enclavelabs

enclave-scribe-devanagari-iter4

What it does

Given an image containing Devanagari text (Hindi, Marathi, Sanskrit, Nepali, Pali), returns the Unicode transcription. Unlike iter-3 which was trained on word crops only, iter-4 also handles full document pages without falling into generation loops.

For full-page PDFs, use it via the EnclaveScribe agent pipeline which handles page rasterization and generation-config.

Results — page-level ship-gate

Tested against a 6-page Gazette of India Extraordinary notification (dense Hindi + English) with the runtime bad_words_ids workaround disabled:

Table with columns: Signal, iter-3, iter-4
Signal	iter-3	iter-4
Total chars extracted	7,122	13,646
Pages dead-looped	2/6	0/6
Max page chars (of 4,096 ceiling)	2,430	4,077
Wallclock (6 pages, g5.xlarge)	20 min	12 min

Iter-3 pages 4 and 5 exhausted the full generation budget emitting <tool_call> blocks and produced empty output after the post-hoc regex strip. Iter-4 completes every page inside the budget with real document text.

Iter-4 is also more faithful to source script conventions — for example, it preserves Arabic numerals inside English sections (where iter-3 hallucinates Devanagari digits like २११३ instead of 2113).

See the full head-to-head in reports/iter4/GAZETTE_TEST.md.

How to use

python
import torch
from transformers import AutoProcessor, AutoModelForImageTextToText
from peft import PeftModel
from PIL import Image

BASE = "allenai/olmOCR-2-7B-1025"
ADAPTER = "enclavelabs/enclave-scribe-devanagari-iter4"

processor = AutoProcessor.from_pretrained(BASE)
model = AutoModelForImageTextToText.from_pretrained(
    BASE, dtype=torch.bfloat16, device_map="auto"
)
model = PeftModel.from_pretrained(model, ADAPTER)
model.eval()

image = Image.open("hindi_page.png").convert("RGB")
messages = [{
    "role": "user",
    "content": [
        {"type": "image", "image": image},
        {"type": "text",  "text": "Extract this document page as clean markdown. "
                                   "Preserve the original script exactly."},
    ],
}]
text = processor.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
inputs = processor(text=[text], images=[image], return_tensors="pt").to(model.device)

with torch.no_grad():
    out = model.generate(
        **inputs,
        max_new_tokens=4096,
        do_sample=False,
        repetition_penalty=1.1,   # belt-and-suspenders; iter-4 no longer strictly needs it
    )

print(processor.batch_decode(
    out[:, inputs.input_ids.shape[1]:], skip_special_tokens=True
)[0].strip())

Note: iter-4 no longer dead-loops on the pages that broke iter-3, but we recommend keeping repetition_penalty=1.1 in production. It costs nothing and stays as insurance.

Training details

Base model: allenai/olmOCR-2-7B-1025
Starting point: iter-3 LoRA adapter (enclavelabs/enclave-scribe-devanagari), loaded via resume_adapter so training continues rather than starting fresh
Method: LoRA (r=32, α=64, inherited from iter-3)
Trainable parameters: ~95M of 8.4B (~1.1%)
Training data:
- Pages — 793 pseudo-labeled Devanagari page images from ai4bharat/indicdlp (Hindi + Marathi), labeled by iter-3 through the EnclaveScribe agent pipeline; quality-filtered from 2,000 attempts (41.75% pass rate)
- Word replay — 500 word-level samples from himalaya-ai/devanagari_ocr_dataset mixed in at a 5:1 pages:words ratio to prevent catastrophic forgetting on iter-3's word-level competency

Limitations

Pseudo-label ceiling: training labels came from iter-3, which has 17.4% CER on word crops. Iter-4 inherited character-level errors from those labels. Real human labels would break this ceiling in a future iteration.
Training loss stayed flat at ~4.2 across all 82 steps. Iter-4's improvements are class-of-failure fixes (no more dead-looping) rather than a broad character-level uplift. Iter-4 is "iter-3 that doesn't break on long pages", not "iter-3 but sharper".
English regression: not measured against a held-out English benchmark. Qualitative testing on a bilingual Hindi/English gazette PDF suggests English is preserved and Arabic numerals are handled correctly, but this is not a formal claim.
Not evaluated on: Tamil, Telugu, Kannada, Bengali, Gujarati, Punjabi, other Indic scripts. This adapter is Devanagari-family only.
Ship-gate scope: verified on a 6-page bilingual gazette PDF. Longer documents, tables, and structure preservation not exhaustively benchmarked.

Iter-5 plans

Real human-labeled pages (300–500 samples) to break the pseudo-label ceiling, OR
Bigger base model (e.g. allenai/olmOCR-2-32B-1025) with the same corpus for more LoRA capacity
English regression benchmark for both iter-3 and iter-4
Regression gate: every future iteration runs against tests/fixtures/pdfs/gazette_moef_2024_06_07.pdf before shipping

Follow github.com/Enclave-Labs-Inc/enclave-scribe for updates.

Citation

markdown
@software{enclavescribe_devanagari_iter4_2026,
  title  = {EnclaveScribe: Self-hostable Indic OCR — Devanagari adapter (iter-4)},
  author = {Enclave Labs},
  year   = {2026},
  url    = {https://huggingface.co/enclavelabs/enclave-scribe-devanagari-iter4},
}

License

MIT — same as the base model and the EnclaveScribe repo.

Available on FriendliAI

Dedicated Endpoints

Run this model inference on single tenant GPU with unmatched speed and reliability at scale.

