opus-mt-lg-en-informal

This model is a fine-tuned version of Helsinki-NLP/opus-mt-lg-en on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.1343
• Bleu: 0.0
• Bleu Precision: [0.019908987485779295, 0.0006461339651087659, 0.0, 0.0]
• Bleu Brevity Penalty: 1.0
• Bleu Length Ratio: 1.2563
• Bleu Translation Length: 5274
• Bleu Reference Length: 4198

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
• lr_scheduler_warmup_steps: 500
• num_epochs: 20
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Bleu	Bleu Precision	Bleu Brevity Penalty	Bleu Length Ratio	Bleu Translation Length	Bleu Reference Length
4.568	1.0	119	0.8525	0.0	[0.011322534989778267, 0.00017458100558659218, 0.0, 0.0]	1.0	1.5148	6359	4198
0.6495	2.0	238	0.1701	0.0	[0.012054948135688253, 0.0003405994550408719, 0.0, 0.0]	0.8379	0.8497	3567	4198
0.1889	3.0	357	0.1443	0.0	[0.0408483896307934, 0.0010443864229765013, 0.0, 0.0]	0.5226	0.6065	2546	4198
0.1513	4.0	476	0.1384	0.0	[0.03887070376432079, 0.0005515719801434088, 0.0, 0.0]	0.4879	0.5822	2444	4198
0.1424	5.0	595	0.1357	0.0	[0.027095148078134845, 0.0012106537530266344, 0.0, 0.0]	1.0	1.1341	4761	4198
0.1331	6.0	714	0.1346	0.0	[0.016541609822646658, 0.0005732849226065354, 0.0, 0.0]	1.0	1.3969	5864	4198
0.1265	7.0	833	0.1340	0.0	[0.03237891356703238, 0.0016097875080489374, 0.0, 0.0]	0.8839	0.8902	3737	4198
0.1296	8.0	952	0.1339	0.0	[0.026692456479690523, 0.0013218770654329147, 0.0, 0.0]	1.0	1.2315	5170	4198
0.123	9.0	1071	0.1340	0.0	[0.025897226753670472, 0.001404165691551603, 0.0, 0.0]	1.0	1.1682	4904	4198
0.1227	10.0	1190	0.1339	0.0	[0.014839915868193504, 0.0008830579033682352, 0.0, 0.0]	1.0	2.0386	8558	4198
0.117	11.0	1309	0.1343	0.0	[0.019908987485779295, 0.0006461339651087659, 0.0, 0.0]	1.0	1.2563	5274	4198

Framework versions

• Transformers 4.41.1
• Pytorch 2.1.2
• Datasets 2.19.1
• Tokenizers 0.19.1