Upload folder using huggingface_hub
Browse files- .gitattributes +1 -0
- LICENSE-MODEL.md +222 -0
- NOTICE +17 -0
- README.md +130 -0
- SHA256SUMS +14 -0
- evaluation/pii.adapter.jsonl +16 -0
- evaluation/pii.base.jsonl +16 -0
- evaluation/published-f32-adapter.jsonl +16 -0
- evaluation/published-f32-base.jsonl +16 -0
- evaluation/published-q8_0-adapter.jsonl +16 -0
- evaluation/published-q8_0-base.jsonl +16 -0
- evaluation/summary.json +279 -0
- peft/adapter_config.json +50 -0
- peft/adapter_model.safetensors +3 -0
- statim-decide-multilingual-base-pii.lora.gguf +3 -0
- training/train_lora.json +165 -0
.gitattributes
CHANGED
|
@@ -33,3 +33,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
|
|
| 33 |
*.zip filter=lfs diff=lfs merge=lfs -text
|
| 34 |
*.zst filter=lfs diff=lfs merge=lfs -text
|
| 35 |
*tfevents* filter=lfs diff=lfs merge=lfs -text
|
|
|
|
|
|
| 33 |
*.zip filter=lfs diff=lfs merge=lfs -text
|
| 34 |
*.zst filter=lfs diff=lfs merge=lfs -text
|
| 35 |
*tfevents* filter=lfs diff=lfs merge=lfs -text
|
| 36 |
+
statim-decide-multilingual-base-pii.lora.gguf filter=lfs diff=lfs merge=lfs -text
|
LICENSE-MODEL.md
ADDED
|
@@ -0,0 +1,222 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# Statim model weights licence
|
| 2 |
+
|
| 3 |
+
This licence applies to **model weights** published by the Statim project (fine-tuned checkpoints and
|
| 4 |
+
their GGUF and safetensors files), not to the source code, which is licensed under the Apache
|
| 5 |
+
License 2.0 (see [LICENSE](LICENSE)).
|
| 6 |
+
|
| 7 |
+
You may use the weights under **any one** of the following licences, whichever fits you:
|
| 8 |
+
|
| 9 |
+
| Licence | For |
|
| 10 |
+
|---|---|
|
| 11 |
+
| [PolyForm Noncommercial 1.0.0](#polyform-noncommercial-license-100) | Personal use, research, experiments, hobby projects and noncommercial organisations such as charities, schools, public research organisations and government institutions. |
|
| 12 |
+
| [PolyForm Small Business 1.0.0](#polyform-small-business-license-100) | Companies with fewer than 100 people (employees and independent contractors) and less than 1,000,000 USD (2019, adjusted for inflation) total revenue in the prior tax year, **including commercial use**. |
|
| 13 |
+
| [PolyForm Free Trial 1.0.0](#polyform-free-trial-license-100) | Any company evaluating whether the weights suit a particular application, for fewer than 32 consecutive calendar days. |
|
| 14 |
+
| Statim commercial licence | Every other commercial use. See [COMMERCIAL.md](COMMERCIAL.md). |
|
| 15 |
+
|
| 16 |
+
The summaries above are for orientation only; the licence texts below, reproduced verbatim from
|
| 17 |
+
<https://github.com/polyformproject/polyform-licenses>, are authoritative.
|
| 18 |
+
|
| 19 |
+
Required Notice: Copyright 2026 Belkis Aslani (https://github.com/BEKO2210/statim)
|
| 20 |
+
Required Notice: Statim weights are fine-tuned from Laya checkpoints (https://huggingface.co/convaiinnovations), Copyright their authors, licensed under the Apache License 2.0.
|
| 21 |
+
|
| 22 |
+
---
|
| 23 |
+
|
| 24 |
+
# PolyForm Noncommercial License 1.0.0
|
| 25 |
+
|
| 26 |
+
<https://polyformproject.org/licenses/noncommercial/1.0.0>
|
| 27 |
+
|
| 28 |
+
## Acceptance
|
| 29 |
+
|
| 30 |
+
In order to get any license under these terms, you must agree to them as both strict obligations and conditions to all your licenses.
|
| 31 |
+
|
| 32 |
+
## Copyright License
|
| 33 |
+
|
| 34 |
+
The licensor grants you a copyright license for the software to do everything you might do with the software that would otherwise infringe the licensor's copyright in it for any permitted purpose. However, you may only distribute the software according to [Distribution License](#distribution-license) and make changes or new works based on the software according to [Changes and New Works License](#changes-and-new-works-license).
|
| 35 |
+
|
| 36 |
+
## Distribution License
|
| 37 |
+
|
| 38 |
+
The licensor grants you an additional copyright license to distribute copies of the software. Your license to distribute covers distributing the software with changes and new works permitted by [Changes and New Works License](#changes-and-new-works-license).
|
| 39 |
+
|
| 40 |
+
## Notices
|
| 41 |
+
|
| 42 |
+
You must ensure that anyone who gets a copy of any part of the software from you also gets a copy of these terms or the URL for them above, as well as copies of any plain-text lines beginning with `Required Notice:` that the licensor provided with the software. For example:
|
| 43 |
+
|
| 44 |
+
> Required Notice: Copyright Yoyodyne, Inc. (http://example.com)
|
| 45 |
+
|
| 46 |
+
## Changes and New Works License
|
| 47 |
+
|
| 48 |
+
The licensor grants you an additional copyright license to make changes and new works based on the software for any permitted purpose.
|
| 49 |
+
|
| 50 |
+
## Patent License
|
| 51 |
+
|
| 52 |
+
The licensor grants you a patent license for the software that covers patent claims the licensor can license, or becomes able to license, that you would infringe by using the software.
|
| 53 |
+
|
| 54 |
+
## Noncommercial Purposes
|
| 55 |
+
|
| 56 |
+
Any noncommercial purpose is a permitted purpose.
|
| 57 |
+
|
| 58 |
+
## Personal Uses
|
| 59 |
+
|
| 60 |
+
Personal use for research, experiment, and testing for the benefit of public knowledge, personal study, private entertainment, hobby projects, amateur pursuits, or religious observance, without any anticipated commercial application, is use for a permitted purpose.
|
| 61 |
+
|
| 62 |
+
## Noncommercial Organizations
|
| 63 |
+
|
| 64 |
+
Use by any charitable organization, educational institution, public research organization, public safety or health organization, environmental protection organization, or government institution is use for a permitted purpose regardless of the source of funding or obligations resulting from the funding.
|
| 65 |
+
|
| 66 |
+
## Fair Use
|
| 67 |
+
|
| 68 |
+
You may have "fair use" rights for the software under the law. These terms do not limit them.
|
| 69 |
+
|
| 70 |
+
## No Other Rights
|
| 71 |
+
|
| 72 |
+
These terms do not allow you to sublicense or transfer any of your licenses to anyone else, or prevent the licensor from granting licenses to anyone else. These terms do not imply any other licenses.
|
| 73 |
+
|
| 74 |
+
## Patent Defense
|
| 75 |
+
|
| 76 |
+
If you make any written claim that the software infringes or contributes to infringement of any patent, your patent license for the software granted under these terms ends immediately. If your company makes such a claim, your patent license ends immediately for work on behalf of your company.
|
| 77 |
+
|
| 78 |
+
## Violations
|
| 79 |
+
|
| 80 |
+
The first time you are notified in writing that you have violated any of these terms, or done anything with the software not covered by your licenses, your licenses can nonetheless continue if you come into full compliance with these terms, and take practical steps to correct past violations, within 32 days of receiving notice. Otherwise, all your licenses end immediately.
|
| 81 |
+
|
| 82 |
+
## No Liability
|
| 83 |
+
|
| 84 |
+
***As far as the law allows, the software comes as is, without any warranty or condition, and the licensor will not be liable to you for any damages arising out of these terms or the use or nature of the software, under any kind of legal claim.***
|
| 85 |
+
|
| 86 |
+
## Definitions
|
| 87 |
+
|
| 88 |
+
The **licensor** is the individual or entity offering these terms, and the **software** is the software the licensor makes available under these terms.
|
| 89 |
+
|
| 90 |
+
**You** refers to the individual or entity agreeing to these terms.
|
| 91 |
+
|
| 92 |
+
**Your company** is any legal entity, sole proprietorship, or other kind of organization that you work for, plus all organizations that have control over, are under the control of, or are under common control with that organization. **Control** means ownership of substantially all the assets of an entity, or the power to direct its management and policies by vote, contract, or otherwise. Control can be direct or indirect.
|
| 93 |
+
|
| 94 |
+
**Your licenses** are all the licenses granted to you for the software under these terms.
|
| 95 |
+
|
| 96 |
+
**Use** means anything you do with the software requiring one of your licenses.
|
| 97 |
+
|
| 98 |
+
---
|
| 99 |
+
|
| 100 |
+
# PolyForm Small Business License 1.0.0
|
| 101 |
+
|
| 102 |
+
<https://polyformproject.org/licenses/small-business/1.0.0>
|
| 103 |
+
|
| 104 |
+
## Acceptance
|
| 105 |
+
|
| 106 |
+
In order to get any license under these terms, you must agree to them as both strict obligations and conditions to all your licenses.
|
| 107 |
+
|
| 108 |
+
## Copyright License
|
| 109 |
+
|
| 110 |
+
The licensor grants you a copyright license for the software to do everything you might do with the software that would otherwise infringe the licensor's copyright in it for any permitted purpose. However, you may only distribute the software according to [Distribution License](#distribution-license) and make changes or new works based on the software according to [Changes and New Works License](#changes-and-new-works-license).
|
| 111 |
+
|
| 112 |
+
## Distribution License
|
| 113 |
+
|
| 114 |
+
The licensor grants you an additional copyright license to distribute copies of the software. Your license to distribute covers distributing the software with changes and new works permitted by [Changes and New Works License](#changes-and-new-works-license).
|
| 115 |
+
|
| 116 |
+
## Notices
|
| 117 |
+
|
| 118 |
+
You must ensure that anyone who gets a copy of any part of the software from you also gets a copy of these terms or the URL for them above, as well as copies of any plain-text lines beginning with `Required Notice:` that the licensor provided with the software. For example:
|
| 119 |
+
|
| 120 |
+
> Required Notice: Copyright Yoyodyne, Inc. (http://example.com)
|
| 121 |
+
|
| 122 |
+
## Changes and New Works License
|
| 123 |
+
|
| 124 |
+
The licensor grants you an additional copyright license to make changes and new works based on the software for any permitted purpose.
|
| 125 |
+
|
| 126 |
+
## Patent License
|
| 127 |
+
|
| 128 |
+
The licensor grants you a patent license for the software that covers patent claims the licensor can license, or becomes able to license, that you would infringe by using the software.
|
| 129 |
+
|
| 130 |
+
## Fair Use
|
| 131 |
+
|
| 132 |
+
You may have "fair use" rights for the software under the law. These terms do not limit them.
|
| 133 |
+
|
| 134 |
+
## Small Business
|
| 135 |
+
|
| 136 |
+
Use of the software for the benefit of your company is use for a permitted purpose if your company has fewer than 100 total individuals working as employees and independent contractors, and less than 1,000,000 USD (2019) total revenue in the prior tax year. Adjust this revenue threshold for inflation according to the United States Bureau of Labor Statistics' consumer price index for all urban consumers, U.S. city average, for all items, not seasonally adjusted, with 1982–1984=100 reference base.
|
| 137 |
+
|
| 138 |
+
## No Other Rights
|
| 139 |
+
|
| 140 |
+
These terms do not allow you to sublicense or transfer any of your licenses to anyone else, or prevent the licensor from granting licenses to anyone else. These terms do not imply any other licenses.
|
| 141 |
+
|
| 142 |
+
## Patent Defense
|
| 143 |
+
|
| 144 |
+
If you make any written claim that the software infringes or contributes to infringement of any patent, your patent license for the software granted under these terms ends immediately. If your company makes such a claim, your patent license ends immediately for work on behalf of your company.
|
| 145 |
+
|
| 146 |
+
## Violations
|
| 147 |
+
|
| 148 |
+
The first time you are notified in writing that you have violated any of these terms, or done anything with the software not covered by your licenses, your licenses can nonetheless continue if you come into full compliance with these terms, and take practical steps to correct past violations, within 32 days of receiving notice. Otherwise, all your licenses end immediately.
|
| 149 |
+
|
| 150 |
+
## No Liability
|
| 151 |
+
|
| 152 |
+
***As far as the law allows, the software comes as is, without any warranty or condition, and the licensor will not be liable to you for any damages arising out of these terms or the use or nature of the software, under any kind of legal claim.***
|
| 153 |
+
|
| 154 |
+
## Definitions
|
| 155 |
+
|
| 156 |
+
The **licensor** is the individual or entity offering these terms, and the **software** is the software the licensor makes available under these terms.
|
| 157 |
+
|
| 158 |
+
**You** refers to the individual or entity agreeing to these terms.
|
| 159 |
+
|
| 160 |
+
**Your company** is any legal entity, sole proprietorship, or other kind of organization that you work for, plus all organizations that have control over, are under the control of, or are under common control with that organization. **Control** means ownership of substantially all the assets of an entity, or the power to direct its management and policies by vote, contract, or otherwise. Control can be direct or indirect.
|
| 161 |
+
|
| 162 |
+
**Your licenses** are all the licenses granted to you for the software under these terms.
|
| 163 |
+
|
| 164 |
+
**Use** means anything you do with the software requiring one of your licenses.
|
| 165 |
+
|
| 166 |
+
---
|
| 167 |
+
|
| 168 |
+
# PolyForm Free Trial License 1.0.0
|
| 169 |
+
|
| 170 |
+
<https://polyformproject.org/licenses/free-trial/1.0.0>
|
| 171 |
+
|
| 172 |
+
## Acceptance
|
| 173 |
+
|
| 174 |
+
In order to get any license under these terms, you must agree to them as both strict obligations and conditions to all your licenses.
|
| 175 |
+
|
| 176 |
+
## Copyright License
|
| 177 |
+
|
| 178 |
+
The licensor grants you a copyright license for the software to do everything you might do with the software that would otherwise infringe the licensor's copyright in it for any permitted purpose. However, you may only make changes or new works based on the software according to [Changes and New Works License](#changes-and-new-works-license), and you may not distribute copies of the software.
|
| 179 |
+
|
| 180 |
+
## Changes and New Works License
|
| 181 |
+
|
| 182 |
+
The licensor grants you an additional copyright license to make changes and new works based on the software for any permitted purpose.
|
| 183 |
+
|
| 184 |
+
## Patent License
|
| 185 |
+
|
| 186 |
+
The licensor grants you a patent license for the software that covers patent claims the licensor can license, or becomes able to license, that you would infringe by using the software.
|
| 187 |
+
|
| 188 |
+
## Fair Use
|
| 189 |
+
|
| 190 |
+
You may have "fair use" rights for the software under the law. These terms do not limit them.
