opus-mt-en-bkm

This model is a fine-tuned version of Helsinki-NLP/opus-mt-en-ro on the arrow dataset. It achieves the following results on the evaluation set:

• Loss: 1.1597
• Bleu: 14.5684
• Gen Len: 58.4294

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
• train_batch_size: 16
• eval_batch_size: 16
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 25

Training results

Training Loss	Epoch	Step	Validation Loss	Bleu	Gen Len
3.3983	1.0	974	1.9251	3.7894	60.1579
1.9429	2.0	1948	1.6720	5.7083	58.6443
1.7118	3.0	2922	1.5389	7.1977	58.8536
1.5647	4.0	3896	1.4484	8.4631	57.9068
1.4611	5.0	4870	1.3836	9.5314	59.3106
1.3735	6.0	5844	1.3357	10.1879	59.5501
1.3078	7.0	6818	1.3014	10.9172	59.4968
1.245	8.0	7792	1.2737	11.445	59.585
1.2048	9.0	8766	1.2485	11.9346	58.3275
1.1648	10.0	9740	1.2298	12.3049	58.7768
1.1272	11.0	10714	1.2176	12.7287	58.1549
1.086	12.0	11688	1.2043	13.0962	59.2217
1.0595	13.0	12662	1.1973	13.3375	58.6736
1.0343	14.0	13636	1.1844	13.3963	58.2763
1.0174	15.0	14610	1.1797	13.7067	58.1738
0.9923	16.0	15584	1.1757	13.9467	59.3246
0.9703	17.0	16558	1.1704	14.1023	58.9813
0.9589	18.0	17532	1.1663	14.2842	58.401
0.9472	19.0	18506	1.1662	14.2109	58.4796
0.9262	20.0	19480	1.1635	14.3872	58.1601
0.9147	21.0	20454	1.1606	14.4983	58.7417
0.9162	22.0	21428	1.1630	14.5229	58.4345
0.9012	23.0	22402	1.1607	14.6204	58.0767
0.899	24.0	23376	1.1600	14.5681	58.4357
0.8934	25.0	24350	1.1597	14.5684	58.4294

Framework versions

• Transformers 4.39.3
• Pytorch 2.2.1+cu121
• Datasets 2.18.0
• Tokenizers 0.15.2