metadata

license: apache-2.0
base_model: Samuael/asr-amharic-phoneme-based-37-6
tags:
  - generated_from_trainer
metrics:
  - wer
model-index:
  - name: asr-amharic-phoneme-based-37-6
    results: []

asr-amharic-phoneme-based-37-6

This model is a fine-tuned version of Samuael/asr-amharic-phoneme-based-37-6 on the None dataset. It achieves the following results on the evaluation set:

Loss: 0.2467
Wer: 0.2877
Phoneme Cer: 0.0539
Cer: 0.0777

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: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: linear
lr_scheduler_warmup_steps: 100
num_epochs: 10

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Phoneme Cer	Cer
3.2186	0.25	200	3.2626	1.0	1.0	1.0
3.0459	0.5	400	3.0440	1.0	1.0	1.0
2.9983	0.75	600	2.9807	1.0	1.0	1.0
1.8006	1.0	800	1.4020	0.8218	0.2751	0.4270
0.6145	1.25	1000	0.4420	0.3611	0.0717	0.1040
0.4288	1.5	1200	0.3304	0.3394	0.0649	0.0941
0.3335	1.75	1400	0.2996	0.3314	0.0622	0.0909
0.4583	2.01	1600	0.2815	0.3263	0.0612	0.0890
0.3966	2.26	1800	0.2743	0.3120	0.0584	0.0847
0.3781	2.51	2000	0.2669	0.3241	0.0609	0.0881
0.5792	2.76	2200	0.2597	0.3135	0.0590	0.0857
0.3322	3.01	2400	0.2548	0.3076	0.0575	0.0837
0.4129	3.26	2600	0.2577	0.3051	0.0575	0.0832
0.3453	3.51	2800	0.2520	0.3070	0.0577	0.0828
0.3279	3.76	3000	0.2516	0.2989	0.0563	0.0818
0.3312	4.01	3200	0.2487	0.2947	0.0553	0.0801
0.2818	4.26	3400	0.2530	0.2985	0.0563	0.0816
0.2244	4.51	3600	0.2529	0.3021	0.0565	0.0820
0.3562	4.76	3800	0.2477	0.2984	0.0563	0.0814
0.2922	5.01	4000	0.2530	0.2995	0.0563	0.0813
0.2535	5.26	4200	0.2516	0.2990	0.0556	0.0805
0.2043	5.51	4400	0.2532	0.2966	0.0557	0.0800
0.2653	5.76	4600	0.2471	0.2922	0.0548	0.0792
0.4389	6.02	4800	0.2493	0.2942	0.0555	0.0805
0.2757	6.27	5000	0.2533	0.2928	0.0551	0.0799
0.3063	6.52	5200	0.2459	0.2950	0.0554	0.0800
0.2069	6.77	5400	0.2499	0.2946	0.0555	0.0800
0.2018	7.02	5600	0.2503	0.2934	0.0549	0.0792
0.2068	7.27	5800	0.2469	0.2966	0.0551	0.0798
0.2535	7.52	6000	0.2513	0.2936	0.0548	0.0792
0.285	7.77	6200	0.2492	0.2941	0.0548	0.0795
0.1801	8.02	6400	0.2460	0.2896	0.0540	0.0781
0.1705	8.27	6600	0.2475	0.2933	0.0549	0.0793
0.3344	8.52	6800	0.2501	0.2890	0.0540	0.0777
0.2236	8.77	7000	0.2471	0.2871	0.0536	0.0773
0.349	9.02	7200	0.2480	0.2869	0.0541	0.0779
0.2692	9.27	7400	0.2482	0.2891	0.0540	0.0779
0.2265	9.52	7600	0.2481	0.2879	0.0539	0.0776
0.1649	9.77	7800	0.2467	0.2877	0.0539	0.0777

Framework versions

Transformers 4.35.2
Pytorch 2.1.0+cu121
Datasets 2.16.1
Tokenizers 0.15.1