|
| 191 |
+
|
| 192 |
+
## Free Trial
|
| 193 |
+
|
| 194 |
+
Use to evaluate whether the software suits a particular application for less than 32 consecutive calendar days, on behalf of you or your company, is use for a permitted purpose.
|
| 195 |
+
|
| 196 |
+
## No Other Rights
|
| 197 |
+
|
| 198 |
+
These terms do not allow you to sublicense or transfer any of your licenses to anyone else, or prevent the licensor from granting licenses to anyone else. These terms do not imply any other licenses.
|
| 199 |
+
|
| 200 |
+
## Patent Defense
|
| 201 |
+
|
| 202 |
+
If you make any written claim that the software infringes or contributes to infringement of any patent, your patent license for the software granted under these terms ends immediately. If your company makes such a claim, your patent license ends immediately for work on behalf of your company.
|
| 203 |
+
|
| 204 |
+
## Violations
|
| 205 |
+
|
| 206 |
+
If you violate any of these terms, or do anything with the software not covered by your licenses, all your licenses end immediately.
|
| 207 |
+
|
| 208 |
+
## No Liability
|
| 209 |
+
|
| 210 |
+
***As far as the law allows, the software comes as is, without any warranty or condition, and the licensor will not be liable to you for any damages arising out of these terms or the use or nature of the software, under any kind of legal claim.***
|
| 211 |
+
|
| 212 |
+
## Definitions
|
| 213 |
+
|
| 214 |
+
The **licensor** is the individual or entity offering these terms, and the **software** is the software the licensor makes available under these terms.
|
| 215 |
+
|
| 216 |
+
**You** refers to the individual or entity agreeing to these terms.
|
| 217 |
+
|
| 218 |
+
**Your company** is any legal entity, sole proprietorship, or other kind of organization that you work for, plus all organizations that have control over, are under the control of, or are under common control with that organization. **Control** means ownership of substantially all the assets of an entity, or the power to direct its management and policies by vote, contract, or otherwise. Control can be direct or indirect.
|
| 219 |
+
|
| 220 |
+
**Your licenses** are all the licenses granted to you for the software under these terms.
|
| 221 |
+
|
| 222 |
+
**Use** means anything you do with the software requiring one of your licenses.
|
NOTICE
ADDED
|
@@ -0,0 +1,17 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
Statim Decide Multilingual Base: PII adapter
|
| 2 |
+
|
| 3 |
+
Required Notice: Copyright 2026 Belkis Aslani (https://github.com/BEKO2210/statim)
|
| 4 |
+
Required Notice: Statim weights are fine-tuned from Laya checkpoints (https://huggingface.co/convaiinnovations), Copyright their authors, licensed under the Apache License 2.0.
|
| 5 |
+
|
| 6 |
+
The weights are licensed under PolyForm Noncommercial 1.0.0, PolyForm Small Business 1.0.0 or PolyForm Free Trial 1.0.0, at the user's choice (LICENSE-MODEL.md), or a commercial licence (https://github.com/BEKO2210/statim/blob/main/COMMERCIAL.md).
|
| 7 |
+
|
| 8 |
+
It adapts https://huggingface.co/Beko2210/statim-decide-multilingual-base.
|
| 9 |
+
Laya (https://github.com/NandhaKishorM/laya) is Apache-2.0; Statim is independent and not affiliated with or endorsed by the Laya authors.
|
| 10 |
+
|
| 11 |
+
Training data
|
| 12 |
+
- E3-JSI/synthetic-multi-pii-ner-v1, https://huggingface.co/datasets/E3-JSI/synthetic-multi-pii-ner-v1: licensed under MIT; 2,971 rows used.
|
| 13 |
+
- Wismut/nym-pii-multilingual-data, https://huggingface.co/datasets/Wismut/nym-pii-multilingual-data: licensed under MIT; 6,200 rows used.
|
| 14 |
+
- gretelai/gretel-pii-masking-en-v1, https://huggingface.co/datasets/gretelai/gretel-pii-masking-en-v1: licensed under Apache-2.0; 6,200 rows used.
|
| 15 |
+
- gretelai/synthetic_pii_finance_multilingual, https://huggingface.co/datasets/gretelai/synthetic_pii_finance_multilingual: licensed under Apache-2.0; 6,200 rows used.
|
| 16 |
+
- nvidia/Nemotron-PII, https://huggingface.co/datasets/nvidia/Nemotron-PII: licensed under CC BY 4.0; 6,200 rows used.
|
| 17 |
+
- urchade/synthetic-pii-ner-mistral-v1, https://huggingface.co/datasets/urchade/synthetic-pii-ner-mistral-v1: licensed under Apache-2.0; 6,200 rows used.
|
README.md
ADDED
|
@@ -0,0 +1,130 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: "other"
|
| 3 |
+
license_name: "statim-weights"
|
| 4 |
+
license_link: "https://huggingface.co/Beko2210/statim-decide-multilingual-base-pii/blob/main/LICENSE-MODEL.md"
|
| 5 |
+
base_model: "Beko2210/statim-decide-multilingual-base"
|
| 6 |
+
base_model_relation: "adapter"
|
| 7 |
+
library_name: "gguf"
|
| 8 |
+
language: ["ar", "de", "en", "es", "fr", "it", "ja", "nl", "ru", "sv", "zh"]
|
| 9 |
+
tags: ["statim", "lora", "adapter", "gguf", "pii"]
|
| 10 |
+
pipeline_tag: "zero-shot-classification"
|
| 11 |
+
model-index: [{"name": "statim-decide-multilingual-base-pii", "results": [{"task": {"type": "text-classification"}, "dataset": {"name": "pii ar (n=150, seed=20260927, z=2.0, alpha=0.05)", "type": "pii"}, "metrics": [{"type": "accuracy", "value": 0.9067}]}, {"task": {"type": "text-classification"}, "dataset": {"name": "pii de (n=150, seed=20260927, z=2.0, alpha=0.05)", "type": "pii"}, "metrics": [{"type": "accuracy", "value": 0.8667}]}, {"task": {"type": "text-classification"}, "dataset": {"name": "pii en (n=150, seed=20260927, z=2.0, alpha=0.05)", "type": "pii"}, "metrics": [{"type": "accuracy", "value": 0.9267}]}, {"task": {"type": "text-classification"}, "dataset": {"name": "pii es (n=150, seed=20260927, z=2.0, alpha=0.05)", "type": "pii"}, "metrics": [{"type": "accuracy", "value": 0.8933}]}, {"task": {"type": "text-classification"}, "dataset": {"name": "pii fr (n=150, seed=20260927, z=2.0, alpha=0.05)", "type": "pii"}, "metrics": [{"type": "accuracy", "value": 0.94}]}, {"task": {"type": "text-classification"}, "dataset": {"name": "pii it (n=150, seed=20260927, z=2.0, alpha=0.05)", "type": "pii"}, "metrics": [{"type": "accuracy", "value": 0.92}]}, {"task": {"type": "text-classification"}, "dataset": {"name": "pii ja (n=150, seed=20260927, z=2.0, alpha=0.05)", "type": "pii"}, "metrics": [{"type": "accuracy", "value": 0.9267}]}, {"task": {"type": "text-classification"}, "dataset": {"name": "pii nl (n=150, seed=20260927, z=2.0, alpha=0.05)", "type": "pii"}, "metrics": [{"type": "accuracy", "value": 0.8667}]}, {"task": {"type": "text-classification"}, "dataset": {"name": "pii ru (n=150, seed=20260927, z=2.0, alpha=0.05)", "type": "pii"}, "metrics": [{"type": "accuracy", "value": 0.9467}]}, {"task": {"type": "text-classification"}, "dataset": {"name": "pii sv (n=150, seed=20260927, z=2.0, alpha=0.05)", "type": "pii"}, "metrics": [{"type": "accuracy", "value": 0.8733}]}, {"task": {"type": "text-classification"}, "dataset": {"name": "pii zh (n=150, seed=20260927, z=2.0, alpha=0.05)", "type": "pii"}, "metrics": [{"type": "accuracy", "value": 0.9467}]}]}]
|
| 12 |
+
---
|
| 13 |
+
|
| 14 |
+
# Statim Decide Multilingual Base: PII adapter
|
| 15 |
+
|
| 16 |
+
A LoRA adapter that improves the PII decisions of
|
| 17 |
+
[`Beko2210/statim-decide-multilingual-base`](https://huggingface.co/Beko2210/statim-decide-multilingual-base) **0.7.0**.
|
| 18 |
+
Checked with Statim 0.8.3: the published f32 file loads it merged at load, the q8_0 file as runtime LoRA. LoRA adapters need Statim 0.8.0 or later.
|
| 19 |
+
|
| 20 |
+
## Quick start
|
| 21 |
+
|
| 22 |
+
```sh
|
| 23 |
+
hf download Beko2210/statim-decide-multilingual-base statim-decide-multilingual-base-q8_0.gguf --local-dir models
|
| 24 |
+
hf download Beko2210/statim-decide-multilingual-base-pii statim-decide-multilingual-base-pii.lora.gguf --local-dir models
|
| 25 |
+
statim serve -m multilingual=models/statim-decide-multilingual-base-q8_0.gguf --adapter multilingual:pii=models/statim-decide-multilingual-base-pii.lora.gguf --port 8080
|
| 26 |
+
|
| 27 |
+
curl -s localhost:8080/v1/systemone -d '{"state": {"text": "Hi, this is Anna Schmidt. Call me back at +49 170 1234567."}, "questions": {"pii": {"type": "noul", "instructions": "Does the text contain a phone number?"}}, "adapter": "pii"}'
|
| 28 |
+
curl -s localhost:8080/v1/systemone -d '{"state": {"text": "Hi, this is Anna Schmidt. Call me back at +49 170 1234567."}, "questions": {"pii": {"type": "noul", "instructions": "Does the text contain a phone number?"}}, "adapter": "auto"}'
|
| 29 |
+
```
|
| 30 |
+
|
| 31 |
+
Python (Python SDK 0.8.3 or later):
|
| 32 |
+
|
| 33 |
+
```python
|
| 34 |
+
from statim import Client
|
| 35 |
+
client = Client("http://127.0.0.1:8080")
|
| 36 |
+
client.decide({"text": "Hi, this is Anna Schmidt. Call me back at +49 170 1234567."}, {"pii": {"type": "noul", "instructions": "Does the text contain a phone number?"}}, adapter="pii")
|
| 37 |
+
```
|
| 38 |
+
|
| 39 |
+
## Results (experiment's f32 run)
|
| 40 |
+
|
| 41 |
+
Each cell uses 150 items (seed 20260927).
|
| 42 |
+
|
| 43 |
+
| Language | Base | Adapter | Change (points) | Verdict | Qwen3-8B zero-shot |
|
| 44 |
+
|---|---:|---:|---:|---|---:|
|
| 45 |
+
| ar | 0.8467 | **0.9067** | +6.00 | within noise | 0.893 |
|
| 46 |
+
| de | 0.8400 | **0.8667** | +2.67 | within noise | 0.893 |
|
| 47 |
+
| en | 0.8933 | **0.9267** | +3.34 | within noise | 0.887 |
|
| 48 |
+
| es | 0.8533 | **0.8933** | +4.00 | within noise | 0.867 |
|
| 49 |
+
| fr | 0.8933 | **0.9400** | +4.67 | within noise | 0.880 |
|
| 50 |
+
| it | 0.8400 | **0.9200** | +8.00 | gain (2 SE) | 0.873 |
|
| 51 |
+
| ja | 0.8467 | **0.9267** | +8.00 | gain (2 SE) | 0.893 |
|
| 52 |
+
| nl | 0.7600 | **0.8667** | +10.67 | gain (2 SE) | 0.800 |
|
| 53 |
+
| ru | 0.9067 | **0.9467** | +4.00 | within noise | 0.947 |
|
| 54 |
+
| sv | 0.8333 | **0.8733** | +4.00 | within noise | 0.800 |
|
| 55 |
+
| zh | 0.9000 | **0.9467** | +4.67 | within noise | 0.920 |
|
| 56 |
+
| **Mean** | 0.8558 | **0.9103** | +5.46 | — | 0.878 |
|
| 57 |
+
|
| 58 |
+
The pooled family change is **+5.46 points** with **2 SE = 2.22 points** (gain).
|
| 59 |
+
|
| 60 |
+
A single cell of 150 items rarely clears 2 SE on its own; the decision uses the pooled family.
|
| 61 |
+
|
| 62 |
+
Decision rule: promote when the category family gains more than 2 standard errors, pooled by rows or by suites, and nothing regresses. A regression is a pooled drop beyond 2 standard errors under either pooling, or a drop in one language cell that stays significant after Holm-Bonferroni (family-wise 5 %).
|
| 63 |
+
|
| 64 |
+
Statim is trained on this category, while [Qwen3-8B runs zero-shot](https://github.com/BEKO2210/statim/blob/main/docs/BASELINES.md).
|
| 65 |
+
|
| 66 |
+
## Checked on the published files
|
| 67 |
+
|
| 68 |
+
| Weights | Adapter mode | Base mean | Adapter mean | Change (points) | Cell for cell as in the experiment |
|
| 69 |
+
|---|---|---:|---:|---:|---|
|
| 70 |
+
| f32 | merged at load | 0.8558 | 0.9103 | +5.46 | yes |
|
| 71 |
+
| q8_0 | runtime LoRA | 0.8558 | 0.9109 | +5.52 | — (experiment: f32) |
|
| 72 |
+
|
| 73 |
+
## Training
|
| 74 |
+
|
| 75 |
+
| Source | Rows | Licence |
|
| 76 |
+
|---|---:|---|
|
| 77 |
+
| [`E3-JSI/synthetic-multi-pii-ner-v1`](https://huggingface.co/datasets/E3-JSI/synthetic-multi-pii-ner-v1) (`default`) | 2,971 | MIT |
|
| 78 |
+
| [`Wismut/nym-pii-multilingual-data`](https://huggingface.co/datasets/Wismut/nym-pii-multilingual-data) (`default`) | 6,200 | MIT |
|
| 79 |
+
| [`gretelai/gretel-pii-masking-en-v1`](https://huggingface.co/datasets/gretelai/gretel-pii-masking-en-v1) (`default`) | 6,200 | Apache-2.0 |
|
| 80 |
+
| [`gretelai/synthetic_pii_finance_multilingual`](https://huggingface.co/datasets/gretelai/synthetic_pii_finance_multilingual) (`default`) | 6,200 | Apache-2.0 |
|
| 81 |
+
| [`nvidia/Nemotron-PII`](https://huggingface.co/datasets/nvidia/Nemotron-PII) (`default`) | 6,200 | CC-BY-4.0 |
|
| 82 |
+
| [`urchade/synthetic-pii-ner-mistral-v1`](https://huggingface.co/datasets/urchade/synthetic-pii-ner-mistral-v1) (`data.json`) | 6,200 | Apache-2.0 |
|
| 83 |
+
|
| 84 |
+
LoRA rank **16**, alpha **32.0**, dropout **0.05**;
|
| 85 |
+
target modules `Wqkv`, `Wo`, `Wi`; **88** wrapped modules and
|
| 86 |
+
**3,379,200** trainable parameters.
|
| 87 |
+
|
| 88 |
+
- Items: **33,571 train**, **400 dev**.
