asr_model

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: 0.2363
• Wer: 0.5153

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: 1e-05
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 16
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• num_epochs: 30
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.4621	2.0	1000	0.4702	0.9741
0.4612	4.0	2000	0.4621	0.9741
0.4458	6.0	3000	0.4464	0.9714
0.384	8.0	4000	0.3853	0.8235
0.3065	10.0	5000	0.3166	0.7829
0.2861	12.0	6000	0.2809	0.6802
0.248	14.0	7000	0.2677	0.6051
0.2449	16.0	8000	0.2541	0.5778
0.2298	18.0	9000	0.2480	0.5710
0.2281	20.0	10000	0.2418	0.5505
0.216	22.0	11000	0.2420	0.5340
0.2083	24.0	12000	0.2380	0.5253
0.1957	26.0	13000	0.2380	0.5209
0.1985	28.0	14000	0.2360	0.5181
0.2078	30.0	15000	0.2363	0.5153

Framework versions

• Transformers 4.35.2
• Pytorch 2.1.0+cu118
• Datasets 2.15.0
• Tokenizers 0.15.0