metadata

license: cc-by-nc-4.0
tags:
  - generated_from_trainer
model-index:
  - name: wav2vec2-base-fi-voxpopuli-v2-finetuned
    results: []

wav2vec2-base-fi-voxpopuli-v2-finetuned

This model is a fine-tuned version of facebook/wav2vec2-base-fi-voxpopuli-v2 on the None dataset. It achieves the following results on the evaluation set:

Loss: 0.1316
Wer: 0.1498

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: 0.0001
train_batch_size: 64
eval_batch_size: 64
seed: 42
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: linear
lr_scheduler_warmup_steps: 500
num_epochs: 10
mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
1.575	0.33	500	0.7454	0.7048
0.5838	0.66	1000	0.2377	0.2608
0.5692	1.0	1500	0.2014	0.2244
0.5112	1.33	2000	0.1885	0.2013
0.4857	1.66	2500	0.1881	0.2120
0.4821	1.99	3000	0.1603	0.1894
0.4531	2.32	3500	0.1594	0.1865
0.4411	2.65	4000	0.1641	0.1874
0.4437	2.99	4500	0.1545	0.1874
0.4191	3.32	5000	0.1565	0.1770
0.4158	3.65	5500	0.1696	0.1867
0.4032	3.98	6000	0.1561	0.1746
0.4003	4.31	6500	0.1432	0.1749
0.4059	4.64	7000	0.1390	0.1690
0.4019	4.98	7500	0.1291	0.1646
0.3811	5.31	8000	0.1485	0.1755
0.3955	5.64	8500	0.1351	0.1659
0.3562	5.97	9000	0.1328	0.1614
0.3646	6.3	9500	0.1329	0.1584
0.351	6.64	10000	0.1342	0.1554
0.3408	6.97	10500	0.1422	0.1509
0.3562	7.3	11000	0.1309	0.1528
0.3335	7.63	11500	0.1305	0.1506
0.3491	7.96	12000	0.1365	0.1560
0.3538	8.29	12500	0.1293	0.1512
0.3338	8.63	13000	0.1328	0.1511
0.3509	8.96	13500	0.1304	0.1520
0.3431	9.29	14000	0.1360	0.1517
0.3309	9.62	14500	0.1328	0.1514
0.3252	9.95	15000	0.1316	0.1498

Framework versions

Transformers 4.19.1
Pytorch 1.11.0+cu102
Datasets 2.2.1
Tokenizers 0.11.0