|
| 89 |
+
- Updates: **1,172**.
|
| 90 |
+
- Dev accuracy: **0.8375** before, **0.8725** after.
|
| 91 |
+
- Time: **1631.7 seconds**; peak memory: **2,332 MB**.
|
| 92 |
+
|
| 93 |
+
## Provenance
|
| 94 |
+
|
| 95 |
+
- Adapter GGUF SHA-256: `2991a33b5d9db4f5b679081885faab1f84289d6eae7660830447e790bd168b29`
|
| 96 |
+
- PEFT safetensors SHA-256: `f8e7e674ab868b312ff32f403552daa436e7437411e83fa0ece60187431a2d6f`
|
| 97 |
+
- Training checkpoint `model.safetensors` SHA-256: `c44425f14ac9d55508f73a6f371e4e2802ed59e646287c3abbec10b653a19840`
|
| 98 |
+
- Base fingerprint: `e7a8fa743b4920850167b587226e339d651be0244fcc6a9e24fff8ed5073bae5`
|
| 99 |
+
- Training mixture SHA-256: `b8a87e85f3bf72b61509555aba9148358760d3329ecf37a36cebd94f9358cb03`
|
| 100 |
+
- Source registry SHA-256: `309ce8ec262886b3bfaa529934151ae0c1879595c26301e4269da276d95cbfcd`
|
| 101 |
+
|
| 102 |
+
Experiment commands (paths relative to the Statim repository):
|
| 103 |
+
|
| 104 |
+
- `train`: `'.venv-train/bin/python' 'tools/finetune/train_lora.py' 'models/laya-multilingual-v9' --mixture 'data/mixture-v8.jsonl.gz' --category pii --registry 'tools/finetune/sources/v6-keep.json' --out 'models/lora-exp1/pii' --device cuda --epochs 2`
|
| 105 |
+
- `convert`: `'.venv/bin/python' 'tools/convert_lora.py' 'models/lora-exp1/pii' -o 'models/lora-exp1/pii.lora.gguf' --base 'models/laya-multilingual-v9-f32.gguf' --category pii --name pii`
|
| 106 |
+
- `serve`: `'build-vk/statim' serve -m 'multilingual=models/laya-multilingual-v9-f32.gguf' --adapter 'multilingual:pii=models/lora-exp1/pii.lora.gguf' --device vulkan --threads 16 --port 8098 --no-access-log --inference-timeout 600`
|
| 107 |
+
- `eval_base`: `'.venv/bin/python' 'bench/eval_categories.py' --strict --url http://127.0.0.1:8098 --model multilingual --suites pii --n 150 --seed 20260927 --exclude-mixture 'data/mixture-v8.jsonl.gz' --out 'models/lora-exp1/pii.base.jsonl'`
|
| 108 |
+
- `eval_adapter`: `'.venv/bin/python' 'bench/eval_categories.py' --strict --url http://127.0.0.1:8098 --model multilingual --suites pii --n 150 --seed 20260927 --exclude-mixture 'data/mixture-v8.jsonl.gz' --adapter pii --out 'models/lora-exp1/pii.adapter.jsonl'`
|
| 109 |
+
|
| 110 |
+
Protocol and experiment: [docs/ADAPTERS.md](https://github.com/BEKO2210/statim/blob/main/docs/ADAPTERS.md). Full reproduction
|
| 111 |
+
instructions: [REPRODUCE.md](https://github.com/BEKO2210/statim/blob/main/REPRODUCE.md).
|
| 112 |
+
|
| 113 |
+
## Intended use and limits
|
| 114 |
+
|
| 115 |
+
- This adapter only helps its category; route requests with `"pii"` or `"auto"`.
|
| 116 |
+
- Bound to `Beko2210/statim-decide-multilingual-base` 0.7.0 by the base fingerprint `e7a8fa743b492085…` (`statim.lora.base_fingerprint`, SHA-256 over the checkpoint's norm and bias tensors); Statim refuses the adapter on a base whose fingerprint differs.
|
| 117 |
+
- Languages outside the evaluated list are untested.
|
| 118 |
+
- Each language cell has 150 items.
|
| 119 |
+
- Do not automate decisions about people without human review.
|
| 120 |
+
|
| 121 |
+
## Licence
|
| 122 |
+
|
| 123 |
+
The weights may be used under any one of: PolyForm Noncommercial 1.0.0, PolyForm Small Business
|
| 124 |
+
1.0.0 (free commercial use below 100 people and 1 M USD revenue), PolyForm Free Trial 1.0.0 (any
|
| 125 |
+
company, fewer than 32 days), or a Statim commercial licence
|
| 126 |
+
([COMMERCIAL.md](https://github.com/BEKO2210/statim/blob/main/COMMERCIAL.md)). Texts in [LICENSE-MODEL.md](LICENSE-MODEL.md).
|
| 127 |
+
The Statim engine is Apache-2.0.
|
| 128 |
+
|
| 129 |
+
Training data attribution is listed source by source above, with the row count and licence read
|
| 130 |
+
from the experiment registry.
|
SHA256SUMS
ADDED
|
@@ -0,0 +1,14 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
1106ca74eec5e589cec780e4f19f5218ea84ceb6b0d23594d7d9b5b30a2b2322 LICENSE-MODEL.md
|
| 2 |
+
53eb1a64b7123a906b3261d37f467e40702e05239a1137d270202397eb203e87 NOTICE
|
| 3 |
+
8c622161e9a064a85732dc8526ce019e39118c80ae05218062373c2de30da4b6 README.md
|
| 4 |
+
335c3a21ff0599b2e521f055b31c0e6e64b64fb8db8796e96eafd4c57389b00d evaluation/pii.adapter.jsonl
|
| 5 |
+
47d0ff7024199b9c7e9832cbbe336f6a4ec5f0364bfd1cb0a2a4adff387f0821 evaluation/pii.base.jsonl
|
| 6 |
+
2dca26b1bc42e41e4792c478ad2cfc832c9d9fe80454fbac7d52835a63b56502 evaluation/published-f32-adapter.jsonl
|
| 7 |
+
4e59da1a13489e6edd5bbbdf24f8b505f8ba7b74468a4269cd097d608f42cc30 evaluation/published-f32-base.jsonl
|
| 8 |
+
133c5072dd229211c44ff2dac02302d799ea1d723c9c67e910681ce8ff3bc4a4 evaluation/published-q8_0-adapter.jsonl
|
| 9 |
+
4856e5a8c45650beb9fc4cd76948a9c7c184c9082e5902c2060897f0940df98e evaluation/published-q8_0-base.jsonl
|
| 10 |
+
f9218a9b7871e0f850b2264ca163ffff8880f8e0024b9b4d2c6f5bb8d7e673ab evaluation/summary.json
|
| 11 |
+
cdf68a8aee322bc2841047525fec52b8f89d4310867fd6458367b9378320091b peft/adapter_config.json
|
| 12 |
+
f8e7e674ab868b312ff32f403552daa436e7437411e83fa0ece60187431a2d6f peft/adapter_model.safetensors
|
| 13 |
+
2991a33b5d9db4f5b679081885faab1f84289d6eae7660830447e790bd168b29 statim-decide-multilingual-base-pii.lora.gguf
|
| 14 |
+
48dcc90dfc3d32c9129f50efa29397d7302a90d8051d1ae0d36277e2b98142f5 training/train_lora.json
|
evaluation/pii.adapter.jsonl
ADDED
|
@@ -0,0 +1,16 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{"family": "categories", "suite": "pii", "lang": "ar", "model": "multilingual:pii", "adapter": "pii", "n": 150, "seed": 20260927, "pool": "a6d43a981c13a016+b8a87e85f3bf72b6", "pool_items_sha256": "323112f111aa8c9c401e70d0f081ea6628b5230e57d741ac291694e16708c2b9", "sources": {"Wismut/nym-pii-multilingual-data::default::default/test": 150}, "question_types": {"noul": 150}, "gold_counts": {"false": 55, "true": 95}, "accuracy": 0.9067, "nll": 0.2518, "brier": 0.1512, "ece": 0.0472, "balanced_accuracy": 0.9187, "seconds": 2.1, "gate": true}
|
| 2 |
+
{"family": "categories", "suite": "pii", "lang": "de", "model": "multilingual:pii", "adapter": "pii", "n": 150, "seed": 20260927, "pool": "a6d43a981c13a016+b8a87e85f3bf72b6", "pool_items_sha256": "52145c7106ce9f85fa46c267cdace80c42181e148d9cb45594263718f5a49d92", "sources": {"Wismut/nym-pii-multilingual-data::default::default/test": 72, "gretelai/synthetic_pii_finance_multilingual::default::default/test": 78}, "question_types": {"noul": 150}, "gold_counts": {"false": 71, "true": 79}, "accuracy": 0.8667, "nll": 0.2554, "brier": 0.1635, "ece": 0.0494, "balanced_accuracy": 0.8656, "seconds": 2.8, "gate": true}
|
| 3 |
+
{"family": "categories", "suite": "pii", "lang": "en", "model": "multilingual:pii", "adapter": "pii", "n": 150, "seed": 20260927, "pool": "a6d43a981c13a016+b8a87e85f3bf72b6", "pool_items_sha256": "10b8d7d2c36445bbeecdfded7c937f4ca228c43ab80c515dd038be0f658a2cbe", "sources": {"Wismut/nym-pii-multilingual-data::default::default/test": 76, "gretelai/synthetic_pii_finance_multilingual::default::default/test": 74}, "question_types": {"noul": 150}, "gold_counts": {"false": 75, "true": 75}, "accuracy": 0.9267, "nll": 0.1618, "brier": 0.0961, "ece": 0.039, "balanced_accuracy": 0.9267, "seconds": 3.1, "gate": true}
|
| 4 |
+
{"family": "categories", "suite": "pii", "lang": "es", "model": "multilingual:pii", "adapter": "pii", "n": 150, "seed": 20260927, "pool": "a6d43a981c13a016+b8a87e85f3bf72b6", "pool_items_sha256": "853091fcb462b7b89e00759c1085ad942fb99eb455c18f4c5ecd8db63cb20e6e", "sources": {"Wismut/nym-pii-multilingual-data::default::default/test": 76, "gretelai/synthetic_pii_finance_multilingual::default::default/test": 74}, "question_types": {"noul": 150}, "gold_counts": {"false": 75, "true": 75}, "accuracy": 0.8933, "nll": 0.2465, "brier": 0.1489, "ece": 0.0492, "balanced_accuracy": 0.8933, "seconds": 3.1, "gate": true}
|
| 5 |
+
{"family": "categories", "suite": "pii", "lang": "fr", "model": "multilingual:pii", "adapter": "pii", "n": 150, "seed": 20260927, "pool": "a6d43a981c13a016+b8a87e85f3bf72b6", "pool_items_sha256": "2be7c20bc628fd30fe60af29d6e457b9ce63cba3d8961d3528007cacba4fcbb0", "sources": {"Wismut/nym-pii-multilingual-data::default::default/test": 74, "gretelai/synthetic_pii_finance_multilingual::default::default/test": 76}, "question_types": {"noul": 150}, "gold_counts": {"false": 73, "true": 77}, "accuracy": 0.94, "nll": 0.2037, "brier": 0.107, "ece": 0.0308, "balanced_accuracy": 0.9394, "seconds": 3.0, "gate": true}
|
| 6 |
+
{"family": "categories", "suite": "pii", "lang": "it", "model": "multilingual:pii", "adapter": "pii", "n": 150, "seed": 20260927, "pool": "a6d43a981c13a016+b8a87e85f3bf72b6", "pool_items_sha256": "632ad4c14c949eb65c7261d70f3532ea54f0d930e8df47b6da1a69bd294dc5a3", "sources": {"Wismut/nym-pii-multilingual-data::default::default/test": 70, "gretelai/synthetic_pii_finance_multilingual::default::default/test": 80}, "question_types": {"noul": 150}, "gold_counts": {"false": 70, "true": 80}, "accuracy": 0.92, "nll": 0.1802, "brier": 0.1074, "ece": 0.043, "balanced_accuracy": 0.9214, "seconds": 3.1, "gate": true}
|
| 7 |
+
{"family": "categories", "suite": "pii", "lang": "ja", "model": "multilingual:pii", "adapter": "pii", "n": 150, "seed": 20260927, "pool": "a6d43a981c13a016+b8a87e85f3bf72b6", "pool_items_sha256": "9d0a6eeb1fe27e2e5fcb7d7aa058c3bebfaa4e0fd650fe272e96049037cf03f6", "sources": {"Wismut/nym-pii-multilingual-data::default::default/test": 150}, "question_types": {"noul": 150}, "gold_counts": {"false": 66, "true": 84}, "accuracy": 0.9267, "nll": 0.2702, "brier": 0.145, "ece": 0.0549, "balanced_accuracy": 0.9264, "seconds": 1.7, "gate": true}
|
| 8 |
+
{"family": "categories", "suite": "pii", "lang": "nl", "model": "multilingual:pii", "adapter": "pii", "n": 150, "seed": 20260927, "pool": "a6d43a981c13a016+b8a87e85f3bf72b6", "pool_items_sha256": "72e3d46617d1027c2370304fbfea819c14e7d9e7be8a1e7e21e05137680472bd", "sources": {"Wismut/nym-pii-multilingual-data::default::default/test": 64, "gretelai/synthetic_pii_finance_multilingual::default::default/test": 86}, "question_types": {"noul": 150}, "gold_counts": {"false": 63, "true": 87}, "accuracy": 0.8667, "nll": 0.3322, "brier": 0.2051, "ece": 0.0793, "balanced_accuracy": 0.861, "seconds": 3.1, "gate": true}
|
| 9 |
+
{"family": "categories", "suite": "pii", "lang": "ru", "model": "multilingual:pii", "adapter": "pii", "n": 150, "seed": 20260927, "pool": "a6d43a981c13a016+b8a87e85f3bf72b6", "pool_items_sha256": "dde80741bffb81fecd80f3a66d88ea65ff9c6b15c050ece7db985ed910ef2f36", "sources": {"Wismut/nym-pii-multilingual-data::default::default/test": 150}, "question_types": {"noul": 150}, "gold_counts": {"false": 75, "true": 75}, "accuracy": 0.9467, "nll": 0.1364, "brier": 0.0785, "ece": 0.0323, "balanced_accuracy": 0.9467, "seconds": 1.9, "gate": true}
