Modelpublic

alpcansoydas/whisper-large-v2-tr-ft-12-04-26-encoder-only-100ksamples-simulated-data

sourceHugging Faceapache-2.0updated 6mo agoView on Hugging Face

0likes13downloads

Model Card

Whisper for Turkish Call Centers

This model is a fine-tuned version of openai/whisper-large-v2 on the Custom turkish call center simulated data dataset. It achieves the following results on the evaluation set:

—Loss: 0.2557

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
—trainbatchsize: 32
—evalbatchsize: 16
—seed: 42
—optimizer: Use OptimizerNames.ADAMWTORCHFUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
—lrschedulertype: linear
—lrschedulerwarmup_steps: 500
—training_steps: 3200
—mixedprecisiontraining: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss
0.2784	0.0889	250	0.2990
0.3138	0.1777	500	0.2876
0.2617	0.2666	750	0.2801
0.2865	0.3555	1000	0.2761
0.2546	0.4444	1250	0.2710
0.2946	0.5332	1500	0.2692
0.2848	0.6221	1750	0.2653
0.2591	0.7110	2000	0.2629
0.2685	0.7999	2250	0.2604
0.2547	0.8887	2500	0.2586
0.2350	0.9776	2750	0.2568
0.2371	1.0665	3000	0.2560
0.2368	1.1376	3200	0.2557

Framework versions

—Transformers 5.5.3
—Pytorch 2.10.0+cu128
—Datasets 4.8.4
—Tokenizers 0.22.2