wav2vec2-xls-r-300m-5e-sw-asr

This model is a fine-tuned version of facebook/wav2vec2-xls-r-300m on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.2786
• Wer: 0.3503

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: 8
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 32
• 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: cosine
• lr_scheduler_warmup_ratio: 0.03
• num_epochs: 5
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
2.5958	0.5481	800	0.5079	0.5467
0.4353	1.0959	1600	0.3661	0.4409
0.3382	1.6441	2400	0.3355	0.4133
0.2927	2.1918	3200	0.3111	0.3885
0.2647	2.7400	4000	0.2925	0.3700
0.2381	3.2878	4800	0.2874	0.3629
0.2242	3.8359	5600	0.2812	0.3539
0.2108	4.3837	6400	0.2795	0.3511
0.2092	4.9318	7200	0.2786	0.3503

Framework versions

• Transformers 4.57.3
• Pytorch 2.9.0+cu126
• Datasets 4.4.2
• Tokenizers 0.22.1