|
| 10 |
+
{"family": "categories", "suite": "pii", "lang": "sv", "model": "multilingual:pii", "adapter": "pii", "n": 150, "seed": 20260927, "pool": "a6d43a981c13a016+b8a87e85f3bf72b6", "pool_items_sha256": "f0e3e43965db3fb821f3191ba6572b783dc15cdc927dc4e9680da8c2d07e9c5f", "sources": {"gretelai/synthetic_pii_finance_multilingual::default::default/test": 150}, "question_types": {"noul": 150}, "gold_counts": {"false": 75, "true": 75}, "accuracy": 0.8733, "nll": 0.2882, "brier": 0.1747, "ece": 0.0397, "balanced_accuracy": 0.8733, "seconds": 3.4, "gate": true}
|
| 11 |
+
{"family": "categories", "suite": "pii", "lang": "zh", "model": "multilingual:pii", "adapter": "pii", "n": 150, "seed": 20260927, "pool": "a6d43a981c13a016+b8a87e85f3bf72b6", "pool_items_sha256": "6a85c938e961cc68abc1e79faf75d91c5af55c9a7c38176f0313bd23bf22b1dc", "sources": {"Wismut/nym-pii-multilingual-data::default::default/test": 150}, "question_types": {"noul": 150}, "gold_counts": {"false": 74, "true": 76}, "accuracy": 0.9467, "nll": 0.1717, "brier": 0.0899, "ece": 0.0139, "balanced_accuracy": 0.9468, "seconds": 1.8, "gate": true}
|
| 12 |
+
{"family": "categories", "suite": "pii", "lang": "macro", "model": "multilingual:pii", "adapter": "pii", "n": 1650, "n_langs": 11, "accuracy": 0.9103, "balanced_accuracy": 0.9108, "ece": 0.0435, "nll": 0.2271, "brier": 0.1334}
|
| 13 |
+
{"family": "categories", "suite": "pii", "lang": "hi", "skipped": "only 132 pooled items, need 150"}
|
| 14 |
+
{"family": "categories", "suite": "pii", "lang": "pl", "skipped": "only 42 pooled items, need 150"}
|
| 15 |
+
{"family": "categories", "suite": "pii", "lang": "pt", "skipped": "only 54 pooled items, need 150"}
|
| 16 |
+
{"family": "categories", "suite": "pii", "lang": "tr", "skipped": "only 58 pooled items, need 150"}
|
evaluation/pii.base.jsonl
ADDED
|
@@ -0,0 +1,16 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{"family": "categories", "suite": "pii", "lang": "ar", "model": "multilingual", "n": 150, "seed": 20260927, "pool": "a6d43a981c13a016+b8a87e85f3bf72b6", "pool_items_sha256": "323112f111aa8c9c401e70d0f081ea6628b5230e57d741ac291694e16708c2b9", "sources": {"Wismut/nym-pii-multilingual-data::default::default/test": 150}, "question_types": {"noul": 150}, "gold_counts": {"false": 55, "true": 95}, "accuracy": 0.8467, "nll": 0.3589, "brier": 0.2236, "ece": 0.0645, "balanced_accuracy": 0.8598, "seconds": 2.1, "gate": true}
|
| 2 |
+
{"family": "categories", "suite": "pii", "lang": "de", "model": "multilingual", "n": 150, "seed": 20260927, "pool": "a6d43a981c13a016+b8a87e85f3bf72b6", "pool_items_sha256": "52145c7106ce9f85fa46c267cdace80c42181e148d9cb45594263718f5a49d92", "sources": {"Wismut/nym-pii-multilingual-data::default::default/test": 72, "gretelai/synthetic_pii_finance_multilingual::default::default/test": 78}, "question_types": {"noul": 150}, "gold_counts": {"false": 71, "true": 79}, "accuracy": 0.84, "nll": 0.3724, "brier": 0.2268, "ece": 0.0387, "balanced_accuracy": 0.8403, "seconds": 2.8, "gate": true}
|
| 3 |
+
{"family": "categories", "suite": "pii", "lang": "en", "model": "multilingual", "n": 150, "seed": 20260927, "pool": "a6d43a981c13a016+b8a87e85f3bf72b6", "pool_items_sha256": "10b8d7d2c36445bbeecdfded7c937f4ca228c43ab80c515dd038be0f658a2cbe", "sources": {"Wismut/nym-pii-multilingual-data::default::default/test": 76, "gretelai/synthetic_pii_finance_multilingual::default::default/test": 74}, "question_types": {"noul": 150}, "gold_counts": {"false": 75, "true": 75}, "accuracy": 0.8933, "nll": 0.28, "brier": 0.1697, "ece": 0.0967, "balanced_accuracy": 0.8933, "seconds": 3.2, "gate": true}
|
| 4 |
+
{"family": "categories", "suite": "pii", "lang": "es", "model": "multilingual", "n": 150, "seed": 20260927, "pool": "a6d43a981c13a016+b8a87e85f3bf72b6", "pool_items_sha256": "853091fcb462b7b89e00759c1085ad942fb99eb455c18f4c5ecd8db63cb20e6e", "sources": {"Wismut/nym-pii-multilingual-data::default::default/test": 76, "gretelai/synthetic_pii_finance_multilingual::default::default/test": 74}, "question_types": {"noul": 150}, "gold_counts": {"false": 75, "true": 75}, "accuracy": 0.8533, "nll": 0.3837, "brier": 0.2297, "ece": 0.06, "balanced_accuracy": 0.8533, "seconds": 3.1, "gate": true}
|
| 5 |
+
{"family": "categories", "suite": "pii", "lang": "fr", "model": "multilingual", "n": 150, "seed": 20260927, "pool": "a6d43a981c13a016+b8a87e85f3bf72b6", "pool_items_sha256": "2be7c20bc628fd30fe60af29d6e457b9ce63cba3d8961d3528007cacba4fcbb0", "sources": {"Wismut/nym-pii-multilingual-data::default::default/test": 74, "gretelai/synthetic_pii_finance_multilingual::default::default/test": 76}, "question_types": {"noul": 150}, "gold_counts": {"false": 73, "true": 77}, "accuracy": 0.8933, "nll": 0.3102, "brier": 0.1746, "ece": 0.0387, "balanced_accuracy": 0.8922, "seconds": 3.0, "gate": true}
|
| 6 |
+
{"family": "categories", "suite": "pii", "lang": "it", "model": "multilingual", "n": 150, "seed": 20260927, "pool": "a6d43a981c13a016+b8a87e85f3bf72b6", "pool_items_sha256": "632ad4c14c949eb65c7261d70f3532ea54f0d930e8df47b6da1a69bd294dc5a3", "sources": {"Wismut/nym-pii-multilingual-data::default::default/test": 70, "gretelai/synthetic_pii_finance_multilingual::default::default/test": 80}, "question_types": {"noul": 150}, "gold_counts": {"false": 70, "true": 80}, "accuracy": 0.84, "nll": 0.3676, "brier": 0.2255, "ece": 0.0638, "balanced_accuracy": 0.8411, "seconds": 3.1, "gate": true}
|
| 7 |
+
{"family": "categories", "suite": "pii", "lang": "ja", "model": "multilingual", "n": 150, "seed": 20260927, "pool": "a6d43a981c13a016+b8a87e85f3bf72b6", "pool_items_sha256": "9d0a6eeb1fe27e2e5fcb7d7aa058c3bebfaa4e0fd650fe272e96049037cf03f6", "sources": {"Wismut/nym-pii-multilingual-data::default::default/test": 150}, "question_types": {"noul": 150}, "gold_counts": {"false": 66, "true": 84}, "accuracy": 0.8467, "nll": 0.331, "brier": 0.2079, "ece": 0.0413, "balanced_accuracy": 0.8501, "seconds": 1.7, "gate": true}
|
| 8 |
+
{"family": "categories", "suite": "pii", "lang": "nl", "model": "multilingual", "n": 150, "seed": 20260927, "pool": "a6d43a981c13a016+b8a87e85f3bf72b6", "pool_items_sha256": "72e3d46617d1027c2370304fbfea819c14e7d9e7be8a1e7e21e05137680472bd", "sources": {"Wismut/nym-pii-multilingual-data::default::default/test": 64, "gretelai/synthetic_pii_finance_multilingual::default::default/test": 86}, "question_types": {"noul": 150}, "gold_counts": {"false": 63, "true": 87}, "accuracy": 0.76, "nll": 0.5078, "brier": 0.3349, "ece": 0.1127, "balanced_accuracy": 0.7559, "seconds": 3.1, "gate": true}
|
| 9 |
+
{"family": "categories", "suite": "pii", "lang": "ru", "model": "multilingual", "n": 150, "seed": 20260927, "pool": "a6d43a981c13a016+b8a87e85f3bf72b6", "pool_items_sha256": "dde80741bffb81fecd80f3a66d88ea65ff9c6b15c050ece7db985ed910ef2f36", "sources": {"Wismut/nym-pii-multilingual-data::default::default/test": 150}, "question_types": {"noul": 150}, "gold_counts": {"false": 75, "true": 75}, "accuracy": 0.9067, "nll": 0.2355, "brier": 0.137, "ece": 0.0725, "balanced_accuracy": 0.9067, "seconds": 2.0, "gate": true}
|
| 10 |
+
{"family": "categories", "suite": "pii", "lang": "sv", "model": "multilingual", "n": 150, "seed": 20260927, "pool": "a6d43a981c13a016+b8a87e85f3bf72b6", "pool_items_sha256": "f0e3e43965db3fb821f3191ba6572b783dc15cdc927dc4e9680da8c2d07e9c5f", "sources": {"gretelai/synthetic_pii_finance_multilingual::default::default/test": 150}, "question_types": {"noul": 150}, "gold_counts": {"false": 75, "true": 75}, "accuracy": 0.8333, "nll": 0.3914, "brier": 0.244, "ece": 0.0326, "balanced_accuracy": 0.8333, "seconds": 3.5, "gate": true}
|
| 11 |
+
{"family": "categories", "suite": "pii", "lang": "zh", "model": "multilingual", "n": 150, "seed": 20260927, "pool": "a6d43a981c13a016+b8a87e85f3bf72b6", "pool_items_sha256": "6a85c938e961cc68abc1e79faf75d91c5af55c9a7c38176f0313bd23bf22b1dc", "sources": {"Wismut/nym-pii-multilingual-data::default::default/test": 150}, "question_types": {"noul": 150}, "gold_counts": {"false": 74, "true": 76}, "accuracy": 0.9, "nll": 0.2604, "brier": 0.1538, "ece": 0.0608, "balanced_accuracy": 0.9006, "seconds": 1.8, "gate": true}
|
| 12 |
+
{"family": "categories", "suite": "pii", "lang": "macro", "model": "multilingual", "n": 1650, "n_langs": 11, "accuracy": 0.8558, "balanced_accuracy": 0.857, "ece": 0.062, "nll": 0.3454, "brier": 0.2116}
|
| 13 |
+
{"family": "categories", "suite": "pii", "lang": "hi", "skipped": "only 132 pooled items, need 150"}
|
| 14 |
+
{"family": "categories", "suite": "pii", "lang": "pl", "skipped": "only 42 pooled items, need 150"}
|
| 15 |
+
{"family": "categories", "suite": "pii", "lang": "pt", "skipped": "only 54 pooled items, need 150"}
|
| 16 |
+
{"family": "categories", "suite": "pii", "lang": "tr", "skipped": "only 58 pooled items, need 150"}
|
evaluation/published-f32-adapter.jsonl
ADDED
|
@@ -0,0 +1,16 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{"family": "categories", "suite": "pii", "lang": "ar", "model": "multilingual:pii", "adapter": "pii", "n": 150, "seed": 20260927, "pool": "cd56ab1fa3c1ee43+b8a87e85f3bf72b6", "pool_items_sha256": "323112f111aa8c9c401e70d0f081ea6628b5230e57d741ac291694e16708c2b9", "sources": {"Wismut/nym-pii-multilingual-data::default::default/test": 150}, "question_types": {"noul": 150}, "gold_counts": {"false": 55, "true": 95}, "accuracy": 0.9067, "nll": 0.2518, "brier": 0.1512, "ece": 0.0472, "balanced_accuracy": 0.9187, "seconds": 2.2, "gate": true}
|
| 2 |
+
{"family": "categories", "suite": "pii", "lang": "de", "model": "multilingual:pii", "adapter": "pii", "n": 150, "seed": 20260927, "pool": "cd56ab1fa3c1ee43+b8a87e85f3bf72b6", "pool_items_sha256": "52145c7106ce9f85fa46c267cdace80c42181e148d9cb45594263718f5a49d92", "sources": {"Wismut/nym-pii-multilingual-data::default::default/test": 72, "gretelai/synthetic_pii_finance_multilingual::default::default/test": 78}, "question_types": {"noul": 150}, "gold_counts": {"false": 71, "true": 79}, "accuracy": 0.8667, "nll": 0.2554, "brier": 0.1635, "ece": 0.0494, "balanced_accuracy": 0.8656, "seconds": 2.8, "gate": true}
|
| 3 |
+
{"family": "categories", "suite": "pii", "lang": "en", "model": "multilingual:pii", "adapter": "pii", "n": 150, "seed": 20260927, "pool": "cd56ab1fa3c1ee43+b8a87e85f3bf72b6", "pool_items_sha256": "10b8d7d2c36445bbeecdfded7c937f4ca228c43ab80c515dd038be0f658a2cbe", "sources": {"Wismut/nym-pii-multilingual-data::default::default/test": 76, "gretelai/synthetic_pii_finance_multilingual::default::default/test": 74}, "question_types": {"noul": 150}, "gold_counts": {"false": 75, "true": 75}, "accuracy": 0.9267, "nll": 0.1618, "brier": 0.0961, "ece": 0.039, "balanced_accuracy": 0.9267, "seconds": 3.1, "gate": true}
|
| 4 |
+
