whisper-small-dari

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

• Loss: 0.3171
• Wer: 30.1313

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: 1e-05
• train_batch_size: 64
• eval_batch_size: 64
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 128
• 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: 0.1
• num_epochs: 10

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
No log	1.0	7	0.6818	48.0304
3.2896	2.0	14	0.4796	38.9772
1.7258	3.0	21	0.3972	35.3836
1.7258	4.0	28	0.3604	32.1355
1.1070	5.0	35	0.3359	30.6151
0.8097	6.0	42	0.3274	31.2370
0.8097	7.0	49	0.3196	29.9931
0.6274	8.0	56	0.3194	29.9240
0.5292	9.0	63	0.3163	29.5784
0.4368	10.0	70	0.3171	30.1313

Framework versions

• Transformers 5.14.1
• Pytorch 2.10.0+cu128
• Datasets 5.0.1
• Tokenizers 0.22.2