6000

This model is a fine-tuned version of Helsinki-NLP/opus-mt-es-es on the English-ASL glosses dataset and spanish-MSL glosses dataset. It achieves the following results on the evaluation set:

• Loss: 0.2172
• Model Preparation Time: 0.0055
• Bleu Msl: 90.1233
• Bleu Asl: 0
• Ter Msl: 6.3523
• Ter Asl: 100

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: 1e-05
• train_batch_size: 32
• eval_batch_size: 64
• seed: 42
• optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• num_epochs: 30
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Model Preparation Time	Bleu Msl	Bleu Asl	Ter Msl	Ter Asl
No log	1.0	150	2.1707	0.0055	6.2392	23.0477	102.9488	79.8599
No log	2.0	300	1.5545	0.0055	12.5922	39.9535	98.5256	65.6743
No log	3.0	450	1.1397	0.0055	53.7526	56.8191	35.1258	34.5009
2.0124	4.0	600	0.8560	0.0055	61.3665	58.7258	26.5395	33.1874
2.0124	5.0	750	0.6526	0.0055	66.5262	60.0265	23.0703	32.7496
2.0124	6.0	900	0.5405	0.0055	42.9331	69.2419	32.0035	20.7531
0.7028	7.0	1050	0.4769	0.0055	62.3002	73.9626	22.4631	16.9002
0.7028	8.0	1200	0.4427	0.0055	72.1107	84.5297	16.9991	8.0560
0.7028	9.0	1350	0.4153	0.0055	74.8931	83.3473	16.1318	9.3695
0.3613	10.0	1500	0.3961	0.0055	74.8480	85.2817	15.0911	8.4063
0.3613	11.0	1650	0.3794	0.0055	75.4490	84.4129	15.1778	8.4939
0.3613	12.0	1800	0.3563	0.0055	76.8164	86.0871	13.7901	7.6182
0.3613	13.0	1950	0.3277	0.0055	78.1801	85.5103	13.5299	7.9685
0.2432	14.0	2100	0.3070	0.0055	78.6695	86.9448	13.2697	7.5306
0.2432	15.0	2250	0.2972	0.0055	77.6165	86.1077	13.5299	7.7058
0.2432	16.0	2400	0.2917	0.0055	78.0856	86.5549	13.0095	7.5306
0.1592	17.0	2550	0.2855	0.0055	77.3656	86.8329	13.2697	7.2680
0.1592	18.0	2700	0.2804	0.0055	78.3336	54.3313	12.9228	50.6130
0.1592	19.0	2850	0.2791	0.0055	77.4559	87.1498	13.2697	7.2680
0.1188	20.0	3000	0.2749	0.0055	78.0130	54.3314	12.9228	51.0508
0.1188	21.0	3150	0.2717	0.0055	78.2950	53.8351	12.5759	51.3135
0.1188	22.0	3300	0.2710	0.0055	78.2663	54.3637	12.6626	50.8757
0.1188	23.0	3450	0.2686	0.0055	77.8997	54.1712	13.0095	50.8757
0.0992	24.0	3600	0.2669	0.0055	79.2314	54.3758	12.2290	51.0508
0.0992	25.0	3750	0.2656	0.0055	78.2862	54.2965	12.5759	51.1384
0.0992	26.0	3900	0.2643	0.0055	78.9745	54.2624	12.4024	50.8757
0.0889	27.0	4050	0.2651	0.0055	79.3330	54.3778	12.1422	50.9632
0.0889	28.0	4200	0.2643	0.0055	78.4405	54.3716	12.3157	50.9632
0.0889	29.0	4350	0.2639	0.0055	78.5866	54.3716	12.3157	50.9632
0.083	30.0	4500	0.2639	0.0055	78.5866	54.3716	12.3157	50.9632

Framework versions

• Transformers 4.47.0
• Pytorch 2.5.1+cu121
• Datasets 3.2.0
• Tokenizers 0.21.0