byt5-base-es_sja

This model is a fine-tuned version of google/byt5-base on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.7290
• Bleu: 4.2147
• Gen Len: 99.2837

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: 16
• eval_batch_size: 16
• seed: 65
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 10.0

Training results

Training Loss	Epoch	Step	Validation Loss	Bleu	Gen Len
No log	1.0	397	1.0329	0.0222	19.0
1.5198	2.0	794	0.9034	0.0395	19.0
1.0351	3.0	1191	0.8492	0.0686	19.0
0.9326	4.0	1588	0.8077	0.0832	19.0
0.9326	5.0	1985	0.7786	0.0983	19.0
0.8785	6.0	2382	0.7610	0.0606	19.0
0.8436	7.0	2779	0.7446	0.065	19.0
0.8153	8.0	3176	0.7361	0.0618	19.0
0.7987	9.0	3573	0.7306	0.068	19.0
0.7987	10.0	3970	0.7290	0.0616	19.0

Framework versions

• Transformers 4.27.4
• Pytorch 2.0.0+cu118
• Datasets 2.11.0
• Tokenizers 0.13.3