wav2vec2-base-common-voice-50p-persian-colab

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: 1.0939
• Wer: 0.6537

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: 7.5e-05
• train_batch_size: 16
• eval_batch_size: 1
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 2000
• num_epochs: 40
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
3.0437	2.52	600	3.0170	1.0
2.3667	5.04	1200	2.1575	0.9988
0.9565	7.56	1800	1.0801	0.8410
0.603	10.08	2400	0.9680	0.7678
0.507	12.61	3000	0.9554	0.7470
0.3754	15.13	3600	0.9524	0.7157
0.4267	17.65	4200	0.9290	0.6980
0.3308	20.17	4800	0.9557	0.7061
0.2259	22.69	5400	0.9864	0.6830
0.2486	25.21	6000	1.1086	0.6812
0.1956	27.73	6600	1.0497	0.6805
0.1835	30.25	7200	1.0660	0.6596
0.1926	32.77	7800	1.1274	0.6600
0.2765	35.29	8400	1.0882	0.6603
0.2397	37.82	9000	1.0939	0.6537

Framework versions

• Transformers 4.11.3
• Pytorch 1.10.0+cu113
• Datasets 1.18.3
• Tokenizers 0.10.3