Modelpublic

cyttic/trocr-ara-fonts64

sourceHugging Faceupdated 2mo agoView on Hugging Face

0likes14downloads

Model Card

trocr-ara-fonts64

This model is a fine-tuned version of cyttic/hatformer-synthetic-arabic on an unknown dataset. It achieves the following results on the evaluation set:

—Loss: 0.0265
—Cer: 0.0006

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

—learning_rate: 2e-05
—trainbatchsize: 8
—evalbatchsize: 8
—seed: 42
—gradientaccumulationsteps: 2
—totaltrainbatch_size: 16
—optimizer: Use OptimizerNames.ADAMWTORCHFUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
—lrschedulertype: linear
—lrschedulerwarmup_steps: 0.1
—training_steps: 4000

Training results

Training Loss	Epoch	Step	Validation Loss	Cer
0.2364	0.0161	250	0.0717	0.0034
0.2522	0.0323	500	0.0785	0.0034
0.1697	0.0484	750	0.0696	0.0033
0.1362	0.0645	1000	0.0782	0.0044
0.1461	0.0806	1250	0.0586	0.0021
0.1376	0.0968	1500	0.0537	0.0016
0.1009	0.1129	1750	0.0551	0.0017
0.1136	0.1290	2000	0.0467	0.0015
0.1697	0.1452	2250	0.0537	0.0099
0.0766	0.1613	2500	0.0370	0.0010
0.0890	0.1774	2750	0.0337	0.0007
0.0805	0.1935	3000	0.0312	0.0009
0.0539	0.2097	3250	0.0299	0.0007
0.0677	0.2258	3500	0.0280	0.0006
0.0694	0.2419	3750	0.0270	0.0007
0.0591	0.2581	4000	0.0265	0.0006

Framework versions

—Transformers 5.14.1
—Pytorch 2.11.0+cu128
—Datasets 5.0.1
—Tokenizers 0.22.2