Upload Taurscribe Granite portable INT4 DirectML and CPU artifact

.gitattributes

@@ -33,3 +33,6 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+editor.onnx.data filter=lfs diff=lfs merge=lfs -text
+encoder.onnx.data filter=lfs diff=lfs merge=lfs -text
+projector.onnx.data filter=lfs diff=lfs merge=lfs -text

BENCHMARKS.md

@@ -0,0 +1,42 @@
+# Benchmark Notes
+
+## End-to-End Route Comparison
+
+All routes were measured with the current Taurscribe Rust runtime and the same
+first 30 entries from `eval_manifest_30.jsonl`. Audio was decoded, converted to
+16 kHz mono, edge-trimmed, and processed through the same Granite pipeline.
+
+Test machine:
+
+- CPU: AMD Ryzen 7 8845HS
+- integrated GPU: AMD Radeon 780M
+- discrete GPU: NVIDIA GeForce RTX 4070 Laptop GPU
+- sample count: 30 LibriSpeech `test-clean` utterances
+- mean processed audio duration: 8.47 seconds
+
+| Route | Load | Mean | Median | P95 | Mean RTF | Mean WER |
+|---|---:|---:|---:|---:|---:|---:|
+| CUDA artifact, RTX 4070 | 16.610 s | 0.250 s | 0.245 s | 0.303 s | 0.040 | 4.31% |
+| Portable, full DirectML, Radeon 780M | 12.770 s | 4.045 s | 3.513 s | 4.518 s | 0.643 | 4.31% |
+| Portable, CPU, 8 intra-op threads | 4.698 s | 8.823 s | 8.840 s | 10.168 s | 1.415 | 4.31% |
+| Historical hybrid, 1-thread CPU encoder + DirectML | 8.421 s | 13.135 s | 12.874 s | 14.161 s | 2.194 | 4.31% |
+
+CUDA load time includes a 5.594-second performance warmup. DirectML was forced
+to DXGI device 0, identified as the Radeon 780M on this laptop. Every portable
+graph was loaded on DirectML and runtime fallback was disabled for that run.
+
+## Interpretation
+
+- Full DirectML was approximately 2.18 times faster than eight-thread CPU on
+  this end-to-end workload.
+- The earlier 5.2-second CPU prediction came from adding isolated per-graph
+  probes and was not representative of full transcription. The measured
+  end-to-end CPU mean is 8.823 seconds.
+- All routes produced the same 4.31% mean WER on this subset, so the execution
+  route did not change recognition accuracy here.
+- Thirty utterances are sufficient for route validation but not for publishing
+  a general model-quality claim. Consult the IBM base model card for the full
+  Granite evaluation.
+
+RTF is transcription time divided by processed audio duration. Values below
+1.0 are faster than real time.