{"family": "categories", "suite": "pii", "lang": "es", "model": "multilingual:pii", "adapter": "pii", "n": 150, "seed": 20260927, "pool": "cd56ab1fa3c1ee43+b8a87e85f3bf72b6", "pool_items_sha256": "853091fcb462b7b89e00759c1085ad942fb99eb455c18f4c5ecd8db63cb20e6e", "sources": {"Wismut/nym-pii-multilingual-data::default::default/test": 76, "gretelai/synthetic_pii_finance_multilingual::default::default/test": 74}, "question_types": {"noul": 150}, "gold_counts": {"false": 75, "true": 75}, "accuracy": 0.8933, "nll": 0.2465, "brier": 0.1489, "ece": 0.0492, "balanced_accuracy": 0.8933, "seconds": 3.0, "gate": true}
|
| 5 |
+
{"family": "categories", "suite": "pii", "lang": "fr", "model": "multilingual:pii", "adapter": "pii", "n": 150, "seed": 20260927, "pool": "cd56ab1fa3c1ee43+b8a87e85f3bf72b6", "pool_items_sha256": "2be7c20bc628fd30fe60af29d6e457b9ce63cba3d8961d3528007cacba4fcbb0", "sources": {"Wismut/nym-pii-multilingual-data::default::default/test": 74, "gretelai/synthetic_pii_finance_multilingual::default::default/test": 76}, "question_types": {"noul": 150}, "gold_counts": {"false": 73, "true": 77}, "accuracy": 0.94, "nll": 0.2037, "brier": 0.107, "ece": 0.0308, "balanced_accuracy": 0.9394, "seconds": 3.0, "gate": true}
|
| 6 |
+
{"family": "categories", "suite": "pii", "lang": "it", "model": "multilingual:pii", "adapter": "pii", "n": 150, "seed": 20260927, "pool": "cd56ab1fa3c1ee43+b8a87e85f3bf72b6", "pool_items_sha256": "632ad4c14c949eb65c7261d70f3532ea54f0d930e8df47b6da1a69bd294dc5a3", "sources": {"Wismut/nym-pii-multilingual-data::default::default/test": 70, "gretelai/synthetic_pii_finance_multilingual::default::default/test": 80}, "question_types": {"noul": 150}, "gold_counts": {"false": 70, "true": 80}, "accuracy": 0.92, "nll": 0.1802, "brier": 0.1074, "ece": 0.043, "balanced_accuracy": 0.9214, "seconds": 3.0, "gate": true}
|
| 7 |
+
{"family": "categories", "suite": "pii", "lang": "ja", "model": "multilingual:pii", "adapter": "pii", "n": 150, "seed": 20260927, "pool": "cd56ab1fa3c1ee43+b8a87e85f3bf72b6", "pool_items_sha256": "9d0a6eeb1fe27e2e5fcb7d7aa058c3bebfaa4e0fd650fe272e96049037cf03f6", "sources": {"Wismut/nym-pii-multilingual-data::default::default/test": 150}, "question_types": {"noul": 150}, "gold_counts": {"false": 66, "true": 84}, "accuracy": 0.9267, "nll": 0.2702, "brier": 0.145, "ece": 0.0549, "balanced_accuracy": 0.9264, "seconds": 1.7, "gate": true}
|
| 8 |
+
{"family": "categories", "suite": "pii", "lang": "nl", "model": "multilingual:pii", "adapter": "pii", "n": 150, "seed": 20260927, "pool": "cd56ab1fa3c1ee43+b8a87e85f3bf72b6", "pool_items_sha256": "72e3d46617d1027c2370304fbfea819c14e7d9e7be8a1e7e21e05137680472bd", "sources": {"Wismut/nym-pii-multilingual-data::default::default/test": 64, "gretelai/synthetic_pii_finance_multilingual::default::default/test": 86}, "question_types": {"noul": 150}, "gold_counts": {"false": 63, "true": 87}, "accuracy": 0.8667, "nll": 0.3322, "brier": 0.2051, "ece": 0.0793, "balanced_accuracy": 0.861, "seconds": 3.1, "gate": true}
|
| 9 |
+
{"family": "categories", "suite": "pii", "lang": "ru", "model": "multilingual:pii", "adapter": "pii", "n": 150, "seed": 20260927, "pool": "cd56ab1fa3c1ee43+b8a87e85f3bf72b6", "pool_items_sha256": "dde80741bffb81fecd80f3a66d88ea65ff9c6b15c050ece7db985ed910ef2f36", "sources": {"Wismut/nym-pii-multilingual-data::default::default/test": 150}, "question_types": {"noul": 150}, "gold_counts": {"false": 75, "true": 75}, "accuracy": 0.9467, "nll": 0.1364, "brier": 0.0785, "ece": 0.0323, "balanced_accuracy": 0.9467, "seconds": 1.9, "gate": true}
|
| 10 |
+
{"family": "categories", "suite": "pii", "lang": "sv", "model": "multilingual:pii", "adapter": "pii", "n": 150, "seed": 20260927, "pool": "cd56ab1fa3c1ee43+b8a87e85f3bf72b6", "pool_items_sha256": "f0e3e43965db3fb821f3191ba6572b783dc15cdc927dc4e9680da8c2d07e9c5f", "sources": {"gretelai/synthetic_pii_finance_multilingual::default::default/test": 150}, "question_types": {"noul": 150}, "gold_counts": {"false": 75, "true": 75}, "accuracy": 0.8733, "nll": 0.2882, "brier": 0.1747, "ece": 0.0397, "balanced_accuracy": 0.8733, "seconds": 3.4, "gate": true}
|
| 11 |
+
{"family": "categories", "suite": "pii", "lang": "zh", "model": "multilingual:pii", "adapter": "pii", "n": 150, "seed": 20260927, "pool": "cd56ab1fa3c1ee43+b8a87e85f3bf72b6", "pool_items_sha256": "6a85c938e961cc68abc1e79faf75d91c5af55c9a7c38176f0313bd23bf22b1dc", "sources": {"Wismut/nym-pii-multilingual-data::default::default/test": 150}, "question_types": {"noul": 150}, "gold_counts": {"false": 74, "true": 76}, "accuracy": 0.9467, "nll": 0.1717, "brier": 0.0899, "ece": 0.0139, "balanced_accuracy": 0.9468, "seconds": 1.8, "gate": true}
|
| 12 |
+
{"family": "categories", "suite": "pii", "lang": "macro", "model": "multilingual:pii", "adapter": "pii", "n": 1650, "n_langs": 11, "accuracy": 0.9103, "balanced_accuracy": 0.9108, "ece": 0.0435, "nll": 0.2271, "brier": 0.1334}
|
| 13 |
+
{"family": "categories", "suite": "pii", "lang": "hi", "skipped": "only 132 pooled items, need 150"}
|
| 14 |
+
{"family": "categories", "suite": "pii", "lang": "pl", "skipped": "only 42 pooled items, need 150"}
|
| 15 |
+
{"family": "categories", "suite": "pii", "lang": "pt", "skipped": "only 54 pooled items, need 150"}
|
| 16 |
+
{"family": "categories", "suite": "pii", "lang": "tr", "skipped": "only 58 pooled items, need 150"}
|
evaluation/published-f32-base.jsonl
ADDED
|
@@ -0,0 +1,16 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{"family": "categories", "suite": "pii", "lang": "ar", "model": "multilingual", "n": 150, "seed": 20260927, "pool": "cd56ab1fa3c1ee43+b8a87e85f3bf72b6", "pool_items_sha256": "323112f111aa8c9c401e70d0f081ea6628b5230e57d741ac291694e16708c2b9", "sources": {"Wismut/nym-pii-multilingual-data::default::default/test": 150}, "question_types": {"noul": 150}, "gold_counts": {"false": 55, "true": 95}, "accuracy": 0.8467, "nll": 0.3589, "brier": 0.2236, "ece": 0.0645, "balanced_accuracy": 0.8598, "seconds": 2.1, "gate": true}
|
| 2 |
+
{"family": "categories", "suite": "pii", "lang": "de", "model": "multilingual", "n": 150, "seed": 20260927, "pool": "cd56ab1fa3c1ee43+b8a87e85f3bf72b6", "pool_items_sha256": "52145c7106ce9f85fa46c267cdace80c42181e148d9cb45594263718f5a49d92", "sources": {"Wismut/nym-pii-multilingual-data::default::default/test": 72, "gretelai/synthetic_pii_finance_multilingual::default::default/test": 78}, "question_types": {"noul": 150}, "gold_counts": {"false": 71, "true": 79}, "accuracy": 0.84, "nll": 0.3724, "brier": 0.2268, "ece": 0.0387, "balanced_accuracy": 0.8403, "seconds": 2.8, "gate": true}
|
| 3 |
+
{"family": "categories", "suite": "pii", "lang": "en", "model": "multilingual", "n": 150, "seed": 20260927, "pool": "cd56ab1fa3c1ee43+b8a87e85f3bf72b6", "pool_items_sha256": "10b8d7d2c36445bbeecdfded7c937f4ca228c43ab80c515dd038be0f658a2cbe", "sources": {"Wismut/nym-pii-multilingual-data::default::default/test": 76, "gretelai/synthetic_pii_finance_multilingual::default::default/test": 74}, "question_types": {"noul": 150}, "gold_counts": {"false": 75, "true": 75}, "accuracy": 0.8933, "nll": 0.28, "brier": 0.1697, "ece": 0.0967, "balanced_accuracy": 0.8933, "seconds": 3.1, "gate": true}
|
| 4 |
+
{"family": "categories", "suite": "pii", "lang": "es", "model": "multilingual", "n": 150, "seed": 20260927, "pool": "cd56ab1fa3c1ee43+b8a87e85f3bf72b6", "pool_items_sha256": "853091fcb462b7b89e00759c1085ad942fb99eb455c18f4c5ecd8db63cb20e6e", "sources": {"Wismut/nym-pii-multilingual-data::default::default/test": 76, "gretelai/synthetic_pii_finance_multilingual::default::default/test": 74}, "question_types": {"noul": 150}, "gold_counts": {"false": 75, "true": 75}, "accuracy": 0.8533, "nll": 0.3837, "brier": 0.2297, "ece": 0.06, "balanced_accuracy": 0.8533, "seconds": 3.1, "gate": true}
|
| 5 |
+
{"family": "categories", "suite": "pii", "lang": "fr", "model": "multilingual", "n": 150, "seed": 20260927, "pool": "cd56ab1fa3c1ee43+b8a87e85f3bf72b6", "pool_items_sha256": "2be7c20bc628fd30fe60af29d6e457b9ce63cba3d8961d3528007cacba4fcbb0", "sources": {"Wismut/nym-pii-multilingual-data::default::default/test": 74, "gretelai/synthetic_pii_finance_multilingual::default::default/test": 76}, "question_types": {"noul": 150}, "gold_counts": {"false": 73, "true": 77}, "accuracy": 0.8933, "nll": 0.3102, "brier": 0.1746, "ece": 0.0387, "balanced_accuracy": 0.8922, "seconds": 3.0, "gate": true}
|
| 6 |
+
{"family": "categories", "suite": "pii", "lang": "it", "model": "multilingual", "n": 150, "seed": 20260927, "pool": "cd56ab1fa3c1ee43+b8a87e85f3bf72b6", "pool_items_sha256": "632ad4c14c949eb65c7261d70f3532ea54f0d930e8df47b6da1a69bd294dc5a3", "sources": {"Wismut/nym-pii-multilingual-data::default::default/test": 70, "gretelai/synthetic_pii_finance_multilingual::default::default/test": 80}, "question_types": {"noul": 150}, "gold_counts": {"false": 70, "true": 80}, "accuracy": 0.84, "nll": 0.3676, "brier": 0.2255, "ece": 0.0638, "balanced_accuracy": 0.8411, "seconds": 3.1, "gate": true}
|
| 7 |
+
{"family": "categories", "suite": "pii", "lang": "ja", "model": "multilingual", "n": 150, "seed": 20260927, "pool": "cd56ab1fa3c1ee43+b8a87e85f3bf72b6", "pool_items_sha256": "9d0a6eeb1fe27e2e5fcb7d7aa058c3bebfaa4e0fd650fe272e96049037cf03f6", "sources": {"Wismut/nym-pii-multilingual-data::default::default/test": 150}, "question_types": {"noul": 150}, "gold_counts": {"false": 66, "true": 84}, "accuracy": 0.8467, "nll": 0.331, "brier": 0.2079, "ece": 0.0413, "balanced_accuracy": 0.8501, "seconds": 1.7, "gate": true}
|
| 8 |
+
{"family": "categories", "suite": "pii", "lang": "nl", "model": "multilingual", "n": 150, "seed": 20260927, "pool": "cd56ab1fa3c1ee43+b8a87e85f3bf72b6", "pool_items_sha256": "72e3d46617d1027c2370304fbfea819c14e7d9e7be8a1e7e21e05137680472bd", "sources": {"Wismut/nym-pii-multilingual-data::default::default/test": 64, "gretelai/synthetic_pii_finance_multilingual::default::default/test": 86}, "question_types": {"noul": 150}, "gold_counts": {"false": 63, "true": 87}, "accuracy": 0.76, "nll": 0.5078, "brier": 0.3349, "ece": 0.1127, "balanced_accuracy": 0.7559, "seconds": 3.1, "gate": true}
|
| 9 |
+
{"family": "categories", "suite": "pii", "lang": "ru", "model": "multilingual", "n": 150, "seed": 20260927, "pool": "cd56ab1fa3c1ee43+b8a87e85f3bf72b6", "pool_items_sha256": "dde80741bffb81fecd80f3a66d88ea65ff9c6b15c050ece7db985ed910ef2f36", "sources": {"Wismut/nym-pii-multilingual-data::default::default/test": 150}, "question_types": {"noul": 150}, "gold_counts": {"false": 75, "true": 75}, "accuracy": 0.9067, "nll": 0.2355, "brier": 0.137, "ece": 0.0725, "balanced_accuracy": 0.9067, "seconds": 1.9, "gate": true}
|
| 10 |
+
