wav2vec2-large-xlsr-53-English-Version1

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

• Loss: 0.3182
• Wer: 0.2342

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: 3e-05
• 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: 500
• num_epochs: 100
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
3.1237	5.3191	500	3.0432	1.0
2.9376	10.6383	1000	2.9187	1.0
2.9119	15.9574	1500	2.9069	1.0
2.8653	21.2766	2000	2.8018	1.0
1.0039	26.5957	2500	0.6510	0.5201
0.5604	31.9149	3000	0.4490	0.3812
0.4113	37.2340	3500	0.3902	0.3201
0.4361	42.5532	4000	0.3471	0.2856
0.3635	47.8723	4500	0.3422	0.2767
0.3573	53.1915	5000	0.3322	0.2627
0.2864	58.5106	5500	0.3288	0.2537
0.333	63.8298	6000	0.3220	0.2475
0.2881	69.1489	6500	0.3190	0.2439
0.2716	74.4681	7000	0.3189	0.2427
0.281	79.7872	7500	0.3176	0.2395
0.268	85.1064	8000	0.3200	0.2364
0.2713	90.4255	8500	0.3163	0.2349
0.2307	95.7447	9000	0.3182	0.2342

Framework versions

• Transformers 4.45.0.dev0
• Pytorch 2.4.1+cu121
• Datasets 2.21.0
• Tokenizers 0.19.1