LICENSE

@@ -0,0 +1,176 @@
+                                 Apache License
+                           Version 2.0, January 2004
+                        http://www.apache.org/licenses/
+
+   TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
+
+   1. Definitions.
+
+      "License" shall mean the terms and conditions for use, reproduction,
+      and distribution as defined by Sections 1 through 9 of this document.
+
+      "Licensor" shall mean the copyright owner or entity authorized by
+      the copyright owner that is granting the License.
+
+      "Legal Entity" shall mean the union of the acting entity and all
+      other entities that control, are controlled by, or are under common
+      control with that entity. For the purposes of this definition,
+      "control" means (i) the power, direct or indirect, to cause the
+      direction or management of such entity, whether by contract or
+      otherwise, or (ii) ownership of fifty percent (50%) or more of the
+      outstanding shares, or (iii) beneficial ownership of such entity.
+
+      "You" (or "Your") shall mean an individual or Legal Entity
+      exercising permissions granted by this License.
+
+      "Source" form shall mean the preferred form for making modifications,
+      including but not limited to software source code, documentation
+      source, and configuration files.
+
+      "Object" form shall mean any form resulting from mechanical
+      transformation or translation of a Source form, including but
+      not limited to compiled object code, generated documentation,
+      and conversions to other media types.
+
+      "Work" shall mean the work of authorship, whether in Source or
+      Object form, made available under the License, as indicated by a
+      copyright notice that is included in or attached to the work
+      (an example is provided in the Appendix below).
+
+      "Derivative Works" shall mean any work, whether in Source or Object
+      form, that is based on (or derived from) the Work and for which the
+      editorial revisions, annotations, elaborations, or other modifications
+      represent, as a whole, an original work of authorship. For the purposes
+      of this License, Derivative Works shall not include works that remain
+      separable from, or merely link (or bind by name) to the interfaces of,
+      the Work and Derivative Works thereof.
+
+      "Contribution" shall mean any work of authorship, including
+      the original version of the Work and any modifications or additions
+      to that Work or Derivative Works thereof, that is intentionally
+      submitted to Licensor for inclusion in the Work by the copyright owner
+      or by an individual or Legal Entity authorized to submit on behalf of
+      the copyright owner. For the purposes of this definition, "submitted"
+      means any form of electronic, verbal, or written communication sent
+      to the Licensor or its representatives, including but not limited to
+      communication on electronic mailing lists, source code control systems,
+      and issue tracking systems that are managed by, or on behalf of, the
+      Licensor for the purpose of discussing and improving the Work, but
+      excluding communication that is conspicuously marked or otherwise
+      designated in writing by the copyright owner as "Not a Contribution."
+
+      "Contributor" shall mean Licensor and any individual or Legal Entity
+      on behalf of whom a Contribution has been received by Licensor and
+      subsequently incorporated within the Work.
+
+   2. Grant of Copyright License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      copyright license to reproduce, prepare Derivative Works of,
+      publicly display, publicly perform, sublicense, and distribute the
+      Work and such Derivative Works in Source or Object form.
+
+   3. Grant of Patent License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      (except as stated in this section) patent license to make, have made,
+      use, offer to sell, sell, import, and otherwise transfer the Work,
+      where such license applies only to those patent claims licensable
+      by such Contributor that are necessarily infringed by their
+      Contribution(s) alone or by combination of their Contribution(s)
+      with the Work to which such Contribution(s) was submitted. If You
+      institute patent litigation against any entity (including a
+      cross-claim or counterclaim in a lawsuit) alleging that the Work
+      or a Contribution incorporated within the Work constitutes direct
+      or contributory patent infringement, then any patent licenses
+      granted to You under this License for that Work shall terminate
+      as of the date such litigation is filed.
+
+   4. Redistribution. You may reproduce and distribute copies of the
+      Work or Derivative Works thereof in any medium, with or without
+      modifications, and in Source or Object form, provided that You
+      meet the following conditions:
