xls-r-1b-bemgen-female-model

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

• Loss: 0.2246
• Wer: 0.3799

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
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
No log	0.3527	100	4.0290	1.0090
No log	0.7055	200	2.8473	1.0
No log	1.0564	300	0.6012	0.9269
No log	1.4092	400	0.4191	0.8378
4.9128	1.7619	500	0.3320	0.6424
4.9128	2.1129	600	0.2716	0.5178
4.9128	2.4656	700	0.2724	0.4929
4.9128	2.8183	800	0.2516	0.4788
4.9128	3.1693	900	0.2385	0.4438
0.4407	3.5220	1000	0.2374	0.4345
0.4407	3.8748	1100	0.2354	0.4097
0.4407	4.2257	1200	0.2205	0.3961
0.4407	4.5785	1300	0.2202	0.3897
0.4407	4.9312	1400	0.2246	0.3897
0.2698	5.2822	1500	0.2339	0.3666
0.2698	5.6349	1600	0.2358	0.3735
0.2698	5.9877	1700	0.2246	0.3807

Framework versions

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