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: 0.5335
• Wer: 0.4939

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: 0.0002
• train_batch_size: 32
• eval_batch_size: 16
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 64
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• num_epochs: 20.0
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
No log	1.83	100	4.1084	1.0
No log	3.67	200	3.1519	1.0
No log	5.5	300	1.9348	0.9799
No log	7.34	400	0.7185	0.7490
3.6165	9.17	500	0.6041	0.6368
3.6165	11.01	600	0.5610	0.5771
3.6165	12.84	700	0.5292	0.5398
3.6165	14.68	800	0.5242	0.5083
3.6165	16.51	900	0.5443	0.5037
0.1894	18.35	1000	0.5314	0.4944

Framework versions

• Transformers 4.29.2
• Pytorch 2.0.1
• Datasets 2.13.1
• Tokenizers 0.13.2