+
+      (a) You must give any other recipients of the Work or
+          Derivative Works a copy of this License; and
+
+      (b) You must cause any modified files to carry prominent notices
+          stating that You changed the files; and
+
+      (c) You must retain, in the Source form of any Derivative Works
+          that You distribute, all copyright, patent, trademark, and
+          attribution notices from the Source form of the Work,
+          excluding those notices that do not pertain to any part of
+          the Derivative Works; and
+
+      (d) If the Work includes a "NOTICE" text file as part of its
+          distribution, then any Derivative Works that You distribute must
+          include a readable copy of the attribution notices contained
+          within such NOTICE file, excluding those notices that do not
+          pertain to any part of the Derivative Works, in at least one
+          of the following places: within a NOTICE text file distributed
+          as part of the Derivative Works; within the Source form or
+          documentation, if provided along with the Derivative Works; or,
+          within a display generated by the Derivative Works, if and
+          wherever such third-party notices normally appear. The contents
+          of the NOTICE file are for informational purposes only and
+          do not modify the License. You may add Your own attribution
+          notices within Derivative Works that You distribute, alongside
+          or as an addendum to the NOTICE text from the Work, provided
+          that such additional attribution notices cannot be construed
+          as modifying the License.
+
+      You may add Your own copyright statement to Your modifications and
+      may provide additional or different license terms and conditions
+      for use, reproduction, or distribution of Your modifications, or
+      for any such Derivative Works as a whole, provided Your use,
+      reproduction, and distribution of the Work otherwise complies with
+      the conditions stated in this License.
+
+   5. Submission of Contributions. Unless You explicitly state otherwise,
+      any Contribution intentionally submitted for inclusion in the Work
+      by You to the Licensor shall be under the terms and conditions of
+      this License, without any additional terms or conditions.
+      Notwithstanding the above, nothing herein shall supersede or modify
+      the terms of any separate license agreement you may have executed
+      with Licensor regarding such Contributions.
+
+   6. Trademarks. This License does not grant permission to use the trade
+      names, trademarks, service marks, or product names of the Licensor,
+      except as required for reasonable and customary use in describing the
+      origin of the Work and reproducing the content of the NOTICE file.
+
+   7. Disclaimer of Warranty. Unless required by applicable law or
+      agreed to in writing, Licensor provides the Work (and each
+      Contributor provides its Contributions) on an "AS IS" BASIS,
+      WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
+      implied, including, without limitation, any warranties or conditions
+      of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
+      PARTICULAR PURPOSE. You are solely responsible for determining the
+      appropriateness of using or redistributing the Work and assume any
+      risks associated with Your exercise of permissions under this License.
+
+   8. Limitation of Liability. In no event and under no legal theory,
+      whether in tort (including negligence), contract, or otherwise,
+      unless required by applicable law (such as deliberate and grossly
+      negligent acts) or agreed to in writing, shall any Contributor be
+      liable to You for damages, including any direct, indirect, special,
+      incidental, or consequential damages of any character arising as a
+      result of this License or out of the use or inability to use the
+      Work (including but not limited to damages for loss of goodwill,
+      work stoppage, computer failure or malfunction, or any and all
+      other commercial damages or losses), even if such Contributor
+      has been advised of the possibility of such damages.
+
+   9. Accepting Warranty or Additional Liability. While redistributing
+      the Work or Derivative Works thereof, You may choose to offer,
+      and charge a fee for, acceptance of support, warranty, indemnity,
+      or other liability obligations and/or rights consistent with this
+      License. However, in accepting such obligations, You may act only
+      on Your own behalf and on Your sole responsibility, not on behalf
+      of any other Contributor, and only if You agree to indemnify,
+      defend, and hold each Contributor harmless for any liability
+      incurred by, or claims asserted against, such Contributor by reason
+      of your accepting any such warranty or additional liability.
+
+   END OF TERMS AND CONDITIONS

