Zarnabh/whisper-small-ps

Whisper small Ps - ZFA

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

• Loss: 0.7353
• Wer: 26.1787

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: 2e-05
• train_batch_size: 1
• eval_batch_size: 1
• seed: 42
• gradient_accumulation_steps: 8
• total_train_batch_size: 8
• optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED 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: 200
• training_steps: 1500
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
No log	2.8341	300	0.5501	32.4650
5.4773	5.6635	600	0.5894	28.7836
5.4773	8.4929	900	0.6551	27.0398
0.1933	11.3223	1200	0.7140	26.7600
0.0125	14.1517	1500	0.7353	26.1787

Framework versions

• Transformers 5.0.0
• Pytorch 2.9.0+cu126
• Datasets 4.0.0
• Tokenizers 0.22.2