README.md

metadata

license: mit
library_name: transformers
tags:
  - donut
  - classification
  - irs
  - tax
  - document AI

Donut - model fine-tuned for US IRS tax documents classification

This donut model has been fine-tuned for IRS (US) tax document classification. It can classify up to 28 different types of IRS documents, targeting common set of documents used for tax returns.

1. 1040 U.S. Individual Income Tax Return
2. 1040-NR U.S. Nonresident Alien Income Tax Return
3. 1040-NR SCHEDULE OI Other Information
4. 1040 SCHEDULE 1 Additional Income and Adjustments to Income
5. 1040 SCHEDULE 2 Additional Taxes
6. 1040 SCHEDULE 3 Additional Credits and Payments
7. 1040 SCHEDULE 8812 Credits for Qualifying Children and Other Dependents
8. 1040 SCHEDULE A Itemized Deductions
9. 1040 SCHEDULE B Interest and Ordinary Dividends
10. 1040 SCHEDULE C Profit or Loss From Business
11. 1040 SCHEDULE D Capital Gains and Losses
12. 1040 SCHEDULE E Supplemental Income and Loss
13. 1040 SCHEDULE SE Self-Employment Tax
14. Form 1125-A Cost of Goods Sold
15. Form 8949 Sales and Other Dispositions of Capital Assets
16. Form 8959 Additional Medicare Tax
17. Form 8960 Net Investment Income Tax — Individuals, Estates, and Trusts
18. Form 8995 Qualified Business Income Deduction Simplified Computation
19. Form 8995-A SCHEDULE A Specified Service Trades or Businesses
20. Form W-2 Wage and Tax Statement

Model Details & Description

The base model is 'naver-clova-ix/donut-base-finetuned-rvlcdip', the model is finetuned using training data set of over 3000+ documents. The config.json file has assocociated label2id updated to reflect all labels that can be classified vi the model.