Image-Text-to-Text
Transformers
Safetensors
German
mistral3
text-generation
ocr
htr
handwritten-text-recognition
page-level-htr
kurrent
suetterlin
german
handwriting
historical-documents
lighton-ocr
conversational
Instructions to use 13fragments/kurrent-transcriber-1b with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use 13fragments/kurrent-transcriber-1b with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("image-text-to-text", model="13fragments/kurrent-transcriber-1b") messages = [ { "role": "user", "content": [ {"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/p-blog/candy.JPG"}, {"type": "text", "text": "What animal is on the candy?"} ] }, ] pipe(text=messages)# pip install -U transformers accelerate # Load model directly from transformers import AutoProcessor, AutoModelForSeq2SeqLM processor = AutoProcessor.from_pretrained("13fragments/kurrent-transcriber-1b") model = AutoModelForSeq2SeqLM.from_pretrained("13fragments/kurrent-transcriber-1b", device_map="auto") messages = [ { "role": "user", "content": [ {"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/p-blog/candy.JPG"}, {"type": "text", "text": "What animal is on the candy?"} ] }, ] inputs = processor.apply_chat_template( messages, add_generation_prompt=True, tokenize=True, return_dict=True, return_tensors="pt", ).to(model.device) outputs = model.generate(**inputs, max_new_tokens=256) print(processor.decode(outputs[0][inputs["input_ids"].shape[-1]:])) - Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- vLLM
How to use 13fragments/kurrent-transcriber-1b with vLLM:
Install from pip and serve model
# Install vLLM from pip: pip install vllm # Start the vLLM server: vllm serve "13fragments/kurrent-transcriber-1b" # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:8000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "13fragments/kurrent-transcriber-1b", "messages": [ { "role": "user", "content": [ { "type": "text", "text": "Describe this image in one sentence." }, { "type": "image_url", "image_url": { "url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg" } } ] } ] }'Use Docker
docker model run hf.co/13fragments/kurrent-transcriber-1b
- SGLang
How to use 13fragments/kurrent-transcriber-1b with SGLang:
Install from pip and serve model
# Install SGLang from pip: pip install sglang # Start the SGLang server: python3 -m sglang.launch_server \ --model-path "13fragments/kurrent-transcriber-1b" \ --host 0.0.0.0 \ --port 30000 # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:30000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "13fragments/kurrent-transcriber-1b", "messages": [ { "role": "user", "content": [ { "type": "text", "text": "Describe this image in one sentence." }, { "type": "image_url", "image_url": { "url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg" } } ] } ] }'Use Docker images
docker run --gpus all \ --shm-size 32g \ -p 30000:30000 \ -v ~/.cache/huggingface:/root/.cache/huggingface \ --env "HF_TOKEN=<secret>" \ --ipc=host \ lmsysorg/sglang:latest \ python3 -m sglang.launch_server \ --model-path "13fragments/kurrent-transcriber-1b" \ --host 0.0.0.0 \ --port 30000 # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:30000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "13fragments/kurrent-transcriber-1b", "messages": [ { "role": "user", "content": [ { "type": "text", "text": "Describe this image in one sentence." }, { "type": "image_url", "image_url": { "url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg" } } ] } ] }' - Docker Model Runner
How to use 13fragments/kurrent-transcriber-1b with Docker Model Runner:
docker model run hf.co/13fragments/kurrent-transcriber-1b
Download training_data.json from 13fragments/kurrent-transcriber-1b: direct link, hf CLI and curl.