NOTICE

@@ -0,0 +1,22 @@
+Granite Speech 4.1 2B NAR Portable INT4 for Taurscribe
+
+This artifact is derived from IBM Granite Speech 4.1 2B NAR:
+
+  https://huggingface.co/ibm-granite/granite-speech-4.1-2b-nar
+
+Original model provider:
+  IBM Granite Team
+
+Original license:
+  Apache License 2.0
+
+Taurscribe modifications:
+  - Copied the Granite ONNX model bundle and tokenizer assets.
+  - Applied INT4 weight-only MatMulNBits quantization to selected MatMul weights.
+  - Patched editor.onnx to output token_ids through in-graph ArgMax.
+  - Flattened rank-5 windowed-attention MatMul operations to equivalent rank-3 operations for DirectML.
+  - Baked fixed-bucket shape calculations as ONNX initializers and pruned dead shape nodes.
+  - Replaced two-output GLU Split operations with equivalent Slice pairs for DirectML correctness.
+  - Packaged the artifact for Taurscribe's ONNX Runtime DirectML and CPU paths.
+
+No additional training or fine-tuning was performed by Taurscribe.

README.md

@@ -0,0 +1,176 @@
+---
+license: apache-2.0
+base_model: ibm-granite/granite-speech-4.1-2b-nar
+library_name: onnx
+pipeline_tag: automatic-speech-recognition
+language:
+  - en
+  - fr
+  - de
+  - es
+  - pt
+tags:
+  - automatic-speech-recognition
+  - granite-speech
+  - onnx
+  - onnxruntime
+  - directml
+  - cpu
+  - int4
+  - matmulnbits
+  - taurscribe
+---
+
+# Granite Speech 4.1 2B NAR Portable INT4 for Taurscribe
+
+This repository contains a portable ONNX Runtime artifact derived from IBM's
+[Granite Speech 4.1 2B NAR](https://huggingface.co/ibm-granite/granite-speech-4.1-2b-nar)
+model for local transcription in Taurscribe.
+
+It is designed for Windows systems without an NVIDIA CUDA GPU. It supports:
+
+- full GPU inference through ONNX Runtime DirectML on compatible AMD, Intel,
+  and NVIDIA DirectX 12 GPUs;
+- multi-threaded ONNX Runtime CPU inference as a portable fallback.
+
+This is not a newly trained or fine-tuned model. The learned Granite weights,
+tokenizer, vocabulary, and supported languages are inherited from IBM's base
+model. Taurscribe changed the ONNX representation and host runtime path.
+
+> This is a multi-graph Taurscribe runtime bundle, not a single-file
+> Transformers replacement. Correct inference requires Taurscribe's feature
+> extraction, CTC, projection, edit-slot, and tokenizer pipeline.
+
+## What Changed
+
+The portable artifact builds on Taurscribe's INT4/token-ID Granite bundle:
+
+1. Selected large `MatMul` weights use INT4 weight-only quantization through
+   ONNX `MatMulNBits`.
+2. `editor.onnx` performs `ArgMax` in the graph and returns `token_ids` instead
+   of transferring the full sequence-by-vocabulary logits tensor to Rust.
+3. Thirty-two rank-5 windowed-attention `MatMul` operations in `encoder.onnx`
+   are reshaped to equivalent rank-3 batched operations supported by DirectML.
+4. Runtime shape chains are evaluated for Taurscribe's fixed `[1, 800, 160]`
+   encoder bucket, stored as constants, and approximately 950 dead shape nodes
+   are removed.
+5. Sixteen two-output GLU `Split` operations are replaced with equivalent
+   `Slice` pairs to avoid incorrect output inside DirectML fused partitions.
+
+Only `encoder.onnx` receives the DirectML compatibility rewrites. The
+projector, token embedding graph, editor, tokenizer, and model weights are
+copied from the INT4/token-ID parent artifact.
+
+## Runtime Placement
+
+Full DirectML:
+
+```text
+encoder.onnx      -> DirectML
+projector.onnx    -> DirectML
