650c49b510b81e96c0b889605e9f972a

This model is a fine-tuned version of facebook/mbart-large-50-many-to-one-mmt on the Helsinki-NLP/opus_books [en-pl] dataset. It achieves the following results on the evaluation set:

• Loss: 4.3108
• Data Size: 1.0
• Epoch Runtime: 22.8897
• Bleu: 3.5178

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	6.9429	0	2.1015	0.2943
No log	1	70	6.4868	0.0078	2.8303	0.3469
No log	2	140	5.9445	0.0156	4.2468	0.3256
No log	3	210	5.6516	0.0312	5.5153	0.4651
No log	4	280	5.3119	0.0625	6.6668	0.6654
No log	5	350	4.9900	0.125	9.2598	0.7085
No log	6	420	4.6067	0.25	11.1137	0.8583
0.725	7	490	4.2358	0.5	13.9405	1.1236
3.7151	8.0	560	3.8947	1.0	23.5088	1.2999
3.1297	9.0	630	3.8078	1.0	23.0871	1.5138
2.414	10.0	700	3.8846	1.0	21.5226	2.0615
1.9158	11.0	770	3.9866	1.0	21.9196	2.2940
1.6974	12.0	840	4.1980	1.0	22.1827	3.9032
1.1698	13.0	910	4.3108	1.0	22.8897	3.5178

Framework versions

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