Publish verified Orukeet r3 Core ML preview bundles for TapTalk

README.md

@@ -153,6 +153,18 @@ times; it does not return emotion, speaking-style or speaker-diarization scores.
 
 This file has a different tensor layout from the native NeMo-Speech.cpp GGUFs above. Select the export for your runtime. [Conversion, checksums and validation](transcribe-cpp/README.md).
 
+## Core ML preview for Apple Silicon
+
+The [Core ML download and integration guide](coreml/README.md) includes portable
+bundles for FluidAudio 0.15.5 and TapTalk, with checksums and a Swift loader.
+Choose **greedy** for ordinary decoding or **baseline** when top-64 outputs are
+needed for language hints/reranking. Compile the packages on the destination Mac.
+
+The greedy profile reduced paired batch latency by 9.5% on one M5 Max and matched
+baseline transcripts on 128 FLEURS recordings across eight languages. This is a
+preview with separate [Core ML validation](https://github.com/Oruk-AI/orukeet/tree/347f646cacda2e001865b6ac40ba5cbc7e90d1c9/evidence/coreml-taptalk-20260915);
+it does not convert the EOU 120M live-typing model.
+
 ## Model files
 
 | Format | File | Bytes |
@@ -162,6 +174,8 @@ This file has a different tensor layout from the native NeMo-Speech.cpp GGUFs ab
 | transcribe.cpp Q8 | `orukeet-transcribe-cpp-Q8_0.gguf` | 739,508,608 |
 | Native F16 | `orukeet-v0.1.0-f16.gguf` | 1,296,681,088 |
 | ONNX INT8 archive | `onnx/sherpa-onnx-orukeet-v0.1.0-int8.tar.bz2` | 486,807,585 |
+| Core ML greedy preview | `coreml/orukeet-r3-coreml-greedy.zip` | 466,579,943 |
+| Core ML baseline preview | `coreml/orukeet-r3-coreml-baseline.zip` | 466,579,851 |
 
 All formats derive from **r3**. NeMo and native files are pinned to revision `555136b50265a132d4cea0d35560c26fc4f657ab`; the ONNX archive is pinned to `55a984d46f68323301837194ce647c702f55facc`. The ONNX package occupies 671,619,800 bytes after extraction.
 

coreml/README.md

@@ -0,0 +1,107 @@
+# Orukeet r3 Core ML preview
+
+Portable Core ML bundles for Apple Silicon, macOS 14+, and **FluidAudio 0.15.5**.
+These are the same archives as the
+[GitHub preview release](https://github.com/Oruk-AI/orukeet/releases/tag/coreml-taptalk-preview-20260915),
+published here for TapTalk and other applications to download.
+
+| Archive | Profile | Bytes |
+|:--|:--|--:|
+| [orukeet-r3-coreml-greedy.zip](https://huggingface.co/oruk/orukeet/resolve/coreml-taptalk-preview-20260915/coreml/orukeet-r3-coreml-greedy.zip?download=true) | Recommended for ordinary greedy decoding | 466,579,943 |
+| [orukeet-r3-coreml-baseline.zip](https://huggingface.co/oruk/orukeet/resolve/coreml-taptalk-preview-20260915/coreml/orukeet-r3-coreml-baseline.zip?download=true) | Retains top-64 outputs for language hints/reranking | 466,579,851 |
+
+[SHA256SUMS.txt](https://huggingface.co/oruk/orukeet/resolve/coreml-taptalk-preview-20260915/coreml/SHA256SUMS.txt)
+contains the archive hashes. Each archive includes a `bundle.json` with per-file
+SHA-256 hashes, conversion receipts, attribution, and the weight license.
+
+## Download
+
+```sh
+python -m pip install 'huggingface-hub>=0.34,<2'
+```
+
+```python
+import hashlib
+from pathlib import Path
+from zipfile import ZipFile
+from huggingface_hub import hf_hub_download
+
+archive = Path(hf_hub_download(
+    repo_id="oruk/orukeet",
+    filename="coreml/orukeet-r3-coreml-greedy.zip",
+    revision="coreml-taptalk-preview-20260915",
+))
+expected = "beccdc6f18c4b10527a764f6e3ab12e3e11b969220c0cee175b3bb7eaa94290e"
+with archive.open("rb") as stream:
+    digest = hashlib.sha256()
+    for block in iter(lambda: stream.read(8 * 1024 * 1024), b""):
+        digest.update(block)
+if digest.hexdigest() != expected:
+    raise ValueError("Core ML archive checksum mismatch")
+with ZipFile(archive) as bundle:
+    bundle.extractall("models")
+print(Path("models/orukeet-r3-coreml-greedy").resolve())
+```
+
+For an application downloader, use the archive URL above and verify its hash
+before extraction. The archive has one top-level directory,
+`orukeet-r3-coreml-greedy/` (or `orukeet-r3-coreml-baseline/`).
+
+## Load with Swift
+
+Each bundle contains `Preprocessor.mlpackage`, `Encoder.mlpackage`,
+`Decoder.mlpackage`, `JointDecisionv3.mlpackage`, and `parakeet_vocab.json`.
+Compile all four packages on the destination Mac during installation;
+machine-specific `.mlmodelc` caches are deliberately excluded.
+
+The `OrukeetCoreML` library is in `export/coreml/benchmark` at
+[Oruk-AI/orukeet commit 347f646](https://github.com/Oruk-AI/orukeet/tree/347f646cacda2e001865b6ac40ba5cbc7e90d1c9/export/coreml/benchmark).
+Add that directory as a local Swift package dependency, then:
+
+```swift
+import OrukeetCoreML
+
+try OrukeetLocalModels.compilePackages(from: downloadedBundle, to: installedCache)
+let engine = OrukeetEngine(modelDirectory: installedCache)
+try await engine.ensureLoaded()
+let result = try await engine.transcribe(samples: mono16kSamples)
+print(result.text)
+```
+
+`downloadedBundle` is the extracted profile directory; `installedCache` is a
+fresh app-owned directory. Audio must be mono 16 kHz PCM. Keep the engine loaded
+between recordings and keep the Orukeet cache separate from Parakeet's.
+Use the explicit local loader rather than FluidAudio's NVIDIA model downloader.
+
+[Full conversion and integration guide](https://github.com/Oruk-AI/orukeet/blob/347f646cacda2e001865b6ac40ba5cbc7e90d1c9/export/coreml/README.md)
+· [Source PR #6](https://github.com/Oruk-AI/orukeet/pull/6)
+· [TapTalk](https://github.com/vakharwalad23/tap-talk)
+
+## Measurements and scope
+
+On one M5 Max, greedy decoding reduced paired batch latency by 9.5% relative
+to the Core ML baseline. Baseline and greedy returned identical text on 128
+FLEURS recordings across eight languages. Core ML WER was 7.64%, compared with
+8.55% for Parakeet Core ML and 7.29% for the uncompressed Orukeet source.
+[Raw evidence and methodology](https://github.com/Oruk-AI/orukeet/tree/347f646cacda2e001865b6ac40ba5cbc7e90d1c9/evidence/coreml-taptalk-20260915).
+
+The baseline uses 6-bit LUT/FP16 weights; the greedy profile removes unused
+top-64 joint calculations. It is unsuitable for decoding paths that require
+those outputs. The experimental precision profile is not included.
+
+This is a preview. Other Macs, older OS versions, power use, and the full
+25-language Core ML evaluation remain unqualified. FluidAudio's default
+sliding-window TDT path buffers roughly 13 seconds before first text. This is
+not the separate Parakeet EOU 120M live-typing model.
+
+## License and attribution
+
+Orukeet modified weights: CC BY-SA 4.0, retaining NVIDIA Parakeet attribution.
+Core ML graphs derive from Fluid Inference's pinned Parakeet conversion.
+FluidAudio runtime: Apache-2.0. Conversion and integration code: MIT.
+The bundles retain `LICENSE-WEIGHTS`, `NOTICE.md`, and `COREML-NOTICE.txt`.
+
+All weights derive from the r3 checkpoint with SHA-256
+`031c8ddab4845aeced904a7cde8e8aa57993b2e344716cf83a545b079c473b56`.
+Baseline graph metadata retains the upstream names for graph identity; use
+`bundle.json` to identify the weights as Orukeet.