{"family": "categories", "suite": "pii", "lang": "sv", "model": "multilingual", "n": 150, "seed": 20260927, "pool": "cd56ab1fa3c1ee43+b8a87e85f3bf72b6", "pool_items_sha256": "f0e3e43965db3fb821f3191ba6572b783dc15cdc927dc4e9680da8c2d07e9c5f", "sources": {"gretelai/synthetic_pii_finance_multilingual::default::default/test": 150}, "question_types": {"noul": 150}, "gold_counts": {"false": 75, "true": 75}, "accuracy": 0.8333, "nll": 0.3914, "brier": 0.244, "ece": 0.0326, "balanced_accuracy": 0.8333, "seconds": 3.4, "gate": true}
|
| 11 |
+
{"family": "categories", "suite": "pii", "lang": "zh", "model": "multilingual", "n": 150, "seed": 20260927, "pool": "cd56ab1fa3c1ee43+b8a87e85f3bf72b6", "pool_items_sha256": "6a85c938e961cc68abc1e79faf75d91c5af55c9a7c38176f0313bd23bf22b1dc", "sources": {"Wismut/nym-pii-multilingual-data::default::default/test": 150}, "question_types": {"noul": 150}, "gold_counts": {"false": 74, "true": 76}, "accuracy": 0.9, "nll": 0.2604, "brier": 0.1538, "ece": 0.0608, "balanced_accuracy": 0.9006, "seconds": 1.8, "gate": true}
|
| 12 |
+
{"family": "categories", "suite": "pii", "lang": "macro", "model": "multilingual", "n": 1650, "n_langs": 11, "accuracy": 0.8558, "balanced_accuracy": 0.857, "ece": 0.062, "nll": 0.3454, "brier": 0.2116}
|
| 13 |
+
{"family": "categories", "suite": "pii", "lang": "hi", "skipped": "only 132 pooled items, need 150"}
|
| 14 |
+
{"family": "categories", "suite": "pii", "lang": "pl", "skipped": "only 42 pooled items, need 150"}
|
| 15 |
+
{"family": "categories", "suite": "pii", "lang": "pt", "skipped": "only 54 pooled items, need 150"}
|
| 16 |
+
{"family": "categories", "suite": "pii", "lang": "tr", "skipped": "only 58 pooled items, need 150"}
|
evaluation/published-q8_0-adapter.jsonl
ADDED
|
@@ -0,0 +1,16 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{"family": "categories", "suite": "pii", "lang": "ar", "model": "multilingual:pii", "adapter": "pii", "n": 150, "seed": 20260927, "pool": "cd56ab1fa3c1ee43+b8a87e85f3bf72b6", "pool_items_sha256": "323112f111aa8c9c401e70d0f081ea6628b5230e57d741ac291694e16708c2b9", "sources": {"Wismut/nym-pii-multilingual-data::default::default/test": 150}, "question_types": {"noul": 150}, "gold_counts": {"false": 55, "true": 95}, "accuracy": 0.9067, "nll": 0.2534, "brier": 0.1526, "ece": 0.0462, "balanced_accuracy": 0.9187, "seconds": 2.4, "gate": true}
|
| 2 |
+
{"family": "categories", "suite": "pii", "lang": "de", "model": "multilingual:pii", "adapter": "pii", "n": 150, "seed": 20260927, "pool": "cd56ab1fa3c1ee43+b8a87e85f3bf72b6", "pool_items_sha256": "52145c7106ce9f85fa46c267cdace80c42181e148d9cb45594263718f5a49d92", "sources": {"Wismut/nym-pii-multilingual-data::default::default/test": 72, "gretelai/synthetic_pii_finance_multilingual::default::default/test": 78}, "question_types": {"noul": 150}, "gold_counts": {"false": 71, "true": 79}, "accuracy": 0.8667, "nll": 0.256, "brier": 0.1646, "ece": 0.0509, "balanced_accuracy": 0.8656, "seconds": 3.2, "gate": true}
|
| 3 |
+
{"family": "categories", "suite": "pii", "lang": "en", "model": "multilingual:pii", "adapter": "pii", "n": 150, "seed": 20260927, "pool": "cd56ab1fa3c1ee43+b8a87e85f3bf72b6", "pool_items_sha256": "10b8d7d2c36445bbeecdfded7c937f4ca228c43ab80c515dd038be0f658a2cbe", "sources": {"Wismut/nym-pii-multilingual-data::default::default/test": 76, "gretelai/synthetic_pii_finance_multilingual::default::default/test": 74}, "question_types": {"noul": 150}, "gold_counts": {"false": 75, "true": 75}, "accuracy": 0.9267, "nll": 0.1611, "brier": 0.0958, "ece": 0.0407, "balanced_accuracy": 0.9267, "seconds": 3.5, "gate": true}
|
| 4 |
+
{"family": "categories", "suite": "pii", "lang": "es", "model": "multilingual:pii", "adapter": "pii", "n": 150, "seed": 20260927, "pool": "cd56ab1fa3c1ee43+b8a87e85f3bf72b6", "pool_items_sha256": "853091fcb462b7b89e00759c1085ad942fb99eb455c18f4c5ecd8db63cb20e6e", "sources": {"Wismut/nym-pii-multilingual-data::default::default/test": 76, "gretelai/synthetic_pii_finance_multilingual::default::default/test": 74}, "question_types": {"noul": 150}, "gold_counts": {"false": 75, "true": 75}, "accuracy": 0.9, "nll": 0.2445, "brier": 0.1472, "ece": 0.0492, "balanced_accuracy": 0.9, "seconds": 3.5, "gate": true}
|
| 5 |
+
{"family": "categories", "suite": "pii", "lang": "fr", "model": "multilingual:pii", "adapter": "pii", "n": 150, "seed": 20260927, "pool": "cd56ab1fa3c1ee43+b8a87e85f3bf72b6", "pool_items_sha256": "2be7c20bc628fd30fe60af29d6e457b9ce63cba3d8961d3528007cacba4fcbb0", "sources": {"Wismut/nym-pii-multilingual-data::default::default/test": 74, "gretelai/synthetic_pii_finance_multilingual::default::default/test": 76}, "question_types": {"noul": 150}, "gold_counts": {"false": 73, "true": 77}, "accuracy": 0.9333, "nll": 0.2035, "brier": 0.1077, "ece": 0.032, "balanced_accuracy": 0.9329, "seconds": 3.4, "gate": true}
|
| 6 |
+
{"family": "categories", "suite": "pii", "lang": "it", "model": "multilingual:pii", "adapter": "pii", "n": 150, "seed": 20260927, "pool": "cd56ab1fa3c1ee43+b8a87e85f3bf72b6", "pool_items_sha256": "632ad4c14c949eb65c7261d70f3532ea54f0d930e8df47b6da1a69bd294dc5a3", "sources": {"Wismut/nym-pii-multilingual-data::default::default/test": 70, "gretelai/synthetic_pii_finance_multilingual::default::default/test": 80}, "question_types": {"noul": 150}, "gold_counts": {"false": 70, "true": 80}, "accuracy": 0.9267, "nll": 0.1823, "brier": 0.1088, "ece": 0.036, "balanced_accuracy": 0.9286, "seconds": 3.4, "gate": true}
|
| 7 |
+
{"family": "categories", "suite": "pii", "lang": "ja", "model": "multilingual:pii", "adapter": "pii", "n": 150, "seed": 20260927, "pool": "cd56ab1fa3c1ee43+b8a87e85f3bf72b6", "pool_items_sha256": "9d0a6eeb1fe27e2e5fcb7d7aa058c3bebfaa4e0fd650fe272e96049037cf03f6", "sources": {"Wismut/nym-pii-multilingual-data::default::default/test": 150}, "question_types": {"noul": 150}, "gold_counts": {"false": 66, "true": 84}, "accuracy": 0.9267, "nll": 0.27, "brier": 0.1455, "ece": 0.0583, "balanced_accuracy": 0.9264, "seconds": 1.9, "gate": true}
|
| 8 |
+
{"family": "categories", "suite": "pii", "lang": "nl", "model": "multilingual:pii", "adapter": "pii", "n": 150, "seed": 20260927, "pool": "cd56ab1fa3c1ee43+b8a87e85f3bf72b6", "pool_items_sha256": "72e3d46617d1027c2370304fbfea819c14e7d9e7be8a1e7e21e05137680472bd", "sources": {"Wismut/nym-pii-multilingual-data::default::default/test": 64, "gretelai/synthetic_pii_finance_multilingual::default::default/test": 86}, "question_types": {"noul": 150}, "gold_counts": {"false": 63, "true": 87}, "accuracy": 0.8667, "nll": 0.332, "brier": 0.2045, "ece": 0.065, "balanced_accuracy": 0.861, "seconds": 3.5, "gate": true}
|
| 9 |
+
{"family": "categories", "suite": "pii", "lang": "ru", "model": "multilingual:pii", "adapter": "pii", "n": 150, "seed": 20260927, "pool": "cd56ab1fa3c1ee43+b8a87e85f3bf72b6", "pool_items_sha256": "dde80741bffb81fecd80f3a66d88ea65ff9c6b15c050ece7db985ed910ef2f36", "sources": {"Wismut/nym-pii-multilingual-data::default::default/test": 150}, "question_types": {"noul": 150}, "gold_counts": {"false": 75, "true": 75}, "accuracy": 0.9467, "nll": 0.1359, "brier": 0.0785, "ece": 0.0275, "balanced_accuracy": 0.9467, "seconds": 2.2, "gate": true}
|
| 10 |
+
{"family": "categories", "suite": "pii", "lang": "sv", "model": "multilingual:pii", "adapter": "pii", "n": 150, "seed": 20260927, "pool": "cd56ab1fa3c1ee43+b8a87e85f3bf72b6", "pool_items_sha256": "f0e3e43965db3fb821f3191ba6572b783dc15cdc927dc4e9680da8c2d07e9c5f", "sources": {"gretelai/synthetic_pii_finance_multilingual::default::default/test": 150}, "question_types": {"noul": 150}, "gold_counts": {"false": 75, "true": 75}, "accuracy": 0.88, "nll": 0.2874, "brier": 0.1749, "ece": 0.0508, "balanced_accuracy": 0.88, "seconds": 3.9, "gate": true}
|
| 11 |
+
{"family": "categories", "suite": "pii", "lang": "zh", "model": "multilingual:pii", "adapter": "pii", "n": 150, "seed": 20260927, "pool": "cd56ab1fa3c1ee43+b8a87e85f3bf72b6", "pool_items_sha256": "6a85c938e961cc68abc1e79faf75d91c5af55c9a7c38176f0313bd23bf22b1dc", "sources": {"Wismut/nym-pii-multilingual-data::default::default/test": 150}, "question_types": {"noul": 150}, "gold_counts": {"false": 74, "true": 76}, "accuracy": 0.94, "nll": 0.1721, "brier": 0.0905, "ece": 0.0224, "balanced_accuracy": 0.9403, "seconds": 2.0, "gate": true}
|
| 12 |
+
{"family": "categories", "suite": "pii", "lang": "macro", "model": "multilingual:pii", "adapter": "pii", "n": 1650, "n_langs": 11, "accuracy": 0.9109, "balanced_accuracy": 0.9115, "ece": 0.0435, "nll": 0.2271, "brier": 0.1337}
|
| 13 |
+
{"family": "categories", "suite": "pii", "lang": "hi", "skipped": "only 132 pooled items, need 150"}
|
| 14 |
+
{"family": "categories", "suite": "pii", "lang": "pl", "skipped": "only 42 pooled items, need 150"}
|
| 15 |
+
{"family": "categories", "suite": "pii", "lang": "pt", "skipped": "only 54 pooled items, need 150"}
|
| 16 |
+
{"family": "categories", "suite": "pii", "lang": "tr", "skipped": "only 58 pooled items, need 150"}
|
evaluation/published-q8_0-base.jsonl
ADDED
|
@@ -0,0 +1,16 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{"family": "categories", "suite": "pii", "lang": "ar", "model": "multilingual", "n": 150, "seed": 20260927, "pool": "cd56ab1fa3c1ee43+b8a87e85f3bf72b6", "pool_items_sha256": "323112f111aa8c9c401e70d0f081ea6628b5230e57d741ac291694e16708c2b9", "sources": {"Wismut/nym-pii-multilingual-data::default::default/test": 150}, "question_types": {"noul": 150}, "gold_counts": {"false": 55, "true": 95}, "accuracy": 0.8467, "nll": 0.3616, "brier": 0.225, "ece": 0.0474, "balanced_accuracy": 0.8598, "seconds": 3.0, "gate": true}
|
| 2 |
+
{"family": "categories", "suite": "pii", "lang": "de", "model": "multilingual", "n": 150, "seed": 20260927, "pool": "cd56ab1fa3c1ee43+b8a87e85f3bf72b6", "pool_items_sha256": "52145c7106ce9f85fa46c267cdace80c42181e148d9cb45594263718f5a49d92", "sources": {"Wismut/nym-pii-multilingual-data::default::default/test": 72, "gretelai/synthetic_pii_finance_multilingual::default::default/test": 78}, "question_types": {"noul": 150}, "gold_counts": {"false": 71, "true": 79}, "accuracy": 0.84, "nll": 0.3699, "brier": 0.2263, "ece": 0.0444, "balanced_accuracy": 0.841, "seconds": 2.7, "gate": true}
|
| 3 |
+
{"family": "categories", "suite": "pii", "lang": "en", "model": "multilingual", "n": 150, "seed": 20260927, "pool": "cd56ab1fa3c1ee43+b8a87e85f3bf72b6", "pool_items_sha256": "10b8d7d2c36445bbeecdfded7c937f4ca228c43ab80c515dd038be0f658a2cbe", "sources": {"Wismut/nym-pii-multilingual-data::default::default/test": 76, "gretelai/synthetic_pii_finance_multilingual::default::default/test": 74}, "question_types": {"noul": 150}, "gold_counts": {"false": 75, "true": 75}, "accuracy": 0.8933, "nll": 0.2804, "brier": 0.1699, "ece": 0.0938, "balanced_accuracy": 0.8933, "seconds": 3.0, "gate": true}
|
| 4 |
+
