marianmt_ipa_to_bangla

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

• Loss: 1.3673
• Bleu: 24.2746
• Chrf: 40.2174

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
6.0271	1.0	167	2.2740	15.1069	42.5
2.1997	2.0	334	1.9670	16.5158	34.8442
1.9454	3.0	501	1.8395	24.2746	42.5
1.7887	4.0	668	1.7030	19.3049	45.0581
1.6442	5.0	835	1.6269	19.6407	38.1679
1.5604	6.0	1002	1.5627	20.4124	45.0581
1.4686	7.0	1169	1.5193	17.2860	36.3176
1.3927	8.0	1336	1.4721	20.4124	40.2174
1.3269	9.0	1503	1.4565	15.9736	47.9452
1.2602	10.0	1670	1.4393	9.8204	36.3176
1.2216	11.0	1837	1.4321	12.1373	40.7973
1.1577	12.0	2004	1.4098	15.1069	31.7073
1.1119	13.0	2171	1.3935	12.1373	38.1679
1.0807	14.0	2338	1.3921	24.2746	51.8617
1.0433	15.0	2505	1.3676	24.2746	51.8617
1.0115	16.0	2672	1.3978	24.2746	40.2174
0.9770	17.0	2839	1.3756	24.2746	51.8617
0.9637	18.0	3006	1.3610	24.2746	51.8617
0.9291	19.0	3173	1.3760	24.2746	51.8617
0.9146	20.0	3340	1.3665	20.4124	42.5
0.8912	21.0	3507	1.3701	19.3049	51.8617
0.8767	22.0	3674	1.3631	19.3049	49.2941
0.8686	23.0	3841	1.3674	24.2746	51.8617
0.8477	24.0	4008	1.3695	24.2746	51.8617
0.8435	25.0	4175	1.3673	24.2746	40.2174

Framework versions

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