Why did we open-source our inference engine? Read the post

← Catalog

naver-clova-ix/donut-base-finetuned-docvqa

Open comparison →

Primitive: /extract · Extract · Encoder-Decoder

Donut model fine-tuned on DocVQA. It was introduced in the paper OCR-free Document Understanding Transformer by Geewok et al. and first released in this repository.

MultimodalText regions

Overview

Hardware: — drives latency, throughput & cost

Size110M params
Tasks /extract
Licensemit
Latency6.9 s
Throughput87 tok/s
Cost$2.56 /1M tok

Cost is approximate — computed from list GPU prices; your actual price depends on the provider you deploy SIE with.

Extraction

Output kindstext_regions
Inputsimage
Max sequence length

Benchmarks

DocVQA

general kie en

Visual question answering on document images

Corpus: 5,188 Queries: 5,188
Quality
anls 0.6350
Performance L4-SPOT b1 c4
Performance L4 b1 c16
Reference →

Open source inference for agents

Open-source inference for the models behind your agents. Run it yourself, or let us run it for you.

Github 2.3K

Contact us

Tell us about your use case and we'll get back to you shortly.

Apply for an inference grant

Free capacity on our hosted cluster for selected projects. Tell us what you run and we reply by email.