whisper-large-v3-turbo-bulgarian-bulgaria

This model is a fine-tuned version of openai/whisper-large-v3-turbo on the fleurs dataset. It achieves the following results on the evaluation set:

• Loss: 0.4592
• Model Preparation Time: 0.0067
• Wer Ortho: 28.9047
• Wer: 9.9711

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 adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.06
• num_epochs: 1
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Model Preparation Time	Wer Ortho	Wer
0.1485	0.0299	32	0.5022	0.0067	31.5662	13.2030
0.1236	0.0598	64	0.4940	0.0067	31.6233	13.3865
0.1136	0.0897	96	0.4479	0.0067	31.6233	13.3865
0.1071	0.1196	128	0.4630	0.0067	31.0596	13.1536
0.1091	0.1495	160	0.4768	0.0067	31.1880	13.2030
0.1033	0.1794	192	0.4677	0.0067	30.3603	12.4903
0.0942	0.2092	224	0.4513	0.0067	30.6315	12.4832
0.097	0.2391	256	0.4636	0.0067	30.0250	12.0740
0.0955	0.2690	288	0.4449	0.0067	30.5102	12.6314
0.0981	0.2989	320	0.4494	0.0067	30.2747	12.1234
0.089	0.3288	352	0.4999	0.0067	30.5173	11.9258
0.0929	0.3587	384	0.5081	0.0067	30.4174	12.0387
0.0912	0.3886	416	0.4836	0.0067	30.3175	11.5094
0.0856	0.4185	448	0.4557	0.0067	29.6183	11.4247
0.0829	0.4484	480	0.4551	0.0067	29.9394	11.7987
0.0815	0.4783	512	0.4763	0.0067	30.5316	12.8078
0.0874	0.5082	544	0.4603	0.0067	29.5326	11.3824
0.0925	0.5381	576	0.4710	0.0067	29.6682	10.9237
0.0825	0.5680	608	0.4560	0.0067	29.5398	10.7685
0.0819	0.5979	640	0.4704	0.0067	29.4328	10.7614
0.0787	0.6277	672	0.4803	0.0067	29.6539	10.8955
0.0799	0.6576	704	0.4460	0.0067	29.1616	10.7826
0.0812	0.6875	736	0.4476	0.0067	29.1545	10.7261
0.0767	0.7174	768	0.4588	0.0067	29.3471	10.9167
0.082	0.7473	800	0.4658	0.0067	29.2615	10.7332
0.0836	0.7772	832	0.4640	0.0067	29.1188	10.4650
0.0848	0.8071	864	0.4474	0.0067	28.9119	10.2816
0.0813	0.8370	896	0.4662	0.0067	29.1117	10.3451
0.0779	0.8669	928	0.4570	0.0067	28.9975	10.2039
0.0727	0.8968	960	0.4621	0.0067	29.0617	10.2110
0.0728	0.9267	992	0.4592	0.0067	28.9047	9.9711
0.0704	0.9566	1024	0.4536	0.0067	28.9404	10.0064
0.076	0.9865	1056	0.4546	0.0067	28.9404	10.1263

Framework versions

• Transformers 4.51.3
• Pytorch 2.5.1
• Datasets 3.6.0
• Tokenizers 0.21.1