mbart-large-50-many-to-many-mmt-finetuned-hi-to-sn

This model is a fine-tuned version of facebook/mbart-large-50-many-to-many-mmt on the None dataset. It achieves the following results on the evaluation set:

• Loss: 10.2452
• Bleu: 1.9982
• Gen Len: 61.4

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: 2e-05
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 16
• total_train_batch_size: 128
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 5
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Bleu	Gen Len
13.529	0.2581	1	12.2155	1.2723	109.8
13.0664	0.5161	2	11.4316	1.9024	65.0
12.1188	0.7742	3	10.9837	2.258	64.6
11.2461	1.0323	4	10.8421	1.8914	62.6
11.2396	1.2903	5	10.7537	0.9536	93.4
11.1143	1.5484	6	10.6779	1.6956	88.2
10.7726	1.8065	7	10.5891	2.0507	82.6
10.7751	2.0645	8	10.4973	1.6111	85.6
10.6148	2.3226	9	10.4231	2.1115	81.4
10.5548	2.5806	10	10.3634	3.0095	83.6
10.4732	2.8387	11	10.3168	1.939	87.8
10.4961	3.0968	12	10.2865	2.9337	86.6
10.4483	3.3548	13	10.2645	3.7622	60.4
10.3593	3.6129	14	10.2514	2.0713	60.4
10.4614	3.8710	15	10.2452	1.9982	61.4

Framework versions

• Transformers 4.40.1
• Pytorch 2.3.0+cu118
• Datasets 2.19.1
• Tokenizers 0.19.1