+embed_tokens.onnx -> DirectML
+editor.onnx       -> DirectML
+```
+
+CPU fallback:
+
+```text
+encoder.onnx      -> CPU
+projector.onnx    -> CPU
+embed_tokens.onnx -> CPU
+editor.onnx       -> CPU
+```
+
+The ONNX files do not select a GPU vendor. Taurscribe's Rust code asks ONNX
+Runtime to create DirectML or CPU sessions. See
+`taurscribe_granite_nar_manifest.json` for the graph contract.
+
+## Validation
+
+The conversion script compares every rewritten encoder output against the
+source encoder on CPU and aborts if the worst relative difference exceeds
+`1e-3`. The produced artifact measured a worst relative difference of about
+`4.4e-5`. DirectML encoder BPE argmax agreement with CPU was `1.0000` on the
+validated input.
+
+The following end-to-end Taurscribe results use the same 30 utterances from
+LibriSpeech `test-clean` (mean processed duration: 8.47 seconds):
+
+| Route | Hardware | Mean latency | Mean RTF | Mean WER |
+|---|---|---:|---:|---:|
+| CUDA reference artifact | RTX 4070 Laptop | 0.250 s | 0.040 | 4.31% |
+| **This artifact, full DirectML** | **Radeon 780M** | **4.045 s** | **0.643** | **4.31%** |
+| **This artifact, CPU (8 threads)** | **Ryzen 7 8845HS** | **8.823 s** | **1.415** | **4.31%** |
+| Historical hybrid baseline | 1-thread CPU encoder + DirectML | 13.135 s | 2.194 | 4.31% |
+
+These are limited local validation results, not a replacement for the base
+model's full published evaluation. See `BENCHMARKS.md` for methodology and tail
+latency.
+
+## Hardware Compatibility
+
+- **AMD Radeon 780M DirectML:** validated end to end with CPU fallback disabled.
+- **NVIDIA DirectML:** graph-level compatibility validated; use the separate
+  CUDA artifact for substantially better NVIDIA performance.
+- **Intel DirectML:** expected to work on DirectX 12 hardware but not yet
+  validated with this exact artifact.
+- **x86-64 CPU:** supported through ONNX Runtime CPU; actual speed depends
+  heavily on memory bandwidth and thread count.
+
+DirectML requires Windows and a DirectX 12-compatible GPU. A GPU driver or
+execution-provider failure should be handled by the host application with CPU
+fallback. For production hardware certification, run the DirectML probe in a
+separate helper process so a native driver crash cannot terminate the app.
+
+## Taurscribe Selection
+
+Taurscribe model ID:
+
+```text
+granite-speech-4.1-2b-nar-portable
+```
+
+Development overrides:
+
+```powershell
+$env:TAURSCRIBE_GRANITE_BACKEND = "directml"
+$env:TAURSCRIBE_GRANITE_DML_DEVICE_ID = "0"
+$env:TAURSCRIBE_GRANITE_BACKEND = "cpu"
+$env:TAURSCRIBE_GRANITE_CPU_THREADS = "8"
+```
+
+The DirectML device ID follows DXGI adapter order and is machine-specific.
+
+## Files
+
+```text
+encoder.onnx
+encoder.onnx.data
+projector.onnx
+projector.onnx.data
+embed_tokens.onnx
+editor.onnx
+editor.onnx.data
+tokenizer.json
+tokenizer_config.json
+preprocessor_config.json
+processor_config.json
+generation_config.json
+taurscribe_granite_nar_manifest.json
+manifest.json
+BENCHMARKS.md
+NOTICE
+LICENSE
+```
+
+`manifest.json` records artifact metadata, exact file sizes, and SHA-256
+checksums for the runtime payload.
+
+## Source and Attribution
+
+- Base model: [ibm-granite/granite-speech-4.1-2b-nar](https://huggingface.co/ibm-granite/granite-speech-4.1-2b-nar)
+- Original provider: IBM Granite Team
+- Original license: Apache License 2.0
+- ONNX optimization and packaging: Taurscribe project
+
+No additional training or fine-tuning was performed by Taurscribe.

