t5-small-finetuned-chinese-to-hausa

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

• Loss: 1.6751
• Bleu: 12.5282
• Gen Len: 18.5325

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: 32
• eval_batch_size: 64
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine
• num_epochs: 30
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Bleu	Gen Len
2.6717	1.0	846	1.9694	12.1664	18.7954
2.0063	2.0	1692	1.8146	10.5635	18.8736
1.8317	3.0	2538	1.7341	10.7724	18.9023
1.7706	4.0	3384	1.6942	11.676	18.0272
1.6908	5.0	4230	1.6608	11.654	17.8361
1.6333	6.0	5076	1.6336	11.6008	18.0251
1.5922	7.0	5922	1.6249	11.1834	18.7068
1.541	8.0	6768	1.6106	12.827	18.6533
1.5121	9.0	7614	1.6082	10.873	14.6468
1.4769	10.0	8460	1.5994	9.1287	15.2999
1.4358	11.0	9306	1.5943	12.1784	18.0381
1.4141	12.0	10152	1.5960	12.3004	18.6165
1.3879	13.0	10998	1.6087	11.6896	18.6615
1.3526	14.0	11844	1.6015	12.2844	18.6508
1.3365	15.0	12690	1.6085	11.9235	17.9056
1.3142	16.0	13536	1.6165	11.8504	17.6737
1.2846	17.0	14382	1.6198	12.4398	18.5284
1.2654	18.0	15228	1.6252	12.8486	17.7201
1.2532	19.0	16074	1.6363	12.1792	17.5936
1.231	20.0	16920	1.6423	12.3326	17.7331
1.2128	21.0	17766	1.6461	12.5054	18.668
1.2029	22.0	18612	1.6509	12.5588	18.5612
1.1899	23.0	19458	1.6570	12.1469	17.7074
1.1804	24.0	20304	1.6620	12.4007	17.7623
1.1728	25.0	21150	1.6681	12.6017	18.5794
1.1697	26.0	21996	1.6688	12.4686	18.5296
1.1661	27.0	22842	1.6720	12.5375	18.5321
1.1617	28.0	23688	1.6744	12.5354	18.5309
1.1606	29.0	24534	1.6752	12.529	18.533
1.1599	30.0	25380	1.6751	12.5282	18.5325

Framework versions

• Transformers 4.44.2
• Pytorch 2.4.0+cu121
• Datasets 2.21.0
• Tokenizers 0.19.1