License for OLM‑OCR: Apache 2.0 Terms and Usage Guide

The OLM‑OCR repository is released under the Apache License 2.0, a permissive open-source license that permits commercial use, modification, and distribution provided you include the original copyright notice and license text.

The allenai/olmocr repository provides a modular OCR pipeline framework designed for large-scale scientific document processing. Understanding the license for OLM‑OCR is essential for developers who want to integrate this tool into commercial applications or distribute modified versions. The Apache 2.0 license found in the repository's top-level LICENSE file offers significant flexibility while maintaining clear attribution requirements.

Where to Find the License for OLM‑OCR

The complete license text resides in the LICENSE file at the root of the allenai/olmocr repository. This file contains the full Apache License, Version 2.0 (2024) legal wording, which governs all source code in the project including the pipeline orchestration, training modules, and evaluation benchmarks.

Apache License 2.0 Permissions and Requirements

The Apache License 2.0 is a permissive license that provides broad rights while imposing minimal requirements on users and contributors.

Commercial Use and Modification Rights

Under the terms of the license for OLM‑OCR, you are free to:

• Use the software for any purpose, including commercial applications.
• Modify the source code and create derivative works based on the OCR pipeline components found in olmocr/pipeline.py or any other module.
• Redistribute the original or modified versions of the codebase.

Attribution Requirements

When distributing OLM‑OCR or derivative works, you must:

• Retain the original copyright notice found in the LICENSE file.
• Include a copy of the Apache License 2.0 text.
• State any significant changes made to the files.

Licensed Components in the OLM‑OCR Repository

The Apache 2.0 license governs all modules in the repository. The following components are covered under these terms:

Component	Purpose	Key Files
Pipeline orchestration	End-to-end flow: loading PDFs → preprocessing → OCR → post-processing	olmocr/pipeline.py
Data utilities	Training data preparation and PDF rendering	olmocr/data/prepare_olmocrmix.py, olmocr/data/renderpdf.py
Training modules	Model fine-tuning and GRPO training scripts	olmocr/train/train.py, olmocr/train/grpo_train.py
Benchmarks	OCR accuracy evaluation and table parsing	olmocr/bench/table_parsing.py, olmocr/bench/report.py
Filtering	Post-processing noise reduction and coherency checks	olmocr/filter/filter.py, olmocr/filter/coherency.py
S3 utilities	Cloud storage operations for large datasets	olmocr/s3_utils.py
Version metadata	Package versioning information	olmocr/version.py

Practical Examples: Using OLM‑OCR Under Apache 2.0

The following examples demonstrate typical usage patterns that comply with the license for OLM‑OCR. Remember to include the required attribution when using these in your own projects.

Checking the Version

from olmocr.version import __version__

print(f"OLM‑OCR version: {__version__}")

Running the OCR Pipeline

from olmocr.pipeline import run_pipeline

# Process a local PDF through the complete pipeline

result = run_pipeline(pdf_path="sample.pdf", output_dir="out")

print("Extracted text:", result["text"][:200])

Building Training Batches

from olmocr.data.build_openai_batch_from_olmocrmix import build_batch

batch = build_batch(
    pdf_dir="data/pdfs",
    model="gpt-4o-mini",
    max_tokens=2048,
)

print(f"Created batch with {len(batch)} entries")

Evaluating OCR Accuracy

from olmocr.bench.report import generate_report

report = generate_report(
    predictions_dir="preds",
    ground_truth_dir="gt",
    metrics=["rouge", "wer", "bleu"],
)

print(report.summary())

Summary

• The license for OLM‑OCR is the Apache License 2.0, located in the repository's root LICENSE file.
• This permissive license allows commercial use, modification, and redistribution without imposing copyleft requirements.
• You must retain copyright notices and include the license text when distributing the code or derivative works.
• All components—including olmocr/pipeline.py, training scripts, and benchmark utilities—are covered under the same Apache 2.0 terms.

Frequently Asked Questions

Can I use OLM‑OCR in a commercial product?

Yes. The Apache License 2.0 explicitly permits commercial use. You can integrate the OCR pipeline into proprietary applications, including those referencing olmocr/pipeline.py, provided you include the required attribution and license notice.

Do I need to open-source my modifications to OLM‑OCR?

No. Unlike copyleft licenses such as GPL, Apache 2.0 does not require you to distribute your modifications. You can keep changes to files like olmocr/train/train.py or olmocr/filter/filter.py private, though you must still include the original license and copyright notices.

Where is the license file located in the repository?

The full license text is in the LICENSE file at the root of the allenai/olmocr repository. This file applies to all source code, including modules in olmocr/data/, olmocr/bench/, and olmocr/train/.

What attribution is required when using OLM‑OCR code?

You must preserve the copyright notice found in the LICENSE file and include a copy of the Apache License 2.0. If you modify files such as olmocr/pipeline.py or olmocr/version.py, you should document the changes you made while maintaining the original attribution.

{
  "mcpServers": {
    "instagit": {
      "command": "npx",
      "args": ["-y", "instagit@latest"]
    }
  }
}