jako_mbartLarge_6p_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: 1.1539
• Bleu: 26.2658
• Gen Len: 19.3961

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: 300
• num_epochs: 10

Training results

Training Loss	Epoch	Step	Validation Loss	Bleu	Gen Len
1.4641	0.48	1000	1.3276	21.6162	19.4434
1.2615	0.96	2000	1.1866	24.346	19.4734
0.9103	1.44	3000	1.1638	25.4249	19.0086
0.8285	1.92	4000	1.1539	26.2658	19.3961
0.5977	2.4	5000	1.1978	25.5651	19.6248
0.5423	2.88	6000	1.1830	26.8441	19.1349
0.3816	3.36	7000	1.2670	26.1301	19.1207
0.3412	3.84	8000	1.2870	26.7783	19.2417

Framework versions

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