This model is a fine-tuned version of Samuael/ethiopic-asr-characters on the alffa_amharic dataset. It achieves the following results on the evaluation set:

Loss: 0.4428
Wer: 0.2996
Phoneme Cer: 0.1421

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: 3e-05
train_batch_size: 32
eval_batch_size: 8
seed: 42
optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
lr_scheduler_type: linear
lr_scheduler_warmup_steps: 100
num_epochs: 4

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Phoneme Cer
0.1559	0.2312	200	0.5540	0.3212	0.1475
0.1228	0.4624	400	0.5394	0.3142	0.1466
0.069	0.6936	600	0.5525	0.3139	0.1466
0.0997	0.9249	800	0.5531	0.3118	0.1462
0.0732	1.1561	1000	0.5644	0.3196	0.1465
0.1637	1.3873	1200	0.5367	0.3185	0.1468
0.1145	1.6185	1400	0.5215	0.3180	0.1468
0.1499	1.8497	1600	0.4985	0.3141	0.1455
0.1975	2.0809	1800	0.4814	0.3114	0.1446
0.2116	2.3121	2000	0.4855	0.3085	0.1446
0.2384	2.5434	2200	0.4702	0.3083	0.1441
0.4346	2.7746	2400	0.4762	0.3063	0.1435
0.3156	3.0058	2600	0.4677	0.3044	0.1432
0.5558	3.2370	2800	0.4574	0.2993	0.1425
0.2787	3.4682	3000	0.4478	0.2989	0.1420
0.3268	3.6994	3200	0.4466	0.2978	0.1418
0.3461	3.9306	3400	0.4428	0.2996	0.1421

Framework versions

Transformers 4.46.2
Pytorch 2.5.1+cu121
Datasets 3.1.0
Tokenizers 0.20.3