nmt-mpst-id-en-lr_0.001-ep_20-seq_128_bs-16

This model is a fine-tuned version of t5-small on the None dataset. It achieves the following results on the evaluation set:

• Loss: 1.9735
• Bleu: 0.2017
• Meteor: 0.3744

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: 0.001
• train_batch_size: 16
• eval_batch_size: 16
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 20

Training results

Training Loss	Epoch	Step	Validation Loss	Bleu	Meteor
No log	1.0	404	2.0539	0.1063	0.2561
2.5624	2.0	808	1.7616	0.1363	0.2966
1.7817	3.0	1212	1.6548	0.1612	0.325
1.4215	4.0	1616	1.6071	0.1725	0.3388
1.2099	5.0	2020	1.5886	0.1841	0.3517
1.2099	6.0	2424	1.5992	0.1882	0.3571
1.0117	7.0	2828	1.5909	0.1892	0.358
0.878	8.0	3232	1.6242	0.1947	0.3652
0.7881	9.0	3636	1.6626	0.197	0.3673
0.6932	10.0	4040	1.6918	0.196	0.3679
0.6932	11.0	4444	1.7347	0.1956	0.3683
0.6069	12.0	4848	1.7718	0.2008	0.3711
0.5426	13.0	5252	1.8005	0.2011	0.372
0.4946	14.0	5656	1.8383	0.201	0.3732
0.4414	15.0	6060	1.8713	0.2031	0.3743
0.4414	16.0	6464	1.9040	0.2025	0.3748
0.4039	17.0	6868	1.9381	0.202	0.3736
0.3661	18.0	7272	1.9497	0.2006	0.3732
0.3424	19.0	7676	1.9680	0.2009	0.3744
0.3253	20.0	8080	1.9735	0.2017	0.3744

Framework versions

• Transformers 4.24.0
• Pytorch 1.12.1+cu113
• Datasets 2.7.0
• Tokenizers 0.13.2