This model is a fine-tuned version of facebook/wav2vec2-base-960h on the audiofolder dataset. It achieves the following results on the evaluation set:

Loss: 0.3282
Wer: 0.1501

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: 1e-05
train_batch_size: 8
eval_batch_size: 8
seed: 42
gradient_accumulation_steps: 2
total_train_batch_size: 16
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: linear
lr_scheduler_warmup_steps: 500
num_epochs: 300
mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
2.3444	30.77	200	2.1940	0.9841
1.972	61.54	400	1.4582	0.8167
1.3875	92.31	600	0.8476	0.5902
0.9092	123.08	800	0.5445	0.3636
0.6382	153.85	1000	0.4129	0.2641
0.5789	184.62	1200	0.3497	0.1876
0.4632	215.38	1400	0.3478	0.1616
0.4474	246.15	1600	0.3394	0.1486
0.429	276.92	1800	0.3282	0.1501

Framework versions

Transformers 4.38.2
Pytorch 2.2.1+cu121
Datasets 2.18.0
Tokenizers 0.15.2

St4n/wav2vec2-base-self-519-colab-3-grams

wav2vec2-base-self-331-colab

Model description

Intended uses & limitations

Training and evaluation data

Training procedure

Training hyperparameters

Training results

Framework versions

St4n/
wav2vec2-base-self-519-colab-3-grams