wav2vec2-common_voice-tr-demo

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

• Loss: 0.3920
• Wer: 0.3511

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: 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: 15.0
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
No log	0.92	100	3.5898	1.0
No log	1.83	200	3.0073	0.9999
No log	2.75	300	0.9230	0.7813
No log	3.67	400	0.5698	0.6135
3.1746	4.59	500	0.5274	0.5653
3.1746	5.5	600	0.4778	0.5123
3.1746	6.42	700	0.4359	0.4725
3.1746	7.34	800	0.4289	0.4485
3.1746	8.26	900	0.4121	0.4288
0.2282	9.17	1000	0.4249	0.4034
0.2282	10.09	1100	0.4106	0.3976
0.2282	11.01	1200	0.4099	0.3935
0.2282	11.93	1300	0.3970	0.3771
0.2282	12.84	1400	0.4037	0.3726
0.1043	13.76	1500	0.3953	0.3636
0.1043	14.68	1600	0.3917	0.3532

Framework versions

• Transformers 4.27.4
• Pytorch 2.0.0+cu117
• Datasets 2.10.1
• Tokenizers 0.13.2