xls-r-base-300m-toratan-240

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: 0.0093
• Cer: 0.0027

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: 16
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 32
• optimizer: Use OptimizerNames.ADAMW_TORCH 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: 500
• num_epochs: 30
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Cer
4.1579	2.0833	400	1.6854	0.4430
1.5758	4.1667	800	1.0236	0.2938
1.2033	6.25	1200	0.7921	0.2512
0.9625	8.3333	1600	0.5875	0.1755
0.8087	10.4167	2000	0.4633	0.1577
0.6701	12.5	2400	0.3454	0.1186
0.5412	14.5833	2800	0.2316	0.0844
0.4432	16.6667	3200	0.1534	0.0565
0.3654	18.75	3600	0.0914	0.0379
0.288	20.8333	4000	0.0501	0.0194
0.2336	22.9167	4400	0.0333	0.0117
0.2041	25.0	4800	0.0197	0.0069
0.1659	27.0833	5200	0.0108	0.0033
0.1436	29.1667	5600	0.0093	0.0027

Framework versions

• Transformers 4.52.0.dev0
• Pytorch 2.6.0+cu124
• Datasets 3.5.1
• Tokenizers 0.21.1