wav2vec2-xls-r-common_voice-tr-ft-stream

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

• Loss: 0.3519
• Wer: 0.2927
• Cer: 0.0694

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: 0.0005
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• distributed_type: multi-GPU
• num_devices: 4
• gradient_accumulation_steps: 2
• total_train_batch_size: 64
• 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
• training_steps: 5000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
0.6768	9.01	500	0.4220	0.5143	0.1235
0.3801	19.01	1000	0.3303	0.4403	0.1055
0.3616	29.0	1500	0.3540	0.3716	0.0878
0.2334	39.0	2000	0.3666	0.3671	0.0842
0.3141	49.0	2500	0.3407	0.3373	0.0819
0.1926	58.01	3000	0.3886	0.3520	0.0867
0.1372	68.01	3500	0.3415	0.3189	0.0743
0.091	78.0	4000	0.3750	0.3164	0.0757
0.0893	88.0	4500	0.3559	0.2968	0.0712
0.095	98.0	5000	0.3519	0.2927	0.0694

Framework versions

• Transformers 4.16.0.dev0
• Pytorch 1.10.2
• Datasets 1.18.2
• Tokenizers 0.10.3