# stanford-deidentifier-base by Stanford AIMI: Open Model Source: https://savrn.com/models/stanford-deidentifier-base Markdown alternate of the page above; the site index is https://savrn.com/llms.txt --- ## SAVRN's Notes on stanford-deidentifier-base PII de-identification for clinical text: a BERT-shaped token classifier — 12 layers, 768 hidden, 512 tokens of context — trained on radiology and biomedical documents, published at 438 MB in a single pytorch_model.bin. No parameter count is stated, so there is no Runs On table; the model is small enough to colocate with your ETL and run on CPU. MIT license, and Stanford AIMI calls these the recommended weights among its deidentifier checkpoints. The intended home is a de-identification pipeline, not a chat endpoint: the associated MIDRC/Stanford_Penn_Deidentifier repository carries the surrounding work, funded under NIH/MIDRC and ARPA-H. No eval numbers sit on our record, so measure recall on your own note types — a miss is a regulatory problem, not a metric. For clinical NLP in production this is the compliance-first default; for general PII outside healthcare, a broader-domain model is the safer pick. SAVRN Research ## Model Card By Stanford AIMI, published under mit, revision 661b9c1c717d. Stanford de-identifier was trained on a variety of radiology and biomedical documents with the goal of automatising the de-identification process while reaching satisfactory accuracy for use in production. Manuscript in-proceedings. These model weights are the recommended ones among all available deidentifier weights. Associated github repo: https://github.com/MIDRC/Stanford_Penn_Deidentifier ### Acknowledgement This work was supported in part by the Medical Imaging and Data Resource Center (MIDRC), which is funded by the National Institute of Biomedical Imaging and Bioengineering (NIBIB) of the National Institutes of Health under contract 75N92020D00021 and through The Advanced Research Projects Agency for Health (ARPA-H) ### Citation [Read the full model card (412 words)](https://savrn.com/models/stanford-deidentifier-base/card) ## Configuration Architecture AutoModel Context length (tokens) 512 Layers 12 Hidden size 768 Feed-forward size 3,072 Attention heads 12 Vocabulary size 30,522 Model type bert ## Identity and Version Repository StanfordAIMI/stanford-deidentifier-base Publisher Stanford AIMI Task Token classification Modality Text Library transformers Parameters Not stated by the source Languages en Revision 661b9c1c717d3165512d440abc3700c386aefab6 First published 2022-06-16 Last updated 2024-10-09 ## Files and Weights 7 files, 438.3 MB in total. The weights are 1 file totalling 438.0 MB in bin. Weights1 file · 438.0 MB Configuration2 files · 1.3 KB Tokenizer2 files · 226.2 KB Documentation1 file · 3.9 KB Repository1 file · 1.2 KB Every file | File | Type | Size | SHA-256 | | --- | --- | --- | --- | | pytorch_model.bin | Weights | 438.0 MB | fa49ef069171 | | config.json | Configuration | 1.2 KB | — | | special_tokens_map.json | Configuration | 112 B | — | | README.md | Documentation | 3.9 KB | — | | .gitattributes | Repository | 1.2 KB | — | | tokenizer_config.json | Tokenizer | 29 B | — | | vocab.txt | Tokenizer | 226.2 KB | — | ## License and Download License mit Access Open weights, no gate Download size 438.0 MB [Download from Stanford AIMI](https://huggingface.co/StanfordAIMI/stanford-deidentifier-base) Released by Stanford AIMI through its official repository on Hugging Face. [Read the license](https://opensource.org/license/mit). ## Built From - Trained on (disclosed) radreports ## Memory Requirements | Precision | Weights in memory | | --- | --- | | As published | 438.0 MB | Weights only, from the published parameter count; the key-value cache and runtime add to this. ## Questions About stanford-deidentifier-base ### Can I use stanford-deidentifier-base commercially? Yes. stanford-deidentifier-base is released under MIT License. The MIT License is a short permissive license. It permits commercial use, modification and redistribution, provided the copyright notice and permission notice are included. ### What is stanford-deidentifier-base's context length? 512 tokens, from the maximum position embeddings in its published configuration. ## Similar Models Model · Token classification ### [bert-portuguese-ner](https://savrn.com/models/bert-portuguese-ner) [Luís Filipe Cunha](https://savrn.com/model-publishers/lfcc) This model is a fine-tuned version of neuralmind/bert-base-portuguese-cased It achieves the following results on the evaluation set: This model was fine-tunned on token classification task (NER) on Portuguese archival documents. The annotated labels are: Date, Profession, Person, Place, Organization All the training and evaluation data is available at: http://ner.epl.di.uminho.pt/ The following hyperparameters were used during training: - learningrate: 2e-05 - trainbatchsize: 16 - evalbatchsize: 16 - lrschedulertype: linear - numepochs: 4 - Transformers 4.10.0.dev0 - Pytorch 1.9.0+cu111 - Datasets 1.10.2 - Tokenizers 0.10.3 Open weights mit 