{"family": "categories", "suite": "pii", "lang": "es", "model": "multilingual", "n": 150, "seed": 20260927, "pool": "cd56ab1fa3c1ee43+b8a87e85f3bf72b6", "pool_items_sha256": "853091fcb462b7b89e00759c1085ad942fb99eb455c18f4c5ecd8db63cb20e6e", "sources": {"Wismut/nym-pii-multilingual-data::default::default/test": 76, "gretelai/synthetic_pii_finance_multilingual::default::default/test": 74}, "question_types": {"noul": 150}, "gold_counts": {"false": 75, "true": 75}, "accuracy": 0.8533, "nll": 0.3797, "brier": 0.2275, "ece": 0.0654, "balanced_accuracy": 0.8533, "seconds": 3.0, "gate": true}
|
| 5 |
+
{"family": "categories", "suite": "pii", "lang": "fr", "model": "multilingual", "n": 150, "seed": 20260927, "pool": "cd56ab1fa3c1ee43+b8a87e85f3bf72b6", "pool_items_sha256": "2be7c20bc628fd30fe60af29d6e457b9ce63cba3d8961d3528007cacba4fcbb0", "sources": {"Wismut/nym-pii-multilingual-data::default::default/test": 74, "gretelai/synthetic_pii_finance_multilingual::default::default/test": 76}, "question_types": {"noul": 150}, "gold_counts": {"false": 73, "true": 77}, "accuracy": 0.8933, "nll": 0.3089, "brier": 0.1734, "ece": 0.0443, "balanced_accuracy": 0.8922, "seconds": 2.9, "gate": true}
|
| 6 |
+
{"family": "categories", "suite": "pii", "lang": "it", "model": "multilingual", "n": 150, "seed": 20260927, "pool": "cd56ab1fa3c1ee43+b8a87e85f3bf72b6", "pool_items_sha256": "632ad4c14c949eb65c7261d70f3532ea54f0d930e8df47b6da1a69bd294dc5a3", "sources": {"Wismut/nym-pii-multilingual-data::default::default/test": 70, "gretelai/synthetic_pii_finance_multilingual::default::default/test": 80}, "question_types": {"noul": 150}, "gold_counts": {"false": 70, "true": 80}, "accuracy": 0.84, "nll": 0.3704, "brier": 0.2271, "ece": 0.0688, "balanced_accuracy": 0.8411, "seconds": 3.0, "gate": true}
|
| 7 |
+
{"family": "categories", "suite": "pii", "lang": "ja", "model": "multilingual", "n": 150, "seed": 20260927, "pool": "cd56ab1fa3c1ee43+b8a87e85f3bf72b6", "pool_items_sha256": "9d0a6eeb1fe27e2e5fcb7d7aa058c3bebfaa4e0fd650fe272e96049037cf03f6", "sources": {"Wismut/nym-pii-multilingual-data::default::default/test": 150}, "question_types": {"noul": 150}, "gold_counts": {"false": 66, "true": 84}, "accuracy": 0.8467, "nll": 0.3313, "brier": 0.2076, "ece": 0.0431, "balanced_accuracy": 0.8501, "seconds": 1.6, "gate": true}
|
| 8 |
+
{"family": "categories", "suite": "pii", "lang": "nl", "model": "multilingual", "n": 150, "seed": 20260927, "pool": "cd56ab1fa3c1ee43+b8a87e85f3bf72b6", "pool_items_sha256": "72e3d46617d1027c2370304fbfea819c14e7d9e7be8a1e7e21e05137680472bd", "sources": {"Wismut/nym-pii-multilingual-data::default::default/test": 64, "gretelai/synthetic_pii_finance_multilingual::default::default/test": 86}, "question_types": {"noul": 150}, "gold_counts": {"false": 63, "true": 87}, "accuracy": 0.76, "nll": 0.5088, "brier": 0.3352, "ece": 0.1111, "balanced_accuracy": 0.7559, "seconds": 3.0, "gate": true}
|
| 9 |
+
{"family": "categories", "suite": "pii", "lang": "ru", "model": "multilingual", "n": 150, "seed": 20260927, "pool": "cd56ab1fa3c1ee43+b8a87e85f3bf72b6", "pool_items_sha256": "dde80741bffb81fecd80f3a66d88ea65ff9c6b15c050ece7db985ed910ef2f36", "sources": {"Wismut/nym-pii-multilingual-data::default::default/test": 150}, "question_types": {"noul": 150}, "gold_counts": {"false": 75, "true": 75}, "accuracy": 0.9067, "nll": 0.2357, "brier": 0.1371, "ece": 0.077, "balanced_accuracy": 0.9067, "seconds": 1.9, "gate": true}
|
| 10 |
+
{"family": "categories", "suite": "pii", "lang": "sv", "model": "multilingual", "n": 150, "seed": 20260927, "pool": "cd56ab1fa3c1ee43+b8a87e85f3bf72b6", "pool_items_sha256": "f0e3e43965db3fb821f3191ba6572b783dc15cdc927dc4e9680da8c2d07e9c5f", "sources": {"gretelai/synthetic_pii_finance_multilingual::default::default/test": 150}, "question_types": {"noul": 150}, "gold_counts": {"false": 75, "true": 75}, "accuracy": 0.8333, "nll": 0.3908, "brier": 0.2449, "ece": 0.041, "balanced_accuracy": 0.8333, "seconds": 3.3, "gate": true}
|
| 11 |
+
{"family": "categories", "suite": "pii", "lang": "zh", "model": "multilingual", "n": 150, "seed": 20260927, "pool": "cd56ab1fa3c1ee43+b8a87e85f3bf72b6", "pool_items_sha256": "6a85c938e961cc68abc1e79faf75d91c5af55c9a7c38176f0313bd23bf22b1dc", "sources": {"Wismut/nym-pii-multilingual-data::default::default/test": 150}, "question_types": {"noul": 150}, "gold_counts": {"false": 74, "true": 76}, "accuracy": 0.9, "nll": 0.2605, "brier": 0.1537, "ece": 0.0431, "balanced_accuracy": 0.9006, "seconds": 1.7, "gate": true}
|
| 12 |
+
{"family": "categories", "suite": "pii", "lang": "macro", "model": "multilingual", "n": 1650, "n_langs": 11, "accuracy": 0.8558, "balanced_accuracy": 0.857, "ece": 0.0618, "nll": 0.3453, "brier": 0.2116}
|
| 13 |
+
{"family": "categories", "suite": "pii", "lang": "hi", "skipped": "only 132 pooled items, need 150"}
|
| 14 |
+
{"family": "categories", "suite": "pii", "lang": "pl", "skipped": "only 42 pooled items, need 150"}
|
| 15 |
+
{"family": "categories", "suite": "pii", "lang": "pt", "skipped": "only 54 pooled items, need 150"}
|
| 16 |
+
{"family": "categories", "suite": "pii", "lang": "tr", "skipped": "only 58 pooled items, need 150"}
|
evaluation/summary.json
ADDED
|
@@ -0,0 +1,279 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"n": 150,
|
| 3 |
+
"seed": 20260927,
|
| 4 |
+
"z": 2.0,
|
| 5 |
+
"alpha": 0.05,
|
| 6 |
+
"created": "2026-09-29T02:04:37",
|
| 7 |
+
"argv": [
|
| 8 |
+
"tools/finetune/lora_experiment.py",
|
| 9 |
+
"--base-checkpoint",
|
| 10 |
+
"models/laya-multilingual-v9",
|
| 11 |
+
"--base-gguf",
|
| 12 |
+
"models/laya-multilingual-v9-f32.gguf",
|
| 13 |
+
"--mixture",
|
| 14 |
+
"data/mixture-v8.jsonl.gz",
|
| 15 |
+
"--statim",
|
| 16 |
+
"build-vk/statim",
|
| 17 |
+
"--device",
|
| 18 |
+
"cuda",
|
| 19 |
+
"--server-device",
|
| 20 |
+
"vulkan",
|
| 21 |
+
"--work",
|
| 22 |
+
"models/lora-exp1",
|
| 23 |
+
"--train-args",
|
| 24 |
+
"--epochs 2"
|
| 25 |
+
],
|
| 26 |
+
"category": "pii",
|
| 27 |
+
"status": "ok",
|
| 28 |
+
"stage": "done",
|
| 29 |
+
"commands": {
|
| 30 |
+
"train": "'.venv-train/bin/python' 'tools/finetune/train_lora.py' 'models/laya-multilingual-v9' --mixture 'data/mixture-v8.jsonl.gz' --category pii --registry 'tools/finetune/sources/v6-keep.json' --out 'models/lora-exp1/pii' --device cuda --epochs 2",
|
| 31 |
+
"convert": "'.venv/bin/python' 'tools/convert_lora.py' 'models/lora-exp1/pii' -o 'models/lora-exp1/pii.lora.gguf' --base 'models/laya-multilingual-v9-f32.gguf' --category pii --name pii",
|
| 32 |
+
"serve": "'build-vk/statim' serve -m 'multilingual=models/laya-multilingual-v9-f32.gguf' --adapter 'multilingual:pii=models/lora-exp1/pii.lora.gguf' --device vulkan --threads 16 --port 8098 --no-access-log --inference-timeout 600",
|
| 33 |
+
"eval_base": "'.venv/bin/python' 'bench/eval_categories.py' --strict --url http://127.0.0.1:8098 --model multilingual --suites pii --n 150 --seed 20260927 --exclude-mixture 'data/mixture-v8.jsonl.gz' --out 'models/lora-exp1/pii.base.jsonl'",
|
| 34 |
+
"eval_adapter": "'.venv/bin/python' 'bench/eval_categories.py' --strict --url http://127.0.0.1:8098 --model multilingual --suites pii --n 150 --seed 20260927 --exclude-mixture 'data/mixture-v8.jsonl.gz' --adapter pii --out 'models/lora-exp1/pii.adapter.jsonl'"
|
| 35 |
+
},
|
| 36 |
+
"train": {
|
| 37 |
+
"items": {
|
| 38 |
+
"train": 33571,
|
| 39 |
+
"dev": 400,
|
| 40 |
+
"dropped_marker_mismatch": 0
|
| 41 |
+
},
|
| 42 |
+
"best": {
|
| 43 |
+
"dev_acc": 0.8725,
|
| 44 |
+
"epoch": 2,
|
| 45 |
+
"update": 1172,
|
| 46 |
+
"saved_initial": false
|
| 47 |
+
},
|
| 48 |
+
"dev_before": {
|
| 49 |
+
"dev_acc": 0.8375,
|
| 50 |
+
"dev_by_lang": {
|
| 51 |
+
"?": 0.8375
|
| 52 |
+
}
|
| 53 |
+
},
|
| 54 |
+
"updates": 1172,
|
| 55 |
+
"seconds": 1631.7,
|
| 56 |
+
"peak_memory_mb": 2332,
|
| 57 |
+
"device": "cuda",
|
| 58 |
+
"base_sha256": "c44425f14ac9d55508f73a6f371e4e2802ed59e646287c3abbec10b653a19840",
|
| 59 |
+
"mixture": "data/mixture-v8.jsonl.gz",
|
| 60 |
+
"mixture_sha256": "b8a87e85f3bf72b61509555aba9148358760d3329ecf37a36cebd94f9358cb03",
|
| 61 |
+
"lora": {
|
| 62 |
+
"r": 16,
|
| 63 |
+
"lora_alpha": 32.0,
|
| 64 |
+
"lora_dropout": 0.05,
|
| 65 |
+
"init": "true",
|
| 66 |
+
"target_modules": [
|
| 67 |
+
"Wqkv",
|
| 68 |
+
"Wo",
|
| 69 |
+
"Wi"
|
| 70 |
+
],
|
| 71 |
+
"wrapped_modules": 88,
|
| 72 |
+
"trainable_parameters": 3379200
|
| 73 |
+
}
|
| 74 |
+
},
|
| 75 |
+
"convert": "wrote models/lora-exp1/pii.lora.gguf (88 LoRA pairs, rank 16, 13.53 MB)",
|
| 76 |
+
"server_startup_seconds": 1.5,
|
| 77 |
+
"eval_base_seconds": 93.1,
|
| 78 |
+
"eval_adapter_seconds": 92.0,
|
| 79 |
+
"promote": true,
|
| 80 |
+
"reason": "",
|
| 81 |
+
"cells": [
|
| 82 |
+
{
|
| 83 |
+
"cell": "categories:pii/ar",
|
| 84 |
+
"lang": "ar",
|
| 85 |
+
"n_base": 150,
|
| 86 |
+
"n_adapter": 150,
|
| 87 |
+
"base_acc": 0.8467,
|
| 88 |
+
"adapter_acc": 0.9067,
|
| 89 |
+
"delta": 0.06,
|
| 90 |
+
"se": 0.0378,
|
| 91 |
+
"z": 1.59,
|
| 92 |
+
"p_drop": 0.9437492295760375,
|
| 93 |
+
"p_drop_holm": 1.0,
|
| 94 |
+
"verdict": "within noise"
|
| 95 |
+
},
|
| 96 |
+
{
|
| 97 |
+
"cell": "categories:pii/de",
|
| 98 |
+
"lang": "de",
|
| 99 |
+
"n_base": 150,
|
| 100 |
+
"n_adapter": 150,
|
| 101 |
+
"base_acc": 0.84,
|
| 102 |
+
"adapter_acc": 0.8667,
|
| 103 |
+
"delta": 0.0267,
|
| 104 |
+
"se": 0.0408,
|
| 105 |
+
"z": 0.65,
|
| 106 |
+
"p_drop": 0.7434775622497196,
|
| 107 |
+
"p_drop_holm": 1.0,
|
| 108 |
+
"verdict": "within noise"
|
| 109 |
+
},
|
| 110 |
+
{
|
| 111 |
+
"cell": "categories:pii/en",
|
| 112 |
+
"lang": "en",
|
| 113 |
+
"n_base": 150,
|
| 114 |
+
"n_adapter": 150,
|
| 115 |
+
"base_acc": 0.8933,
|
| 116 |
+
"adapter_acc": 0.9267,
|
| 117 |
+
"delta": 0.0334,
|
| 118 |
+
"se": 0.033,
|
| 119 |
+
"z": 1.01,
|
| 120 |
+
"p_drop": 0.8443397910717234,
|
| 121 |
+
"p_drop_holm": 1.0,
|
| 122 |
+
"verdict": "within noise"
|
| 123 |
+
},
|
| 124 |
+
{
|
| 125 |
+
"cell": "categories:pii/es",
|
| 126 |
+
"lang": "es",
|
| 127 |
+
"n_base": 150,
|
| 128 |
+
"n_adapter": 150,
|
| 129 |
+
"base_acc": 0.8533,
|
| 130 |
+
"adapter_acc": 0.8933,
|
| 131 |
+
"delta": 0.04,
|
| 132 |
+
"se": 0.0383,
|
| 133 |
+
"z": 1.04,
|
| 134 |
+
"p_drop": 0.8515940958363684,
|
| 135 |
+
"p_drop_holm": 1.0,
|
| 136 |
+
"verdict": "within noise"
|
| 137 |
+
},
|
| 138 |
+
{
|
| 139 |
+
"cell": "categories:pii/fr",
|
| 140 |
+
"lang": "fr",
|
| 141 |
+
"n_base": 150,
|
| 142 |
+
"n_adapter": 150,
|
| 143 |
+
"base_acc": 0.8933,
|
| 144 |
+
"adapter_acc": 0.94,
|
| 145 |
+
"delta": 0.0467,
|
| 146 |
+
"se": 0.0318,
|
| 147 |
+
"z": 1.47,
|
| 148 |
+
"p_drop": 0.9290039046681648,
|
| 149 |
+
"p_drop_holm": 1.0,
|
| 150 |
+
"verdict": "within noise"
|
| 151 |
+
},
|
