new_model_TTS

This model is a fine-tuned version of facebook/wav2vec2-xls-r-300m on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.2690
• Cer: 0.1077

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: 32
• eval_batch_size: 16
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 20
• training_steps: 2000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Cer
11.0198	0.1218	100	3.8782	1.0
3.7091	0.2436	200	3.7613	1.0
3.6421	0.3654	300	3.7285	1.0
3.6077	0.4872	400	3.5062	1.0
3.4196	0.6090	500	3.3978	0.9436
3.1641	0.7308	600	2.7949	0.8838
2.4817	0.8526	700	1.7425	0.5752
1.6965	0.9744	800	1.0554	0.3422
1.2581	1.0962	900	0.6237	0.1612
1.0314	1.2180	1000	0.5235	0.1448
0.9251	1.3398	1100	0.4387	0.1395
0.8336	1.4616	1200	0.4046	0.1307
0.7885	1.5834	1300	0.3705	0.1331
0.7256	1.7052	1400	0.3563	0.1087
0.709	1.8270	1500	0.3363	0.1428
0.6809	1.9488	1600	0.3244	0.1225
0.6302	2.0706	1700	0.3114	0.1275
0.6304	2.1924	1800	0.3078	0.1165
0.6196	2.3143	1900	0.3005	0.1102
0.6056	2.4361	2000	0.2986	0.1145

Framework versions

• Transformers 4.41.0
• Pytorch 2.10.0+cu128
• Datasets 4.8.5
• Tokenizers 0.19.1