geez_15k_sc_mt5

This model is a fine-tuned version of Samuael/geez_15k_sc_mt5 on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.8105
• Wer: 0.2849
• Cer: 0.1373
• Bleu: 57.6756

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: 0.0002
• train_batch_size: 128
• eval_batch_size: 128
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 30

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer	Bleu
0.9418	1.0	57	0.8984	0.4204	0.1872	42.5936
0.6386	2.0	114	0.6856	0.3535	0.1563	48.7013
0.6169	3.0	171	0.6101	0.3137	0.1256	52.0856
0.428	4.0	228	0.5921	0.3125	0.1365	53.1328
0.354	5.0	285	0.5692	0.2807	0.1096	55.5660
0.2604	6.0	342	0.5947	0.2818	0.1098	55.6749
0.2797	7.0	399	0.5874	0.2896	0.1203	55.5685
0.2372	8.0	456	0.5978	0.2829	0.1206	56.3162
0.1863	9.0	513	0.6046	0.2808	0.1188	56.7817
0.1595	10.0	570	0.6295	0.2615	0.1006	58.1387
0.1269	11.0	627	0.6548	0.2721	0.1157	57.4783
0.1421	12.0	684	0.6572	0.2734	0.1164	57.7264
0.1475	13.0	741	0.6673	0.2809	0.1204	57.1626
0.1201	14.0	798	0.6996	0.2835	0.1271	57.0274
0.0777	15.0	855	0.7227	0.2634	0.1071	58.3844
0.075	16.0	912	0.7295	0.2607	0.1050	58.7280
0.092	17.0	969	0.7404	0.2633	0.1062	58.7387
0.0795	18.0	1026	0.7437	0.2842	0.1297	57.4888
0.0933	19.0	1083	0.7513	0.2765	0.1207	57.8567
0.0794	20.0	1140	0.7620	0.2718	0.1165	58.1063
0.0559	21.0	1197	0.7543	0.2526	0.0999	59.4966
0.0616	22.0	1254	0.7885	0.2664	0.1126	58.7126
0.0666	23.0	1311	0.7734	0.2774	0.1215	57.7048
0.0746	24.0	1368	0.7832	0.2764	0.1225	58.0553
0.0644	25.0	1425	0.7872	0.2799	0.1273	57.9207
0.0596	26.0	1482	0.8184	0.2721	0.1197	58.1888
0.0507	27.0	1539	0.8053	0.2765	0.1226	58.1358
0.0585	28.0	1596	0.8091	0.2806	0.1293	57.9993
0.0595	29.0	1653	0.8147	0.2888	0.1416	57.4399
0.0469	30.0	1710	0.8105	0.2849	0.1373	57.6756

Framework versions

• Transformers 4.38.2
• Pytorch 2.2.1+cu121
• Datasets 2.18.0
• Tokenizers 0.15.2