whisper-large-v3-turbo-slovenian-slovenia-3-percent

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: 0.5694
• Model Preparation Time: 0.0065
• Wer Ortho: 31.9463
• Wer: 12.9700

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: 3
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Model Preparation Time	Wer Ortho	Wer
0.1125	0.0929	50	0.6541	0.0065	36.1317	18.4317
0.0982	0.1857	100	0.5796	0.0065	36.2471	18.4559
0.0897	0.2786	150	0.5838	0.0065	35.0929	17.4447
0.0879	0.3714	200	0.5543	0.0065	35.0322	17.3236
0.0815	0.4643	250	0.5700	0.0065	34.1575	16.5849
0.0778	0.5571	300	0.5816	0.0065	34.1028	16.3912
0.0765	0.6500	350	0.5661	0.0065	34.4733	16.4335
0.0737	0.7428	400	0.5970	0.0065	33.9266	15.4102
0.0742	0.8357	450	0.6086	0.0065	33.6229	15.4163
0.0715	0.9285	500	0.5717	0.0065	33.3678	14.8774
0.0525	1.0204	550	0.5980	0.0065	33.1005	14.6715
0.0514	1.1133	600	0.5748	0.0065	32.7481	14.3566
0.0489	1.2061	650	0.6079	0.0065	33.2949	14.6352
0.0502	1.2990	700	0.6025	0.0065	33.4224	14.7139
0.0487	1.3918	750	0.5779	0.0065	32.5902	13.9570
0.0503	1.4847	800	0.5803	0.0065	32.5598	13.9873
0.0504	1.5775	850	0.5921	0.0065	32.2926	13.8965
0.0492	1.6704	900	0.5524	0.0065	32.2622	13.6361
0.047	1.7632	950	0.5789	0.0065	32.5598	13.8662
0.0486	1.8561	1000	0.5798	0.0065	32.4930	13.9328
0.0491	1.9489	1050	0.5700	0.0065	32.5841	13.8359
0.034	2.0409	1100	0.6116	0.0065	32.4566	13.7390
0.0365	2.1337	1150	0.5827	0.0065	32.4930	13.6119
0.0353	2.2266	1200	0.5735	0.0065	32.2440	13.3394
0.0338	2.3194	1250	0.5869	0.0065	32.2683	13.3878
0.0353	2.4123	1300	0.5642	0.0065	32.1589	13.3636
0.0342	2.5051	1350	0.5871	0.0065	32.0131	13.2062
0.0347	2.5980	1400	0.5866	0.0065	32.1285	13.2365
0.0341	2.6908	1450	0.5689	0.0065	32.0678	13.2425
0.0357	2.7837	1500	0.5731	0.0065	31.9949	13.0911
0.0336	2.8765	1550	0.5742	0.0065	32.1650	13.1759
0.0351	2.9694	1600	0.5694	0.0065	31.9463	12.9700

Framework versions

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