kozh_mbartLarge_50p_run1

This model is a fine-tuned version of facebook/mbart-large-50-many-to-many-mmt on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 1.2170
• Bleu: 14.9441
• Gen Len: 15.5114

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: 4
• eval_batch_size: 4
• seed: 42
• distributed_type: multi-GPU
• num_devices: 4
• gradient_accumulation_steps: 2
• total_train_batch_size: 32
• total_eval_batch_size: 16
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• num_epochs: 10

Training results

Training Loss	Epoch	Step	Validation Loss	Bleu	Gen Len
1.4736	0.4	3000	1.4102	12.2951	15.3131
1.3408	0.8	6000	1.2932	13.0418	15.7409
1.0858	1.2	9000	1.2527	13.6922	15.432
1.0127	1.6	12000	1.2401	14.3889	15.5705
0.9717	2.0	15000	1.2170	14.8665	15.541
0.7866	2.39	18000	1.2569	14.4656	15.4235
0.7634	2.79	21000	1.2687	15.0137	15.3378
0.6205	3.19	24000	1.3149	14.6472	15.5406
0.5796	3.59	27000	1.3323	14.8879	15.3233

Framework versions

• Transformers 4.34.1
• Pytorch 2.1.0+cu121
• Datasets 2.14.6
• Tokenizers 0.14.1