vgaraujov/t5-base-translation-en-es

<!-- This model card has been generated automatically according to the information the Trainer had access to. You should probably proofread and complete it, then remove this comment. -->

T5S (base-sized model) for en-es translation

This model is a fine-tuned version of T5S on a small portion of WMT13 es-en dataset. It achieves the following results on the evaluation set:

• —Loss: 1.7643
• —Bleu: 30.6296
• —Gen Len: 29.2701

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• —learning_rate: 0.005
• —trainbatchsize: 32
• —evalbatchsize: 32
• —seed: 42
• —gradientaccumulationsteps: 12
• —totaltrainbatch_size: 384
• —optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• —lrschedulertype: linear
• —lrschedulerwarmup_steps: 40000
• —training_steps: 10000

Framework versions

• —Transformers 4.33.0.dev0
• —Pytorch 2.0.1+cu117
• —Datasets 2.14.4
• —Tokenizers 0.13.3