whisper-small-shingali-v1

This model is a fine-tuned version of samil24/whisper-small-kurmanji-v10 on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.2576
• Wer: 21.5760

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: 16
• eval_batch_size: 16
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 32
• 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: 100
• training_steps: 2000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.7441	3.2787	200	0.2691	28.8931
0.1799	6.5574	400	0.2095	22.1388
0.1293	9.8361	600	0.2086	22.5141
0.1065	13.1148	800	0.2129	20.0750
0.0907	16.3934	1000	0.2399	20.8255
0.0763	19.6721	1200	0.2567	21.9512
0.0710	22.9508	1400	0.2576	21.5760

Framework versions

• Transformers 5.13.0
• Pytorch 2.10.0+cu128
• Datasets 5.0.0
• Tokenizers 0.22.2