Add gliner-multi-pii-v1-wo8-2026-10-24: weight-only 8-bit redaction model (float32 compute on every CPU)

.gitattributes

@@ -35,3 +35,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
 nemotron-3-diarization-q8-2026-09-23/Nemotron-3-Diarization.q8_0.gguf filter=lfs diff=lfs merge=lfs -text
 gliner-multi-pii-v1-pc8e-2026-10-23/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+gliner-multi-pii-v1-wo8-2026-10-24/tokenizer.json filter=lfs diff=lfs merge=lfs -text

LICENSE.md

@@ -7,5 +7,5 @@ Each model in this repository keeps its original licence.
 | `parakeet-tdt-0.6b-v3-pc8-*/encoder.int8.onnx`, `decoder.onnx`, `joiner.onnx`, `tokens.txt` | NVIDIA Parakeet TDT 0.6B v3 (modified: int8 quantization of the encoder) | [CC-BY-4.0](licenses/CC-BY-4.0.txt) |
 | `parakeet-tdt-0.6b-v3-pc8-*/silero_vad.onnx` | Silero VAD, Copyright (c) 2020-present Silero Team | [MIT](licenses/Silero-VAD-MIT.txt) |
 | `nemotron-3-diarization-q8-*/Nemotron-3-Diarization.q8_0.gguf` | NVIDIA Nemotron 3 Diarization (unchanged) | [OpenMDW-1.1](licenses/OpenMDW-1.1.txt) |
-| `gliner-multi-pii-v1-pc8e-*/model.onnx`, `gliner_config.json` | GLiNER multi PII v1, by Urchade Zaratiana (modified: ONNX export and per-channel int8 quantization) | [Apache-2.0](licenses/Apache-2.0.txt) |
-| `gliner-multi-pii-v1-pc8e-*/spm.model`, `tokenizer.json` | mDeBERTa-v3 base tokenizer, Copyright (c) Microsoft Corporation | [MIT](licenses/mdeberta-v3-MIT.txt) |
+| `gliner-multi-pii-v1-*/model.onnx`, `gliner_config.json` | GLiNER multi PII v1, by Urchade Zaratiana (modified: ONNX export and 8-bit quantization) | [Apache-2.0](licenses/Apache-2.0.txt) |
+| `gliner-multi-pii-v1-*/spm.model`, `tokenizer.json` | mDeBERTa-v3 base tokenizer, Copyright (c) Microsoft Corporation | [MIT](licenses/mdeberta-v3-MIT.txt) |

README.md

@@ -33,8 +33,21 @@ Speech recognition in 25 European languages, for [sherpa-onnx](https://github.co
 
 - `Nemotron-3-Diarization.q8_0.gguf` is **[NVIDIA Nemotron 3 Diarization](https://huggingface.co/nvidia/Nemotron-3-Diarization)**, by NVIDIA, under the [OpenMDW License Agreement 1.1](licenses/OpenMDW-1.1.txt). It is an unchanged copy of NVIDIA's file at commit `f667ed73aee57d40cc39428eb768b4fd87a0a29e`. NVIDIA's model card, with its notices, is kept in [licenses/Nemotron-3-Diarization-README.md](licenses/Nemotron-3-Diarization-README.md).
 
