xls-r-1b-bigcgen-combined-15hrs-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: 0.5857
• Wer: 0.6274

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	0.1410	100	3.8893	1.0
No log	0.2821	200	2.6866	1.0
No log	0.4231	300	1.4204	1.0
No log	0.5642	400	0.8780	0.8847
5.5784	0.7052	500	0.8821	0.9583
5.5784	0.8463	600	0.6898	0.7509
5.5784	0.9873	700	0.6910	0.8690
5.5784	1.1283	800	0.6632	0.6810
5.5784	1.2694	900	0.6165	0.6048
1.2954	1.4104	1000	0.6006	0.6134
1.2954	1.5515	1100	0.6859	0.7684
1.2954	1.6925	1200	0.5857	0.6273
1.2954	1.8336	1300	0.6305	0.6155
1.2954	1.9746	1400	0.6213	0.5856
1.1582	2.1157	1500	0.5891	0.5984
1.1582	2.2567	1600	0.6606	0.7041

Framework versions

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