263ab4cb82d642995e2be12727b922a1

This model is a fine-tuned version of google/mt5-large on the Helsinki-NLP/opus_books [fr-sv] dataset. It achieves the following results on the evaluation set:

• Loss: 2.2564
• Data Size: 1.0
• Epoch Runtime: 36.4600
• Bleu: 5.7704

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: 5e-05
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• distributed_type: multi-GPU
• num_devices: 4
• total_train_batch_size: 32
• total_eval_batch_size: 32
• optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: constant
• num_epochs: 50

Training results

Training Loss	Epoch	Step	Validation Loss	Data Size	Epoch Runtime	Bleu
No log	0	0	24.3890	0	3.3071	0.0025
No log	1	75	23.7084	0.0078	3.9388	0.0017
No log	2	150	21.1689	0.0156	6.5721	0.0040
No log	3	225	19.1706	0.0312	8.5808	0.0039
No log	4	300	19.6652	0.0625	11.0893	0.0036
No log	5	375	16.3233	0.125	14.1852	0.0045
No log	6	450	15.1590	0.25	18.4018	0.0031
No log	7	525	14.0679	0.5	23.5450	0.0052
15.0882	8.0	600	10.7494	1.0	38.4446	0.0085
10.1869	9.0	675	7.4117	1.0	36.7303	0.0055
6.8121	10.0	750	4.8242	1.0	35.4228	0.0132
4.9805	11.0	825	2.7453	1.0	37.8814	0.4070
3.1367	12.0	900	2.4081	1.0	35.7230	3.3449
2.855	13.0	975	2.3234	1.0	35.9308	3.7089
2.6102	14.0	1050	2.2760	1.0	36.5028	4.7021
2.4512	15.0	1125	2.2514	1.0	36.0870	5.0867
2.3104	16.0	1200	2.2298	1.0	37.8982	5.7355
2.2308	17.0	1275	2.2273	1.0	36.9682	5.7351
2.1351	18.0	1350	2.2185	1.0	36.9190	5.4201
2.0345	19.0	1425	2.2267	1.0	37.6474	5.6028
1.9969	20.0	1500	2.2154	1.0	36.1461	5.9516
1.9012	21.0	1575	2.2291	1.0	37.2582	5.7230
1.8474	22.0	1650	2.2239	1.0	35.4654	5.9171
1.7626	23.0	1725	2.2431	1.0	38.6345	5.7564
1.7088	24.0	1800	2.2564	1.0	36.4600	5.7704

Framework versions

• Transformers 4.57.0
• Pytorch 2.8.0+cu128
• Datasets 4.2.0
• Tokenizers 0.22.1