Modelpublic

veronikayu/paligemma-imgtojson-0009

sourceHugging Facegemmaupdated 1y agoView on Hugging Face

0likes3downloads

Model Card

paligemma-imgtojson-0009

This model is a fine-tuned version of google/paligemma-3b-pt-224 on an unknown dataset. It achieves the following results on the evaluation set:

—Loss: 0.1914
—Json Validity: 1.0
—Structure Similarity: 0.8133
—Value Accuracy: 0.7112
—Field F1: 0.8733
—Value Similarity: 0.8717
—Edit Distance: 0.1063
—Overall Score: 0.6751

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: 0.0002
—trainbatchsize: 4
—evalbatchsize: 4
—seed: 42
—gradientaccumulationsteps: 8
—totaltrainbatch_size: 32
—optimizer: Use adamwtorchfused with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
—lrschedulertype: cosine
—lrschedulerwarmup_ratio: 0.03
—num_epochs: 3

Training results

Training Loss	Epoch	Step	Validation Loss	Json Validity	Structure Similarity	Value Accuracy	Field F1	Value Similarity	Edit Distance	Overall Score
10.6694	0.2	5	0.6766	1.0	0.0696	0.0667	0.1184	0.0673	0.8226	0.2264
4.6432	0.4	10	0.3651	1.0	0.2775	0.1939	0.3752	0.3672	0.4105	0.3245
3.2508	0.6	15	0.3025	1.0	0.5016	0.4425	0.6046	0.6259	0.2402	0.4840
2.7727	0.8	20	0.2589	1.0	0.6407	0.5839	0.7359	0.7741	0.1723	0.5833
2.5086	1.0	25	0.2428	1.0	0.5942	0.6444	0.6892	0.7824	0.2291	0.5925
2.3094	1.2	30	0.2291	1.0	0.5798	0.5271	0.6715	0.6693	0.2127	0.5320
1.9765	1.4	35	0.2187	1.0	0.7352	0.6418	0.8149	0.8423	0.1415	0.6365
1.9071	1.6	40	0.2071	1.0	0.6705	0.5740	0.7602	0.7852	0.1680	0.5928
1.8623	1.8	45	0.2039	1.0	0.7995	0.7513	0.8601	0.8824	0.1163	0.6830
1.6788	2.0	50	0.1969	1.0	0.6933	0.7178	0.7784	0.8735	0.1655	0.6505
1.6307	2.2	55	0.1932	1.0	0.8340	0.7429	0.8830	0.8787	0.0850	0.6845
1.3213	2.4	60	0.1940	1.0	0.7847	0.8190	0.8540	0.9365	0.1023	0.7034
1.4182	2.6	65	0.1924	1.0	0.8068	0.6911	0.8656	0.8716	0.0947	0.6663
1.3986	2.8	70	0.1909	1.0	0.7593	0.8142	0.8333	0.9237	0.1205	0.6947
1.554	3.0	75	0.1914	1.0	0.8133	0.7112	0.8733	0.8717	0.1063	0.6751

Framework versions

—PEFT 0.15.2
—Transformers 4.47.1
—Pytorch 2.7.1+cu126
—Datasets 3.6.0
—Tokenizers 0.21.1