wav2vec2LugandaASR20

This model is a fine-tuned version of Gemmar/wav2vec2LugandaASR on the common_voice_13_0 dataset. It achieves the following results on the evaluation set:

Loss: 0.2393
Wer: 0.2322

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: 0.0003
train_batch_size: 32
eval_batch_size: 8
seed: 42
gradient_accumulation_steps: 4
total_train_batch_size: 128
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: linear
lr_scheduler_warmup_steps: 200
num_epochs: 5

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.1093	0.18	100	0.2134	0.2480
0.1141	0.36	200	0.2329	0.2724
0.1224	0.54	300	0.2560	0.2864
0.1345	0.72	400	0.2348	0.2716
0.1271	0.9	500	0.2339	0.2702
0.1232	1.08	600	0.2457	0.2806
0.1149	1.27	700	0.2372	0.2695
0.1129	1.45	800	0.2328	0.2718
0.1196	1.63	900	0.2326	0.2615
0.1185	1.81	1000	0.2249	0.2672
0.1159	1.99	1100	0.2202	0.2559
0.0933	2.17	1200	0.2302	0.2559
0.0947	2.35	1300	0.2306	0.2530
0.0941	2.53	1400	0.2325	0.2509
0.0946	2.71	1500	0.2233	0.2495
0.0949	2.89	1600	0.2320	0.2443
0.0883	3.07	1700	0.2383	0.2463
0.0783	3.25	1800	0.2386	0.2437
0.0753	3.43	1900	0.2329	0.2426
0.0772	3.62	2000	0.2317	0.2392
0.0774	3.8	2100	0.2308	0.2353
0.0764	3.98	2200	0.2293	0.2357
0.0666	4.16	2300	0.2446	0.2388
0.065	4.34	2400	0.2456	0.2359
0.0643	4.52	2500	0.2446	0.2345
0.0652	4.7	2600	0.2430	0.2325
0.0669	4.88	2700	0.2393	0.2322

Framework versions

Transformers 4.30.2
Pytorch 2.0.1+cu118
Datasets 2.13.0
Tokenizers 0.13.3

Gemmar
/

wav2vec2LugandaASR20

wav2vec2LugandaASR20

Model description

Intended uses & limitations

Training and evaluation data

Training procedure

Training hyperparameters

Training results

Framework versions

Evaluation results