1ad6815255504ab67e0d4da874aafe6f

This model is a fine-tuned version of google/umt5-xl on the Helsinki-NLP/opus_books [fi-fr] dataset. It achieves the following results on the evaluation set:

• Loss: 1.8647
• Data Size: 1.0
• Epoch Runtime: 57.8625
• Bleu: 7.4754

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	5.3468	0	3.9628	2.2765
No log	1	88	4.5329	0.0078	4.7275	4.0894
No log	2	176	3.9195	0.0156	8.7672	6.6590
No log	3	264	3.5006	0.0312	17.8613	7.9060
No log	4	352	2.9627	0.0625	24.8823	10.5429
No log	5	440	2.7270	0.125	21.4343	12.6049
0.301	6	528	2.3474	0.25	27.1772	14.3128
0.9863	7	616	2.0092	0.5	40.3340	6.1987
2.1784	8.0	704	1.7975	1.0	63.8669	6.9955
1.9084	9.0	792	1.7758	1.0	61.1703	7.2400
1.6457	10.0	880	1.7758	1.0	53.9690	7.4825
1.4748	11.0	968	1.7778	1.0	58.3039	7.6142
1.3029	12.0	1056	1.8214	1.0	54.3570	7.6211
1.1673	13.0	1144	1.8647	1.0	57.8625	7.4754

Framework versions

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