wav2vec2-base_phoneme-timit_english_timit-4k_001

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

• Loss: 0.6361
• Per: 0.1195

Model description

The wav2vec 2.0 base model is pre-trained on 960 hours of the LibriSpeech dataset.

• 12 Transformer blocks (Each block: 768 dimensions & 8 attention heads)

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: 16
• eval_batch_size: 16
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 5000
• training_steps: 10000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Per
5.2193	3.46	1000	3.5945	0.9617
1.5174	6.92	2000	0.5574	0.1665
0.5246	10.38	3000	0.4228	0.1503
0.3915	13.84	4000	0.4276	0.1512
0.3293	17.3	5000	0.4656	0.1517
0.2757	20.76	6000	0.4719	0.1486
0.209	24.22	7000	0.5314	0.1478
0.1589	27.68	8000	0.6102	0.1484
0.1207	31.14	9000	0.6449	0.1484
0.0951	34.6	10000	0.6579	0.1471

Framework versions

• Transformers 4.28.1
• Pytorch 2.0.1
• Datasets 2.18.0
• Tokenizers 0.13.3