ASR_Whisper_Stroke

This model is a fine-tuned version of openai/whisper-small on the ASR_Preprocess_Stroke_Dataset dataset. It achieves the following results on the evaluation set:

• Loss: 0.2559
• Cer: 14.5144
• Wer: 19.5868

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: 8
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 16
• optimizer: Use OptimizerNames.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: 700
• training_steps: 7000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Cer	Wer
0.2408	1.1390	1000	0.3222	61.9392	59.4850
0.0986	2.2779	2000	0.2687	40.5623	49.2289
0.0566	3.4169	3000	0.2586	39.9808	45.3462
0.03	4.5558	4000	0.2521	21.4272	28.0574
0.0157	5.6948	5000	0.2559	15.1623	19.7908
0.0072	6.8337	6000	0.2517	16.4368	23.1233
0.0048	7.9727	7000	0.2559	14.5144	19.5868

Framework versions

• Transformers 4.53.0.dev0
• Pytorch 2.6.0+cu124
• Datasets 3.6.0
• Tokenizers 0.21.1