+## gliner-multi-pii-v1-wo8-2026-10-24
+
+The redaction model private mode uses, replacing `gliner-multi-pii-v1-pc8e-2026-10-23`. It finds names, dates, phone numbers, places, ID numbers and other personal details in a transcript, so private mode can replace them on the device. It runs with ONNX Runtime 1.28 or later.
+
+- `model.onnx` is derived from **[GLiNER multi PII v1](https://huggingface.co/urchade/gliner_multi_pii-v1)** (commit `1fcf13e85f4eef5394e1fcd406cf2ca9ea82351d`), by Urchade Zaratiana, licensed under the [Apache License 2.0](licenses/Apache-2.0.txt). `gliner_config.json` is that model's configuration, unchanged.
+- **What we changed:** the model was exported to ONNX with the `gliner` 0.2.29 library and quantized with ONNX Runtime 1.31's tools, **weights only**:
+  - the vocabulary table is per-channel int8, dequantized on load (`quantize_dynamic`, `op_types_to_quantize=["Gather"]`);
+  - every MatMul is 8-bit `MatMulNBits`, block size 32, symmetric, `accuracy_level=1`.
+- **Why it replaces the earlier pack:** this model computes in float32 throughout. The earlier pack quantized activations to int8 at run time, which depends on the processor's int8 instructions. On processors without them (Apple M1, for one), it removed only about half the identifiers. On our synthetic test set this pack removes 97.0% of personal identifiers.
+- `spm.model` and `tokenizer.json` are the tokenizer of **[mDeBERTa-v3 base](https://huggingface.co/microsoft/mdeberta-v3-base)** (commit `a0484667b22365f84929a935b5e50a51f71f159d`), by Microsoft, under the [MIT licence](licenses/mdeberta-v3-MIT.txt). `spm.model` is unchanged. `tokenizer.json` was converted from it by the Hugging Face `transformers` library and adds GLiNER's two tokens, `<<ENT>>` and `<<SEP>>`.
+
 ## gliner-multi-pii-v1-pc8e-2026-10-23
 
+Superseded by `gliner-multi-pii-v1-wo8-2026-10-24`, and kept unchanged, as every folder here is.
+
 Finds names, dates, phone numbers, places, ID numbers and other personal details in a transcript, so private mode can replace them on the device. It runs with ONNX Runtime 1.28 or later.
 
 - `model.onnx` is derived from **[GLiNER multi PII v1](https://huggingface.co/urchade/gliner_multi_pii-v1)** (commit `1fcf13e85f4eef5394e1fcd406cf2ca9ea82351d`), by Urchade Zaratiana, licensed under the [Apache License 2.0](licenses/Apache-2.0.txt). `gliner_config.json` is that model's configuration, unchanged.
@@ -51,6 +64,10 @@ Finds names, dates, phone numbers, places, ID numbers and other personal details
 | parakeet-tdt-0.6b-v3-pc8-2026-10-15/tokens.txt | 93939 | d58544679ea4bc6ac563d1f545eb7d474bd6cfa467f0a6e2c1dc1c7d37e3c35d |
 | parakeet-tdt-0.6b-v3-pc8-2026-10-15/silero_vad.onnx | 643854 | 9e2449e1087496d8d4caba907f23e0bd3f78d91fa552479bb9c23ac09cbb1fd6 |
 | nemotron-3-diarization-q8-2026-09-23/Nemotron-3-Diarization.q8_0.gguf | 107012128 | 08456d9e22cd9a323c0364d98375f3746d6e68507ebb705cd46438c534c7a3a1 |
+| gliner-multi-pii-v1-wo8-2026-10-24/model.onnx | 365533791 | b7569cce986a70955ba1b3da7b4c98e9cbef2c7e91b657ccf6498a7077110e8b |
+| gliner-multi-pii-v1-wo8-2026-10-24/tokenizer.json | 16034824 | c5b8041501fcdee792b9b112dd592861f04391a24633a73fd9f05aaaf6e8eff1 |
+| gliner-multi-pii-v1-wo8-2026-10-24/spm.model | 4305025 | 13c8d666d62a7bc4ac8f040aab68e942c861f93303156cc28f5c7e885d86d6e3 |
+| gliner-multi-pii-v1-wo8-2026-10-24/gliner_config.json | 2312 | 5645ea576d26777a204be05a8cb97d9c2b15d04496c7712a6bb094c828ef1d9d |
 | gliner-multi-pii-v1-pc8e-2026-10-23/model.onnx | 354286656 | 92fdb25ad1bcc07476cb86d0e3aeafe38a549ad4f2664328daaab00e26237696 |
 | gliner-multi-pii-v1-pc8e-2026-10-23/tokenizer.json | 16034824 | c5b8041501fcdee792b9b112dd592861f04391a24633a73fd9f05aaaf6e8eff1 |
 | gliner-multi-pii-v1-pc8e-2026-10-23/spm.model | 4305025 | 13c8d666d62a7bc4ac8f040aab68e942c861f93303156cc28f5c7e885d86d6e3 |