coreml/SHA256SUMS.txt

@@ -0,0 +1,2 @@
+beccdc6f18c4b10527a764f6e3ab12e3e11b969220c0cee175b3bb7eaa94290e  orukeet-r3-coreml-greedy.zip
+b2a6efc4ed3280c860f29b3e2e2ea242ade14c6482c94f1c8d3e8551d5edb626  orukeet-r3-coreml-baseline.zip

coreml/manifest.json

@@ -0,0 +1,28 @@
+{
+  "model": "Orukeet",
+  "source": "r3",
+  "source_sha256": "031c8ddab4845aeced904a7cde8e8aa57993b2e344716cf83a545b079c473b56",
+  "repo_id": "oruk/orukeet",
+  "release_tag": "coreml-taptalk-preview-20260915",
+  "source_code_revision": "347f646cacda2e001865b6ac40ba5cbc7e90d1c9",
+  "github_release": "https://github.com/Oruk-AI/orukeet/releases/tag/coreml-taptalk-preview-20260915",
+  "fluid_audio_version": "0.15.5",
+  "archives": {
+    "greedy": {
+      "filename": "orukeet-r3-coreml-greedy.zip",
+      "bytes": 466579943,
+      "sha256": "beccdc6f18c4b10527a764f6e3ab12e3e11b969220c0cee175b3bb7eaa94290e",
+      "verified_files": 18,
+      "compiled_cache_included": false,
+      "archive_root": "orukeet-r3-coreml-greedy/"
+    },
+    "baseline": {
+      "filename": "orukeet-r3-coreml-baseline.zip",
+      "bytes": 466579851,
+      "sha256": "b2a6efc4ed3280c860f29b3e2e2ea242ade14c6482c94f1c8d3e8551d5edb626",
+      "verified_files": 17,
+      "compiled_cache_included": false,
+      "archive_root": "orukeet-r3-coreml-baseline/"
+    }
+  }
+}

coreml/orukeet-r3-coreml-baseline.zip

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b2a6efc4ed3280c860f29b3e2e2ea242ade14c6482c94f1c8d3e8551d5edb626
+size 466579851

coreml/orukeet-r3-coreml-greedy.zip

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:beccdc6f18c4b10527a764f6e3ab12e3e11b969220c0cee175b3bb7eaa94290e
+size 466579943