whisper-tiny-turkish-0

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

• Loss: 0.3266

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: 64
• eval_batch_size: 64
• 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: 22500
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss
0.6204	0.0444	1000	0.6094
0.5242	0.0889	2000	0.5179
0.4348	0.1333	3000	0.4674
0.4048	0.1778	4000	0.4278
0.3708	0.2222	5000	0.4044
0.3489	0.2667	6000	0.3904
0.3276	0.3111	7000	0.3802
0.3164	0.3556	8000	0.3687
0.3036	0.4	9000	0.3586
0.2878	0.4444	10000	0.3548
0.2831	0.4889	11000	0.3476
0.2696	0.5333	12000	0.3411
0.2713	0.5778	13000	0.3383
0.2512	0.6222	14000	0.3356
0.2319	1.0202	15000	0.3329
0.2443	1.0647	16000	0.3288
0.2242	1.1091	17000	0.3299
0.2236	1.1536	18000	0.3291
0.2214	1.198	19000	0.3271
0.2208	1.2424	20000	0.3264
0.2132	1.2869	21000	0.3268
0.2187	1.3313	22000	0.3266

Framework versions

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