- Browser
- Download file 10.1 kB
-
https://huggingface.co/13fragments/kurrent-transcriber-1b/resolve/main/training_data.json
- Command line
-
hf download hf://13fragments/kurrent-transcriber-1b/training_data.json
-
curl -L -o training_data.json https://huggingface.co/13fragments/kurrent-transcriber-1b/resolve/main/training_data.json
10.1 kB
| { | |
| "model": "Kurrent Transcriber 1B, version 1", | |
| "note": "A weight merge of two fine-tunes of LightOnOCR-3-1B. Parent A: a fine-tune by 13fragments; rows are training examples (pages, regions, or lines). Parent B: LightOnOCR-HGH by Bernhard Straub, revision 47931fbe; counts are training pages as its model card states.", | |
| "parent_a": { | |
| "name": "13fragments fine-tune (not published)", | |
| "total_training_rows": 36818, | |
| "note": "Rows are training examples (pages, regions, or lines). The two Dresdner Hofdiarium sets give 716 rows together.", | |
| "sources": [ | |
| { | |
| "source": "icdar_2017_read_train", | |
| "name": "ICDAR 2017 READ, Train-A and Train-B, German pages", | |
| "training_rows": 7875, | |
| "licence": "CC-BY 4.0", | |
| "url": "https://zenodo.org/records/835489", | |
| "licence_checked": "2026-10-21", | |
| "basis": "Zenodo record licence field cc-by-4.0; covers images and PAGE transcriptions" | |
| }, | |
| { | |
| "source": "kurrent_19c_zenodo_ccby_subset", | |
| "name": "Senatsprotokolle subset of 'German Kurrent of the 19th century'", | |
| "training_rows": 7913, | |
| "licence": "CC-BY 4.0", | |
| "url": "https://zenodo.org/records/17252677", | |
| "licence_checked": "2026-07-11", | |
| "basis": "Zenodo record licence CC BY 4.0; written permission of the Universitätsbibliothek Tübingen (2026-10-29), credit UB Tübingen and Universitätsarchiv Tübingen" | |
| }, | |
| { | |
| "source": "ground_truth_set_for_handwritten_text_recognition_htr_ocr_dr_2", | |
| "name": "Dresdner Hofdiarium 1673 (Mscr.Dresd.K.117)", | |
| "training_rows": null, | |
| "licence": "CC-BY 4.0", | |
| "url": "https://doi.org/10.5281/zenodo.15303243", | |
| "licence_checked": "2026-07-11", | |
| "basis": "Zenodo record licence CC BY 4.0" | |
| }, | |
| { | |
| "source": "ground_truth_set_for_handwritten_text_recognition_htr_ocr_dr_3", | |
| "name": "Dresdner Hofdiarium 1653-56 (Mscr.Dresd.K.113)", | |
| "training_rows": null, | |
| "licence": "CC-BY 4.0", | |
| "url": "https://doi.org/10.5281/zenodo.15303398", | |
| "licence_checked": "2026-07-11", | |
| "basis": "Zenodo record licence CC BY 4.0" | |
| }, | |
| { | |
| "source": "scads_german_handwriting", | |
| "name": "ScaDS.AI German Line- and Word-Level Handwriting Dataset", | |
| "training_rows": 4000, | |
| "licence": "CC-BY 4.0", | |
| "url": "https://zenodo.org/records/18301532", | |
| "licence_checked": "2026-07-11", | |
| "basis": "Zenodo record licence CC BY 4.0" | |
| }, | |
| { | |
| "source": "reichsanzeiger_gt", | |
| "name": "Deutscher Reichsanzeiger ground truth (UB Mannheim)", | |
| "training_rows": 4000, | |
| "licence": "CC0 1.0 ground truth", | |
| "url": "https://github.com/UB-Mannheim/reichsanzeiger-gt", | |
| "licence_checked": "2026-07-11", | |
| "basis": "repository states CC0 1.0 for the ground truth and Public Domain Mark 1.0 for the scans" | |
| }, | |
| { | |
| "source": "sbb_dta_kosmos", | |
| "name": "Kosmos lecture notes, Deutsches Textarchiv (five works)", | |
| "training_rows": 1812, | |
| "licence": "CC-BY 4.0", | |
| "url": "https://www.deutschestextarchiv.de/book/show/nn_msgermqu2124_1827", | |
| "licence_checked": "2026-11-01", | |
| "basis": "TEI header licence CC BY 4.0 in each of the five files; scans Public Domain Mark in the Staatsbibliothek zu Berlin IIIF manifests" | |
