marianmt_ipa_to_bangla_bleufixed

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

• Loss: 1.9446
• Bleu: 7.8032
• Chrf: 32.7398

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: 8
• eval_batch_size: 8
• seed: 42
• optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• num_epochs: 25
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Bleu	Chrf
2.7945	1.0	196	2.3575	0.3765	11.6679
2.1609	2.0	392	2.1661	1.0851	15.3662
1.9014	3.0	588	2.0647	1.5562	19.3520
1.6928	4.0	784	1.9622	1.7184	21.2290
1.5327	5.0	980	1.9350	1.9268	22.6112
1.3743	6.0	1176	1.8898	2.9497	25.3277
1.2556	7.0	1372	1.8947	5.2229	26.6343
1.1407	8.0	1568	1.8984	3.9033	26.7643
1.0517	9.0	1764	1.8745	3.4013	28.7089
0.9613	10.0	1960	1.8597	4.5148	29.8895
0.8835	11.0	2156	1.8714	5.6631	29.5201
0.8143	12.0	2352	1.8995	6.3145	30.9947
0.7673	13.0	2548	1.8932	6.1202	31.9430
0.7181	14.0	2744	1.9212	6.9319	31.4740
0.6687	15.0	2940	1.9263	6.6638	31.6561
0.6195	16.0	3136	1.9160	6.9378	31.4958
0.5911	17.0	3332	1.9079	7.6489	32.2981
0.5617	18.0	3528	1.9261	7.0193	31.3876
0.5464	19.0	3724	1.9237	7.6972	33.0673
0.5095	20.0	3920	1.9477	7.0557	32.4604
0.4983	21.0	4116	1.9287	7.4806	32.5537
0.4842	22.0	4312	1.9340	7.6042	32.5836
0.4561	23.0	4508	1.9404	7.6844	32.8472
0.4529	24.0	4704	1.9431	7.9702	33.1332
0.4447	25.0	4900	1.9446	7.8032	32.7398

Framework versions

• Transformers 5.15.0
• Pytorch 2.11.0+cu128
• Datasets 4.0.0
• Tokenizers 0.22.2