editor.onnx

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:40dc0e1eafa152fd6fdcb0ca8030128469906501a63ea4d4ae83df195d23d792
+size 1630677

editor.onnx.data

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:72704971b72bf00c281a859825cb72b25b7768239d009abe52fb4fbc6ddc009b
+size 873816064

embed_tokens.onnx

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:10f33a9646ea843d65c761c5295d8487e1091bf66de587f395ccec5b44c71ad3
+size 822083791

encoder.onnx

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e4e34584c2de0ebbf95d318a386bb6f150ae62eb9505837dae3e82190a72859a
+size 300910

encoder.onnx.data

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:038e259276a1119ad03afcd8fce16ff50cd3a98db535cd68be0b5a71bbcf3318
+size 650685536

generation_config.json

@@ -0,0 +1,3 @@
+{
+  "transformers_version": "5.8.1"
+}

manifest.json

@@ -0,0 +1,97 @@
+{
+  "name": "granite-speech-4.1-2b-nar-portable",
+  "display_name": "Granite Speech 4.1 2B NAR Portable INT4",
+  "artifact_type": "onnxruntime-directml-cpu",
+  "created_at_utc": "2026-07-10T00:03:54.0897852Z",
+  "source_model": "ibm-granite/granite-speech-4.1-2b-nar",
+  "source_license": "Apache-2.0",
+  "target_runtime": "onnxruntime",
+  "target_execution_providers": [
+    "DmlExecutionProvider",
+    "CPUExecutionProvider"
+  ],
+  "intended_app": "Taurscribe",
+  "expected_editor_output": "token_ids",
+  "fixed_encoder_frames": 800,
+  "encoder_dml_safe": true,
+  "size_bytes": 2398658963,
+  "graphs": {
+    "encoder": "encoder.onnx",
+    "projector": "projector.onnx",
+    "embed_tokens": "embed_tokens.onnx",
+    "editor": "editor.onnx"
+  },
+  "modifications": [
+    "Applied INT4 weight-only MatMulNBits quantization to selected ONNX MatMul weights",
+    "Patched editor.onnx with ArgMax so the graph outputs token_ids instead of full logits",
+    "Flattened 32 rank-5 windowed-attention MatMul operations to equivalent rank-3 batched MatMul operations",
+    "Baked fixed 800-frame shape-slot tensors as initializers and pruned approximately 950 dead nodes",
+    "Replaced 16 two-output GLU Split operations with equivalent Slice pairs for DirectML correctness",
+    "Packaged for Taurscribe ONNX Runtime DirectML with CPU fallback"
+  ],
+  "validation": {
+    "encoder_cpu_worst_relative_difference": 0.000044,
+    "directml_bpe_argmax_agreement": 1.0,
+    "librispeech_subset_utterances": 30,
+    "portable_directml_mean_seconds": 4.0454,
+    "portable_directml_mean_rtf": 0.6429,
+    "portable_directml_mean_wer": 0.0431,
+    "portable_cpu_8_threads_mean_seconds": 8.8227,
+    "portable_cpu_8_threads_mean_rtf": 1.4145,
+    "portable_cpu_8_threads_mean_wer": 0.0431
+  },
+  "files": {
+    "editor.onnx": {
+      "size_bytes": 1630677,
+      "sha256": "40dc0e1eafa152fd6fdcb0ca8030128469906501a63ea4d4ae83df195d23d792"
+    },
+    "editor.onnx.data": {
+      "size_bytes": 873816064,
+      "sha256": "72704971b72bf00c281a859825cb72b25b7768239d009abe52fb4fbc6ddc009b"
+    },
+    "embed_tokens.onnx": {
+      "size_bytes": 822083791,
+      "sha256": "10f33a9646ea843d65c761c5295d8487e1091bf66de587f395ccec5b44c71ad3"
+    },
+    "encoder.onnx": {
+      "size_bytes": 300910,
+      "sha256": "e4e34584c2de0ebbf95d318a386bb6f150ae62eb9505837dae3e82190a72859a"
+    },
+    "encoder.onnx.data": {
+      "size_bytes": 650685536,
+      "sha256": "038e259276a1119ad03afcd8fce16ff50cd3a98db535cd68be0b5a71bbcf3318"
+    },
+    "generation_config.json": {
+      "size_bytes": 38,
+      "sha256": "af7cd9bbb214dea4f6a37756c2c133d002f6325ed4f497efc991f02b4d486dad"
+    },
+    "preprocessor_config.json": {
+      "size_bytes": 289,
+      "sha256": "e12be1e9d4ec5c459741328f28d9d8c00c3c688e8fe4230a6ec28df5470db8b3"
+    },
+    "processor_config.json": {
+      "size_bytes": 153,
+      "sha256": "74d21364b507dcbe465420152d48edd9276f94a84d22a78d0da371e716a79374"
+    },
+    "projector.onnx": {
+      "size_bytes": 33527,
+      "sha256": "0641d423a5700d6b16c0d8a47a25b58e750b1e2d065f26a670b72a78930968b2"
+    },
+    "projector.onnx.data": {
+      "size_bytes": 42950656,
+      "sha256": "d660a66a75178c1b111d154de6b6cfe6a8183111c68ada33948e42ba0acefc19"
+    },
+    "taurscribe_granite_nar_manifest.json": {
+      "size_bytes": 3128,
+      "sha256": "bb59c18cdfa4297ff42d7f76d27911030fb118e04f2291b15e488e32dc92a47e"
+    },
+    "tokenizer_config.json": {
+      "size_bytes": 392,
+      "sha256": "c7e48adce9bdf6cfe3524759067b5dfa2428de9ccba2fd257502bf0246161c2f"
+    },
+    "tokenizer.json": {
+      "size_bytes": 7153802,
+      "sha256": "64c10a88b2495872bd7da5a885861a1757d9c23590c40fd378546ae176d280f6"
+    }
+  }
+}

preprocessor_config.json

@@ -0,0 +1,11 @@
+{
+  "auto_map": {
+    "AutoFeatureExtractor": "feature_extraction_granite_speech_nar.GraniteSpeechNarFeatureExtractor"
+  },
+  "feature_extractor_type": "GraniteSpeechNarFeatureExtractor",
+  "hop_length": 160,
+  "n_fft": 512,
+  "n_mels": 80,
+  "sampling_rate": 16000,
+  "win_length": 400
+}

processor_config.json

@@ -0,0 +1,6 @@
+{
+  "processor_class": "GraniteSpeechNarProcessor",
+  "auto_map": {
+    "AutoProcessor": "processing_granite_speech_nar.GraniteSpeechNarProcessor"
+  }
+}