Learn more

Container

Run this model inference with full control and performance in your environment.

Learn more

Model Details

Model Provider

enclavelabs

Model Tree

Base

allenai/olmOCR-2-7B-1025

Adapter

this model

Input Modalities

TextImage

Output Modalities

Text

Supported Functionality

Dedicated EndpointsContainer

Explore FriendliAI today

Get started Talk to an engineer

What it does

For full-page PDFs, use it via the EnclaveScribe agent pipeline which handles page rasterization and generation-config.

Results — page-level ship-gate

Tested against a 6-page Gazette of India Extraordinary notification (dense Hindi + English) with the runtime bad_words_ids workaround disabled:

Table with columns: Signal, iter-3, iter-4
Signal	iter-3	iter-4
Total chars extracted	7,122	13,646
Pages dead-looped	2/6	0/6
Max page chars (of 4,096 ceiling)	2,430	4,077
Wallclock (6 pages, g5.xlarge)	20 min	12 min

See the full head-to-head in reports/iter4/GAZETTE_TEST.md.

How to use

python
import torch
from transformers import AutoProcessor, AutoModelForImageTextToText
from peft import PeftModel
from PIL import Image

BASE = "allenai/olmOCR-2-7B-1025"
ADAPTER = "enclavelabs/enclave-scribe-devanagari-iter4"

processor = AutoProcessor.from_pretrained(BASE)
model = AutoModelForImageTextToText.from_pretrained(
    BASE, dtype=torch.bfloat16, device_map="auto"
)
model = PeftModel.from_pretrained(model, ADAPTER)
model.eval()

image = Image.open("hindi_page.png").convert("RGB")
messages = [{
    "role": "user",
    "content": [
        {"type": "image", "image": image},
        {"type": "text",  "text": "Extract this document page as clean markdown. "
                                   "Preserve the original script exactly."},
    ],
}]
text = processor.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
inputs = processor(text=[text], images=[image], return_tensors="pt").to(model.device)

with torch.no_grad():
    out = model.generate(
        **inputs,
        max_new_tokens=4096,
        do_sample=False,
        repetition_penalty=1.1,   # belt-and-suspenders; iter-4 no longer strictly needs it
    )

print(processor.batch_decode(
    out[:, inputs.input_ids.shape[1]:], skip_special_tokens=True
)[0].strip())

Note: iter-4 no longer dead-loops on the pages that broke iter-3, but we recommend keeping repetition_penalty=1.1 in production. It costs nothing and stays as insurance.

Training details

Base model: allenai/olmOCR-2-7B-1025
Starting point: iter-3 LoRA adapter (enclavelabs/enclave-scribe-devanagari), loaded via resume_adapter so training continues rather than starting fresh
Method: LoRA (r=32, α=64, inherited from iter-3)
Trainable parameters: ~95M of 8.4B (~1.1%)
Training data:
- Pages — 793 pseudo-labeled Devanagari page images from ai4bharat/indicdlp (Hindi + Marathi), labeled by iter-3 through the EnclaveScribe agent pipeline; quality-filtered from 2,000 attempts (41.75% pass rate)
- Word replay — 500 word-level samples from himalaya-ai/devanagari_ocr_dataset mixed in at a 5:1 pages:words ratio to prevent catastrophic forgetting on iter-3's word-level competency

Limitations

Pseudo-label ceiling: training labels came from iter-3, which has 17.4% CER on word crops. Iter-4 inherited character-level errors from those labels. Real human labels would break this ceiling in a future iteration.
Training loss stayed flat at ~4.2 across all 82 steps. Iter-4's improvements are class-of-failure fixes (no more dead-looping) rather than a broad character-level uplift. Iter-4 is "iter-3 that doesn't break on long pages", not "iter-3 but sharper".
English regression: not measured against a held-out English benchmark. Qualitative testing on a bilingual Hindi/English gazette PDF suggests English is preserved and Arabic numerals are handled correctly, but this is not a formal claim.
Not evaluated on: Tamil, Telugu, Kannada, Bengali, Gujarati, Punjabi, other Indic scripts. This adapter is Devanagari-family only.
Ship-gate scope: verified on a 6-page bilingual gazette PDF. Longer documents, tables, and structure preservation not exhaustively benchmarked.

Iter-5 plans

Real human-labeled pages (300–500 samples) to break the pseudo-label ceiling, OR
Bigger base model (e.g. allenai/olmOCR-2-32B-1025) with the same corpus for more LoRA capacity
English regression benchmark for both iter-3 and iter-4
Regression gate: every future iteration runs against tests/fixtures/pdfs/gazette_moef_2024_06_07.pdf before shipping

Follow github.com/Enclave-Labs-Inc/enclave-scribe for updates.

Citation

markdown
@software{enclavescribe_devanagari_iter4_2026,
  title  = {EnclaveScribe: Self-hostable Indic OCR — Devanagari adapter (iter-4)},
  author = {Enclave Labs},
  year   = {2026},
  url    = {https://huggingface.co/enclavelabs/enclave-scribe-devanagari-iter4},
}

License

MIT — same as the base model and the EnclaveScribe repo.

enclave-scribe-devanagari-iter4

README

What it does

Results — page-level ship-gate

How to use

Training details

Limitations

Iter-5 plans

Citation

License

Explore FriendliAI today

README

What it does

Results — page-level ship-gate

How to use

Training details

Limitations

Iter-5 plans

Citation

License