the encoder concatenates mid-layer (idx 3) and
   final-layer hidden states (`cat_hidden_layers=[3]`, doubling the projector
   K/V input from 1024 to 2048), and the LM token embeddings are tied with
-  the lm_head. Takes a 16 kHz mono WAV and produces a transcript, optionally
-  interleaved with `[SS:N]` word-timestamp markers. Transcribes English,
+  the lm_head. Takes a 16 kHz mono WAV and produces a transcript, with
+  `--timestamps word` returning structured per-word timestamps (parsed from the
+  model's `[T:N]` centisecond markers). Transcribes English,
   French, German, Spanish, and Portuguese (no Japanese on this variant).
   This variant is transcription-only: unlike the base granite-speech-4.1-2b,
   it does not perform speech translation.
@@ -66,6 +67,7 @@ perf:
     metal: 11
     cpu: 5
   ryzen-4750u:
+    vulkan: 2.75
     cpu: 1.4
 
 wer:

{transcribe_cpp_native-0.0.11 → transcribe_cpp_native-0.1.0}/scripts/hf_cards/granite-speech-4.1-2b.yaml

@@ -51,7 +51,7 @@ capabilities:
   streaming: false
   translate: true
   lang_detect: false
-  timestamps: word  # none | segment | word | token
+  timestamps: none  # none | segment | word | token
 
 # Speedup-over-realtime (×RT) per rig/backend, from docs/models; published
 # raw as rtf_<machine> in the metadata block.
@@ -63,6 +63,7 @@ perf:
     metal: 12
     cpu: 4
   ryzen-4750u:
+    vulkan: 2.80
     cpu: 1.4
 
 wer:

{transcribe_cpp_native-0.0.11 → transcribe_cpp_native-0.1.0}/scripts/hf_cards/nemotron-3.5-asr-streaming-0.6b.yaml

@@ -90,7 +90,14 @@ wer:
   metadata_key: fleurs_en
   source: FLEURS test en (en-US), offline att_context_size=[56, 13]
   notes: |
-    WER measured on FLEURS test en (647 utterances), greedy RNN-T, --language en-US, whisper-normalizer scoring; the per-quant column is FLEURS en. NeMo reference baseline on the same manifest: 7.99% (NVIDIA self-reports 7.91% en-US). On LibriSpeech test-clean (2620 utterances) the same presets score F32 3.04 / F16 3.04 / Q8_0 3.05 / Q6_K 3.08 / Q5_K_M 3.10 / Q4_K_M 3.30, against a 3.03% NeMo reference.
+    WER measured on FLEURS test en (647 utterances), greedy RNN-T, --language en-US, whisper-normalizer scoring; the per-quant column is FLEURS en. NeMo reference baseline on the same manifest: 7.99% (NVIDIA self-reports 7.91% en-US). On LibriSpeech test-clean (2620 utterances) the same presets score F32 3.04 / F16 3.03 / Q8_0 3.06 / Q6_K 3.07 / Q5_K_M 3.10 / Q4_K_M 3.28, against a 3.03% NeMo reference.
+  librispeech_test_clean:
+    f32: 3.04
+    f16: 3.03
+    q8_0: 3.06
+    q6_k: 3.07
+    q5_k_m: 3.10
+    q4_k_m: 3.28
 
 quants:
   - name: F32

{transcribe_cpp_native-0.0.11 → transcribe_cpp_native-0.1.0}/scripts/hf_cards/template.md.j2

@@ -39,10 +39,10 @@ on {{ validation.date }}.
 
 ## Downloads
 
-| Quantization | Download | Size | {{ metric|default("WER") }} ({{ wer.source }}) |
-| --- | --- | ---: | ---: |
+| Quantization | Download | Size | {{ metric|default("WER") }} ({{ wer.source }}){% if wer.source2 is defined %} | {{ metric2|default("CER") }} ({{ wer.source2 }}){% endif %} |
+| --- | --- | ---: | ---:{% if wer.source2 is defined %} | ---:{% endif %} |
 {%- for q in quants %}
-| {{ q.name }} | [{{ q.filename }}](https://huggingface.co/{{ target_repo }}/resolve/main/{{ q.filename }}) | {{ q.size }} | {{ q.wer }} |
+| {{ q.name }} | [{{ q.filename }}](https://huggingface.co/{{ target_repo }}/resolve/main/{{ q.filename }}) | {{ q.size }} | {{ q.wer }}{% if wer.source2 is defined %} | {{ q.wer2 }}{% endif %} |
 {%- endfor %}
 
 {{ wer.notes }}