w2v2-base-pretrained_lr1e-4_at0.8_da0.7 – AI Model by MelanieKoe | AlphaNeural AI

You can deploy this model and start earning money today!

MelanieKoe/
w2v2-base-pretrained_lr1e-4_at0.8_da0.7

like

0

automatic-speech-recognition

generated_from_trainer

facebook/wav2vec2-base

endpoints_compatible

Views

No views yet

w2v2-base-pretrained_lr1e-4_at0.8_da0.7

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

Loss: 2.1472
Wer: 0.1696

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: 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
training_steps: 3500
mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
15.7341	7.81	250	3.7148	1.0
3.1637	15.62	500	3.1328	1.0
2.8368	23.44	750	1.7455	1.0
0.3387	31.25	1000	1.2956	0.2136
0.0993	39.06	1250	1.5921	0.1991
0.0544	46.88	1500	1.7186	0.1927
0.0379	54.69	1750	1.6757	0.1862
0.0259	62.5	2000	2.0243	0.1768
0.0195	70.31	2250	2.0139	0.1803
0.0158	78.12	2500	2.0300	0.1794
0.0121	85.94	2750	1.8981	0.1734
0.01	93.75	3000	2.1159	0.1739
0.0079	101.56	3250	2.1113	0.1696
0.0076	109.38	3500	2.1472	0.1696

Framework versions

Transformers 4.35.0
Pytorch 2.0.0
Datasets 2.14.6
Tokenizers 0.14.1