voidbagel
/

TrOCR_ar_hw_KHATT

vision-encoder-decoder

Model card Files Files and versions

Dataset Split

Train: 9100 (80%)
Validation: 1137 (10%)
Test: 1138 (10%)

Training Summary

Epochs trained: 10
Best Validation CER: 0.1343
Best Validation WER: 0.3472

Test Results

Test set size: 1138 samples
WER: 0.4025 (40.25%)
CER: 0.1558 (15.58%)

Downloads last month: 10

Safetensors

Model size

0.3B params

Tensor type

F32

·

Inference Providers NEW

This model isn't deployed by any Inference Provider. 🙋 Ask for provider support