wav2vec2-base-checkpoint-9

This model is a fine-tuned version of jiobiala24/wav2vec2-base-checkpoint-8 on the common_voice dataset. It achieves the following results on the evaluation set:

• Loss: 0.9203
• Wer: 0.3258

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
0.2783	1.58	1000	0.5610	0.3359
0.2251	3.16	2000	0.5941	0.3374
0.173	4.74	3000	0.6026	0.3472
0.1475	6.32	4000	0.6750	0.3482
0.1246	7.9	5000	0.6673	0.3414
0.1081	9.48	6000	0.7072	0.3409
0.1006	11.06	7000	0.7413	0.3392
0.0879	12.64	8000	0.7831	0.3394
0.0821	14.22	9000	0.7371	0.3333
0.0751	15.8	10000	0.8321	0.3445
0.0671	17.38	11000	0.8362	0.3357
0.0646	18.96	12000	0.8709	0.3367
0.0595	20.54	13000	0.8352	0.3321
0.0564	22.12	14000	0.8854	0.3323
0.052	23.7	15000	0.9031	0.3315
0.0485	25.28	16000	0.9171	0.3278
0.046	26.86	17000	0.9390	0.3254
0.0438	28.44	18000	0.9203	0.3258

Framework versions

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