metadata

license: apache-2.0
base_model: Samuael/asr-alffamharic-phoneme-based
tags:
  - generated_from_trainer
metrics:
  - wer
model-index:
  - name: asr-alffamharic-phoneme-based
    results: []

asr-alffamharic-phoneme-based

This model is a fine-tuned version of Samuael/asr-alffamharic-phoneme-based on the None dataset. It achieves the following results on the evaluation set:

Loss: 0.4498
Wer: 0.1084

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: 32
eval_batch_size: 8
seed: 42
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: linear
lr_scheduler_warmup_steps: 1000
num_epochs: 20

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
1.2273	0.59	200	1.0408	0.3569
0.7702	1.18	400	0.8009	0.2169
0.6769	1.76	600	0.6978	0.1869
0.5454	2.35	800	0.6163	0.1696
0.5779	2.94	1000	0.5616	0.1554
0.4996	3.53	1200	0.5413	0.1437
0.5648	4.12	1400	0.5111	0.1439
0.4741	4.71	1600	0.5178	0.1371
0.499	5.29	1800	0.4943	0.1324
0.4247	5.88	2000	0.4884	0.1279
0.4008	6.47	2200	0.4667	0.1254
0.2744	7.06	2400	0.4626	0.1288
0.3495	7.65	2600	0.4794	0.1229
0.4016	8.24	2800	0.4548	0.1228
0.3833	8.82	3000	0.4660	0.1209
0.3684	9.41	3200	0.4463	0.1235
0.4149	10.0	3400	0.4697	0.1171
0.3917	10.59	3600	0.4570	0.1188
0.2957	11.18	3800	0.4431	0.1171
0.3054	11.76	4000	0.4530	0.1163
0.2755	12.35	4200	0.4690	0.1138
0.3091	12.94	4400	0.4551	0.1157
0.2617	13.53	4600	0.4557	0.1147
0.2725	14.12	4800	0.4670	0.1088
0.2795	14.71	5000	0.4486	0.1118
0.3493	15.29	5200	0.4471	0.1109
0.2949	15.88	5400	0.4469	0.1090
0.2802	16.47	5600	0.4516	0.1100
0.2655	17.06	5800	0.4418	0.1105
0.3211	17.65	6000	0.4351	0.1095
0.2245	18.24	6200	0.4436	0.1093
0.2244	18.82	6400	0.4493	0.1091
0.2549	19.41	6600	0.4502	0.1082
0.3528	20.0	6800	0.4498	0.1084

Framework versions

Transformers 4.35.2
Pytorch 2.1.0+cu118
Datasets 2.15.0
Tokenizers 0.15.0