xlsr53

This model is a fine-tuned version of facebook/wav2vec2-large-xlsr-53 on the None dataset. It achieves the following results on the evaluation set:

• Loss: 3.6497
• Cer: 1.0

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: 8
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 16
• optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 10
• num_epochs: 30
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Cer
86.7857	1.0	10	13.0096	0.9930
64.0349	2.0	20	5.7889	1.0
20.8761	3.0	30	4.3652	1.0
12.3363	4.0	40	3.8773	1.0
11.0137	5.0	50	3.7471	1.0
10.7147	6.0	60	3.6982	1.0
10.5702	7.0	70	3.6885	1.0
10.1350	8.0	80	3.6772	1.0
10.7296	9.0	90	3.6753	1.0
10.0155	10.0	100	3.6693	1.0
23.2533	11.0	110	3.6945	1.0
11.0381	12.0	120	3.6739	1.0
10.8187	13.0	130	3.6736	1.0
10.4372	14.0	140	3.6692	1.0
10.3977	15.0	150	3.6721	1.0
10.1144	16.0	160	3.6679	1.0
10.2825	17.0	170	3.6669	1.0
10.4510	18.0	180	3.6651	1.0
10.1287	19.0	190	3.6654	1.0
10.1370	20.0	200	3.6608	1.0
10.3325	21.0	210	3.6639	1.0
10.5143	22.0	220	3.6630	1.0
10.3311	23.0	230	3.6666	1.0
10.4097	24.0	240	3.6611	1.0
13.4818	25.0	250	3.6602	1.0
13.6625	26.0	260	3.6586	1.0
10.1651	27.0	270	3.6568	1.0
10.3636	28.0	280	3.6508	1.0
13.1884	29.0	290	3.6498	1.0
9.9916	30.0	300	3.6497	1.0

Framework versions

• Transformers 5.0.0
• Pytorch 2.10.0+cu128
• Datasets 3.6.0
• Tokenizers 0.22.2