512 tokens transformers [View model](https://savrn.com/models/bert-portuguese-ner) Model · Token classification ### [privacy-filter-nemotron-GGUF](https://savrn.com/models/privacy-filter-nemotron-gguf) [LocalAI-io](https://savrn.com/model-publishers/localai-io) GGUF conversion of OpenMed/privacy-filter-nemotron, a fine-grained PII token-classification model — a fine-tune of openai/privacy-filter on the nvidia/Nemotron-PII dataset. It labels every token with a BIOES tag over 55 PII categories (221 classes) in a single forward pass, then decodes coherent spans with a constrained Viterbi procedure — so it can be served locally with no Python as the encoder/NER tier of a PII redactor. Where the base openai/privacy-filter covers 8 coarse categories, this fine-tune trades multilingual breadth for category depth: 55 fine-grained English categories (first/last name, government IDs, financial, healthcare, vehicle, digital, …). For the full model… Open weights apache-2.0 gguf [View model](https://savrn.com/models/privacy-filter-nemotron-gguf) Model · Token classification ### [punctuate-all](https://savrn.com/models/punctuate-all) [KREDOR](https://savrn.com/model-publishers/kredor) This is based on Oliver Guhr's work. The difference is that it is a finetuned xlm-roberta-base instead of an xlm-roberta-large and on twelve languages instead of four. The languages are: English, German, French, Spanish, Bulgarian, Italian, Polish, Dutch, Czech, Portugese, Slovak, Slovenian. precision recall f1-score support accuracy 0.98 84425503 macro avg 0.83 0.74 0.77 84425503 weighted avg 0.98 0.98 0.98 84425503 Open weights mit 514 tokens transformers [View model](https://savrn.com/models/punctuate-all) Model · Token classification ### [privacy-filter-multilingual-GGUF](https://savrn.com/models/privacy-filter-multilingual-gguf) [LocalAI-io](https://savrn.com/model-publishers/localai-io) GGUF conversion of OpenMed/privacy-filter-multilingual, a multilingual PII token-classification model (a fine-tune of openai/privacy-filter). It labels every token with a BIOES tag over 54 PII categories (217 classes) across 16 languages, so it can be served locally with no Python as the encoder/NER tier of a PII redactor. For the full model description, label space, evaluation, limitations, and citations, see the source model card — this card only covers the GGUF packaging and how to run it. This GGUF uses a custom architecture, openai-privacy-filter, that is not (yet) part of 1. privacy-filter.cpp (recommended) — a small standalone GGML engine for exactly this model family, on stock… Open weights apache-2.0 gguf [View model](https://savrn.com/models/privacy-filter-multilingual-gguf) Model · Token classification ### [unbiased-toxic-roberta-onnx](https://savrn.com/models/unbiased-toxic-roberta-onnx) [Protect AI](https://savrn.com/model-publishers/protectai) This model is a conversion of unitary/unbiased-toxic-roberta to ONNX format using the Optimum library. Trained models & code to predict toxic comments on 3 Jigsaw challenges: Toxic comment classification, Unintended Bias in Toxic comments, Multilingual toxic comment classification. Built by Laura Hanu at Unitary. The huggingface models currently give different results to the detoxify library (see issue here). All challenges have a toxicity label. The toxicity labels represent the aggregate ratings of up to 10 annotators according the following schema: - Very Toxic (a very hateful, aggressive, or disrespectful comment that is very likely to make you leave a discussion or give up on sharing… Open weights apache-2.0 514 tokens transformers [View model](https://savrn.com/models/unbiased-toxic-roberta-onnx) Model · Token classification ### [ner-english](https://savrn.com/models/ner-english) [Flair](https://savrn.com/model-publishers/flair) This is the standard 4-class NER model for English that ships with Flair. Based on Flair embeddings and LSTM-CRF. So, the entities "George Washington" (labeled as a person) and "Washington" (labeled as a location) are found in the sentence "George Washington went to Washington". The following Flair script was used to train this model: Please cite the following paper when using this model. The Flair issue tracker is available here. Open weights flair [View model](https://savrn.com/models/ner-english) ## Stanford AIMI [All models and datasets](https://savrn.com/model-publishers/stanfordaimi) ## Versions - [661b9c1c717d](https://savrn.com/models/stanford-deidentifier-base/versions/661b9c1c717d) · current 2026-10-24 ## Explore More - [All token classification models](https://savrn.com/models/tasks/token-classification) - [All models under mit](https://savrn.com/models/licenses/mit) - [Model comparisons](https://savrn.com/models/comparisons) - [The model directory](https://savrn.com/models) - [Open model prices by host](https://savrn.com/ai-index/pricing/open-models) ## Source - Repository metadata, read 2026-10-19. - [Hugging Face record](https://huggingface.co/StanfordAIMI/stanford-deidentifier-base) - [How the hub is built](https://savrn.com/model-hub/methodology)