wav2vec2-xls-r-164m-id

This model is a fine-tuned version of evanarlian/distil-wav2vec2-xls-r-164m-id on the evanarlian/common_voice_11_0_id_filtered dataset. It achieves the following results on the evaluation set:

• Loss: 0.2865
• Wer: 0.2923

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: 24
• eval_batch_size: 24
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.3
• num_epochs: 80.0
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
1.4047	4.59	5000	1.0167	0.9138
0.587	9.18	10000	0.4639	0.5615
0.3782	13.77	15000	0.3375	0.4496
0.2867	18.37	20000	0.2881	0.4022
0.2519	22.96	25000	0.2775	0.3700
0.1941	27.55	30000	0.2701	0.3516
0.1727	32.14	35000	0.2795	0.3486
0.1448	36.73	40000	0.2878	0.3364
0.1251	41.32	45000	0.2649	0.3275
0.113	45.91	50000	0.2862	0.3168
0.0994	50.51	55000	0.2798	0.3091
0.0938	55.1	60000	0.2864	0.3070
0.0853	59.69	65000	0.2860	0.3069
0.0724	64.28	70000	0.2994	0.3003
0.0723	68.87	75000	0.2951	0.2983
0.0666	73.46	80000	0.2886	0.2941
0.0659	78.05	85000	0.2865	0.2923

Framework versions

• Transformers 4.27.0.dev0
• Pytorch 1.13.1+cu117
• Datasets 2.9.1.dev0
• Tokenizers 0.13.2