62a075372abc23e43ad471fc6bd95a12

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

• Loss: 1.8290
• Data Size: 1.0
• Epoch Runtime: 19.3501
• Bleu: 10.0435

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	23.7712	0	2.0270	0.0223
No log	1	29	23.1319	0.0078	2.5604	0.0263
No log	2	58	21.3967	0.0156	3.9394	0.0231
No log	3	87	18.1552	0.0312	5.5615	0.0385
No log	4	116	17.6035	0.0625	7.8372	0.0349
No log	5	145	17.7126	0.125	10.1551	0.0290
2.2446	6	174	14.6419	0.25	12.8700	0.0361
2.2446	7	203	5.5838	0.5	15.0699	0.1095
2.2446	8.0	232	3.3810	1.0	22.4152	0.2806
5.7139	9.0	261	2.6356	1.0	21.6294	0.6042
5.7139	10.0	290	2.2775	1.0	23.0527	0.9270
3.1909	11.0	319	2.0761	1.0	19.3828	1.0265
3.1909	12.0	348	1.9763	1.0	20.4135	7.5631
2.5045	13.0	377	1.8875	1.0	21.3273	6.4945
2.1513	14.0	406	1.8598	1.0	19.1914	6.5918
2.1513	15.0	435	1.8354	1.0	21.4064	6.9259
1.9497	16.0	464	1.8107	1.0	22.2024	7.2488
1.9497	17.0	493	1.8079	1.0	18.5415	7.5164
1.8033	18.0	522	1.7994	1.0	19.4302	7.9296
1.6604	19.0	551	1.8063	1.0	20.6071	8.6236
1.6604	20.0	580	1.8065	1.0	19.6389	8.6423
1.5436	21.0	609	1.7939	1.0	19.3613	9.4241
1.5436	22.0	638	1.8042	1.0	20.4159	9.5937
1.4611	23.0	667	1.8087	1.0	20.9625	9.6160
1.4611	24.0	696	1.8298	1.0	22.8316	9.7628
1.3726	25.0	725	1.8290	1.0	19.3501	10.0435

Framework versions

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