| }, | |
| { | |
| "source": "sbb_fanny_hensel", | |
| "name": "Fanny loves Wilhelm (letters of Fanny Mendelssohn to Wilhelm Hensel)", | |
| "training_rows": 102, | |
| "licence": "CC-BY 4.0", | |
| "url": "https://zenodo.org/records/15223355", | |
| "licence_checked": "2026-11-01", | |
| "basis": "Zenodo record licence CC BY 4.0; scans Public Domain Mark in the Staatsbibliothek zu Berlin IIIF manifests" | |
| }, | |
| { | |
| "source": "oeaw_akademieprotokolle", | |
| "name": "Akademieprotokolle, ÖAW / ACDH-CH", | |
| "training_rows": 1253, | |
| "licence": "CC-BY 4.0", | |
| "url": "https://github.com/acdh-oeaw/akademie-data/blob/c927cb6506c46cbbc3dff2388fb2ec1bfe23c2b6/data/meta/arche_constants.rdf", | |
| "licence_checked": "2026-11-08", | |
| "basis": "repository metadata states CC BY 4.0 for data and images; written permission of the ACDH-CH (2026-10-29); credit the ÖAW" | |
| }, | |
| { | |
| "source": "greifswald_konzilsprotokolle", | |
| "name": "read_dataset_german_konzilsprotokolle (University of Greifswald)", | |
| "training_rows": 173, | |
| "licence": "CC-BY 4.0", | |
| "url": "https://zenodo.org/records/215383", | |
| "licence_checked": "2026-11-05", | |
| "basis": "Zenodo record licence CC BY 4.0; images and PAGE XML in the same archive" | |
| }, | |
| { | |
| "source": "synthetic_kurrent_stage_a", | |
| "name": "Synthetic Kurrent pages", | |
| "training_rows": 4989, | |
| "licence": "generated", | |
| "url": "https://www.peter-wiegel.de/WiegelKurrent.html", | |
| "licence_checked": "2026-10-11", | |
| "basis": "generated from CC0 Reichsanzeiger text with Wiegel Kurrent fonts (SIL OFL 1.1); fonts not distributed" | |
| }, | |
| { | |
| "source": "synthetic_fraktur_run0", | |
| "name": "Synthetic Fraktur pages", | |
| "training_rows": 2985, | |
| "licence": "generated", | |
| "url": "https://github.com/UB-Mannheim/reichsanzeiger-gt", | |
| "licence_checked": "2026-07-11", | |
| "basis": "generated from CC0 Reichsanzeiger text with SIL OFL fonts; fonts not distributed" | |
| }, | |
| { | |
| "source": "synthetic_abstention_run0", | |
| "name": "Synthetic blank and illegible pages", | |
| "training_rows": 1000, | |
| "licence": "generated", | |
| "url": "https://github.com/UB-Mannheim/reichsanzeiger-gt", | |
| "licence_checked": "2026-07-11", | |
| "basis": "generated from CC0 Reichsanzeiger text with SIL OFL fonts; fonts not distributed" | |
| } | |
| ] | |
| }, | |
| "parent_b": { | |
| "name": "LightOnOCR-3-1B · Historic German Handwriting (b-straub), revision 47931fbe", | |
| "licence": "CC BY-SA 4.0", | |
| "url": "https://huggingface.co/b-straub/LightOnOCR-3-1B-Historic-German-Handwriting", | |
| "total_training_pages": 21038, | |
| "sources": [ | |
| { | |
| "source": "escher", | |
| "name": "Digitale Briefedition Alfred Escher (Staatsarchiv des Kantons Zürich)", | |
| "training_pages": 13156, | |
| "licence": "CC BY-SA 4.0 (text and images; the card notes a planned move to CC BY 4.0)", | |
| "url": "https://zenodo.org/records/10927129", | |
| "licence_checked": "2026-11-17", | |
| "basis": "as stated in the LightOnOCR-HGH model card, revision 47931fbe (sha256 5e59c055b94fd111109e8abdef5bcc68bc5bd8269bef0800c26f0a6707b88478)" | |
| }, | |
| { | |
| "source": "kurrent_xix", | |
| "name": "Kurrent XIX, Regierungsratsprotokolle (Staatsarchiv Zürich / DH Uni Bern)", | |
| "training_pages": 1796, | |
| "licence": "MIT", | |
| "url": "https://huggingface.co/datasets/dh-unibe/image-text_kurrent-xix", | |
| "licence_checked": "2026-11-17", | |
| "basis": "as stated in the LightOnOCR-HGH model card, revision 47931fbe (sha256 5e59c055b94fd111109e8abdef5bcc68bc5bd8269bef0800c26f0a6707b88478)" | |
