xls-r-300m-nyagen-combined-hp-tuning-test-model

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

• Loss: 0.2689
• Wer: 0.2677

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: 0.0009268613959558291
• train_batch_size: 4
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 16
• total_train_batch_size: 64
• 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	2.2254	100	2.9001	1.0
No log	4.4507	200	0.8376	0.7849
No log	6.6761	300	0.3087	0.4330
No log	8.9014	400	0.2516	0.3765
30.7137	11.1127	500	0.2612	0.3434
30.7137	13.3380	600	0.2403	0.3230
30.7137	15.5634	700	0.2480	0.3156
30.7137	17.7887	800	0.2546	0.2939
30.7137	20.0	900	0.2638	0.2919
1.8035	22.2254	1000	0.2689	0.2675

Framework versions

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