mimba/opus-mt-fra2ngiemboon

This model is a fine-tuned version of Helsinki-NLP/opus-mt-fr-en on the mimba/text2text dataset. It achieves the following results on the evaluation set:

• Loss: 1.0585
• Bleu: 0.1073
• Chrf: 33.2904

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: 32
• eval_batch_size: 32
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 64
• 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
• lr_scheduler_warmup_steps: 500
• num_epochs: 22
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Bleu	Chrf
2.547	1.0	602	2.0908	0.0082	10.1092
1.8013	2.0	1204	1.6234	0.0254	16.1978
1.5917	3.0	1806	1.4549	0.0445	20.3687
1.4693	4.0	2408	1.3579	0.0618	23.7895
1.3807	5.0	3010	1.2896	0.0708	26.1327
1.3122	6.0	3612	1.2456	0.0782	27.8947
1.2671	7.0	4214	1.2081	0.0826	28.3534
1.2258	8.0	4816	1.1794	0.0863	29.2487
1.1927	9.0	5418	1.1567	0.0904	30.1251
1.1613	10.0	6020	1.1350	0.0926	30.6358
1.1414	11.0	6622	1.1205	0.0944	30.9280
1.1163	12.0	7224	1.1077	0.0982	31.5726
1.0996	13.0	7826	1.0977	0.0984	31.7920
1.0841	14.0	8428	1.0892	0.1006	31.9491
1.0689	15.0	9030	1.0805	0.1010	32.3391
1.0572	16.0	9632	1.0731	0.1024	32.5539
1.0475	17.0	10234	1.0686	0.1043	32.8205
1.0392	18.0	10836	1.0656	0.1054	32.9240
1.0291	19.0	11438	1.0639	0.1060	33.2118
1.0262	20.0	12040	1.0602	0.1071	33.2937
1.0241	21.0	12642	1.0586	0.1070	33.2877
1.0218	22.0	13244	1.0585	0.1073	33.2904

Framework versions

• Transformers 4.57.6
• Pytorch 2.11.0+cu128
• Datasets 4.5.0
• Tokenizers 0.22.2