mbart-large-cc25-finetuned-hi-to-en-v2

This model is a fine-tuned version of facebook/mbart-large-cc25 on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 1.8027
• Bleu: 33.4814
• Gen Len: 21.8974

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: 1
• eval_batch_size: 1
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 4
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 10
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Bleu	Gen Len
1.8971	1.0	3955	1.6015	19.3557	43.7594
1.3266	2.0	7910	1.4917	19.1404	35.3155
0.9906	3.0	11865	1.5354	26.999	26.7497
0.6987	4.0	15820	1.6457	31.9572	23.4565
0.5073	5.0	19775	1.8544	34.1169	22.1507
0.3554	6.0	23730	2.0985	34.0746	22.2396
0.2423	7.0	27685	2.2534	33.2205	22.2184
0.1918	8.0	31640	2.4014	32.2001	22.635
0.1423	9.0	35595	2.5067	32.4074	22.8716
0.1105	10.0	39550	2.5618	33.1965	22.5905

Framework versions

• Transformers 4.18.0
• Pytorch 1.11.0+cu113
• Datasets 2.1.0
• Tokenizers 0.12.1