whisper-small-turkish-1

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

• Loss: 0.2199

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: 32
• eval_batch_size: 32
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine
• lr_scheduler_warmup_ratio: 0.1
• training_steps: 40000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss
0.295	0.05	2000	0.2957
0.2732	0.1	4000	0.2737
0.2334	0.15	6000	0.2588
0.2148	0.2	8000	0.2413
0.194	0.25	10000	0.2314
0.1816	0.3	12000	0.2264
0.1654	0.35	14000	0.2226
0.1537	0.4	16000	0.2174
0.1456	0.45	18000	0.2152
0.1366	0.5	20000	0.2139
0.1291	0.55	22000	0.2136
0.1186	0.6	24000	0.2109
0.1192	0.65	26000	0.2128
0.1056	0.7	28000	0.2129
0.0779	1.0228	30000	0.2182
0.0907	1.0728	32000	0.2144
0.0748	1.1228	34000	0.2189
0.0746	1.1728	36000	0.2199
0.0785	1.2228	38000	0.2206
0.0739	1.2728	40000	0.2199

Framework versions

• Transformers 4.45.2
• Pytorch 2.5.1+cu124
• Datasets 3.2.0
• Tokenizers 0.20.3