| }, | |
| { | |
| "source": "grundbuch_basel", | |
| "name": "Historisches Grundbuch Basel (DH Uni Bern)", | |
| "training_pages": 1907, | |
| "licence": "MIT", | |
| "url": "https://huggingface.co/datasets/dh-unibe/image-text_historisches-grundbuch-basel_xix-xx", | |
| "licence_checked": "2026-11-17", | |
| "basis": "as stated in the LightOnOCR-HGH model card, revision 47931fbe (sha256 5e59c055b94fd111109e8abdef5bcc68bc5bd8269bef0800c26f0a6707b88478)" | |
| }, | |
| { | |
| "source": "dta_hidden_kosmos", | |
| "name": "DTA, Hidden Kosmos (Deutsches Textarchiv, BBAW; images SBB)", | |
| "training_pages": 1640, | |
| "licence": "CC BY 4.0; images Public Domain Mark", | |
| "url": "https://www.deutschestextarchiv.de", | |
| "licence_checked": "2026-11-17", | |
| "basis": "as stated in the LightOnOCR-HGH model card, revision 47931fbe (sha256 5e59c055b94fd111109e8abdef5bcc68bc5bd8269bef0800c26f0a6707b88478)" | |
| }, | |
| { | |
| "source": "edition_humboldt", | |
| "name": "edition humboldt digital v12.0.0 (BBAW; images SBB)", | |
| "training_pages": 925, | |
| "licence": "CC BY-SA 4.0 (the card notes a planned move to CC BY 4.0); images Public Domain Mark", | |
| "url": "https://github.com/telota/edition-humboldt-digital", | |
| "licence_checked": "2026-11-17", | |
| "basis": "as stated in the LightOnOCR-HGH model card, revision 47931fbe (sha256 5e59c055b94fd111109e8abdef5bcc68bc5bd8269bef0800c26f0a6707b88478)" | |
| }, | |
| { | |
| "source": "naegeli_briefe", | |
| "name": "Nägeli-Briefe (Zentralbibliothek Zürich; images e-manuscripta)", | |
| "training_pages": 680, | |
| "licence": "CC BY-SA 4.0; images Public Domain Mark", | |
| "url": "https://naegeli-briefe.ch", | |
| "licence_checked": "2026-11-17", | |
| "basis": "as stated in the LightOnOCR-HGH model card, revision 47931fbe (sha256 5e59c055b94fd111109e8abdef5bcc68bc5bd8269bef0800c26f0a6707b88478)" | |
| }, | |
| { | |
| "source": "frauen_im_fokus", | |
| "name": "Frauen im Fokus (SBB / Universität Potsdam)", | |
| "training_pages": 180, | |
| "licence": "CC0 1.0; images Public Domain Mark", | |
| "url": "https://zenodo.org/records/11038498", | |
| "licence_checked": "2026-11-17", | |
| "basis": "as stated in the LightOnOCR-HGH model card, revision 47931fbe (sha256 5e59c055b94fd111109e8abdef5bcc68bc5bd8269bef0800c26f0a6707b88478)" | |
| }, | |
| { | |
| "source": "fanny_hensel", | |
| "name": "Fanny Hensel (SBB / UdK Berlin)", | |
| "training_pages": 158, | |
| "licence": "CC BY 4.0; images Public Domain Mark", | |
| "url": "https://zenodo.org/records/15223355", | |
| "licence_checked": "2026-11-17", | |
| "basis": "as stated in the LightOnOCR-HGH model card, revision 47931fbe (sha256 5e59c055b94fd111109e8abdef5bcc68bc5bd8269bef0800c26f0a6707b88478)" | |
| }, | |
| { | |
| "source": "german_20th_century", | |
| "name": "German 20th century, 3 subsets (DH Uni Bern)", | |
| "training_pages": 584, | |
| "licence": "MIT", | |
| "url": "https://huggingface.co/datasets/dh-unibe/image-text_german-20th-century", | |
| "licence_checked": "2026-11-17", | |
| "basis": "as stated in the LightOnOCR-HGH model card, revision 47931fbe (sha256 5e59c055b94fd111109e8abdef5bcc68bc5bd8269bef0800c26f0a6707b88478)" | |
| }, | |
| { | |
| "source": "gantner_dta", | |
| "name": "Gantner, DTA part (M. Gantner)", | |
| "training_pages": 12, | |
| "licence": "CC BY 4.0", | |
| "url": "https://zenodo.org/records/17252677", | |
| "licence_checked": "2026-11-17", | |
| "basis": "as stated in the LightOnOCR-HGH model card, revision 47931fbe (sha256 5e59c055b94fd111109e8abdef5bcc68bc5bd8269bef0800c26f0a6707b88478)" | |
| } | |
| ] | |
| } | |
| } | |