gliner-multi-pii-v1-wo8-2026-10-24/gliner_config.json

@@ -0,0 +1,91 @@
+{
+  "class_token_index": 250103,
+  "dropout": 0.4,
+  "embed_ent_token": true,
+  "encoder_config": {
+    "_name_or_path": "microsoft/mdeberta-v3-base",
+    "architectures": null,
+    "attention_probs_dropout_prob": 0.1,
+    "bos_token_id": null,
+    "chunk_size_feed_forward": 0,
+    "dtype": null,
+    "eos_token_id": null,
+    "hidden_act": "gelu",
+    "hidden_dropout_prob": 0.1,
+    "hidden_size": 768,
+    "id2label": {
+      "0": "LABEL_0",
+      "1": "LABEL_1"
+    },
+    "initializer_range": 0.02,
+    "intermediate_size": 3072,
+    "is_encoder_decoder": false,
+    "label2id": {
+      "LABEL_0": 0,
+      "LABEL_1": 1
+    },
+    "layer_norm_eps": 1e-07,
+    "legacy": true,
+    "max_position_embeddings": 512,
+    "max_relative_positions": -1,
+    "model_type": "deberta-v2",
+    "norm_rel_ebd": "layer_norm",
+    "num_attention_heads": 12,
+    "num_hidden_layers": 12,
+    "output_attentions": false,
+    "output_hidden_states": false,
+    "pad_token_id": 0,
+    "pooler_dropout": 0.0,
+    "pooler_hidden_act": "gelu",
+    "pooler_hidden_size": 768,
+    "pos_att_type": [
+      "p2c",
+      "c2p"
+    ],
+    "position_biased_input": false,
+    "position_buckets": 256,
+    "problem_type": null,
+    "relative_attention": true,
+    "return_dict": true,
+    "share_att_key": true,
+    "tie_word_embeddings": true,
+    "type_vocab_size": 0,
+    "vocab_size": 250105
+  },
+  "ent_token": "<<ENT>>",
+  "eval_every": 5000,
+  "fine_tune": true,
+  "fuse_layers": false,
+  "hidden_size": 512,
+  "id_to_classes": null,
+  "lr_encoder": "5e-6",
+  "lr_others": "1e-5",
+  "max_len": 384,
+  "max_neg_type_ratio": 3,
+  "max_types": 25,
+  "max_width": 12,
+  "model_name": "microsoft/mdeberta-v3-base",
+  "model_type": null,
+  "name": "correct",
+  "neg_spans_ratio": 1.0,
+  "num_post_fusion_layers": 1,
+  "num_rnn_layers": 1,
+  "num_steps": 30000,
+  "post_fusion_schema": "",
+  "precomputed_prompts_mode": null,
+  "random_drop": false,
+  "represent_spans": false,
+  "sep_token": "<<SEP>>",
+  "shuffle_types": true,
+  "size_sup": -1,
+  "span_encoder_config": null,
+  "span_loss_coef": 1.0,
+  "span_mode": "markerV0",
+  "subtoken_pooling": "first",
+  "token_loss_coef": 1.0,
+  "train_batch_size": 8,
+  "transformers_version": "5.16.1",
+  "vocab_size": 250105,
+  "warmup_ratio": 5000,
+  "words_splitter_type": "whitespace"
+}

gliner-multi-pii-v1-wo8-2026-10-24/model.onnx

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b7569cce986a70955ba1b3da7b4c98e9cbef2c7e91b657ccf6498a7077110e8b
+size 365533791

gliner-multi-pii-v1-wo8-2026-10-24/spm.model

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:13c8d666d62a7bc4ac8f040aab68e942c861f93303156cc28f5c7e885d86d6e3
+size 4305025

gliner-multi-pii-v1-wo8-2026-10-24/tokenizer.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c5b8041501fcdee792b9b112dd592861f04391a24633a73fd9f05aaaf6e8eff1
+size 16034824