| 152 |
+
{
|
| 153 |
+
"cell": "categories:pii/it",
|
| 154 |
+
"lang": "it",
|
| 155 |
+
"n_base": 150,
|
| 156 |
+
"n_adapter": 150,
|
| 157 |
+
"base_acc": 0.84,
|
| 158 |
+
"adapter_acc": 0.92,
|
| 159 |
+
"delta": 0.08,
|
| 160 |
+
"se": 0.0372,
|
| 161 |
+
"z": 2.15,
|
| 162 |
+
"p_drop": 0.9841568057301032,
|
| 163 |
+
"p_drop_holm": 1.0,
|
| 164 |
+
"verdict": "gain (2 SE)"
|
| 165 |
+
},
|
| 166 |
+
{
|
| 167 |
+
"cell": "categories:pii/ja",
|
| 168 |
+
"lang": "ja",
|
| 169 |
+
"n_base": 150,
|
| 170 |
+
"n_adapter": 150,
|
| 171 |
+
"base_acc": 0.8467,
|
| 172 |
+
"adapter_acc": 0.9267,
|
| 173 |
+
"delta": 0.08,
|
| 174 |
+
"se": 0.0363,
|
| 175 |
+
"z": 2.2,
|
| 176 |
+
"p_drop": 0.9862185271207049,
|
| 177 |
+
"p_drop_holm": 1.0,
|
| 178 |
+
"verdict": "gain (2 SE)"
|
| 179 |
+
},
|
| 180 |
+
{
|
| 181 |
+
"cell": "categories:pii/nl",
|
| 182 |
+
"lang": "nl",
|
| 183 |
+
"n_base": 150,
|
| 184 |
+
"n_adapter": 150,
|
| 185 |
+
"base_acc": 0.76,
|
| 186 |
+
"adapter_acc": 0.8667,
|
| 187 |
+
"delta": 0.1067,
|
| 188 |
+
"se": 0.0446,
|
| 189 |
+
"z": 2.39,
|
| 190 |
+
"p_drop": 0.9916706286055883,
|
| 191 |
+
"p_drop_holm": 1.0,
|
| 192 |
+
"verdict": "gain (2 SE)"
|
| 193 |
+
},
|
| 194 |
+
{
|
| 195 |
+
"cell": "categories:pii/ru",
|
| 196 |
+
"lang": "ru",
|
| 197 |
+
"n_base": 150,
|
| 198 |
+
"n_adapter": 150,
|
| 199 |
+
"base_acc": 0.9067,
|
| 200 |
+
"adapter_acc": 0.9467,
|
| 201 |
+
"delta": 0.04,
|
| 202 |
+
"se": 0.03,
|
| 203 |
+
"z": 1.33,
|
| 204 |
+
"p_drop": 0.9087448714594788,
|
| 205 |
+
"p_drop_holm": 1.0,
|
| 206 |
+
"verdict": "within noise"
|
| 207 |
+
},
|
| 208 |
+
{
|
| 209 |
+
"cell": "categories:pii/sv",
|
| 210 |
+
"lang": "sv",
|
| 211 |
+
"n_base": 150,
|
| 212 |
+
"n_adapter": 150,
|
| 213 |
+
"base_acc": 0.8333,
|
| 214 |
+
"adapter_acc": 0.8733,
|
| 215 |
+
"delta": 0.04,
|
| 216 |
+
"se": 0.0408,
|
| 217 |
+
"z": 0.98,
|
| 218 |
+
"p_drop": 0.8366204616828873,
|
| 219 |
+
"p_drop_holm": 1.0,
|
| 220 |
+
"verdict": "within noise"
|
| 221 |
+
},
|
| 222 |
+
{
|
| 223 |
+
"cell": "categories:pii/zh",
|
| 224 |
+
"lang": "zh",
|
| 225 |
+
"n_base": 150,
|
| 226 |
+
"n_adapter": 150,
|
| 227 |
+
"base_acc": 0.9,
|
| 228 |
+
"adapter_acc": 0.9467,
|
| 229 |
+
"delta": 0.0467,
|
| 230 |
+
"se": 0.0306,
|
| 231 |
+
"z": 1.53,
|
| 232 |
+
"p_drop": 0.9365095146474287,
|
| 233 |
+
"p_drop_holm": 1.0,
|
| 234 |
+
"verdict": "within noise"
|
| 235 |
+
}
|
| 236 |
+
],
|
| 237 |
+
"family": {
|
| 238 |
+
"rows": {
|
| 239 |
+
"delta": 0.0546,
|
| 240 |
+
"se": 0.0111,
|
| 241 |
+
"groups": 11,
|
| 242 |
+
"flag": "gain"
|
| 243 |
+
},
|
| 244 |
+
"suites": {
|
| 245 |
+
"delta": 0.0546,
|
| 246 |
+
"se": 0.0111,
|
| 247 |
+
"groups": 1,
|
| 248 |
+
"flag": "gain"
|
| 249 |
+
}
|
| 250 |
+
},
|
| 251 |
+
"not_compared": [],
|
| 252 |
+
"harms": [],
|
| 253 |
+
"reported": {
|
| 254 |
+
"base": {},
|
| 255 |
+
"adapter": {}
|
| 256 |
+
},
|
| 257 |
+
"skipped": [
|
| 258 |
+
{
|
| 259 |
+
"lang": "hi",
|
| 260 |
+
"why": "only 132 pooled items, need 150"
|
| 261 |
+
},
|
| 262 |
+
{
|
| 263 |
+
"lang": "pl",
|
| 264 |
+
"why": "only 42 pooled items, need 150"
|
| 265 |
+
},
|
| 266 |
+
{
|
| 267 |
+
"lang": "pt",
|
| 268 |
+
"why": "only 54 pooled items, need 150"
|
| 269 |
+
},
|
| 270 |
+
{
|
| 271 |
+
"lang": "tr",
|
| 272 |
+
"why": "only 58 pooled items, need 150"
|
| 273 |
+
}
|
| 274 |
+
],
|
| 275 |
+
"pool": [
|
| 276 |
+
"a6d43a981c13a016+b8a87e85f3bf72b6"
|
| 277 |
+
],
|
| 278 |
+
"seconds": 1819.9
|
| 279 |
+
}
|
peft/adapter_config.json
ADDED
|
@@ -0,0 +1,50 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"alora_invocation_tokens": null,
|
| 3 |
+
"alpha_pattern": {},
|
| 4 |
+
"arrow_config": null,
|
| 5 |
+
"auto_mapping": {
|
| 6 |
+
"base_model_class": "DecisionModel",
|
| 7 |
+
"parent_library": "laya.common"
|
| 8 |
+
},
|
| 9 |
+
"base_model_name_or_path": null,
|
| 10 |
+
"bias": "none",
|
| 11 |
+
"corda_config": null,
|
| 12 |
+
"ensure_weight_tying": false,
|
| 13 |
+
"eva_config": null,
|
| 14 |
+
"exclude_modules": null,
|
| 15 |
+
"fan_in_fan_out": false,
|
| 16 |
+
"inference_mode": true,
|
| 17 |
+
"init_lora_weights": true,
|
| 18 |
+
"kasa_config": null,
|
| 19 |
+
"layer_replication": null,
|
| 20 |
+
"layers_pattern": null,
|
| 21 |
+
"layers_to_transform": null,
|
| 22 |
+
"loftq_config": {},
|
| 23 |
+
"lora_alpha": 32.0,
|
| 24 |
+
"lora_bias": false,
|
| 25 |
+
"lora_dropout": 0.05,
|
| 26 |
+
"lora_ga_config": null,
|
| 27 |
+
"megatron_config": null,
|
| 28 |
+
"megatron_core": "megatron.core",
|
| 29 |
+
"modules_to_save": null,
|
| 30 |
+
"monteclora_config": null,
|
| 31 |
+
"peft_type": "LORA",
|
| 32 |
+
"peft_version": "0.21.0",
|
| 33 |
+
"qalora_group_size": 16,
|
| 34 |
+
"r": 16,
|
| 35 |
+
"rank_pattern": {},
|
| 36 |
+
"revision": null,
|
| 37 |
+
"target_modules": [
|
| 38 |
+
"Wo",
|
| 39 |
+
"Wi",
|
| 40 |
+
"Wqkv"
|
| 41 |
+
],
|
| 42 |
+
"target_parameters": null,
|
| 43 |
+
"task_type": null,
|
| 44 |
+
"trainable_token_indices": null,
|
| 45 |
+
"use_bdlora": null,
|
| 46 |
+
"use_dora": false,
|
| 47 |
+
"use_qalora": false,
|
| 48 |
+
"use_rslora": false,
|
| 49 |
+
"velora_config": null
|
| 50 |
+
}
|
peft/adapter_model.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f8e7e674ab868b312ff32f403552daa436e7437411e83fa0ece60187431a2d6f
|
| 3 |
+
size 13538880
|
statim-decide-multilingual-base-pii.lora.gguf
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:2991a33b5d9db4f5b679081885faab1f84289d6eae7660830447e790bd168b29
|
| 3 |
+
size 13531200
|
training/train_lora.json
ADDED
|
@@ -0,0 +1,165 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"category": "pii",
|
| 3 |
+
"base": "models/laya-multilingual-v9",
|
| 4 |
+
"base_sha256": "c44425f14ac9d55508f73a6f371e4e2802ed59e646287c3abbec10b653a19840",
|
| 5 |
+
"mixture": "data/mixture-v8.jsonl.gz",
|
| 6 |
+
"mixture_sha256": "b8a87e85f3bf72b61509555aba9148358760d3329ecf37a36cebd94f9358cb03",
|
| 7 |
+
"registry": "tools/finetune/sources/v6-keep.json",
|
| 8 |
+
"registry_sha256": "309ce8ec262886b3bfaa529934151ae0c1879595c26301e4269da276d95cbfcd",
|
| 9 |
+
"sources": {
|
| 10 |
+
"v6/E3-JSI/synthetic-multi-pii-ner-v1/default": {
|
| 11 |
+
"rows": 2971,
|
| 12 |
+
"train": 2936,
|
| 13 |
+
"dev": 35
|
| 14 |
+
},
|
| 15 |
+
"v6/Wismut/nym-pii-multilingual-data/default": {
|
| 16 |
+
"rows": 6200,
|
| 17 |
+
"train": 6127,
|
| 18 |
+
"dev": 73
|
| 19 |
+
},
|
| 20 |
+
"v6/gretelai/gretel-pii-masking-en-v1/default": {
|
| 21 |
+
"rows": 6200,
|
| 22 |
+
"train": 6127,
|
| 23 |
+
"dev": 73
|
| 24 |
+
},
|
| 25 |
+
"v6/gretelai/synthetic_pii_finance_multilingual/default": {
|
| 26 |
+
"rows": 6200,
|
| 27 |
+
"train": 6127,
|
| 28 |
+
"dev": 73
|
| 29 |
+
},
|
| 30 |
+
"v6/nvidia/Nemotron-PII/default": {
|
| 31 |
+
"rows": 6200,
|
| 32 |
+
"train": 6127,
|
| 33 |
+
"dev": 73
|
| 34 |
+
},
|
| 35 |
+
"v6/urchade/synthetic-pii-ner-mistral-v1/data.json": {
|
| 36 |
+
"rows": 6200,
|
| 37 |
+
"train": 6127,
|
| 38 |
+
"dev": 73
|
| 39 |
+
}
|
| 40 |
+
},
|
| 41 |
+
"selection": {
|
| 42 |
+
"rows": 33971,
|
| 43 |
+
"other_category_rows": 500250,
|
| 44 |
+
"unknown_src_rows": 191750,
|
| 45 |
+
"unknown_src_top": {
|
| 46 |
+
"benayas/snips/en/train": 3000,
|
| 47 |
+
"commonsense_qa_2.0": 2000,
|
| 48 |
+
"procedural-typed-decisions/partial_observation_calibration": 2000,
|
| 49 |
+
"utilitarianism": 2000,
|
| 50 |
+
"agent_action_safety": 2000,
|
| 51 |
+
"civil_comments/sexual_explicit_share": 2000,
|
| 52 |
+
"civil_comments/severe_toxicity_share": 2000,
|
| 53 |
+
"lex_glue/case_hold": 2000,
|
| 54 |
+
"shell-safety-v2": 2000,
|
| 55 |
+
"social_i_qa": 2000
|
| 56 |
+
},
|
| 57 |
+
"limited": false
|
| 58 |
+
},
|
| 59 |
+
"items": {
|
| 60 |
+
"train": 33571,
|
| 61 |
+
"dev": 400,
|
| 62 |
+
"dropped_marker_mismatch": 0
|
| 63 |
+
},
|
| 64 |
+
"max_len": 1024,
|
| 65 |
+
"head_max_len": 512,
|
| 66 |
+
"lora": {
|
| 67 |
+
"r": 16,
|
| 68 |
+
"lora_alpha": 32.0,
|
| 69 |
+
"lora_dropout": 0.05,
|
| 70 |
+
"init": "true",
|
| 71 |
+
"target_modules": [
|
| 72 |
+
"Wqkv",
|
| 73 |
+
"Wo",
|
| 74 |
+
"Wi"
|
| 75 |
+
],
|
| 76 |
+
"wrapped_modules": 88,
|
| 77 |
+
"trainable_parameters": 3379200
|
| 78 |
+
},
|
| 79 |
+
"device": "cuda",
|
| 80 |
+
"amp": "torch.bfloat16",
|
| 81 |
+
"base_dtype": "fp32",
|
| 82 |
+
"gradient_checkpointing": true,
|
| 83 |
+
"head_checkpointing": true,
|
| 84 |
+
"head_mode": "eval (frozen, dropout off)",
|
| 85 |
+
"dev_before": {
|
| 86 |
+
"dev_acc": 0.8375,
|
| 87 |
+
"dev_by_lang": {
|
| 88 |
+
"?": 0.8375
|
| 89 |
+
}
|
| 90 |
+
},
|
| 91 |
+
"log": [
|
| 92 |
+
{
|
| 93 |
+
"epoch": 1,
|
| 94 |
+
"update": 586,
|
| 95 |
+
"train_ce": 0.5629,
|
| 96 |
+
"lr": 0.00011041148082216535,
|
| 97 |
+
"seconds": 791.6,
|
| 98 |
+
"dev_acc": 0.865,
|
| 99 |
+
"dev_by_lang": {
|
| 100 |
+
"?": 0.865
|
| 101 |
+
}
|
| 102 |
+
},
|
| 103 |
+
{
|
| 104 |
+
"epoch": 2,
|
| 105 |
+
"update": 1172,
|
| 106 |
+
"train_ce": 0.3944,
|
| 107 |
+
"lr": 1e-06,
|
| 108 |
+
"seconds": 1582.3,
|
| 109 |
+
"dev_acc": 0.8725,
|
| 110 |
+
"dev_by_lang": {
|
| 111 |
+
"?": 0.8725
|
| 112 |
+
}
|
| 113 |
+
}
|
| 114 |
+
],
|
| 115 |
+
"updates": 1172,
|
| 116 |
+
"updates_planned": 1172,
|
| 117 |
+
"best": {
|
| 118 |
+
"dev_acc": 0.8725,
|
| 119 |
+
"epoch": 2,
|
| 120 |
+
"update": 1172,
|
| 121 |
+
"saved_initial": false
|
| 122 |
+
},
|
| 123 |
+
"train_seconds": 1585.2,
|
| 124 |
+
"peak_memory_mb": 2332,
|
| 125 |
+
"peak_memory_kind": "cuda max_memory_allocated",
|
| 126 |
+
"seconds": 1631.7,
|
| 127 |
+
"args": {
|
| 128 |
+
"base": "models/laya-multilingual-v9",
|
| 129 |
+
"mixture": "data/mixture-v8.jsonl.gz",
|
| 130 |
+
"category": "pii",
|
| 131 |
+
"registry": "tools/finetune/sources/v6-keep.json",
|
| 132 |
+
"out": "models/lora-exp1/pii",
|
| 133 |
+
"r": 16,
|
| 134 |
+
"lora_alpha": 32.0,
|
| 135 |
+
"lora_dropout": 0.05,
|
| 136 |
+
"init": "true",
|
| 137 |
+
"epochs": 2,
|
| 138 |
+
"lr": 0.0002,
|
| 139 |
+
"weight_decay": 0.0,
|
| 140 |
+
"warmup": 0.06,
|
| 141 |
+
"max_tokens": 8192,
|
| 142 |
+
"max_rows": 64,
|
| 143 |
+
"accum": 2,
|
| 144 |
+
"dev_items": 400,
|
| 145 |
+
"eval_every": 0,
|
| 146 |
+
"max_steps": 0,
|
| 147 |
+
"limit_items": 0,
|
| 148 |
+
"max_len": 0,
|
| 149 |
+
"head_max_len": 512,
|
| 150 |
+
"device": "cuda",
|
| 151 |
+
"amp": "auto",
|
| 152 |
+
"base_dtype": "fp32",
|
| 153 |
+
"grad_checkpointing": true,
|
| 154 |
+
"head_checkpointing": true,
|
| 155 |
+
"seed": 20260926,
|
| 156 |
+
"profile_batch": null
|
| 157 |
+
},
|
| 158 |
+
"versions": {
|
| 159 |
+
"python": "3.12.3",
|
| 160 |
+
"torch": "2.14.0+cu130",
|
| 161 |
+
"peft": "0.21.0",
|
| 162 |
+
"transformers": "5.17.0",
|
| 163 |
+
"laya": "0.3.20"
|
| 164 |
+
}
|
| 165 |
+
}
|