Modelpublic

thunderboltc/mbart50_sanlish_to_english

sourceHugging Faceupdated 2mo agoView on Hugging Face

0likes9downloads

Model Card

mbart50sanlishto_english

This model is a fine-tuned version of facebook/mbart-large-50-many-to-many-mmt on the None dataset. It achieves the following results on the evaluation set:

—Loss: 4.0709
—Bleu: 9.2087
—Chrf: 21.4673

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: 3e-05
—trainbatchsize: 4
—evalbatchsize: 4
—seed: 42
—optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
—lrschedulertype: linear
—num_epochs: 25
—mixedprecisiontraining: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Bleu	Chrf
2.7341	1.0	101	2.5825	2.3707	15.0326
1.5367	2.0	202	2.5427	3.0316	16.8606
0.7792	3.0	303	2.7498	7.8472	21.4154
0.3581	4.0	404	3.0326	8.0444	21.4348
0.1964	5.0	505	3.3518	9.3819	21.2444
0.1269	6.0	606	3.2937	9.6564	21.7801
0.0756	7.0	707	3.6156	9.3955	21.5854
0.0641	8.0	808	3.5715	9.7613	24.5757
0.045	9.0	909	3.7883	9.9786	23.6823
0.0257	10.0	1010	3.8386	8.8550	21.2765
0.0113	11.0	1111	3.8099	9.3571	21.7522
0.0126	12.0	1212	3.9253	9.6292	23.0794
0.0062	13.0	1313	3.8215	8.7199	20.5837
0.0048	14.0	1414	3.9446	9.2593	21.9610
0.0068	15.0	1515	3.9407	9.0051	20.8048
0.0022	16.0	1616	4.0021	9.9346	22.0919
0.0023	17.0	1717	4.0145	9.8565	21.6968
0.0014	18.0	1818	4.0413	9.0774	21.2321
0.0017	19.0	1919	4.0513	9.0965	21.3074
0.0011	20.0	2020	4.0509	9.2023	21.4822
0.0006	21.0	2121	4.0572	9.2087	21.4673
0.0006	22.0	2222	4.0624	9.2087	21.4673
0.0005	23.0	2323	4.0694	9.2150	21.5036
0.0007	24.0	2424	4.0689	9.2087	21.4673
0.0005	25.0	2525	4.0709	9.2087	21.4673

Framework versions

—Transformers 4.44.2
—Pytorch 2.11.0+cu128
—Datasets 4.0.0
—Tokenizers 0.19.1