wav2vec2_common_voice_accents_5

This model is a fine-tuned version of facebook/wav2vec2-xls-r-300m on the common_voice dataset. It achieves the following results on the evaluation set:

• Loss: 0.0027

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: 0.0003
• train_batch_size: 48
• eval_batch_size: 4
• seed: 42
• distributed_type: multi-GPU
• num_devices: 8
• total_train_batch_size: 384
• total_eval_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: 30
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss
4.4163	1.34	400	0.5520
0.3305	2.68	800	0.1698
0.2138	4.03	1200	0.1104
0.1714	5.37	1600	0.0944
0.1546	6.71	2000	0.0700
0.1434	8.05	2400	0.0610
0.1272	9.4	2800	0.0493
0.1183	10.74	3200	0.0371
0.1113	12.08	3600	0.0468
0.1013	13.42	4000	0.0336
0.0923	14.77	4400	0.0282
0.0854	16.11	4800	0.0410
0.0791	17.45	5200	0.0252
0.0713	18.79	5600	0.0128
0.0662	20.13	6000	0.0252
0.0635	21.48	6400	0.0080
0.0607	22.82	6800	0.0098
0.0557	24.16	7200	0.0069
0.0511	25.5	7600	0.0057
0.0474	26.85	8000	0.0046
0.045	28.19	8400	0.0037
0.0426	29.53	8800	0.0027

Framework versions

• Transformers 4.17.0
• Pytorch 1.10.2+cu102
• Datasets 1.18.4
• Tokenizers 0.11.6