anton-l/wav2vec2-xls-r-common_voice-tr-ft-100sh

wav2vec2-xls-r-common_voice-tr-ft

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: 0.5806
• Wer: 0.3998
• Cer: 0.1053

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
0.5369	17.0	500	0.6021	0.6366	0.1727
0.3542	34.0	1000	0.5265	0.4906	0.1278
0.1866	51.0	1500	0.5805	0.4768	0.1261
0.1674	68.01	2000	0.5336	0.4518	0.1186
0.19	86.0	2500	0.5676	0.4427	0.1151
0.0815	103.0	3000	0.5510	0.4268	0.1125
0.0545	120.0	3500	0.5608	0.4175	0.1099
0.0299	137.01	4000	0.5875	0.4222	0.1124
0.0267	155.0	4500	0.5882	0.4026	0.1063
0.025	172.0	5000	0.5806	0.3998	0.1053

Framework versions

• Transformers 4.17.0.dev0
• Pytorch 1.10.2
• Datasets 1.18.2
• Tokenizers 0.10.3