wav2vec2-base-checkpoint-13

This model is a fine-tuned version of jiobiala24/wav2vec2-base-checkpoint-12 on the common_voice dataset. It achieves the following results on the evaluation set:

• Loss: 1.1804
• Wer: 0.3809

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
• num_epochs: 30
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.2688	1.92	1000	0.6518	0.3692
0.1944	3.85	2000	0.7188	0.3808
0.1503	5.77	3000	0.7552	0.3853
0.1218	7.69	4000	0.8155	0.3834
0.1024	9.62	5000	0.8867	0.3779
0.0874	11.54	6000	0.8917	0.3866
0.0775	13.46	7000	1.0320	0.4019
0.0712	15.38	8000	1.0110	0.3922
0.0656	17.31	9000	1.0494	0.3885
0.0578	19.23	10000	1.1054	0.3883
0.053	21.15	11000	1.1285	0.3938
0.0496	23.08	12000	1.1358	0.3884
0.0459	25.0	13000	1.2062	0.3904
0.0445	26.92	14000	1.1811	0.3830
0.0414	28.85	15000	1.1804	0.3809

Framework versions

• Transformers 4.11.3
• Pytorch 1.10.0+cu111
• Datasets 1.13.3
• Tokenizers 0.10.3