xls-r-1b-bigcgen-male-5hrs-model

This model is a fine-tuned version of facebook/wav2vec2-xls-r-1b on the BIGCGEN - NA dataset. It achieves the following results on the evaluation set:

• Loss: 0.5146
• Wer: 0.5626

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: 4
• eval_batch_size: 4
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 8
• optimizer: Use 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: 30.0

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
No log	0.4219	100	4.0011	1.0
No log	0.8439	200	2.7752	1.0
No log	1.2658	300	0.8068	0.8170
No log	1.6878	400	0.7022	0.7646
5.5827	2.1097	500	0.5850	0.6982
5.5827	2.5316	600	0.5745	0.6058
5.5827	2.9536	700	0.5982	0.5846
5.5827	3.3755	800	0.5711	0.6152
5.5827	3.7975	900	0.5385	0.5780
1.0918	4.2194	1000	0.5452	0.6315
1.0918	4.6414	1100	0.5541	0.5615
1.0918	5.0633	1200	0.5146	0.5631
1.0918	5.4852	1300	0.5406	0.5766
1.0918	5.9072	1400	0.5415	0.5562
0.7578	6.3291	1500	0.5689	0.5521
0.7578	6.7511	1600	0.5789	0.5645

Framework versions

• Transformers 4.47.1
• Pytorch 2.5.1+cu124
• Datasets 3.2.0
• Tokenizers 0.21.0