finetune_add_transformer

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: 0.2934
• Cer: 0.0724

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: 3e-05
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 4
• 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: 2000
• training_steps: 10000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Cer
4.2464	5.95	500	0.7569	0.2284
0.5463	11.9	1000	0.3276	0.0814
0.3545	17.86	1500	0.3084	0.0779
0.2841	23.81	2000	0.3200	0.0756
0.232	29.76	2500	0.3181	0.0735
0.1922	35.71	3000	0.3480	0.0731
0.1601	41.67	3500	0.3990	0.0742
0.1362	47.62	4000	0.4304	0.0736
0.1165	53.57	4500	0.4847	0.0746
0.0994	59.52	5000	0.5250	0.0761
0.0876	65.48	5500	0.5628	0.0740
0.0791	71.43	6000	0.5871	0.0742
0.0716	77.38	6500	0.5933	0.0729
0.0661	83.33	7000	0.6238	0.0739
0.0605	89.29	7500	0.6623	0.0742
0.0569	95.24	8000	0.6638	0.0729
0.0535	101.19	8500	0.6681	0.0730
0.0498	107.14	9000	0.6815	0.0733
0.0491	113.1	9500	0.6818	0.0733
0.0471	119.05	10000	0.6810	0.0732

Framework versions

• Transformers 4.17.0
• Pytorch 2.4.0
• Datasets 3.0.1
• Tokenizers 0.20.0