58121f3e1ea38ae2105ac0fa39415789

This model is a fine-tuned version of google-t5/t5-base on the Helsinki-NLP/opus_books [fi-fr] dataset. It achieves the following results on the evaluation set:

• Loss: 2.2311
• Data Size: 1.0
• Epoch Runtime: 25.8838
• Bleu: 4.1377

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	3.0327	0	2.6484	0.7169
No log	1	88	3.0123	0.0078	3.5483	0.7418
No log	2	176	2.9323	0.0156	3.5797	0.8045
No log	3	264	2.8649	0.0312	4.0904	0.9885
No log	4	352	2.7618	0.0625	4.8085	1.9944
No log	5	440	2.6778	0.125	5.9659	2.2098
0.2201	6	528	2.5888	0.25	8.3061	2.5983
0.947	7	616	2.5024	0.5	13.1454	3.4365
2.6005	8.0	704	2.4140	1.0	23.1260	3.2020
2.5269	9.0	792	2.3743	1.0	22.4581	3.4899
2.4391	10.0	880	2.3406	1.0	22.9632	3.6103
2.3775	11.0	968	2.3181	1.0	21.9862	3.3689
2.3111	12.0	1056	2.2969	1.0	22.5946	3.8995
2.254	13.0	1144	2.2771	1.0	24.3770	3.5488
2.1993	14.0	1232	2.2650	1.0	24.5009	3.5215
2.1625	15.0	1320	2.2540	1.0	24.5018	3.7872
2.1073	16.0	1408	2.2507	1.0	23.3027	3.8548
2.063	17.0	1496	2.2424	1.0	24.1523	3.9186
2.0101	18.0	1584	2.2336	1.0	22.7670	3.8923
2.0018	19.0	1672	2.2473	1.0	23.1051	3.8867
1.9415	20.0	1760	2.2292	1.0	23.7347	3.9063
1.9094	21.0	1848	2.2331	1.0	23.6862	3.8730
1.8943	22.0	1936	2.2388	1.0	23.6942	4.0321
1.8465	23.0	2024	2.2375	1.0	23.2658	4.1089
1.8152	24.0	2112	2.2311	1.0	25.8838	4.1377

Framework versions

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