projector.onnx

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0641d423a5700d6b16c0d8a47a25b58e750b1e2d065f26a670b72a78930968b2
+size 33527

projector.onnx.data

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d660a66a75178c1b111d154de6b6cfe6a8183111c68ada33948e42ba0acefc19
+size 42950656

taurscribe_granite_nar_manifest.json

@@ -0,0 +1,81 @@
+{
+  "format": "taurscribe-granite-nar-onnx-bundle",
+  "format_version": 1,
+  "source_model": "ibm-granite/granite-speech-4.1-2b-nar",
+  "export_dtype": "int4-matmulnbits-weights",
+  "fixed_encoder_frames": 800,
+  "opset": 20,
+  "execution_provider_preference": [
+    "DmlExecutionProvider",
+    "CPUExecutionProvider"
+  ],
+  "graphs": {
+    "encoder": {
+      "file": "encoder.onnx",
+      "inputs": [
+        "input_features",
+        "attention_mask"
+      ],
+      "outputs": [
+        "bpe_logits",
+        "hidden_4",
+        "hidden_8",
+        "hidden_12",
+        "hidden_last"
+      ],
+      "notes": "Fixed-shape app bucket: pad input_features to 800 frames. Trim bpe_logits to ceil(valid_frames / bpe_pooling_window) before CTC collapse."
+    },
+    "projector": {
+      "file": "projector.onnx",
+      "inputs": [
+        "multilayer_features"
+      ],
+      "outputs": [
+        "audio_embeds"
+      ]
+    },
+    "embed_tokens": {
+      "file": "embed_tokens.onnx",
+      "inputs": [
+        "token_ids"
+      ],
+      "outputs": [
+        "text_embeds"
+      ]
+    },
+    "editor": {
+      "file": "editor.onnx",
+      "inputs": [
+        "inputs_embeds",
+        "position_ids"
+      ],
+      "outputs": [
+        "token_ids"
+      ],
+      "notes": "Non-causal bidirectional editor. Host code should pass one flattened sample sequence at a time."
+    }
+  },
+  "host_pipeline": [
+    "Extract Granite log-mel features with the copied preprocessor config and pad/truncate to 800 encoder frames.",
+    "Run encoder.onnx.",
+    "Trim padded BPE logits using the valid frame count, argmax, unique-consecutive collapse, and remove blank_token_id.",
+    "Concatenate hidden_4, hidden_8, hidden_12, and hidden_last, then run projector.onnx.",
+    "Trim audio embeddings using floor(valid_frames / projector.downsample_rate).",
+    "Insert blank edit slots around the encoder CTC token IDs.",
+    "Run embed_tokens.onnx for the slotted token IDs.",
+    "Concatenate audio embeddings and text embeddings, create monotonic position_ids, then run editor.onnx.",
+    "Select the text segment logits, argmax, unique-consecutive collapse, remove blank_token_id, and decode with tokenizer.json."
+  ],
+  "validation": {
+    "onnx_checker": "run scripts/granite_nar_export.py validate --model-dir <onnx-bundle>",
+    "ort_cpu_session_load": "run scripts/granite_nar_export.py validate --model-dir <onnx-bundle>"
+  },
+  "notes": [
+    "Fallback is controlled by ONNX Runtime session creation in Rust, not encoded inside the ONNX files.",
+    "INT4 weight-only MatMulNBits bundle with an editor that returns token_ids instead of full vocabulary logits.",
+    "Portable bundle: rank-5 attention MatMuls are flattened to rank 3, fixed 800-frame shape chains are baked as constants, and GLU Split nodes are replaced with Slice pairs so the full encoder runs correctly on DirectML.",
+    "DirectML is preferred on compatible Windows GPUs; ONNX Runtime CPU remains the portable fallback."
+  ],
+  "variant": "int4-argmax-dml-static",
+  "encoder_dml_safe": true
+}

tokenizer_config.json

@@ -0,0 +1,14 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": "<|end_of_text|>",
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|end_of_text|>",
+  "is_local": true,
+  "local_files_only": false,
+  "model_max_length": 1000000000000000019884624838656,
+  "pad_token": "<|pad|>",
+  "padding_side": "left",
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": "<|unk|>"
+}