wav2vec2-11

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: 3.0827
• Wer: 1.0

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: 24
• eval_batch_size: 6
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 400
• num_epochs: 30

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
4.2589	1.18	200	3.1595	1.0
2.8683	2.35	400	3.1270	1.0
2.8692	3.53	600	3.1041	1.0
2.8577	4.71	800	3.0804	1.0
2.8587	5.88	1000	3.0556	1.0
2.8615	7.06	1200	3.1084	1.0
2.8598	8.24	1400	3.0608	1.0
2.8571	9.41	1600	3.0997	1.0
2.8595	10.59	1800	3.1533	1.0
2.8568	11.76	2000	3.0621	1.0
2.8563	12.94	2200	3.1072	1.0
2.8556	14.12	2400	3.1299	1.0
2.8581	15.29	2600	3.0565	1.0
2.8534	16.47	2800	3.0821	1.0
2.857	17.65	3000	3.0734	1.0
2.8545	18.82	3200	3.1392	1.0
2.8568	20.0	3400	3.0541	1.0
2.8519	21.18	3600	3.0856	1.0
2.8542	22.35	3800	3.1477	1.0
2.8565	23.53	4000	3.0433	1.0
2.8525	24.71	4200	3.0826	1.0
2.8538	25.88	4400	3.0972	1.0
2.857	27.06	4600	3.0762	1.0
2.8523	28.24	4800	3.0828	1.0
2.8526	29.41	5000	3.0827	1.0

Framework versions

• Transformers 4.19.2
• Pytorch 1.11.0+cu113
• Datasets 2.2.2
• Tokenizers 0.12.1