wav2vec2-large-xlsr-53-AsanteTwi-04

This model is a fine-tuned version of facebook/wav2vec2-large-xlsr-53 on the common_voice_13_0 dataset. It achieves the following results on the evaluation set:

• Loss: 0.7250
• Wer: 0.625

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: 16
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 32
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 100
• num_epochs: 90

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
13.2642	8.33	50	4.7327	1.0
3.1075	16.67	100	3.1680	1.0
2.8849	25.0	150	2.9745	1.0
2.8553	33.33	200	2.9167	1.0
2.8333	41.67	250	2.8538	1.0
2.6501	50.0	300	2.3417	1.0
1.8966	58.33	350	1.1529	0.875
0.9431	66.67	400	0.8519	0.75
0.5951	75.0	450	0.7970	0.625
0.444	83.33	500	0.7250	0.625

Framework versions

• Transformers 4.30.2
• Pytorch 2.0.1+cu118
• Datasets 2.13.1
• Tokenizers 0.13.3