whisper-small-30-report-sentences

This model is a fine-tuned version of openai/whisper-small on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.2460
• Wer: 11.3636

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
• gradient_accumulation_steps: 2
• total_train_batch_size: 64
• optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 20
• training_steps: 300

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
5.8298	10.0	10	0.5348	4.5455
0.6646	20.0	20	0.1902	9.0909
0.0044	30.0	30	0.2114	4.5455
0.0005	40.0	40	0.2307	9.0909
0.0002	50.0	50	0.2398	9.0909
0.0002	60.0	60	0.2429	9.0909
0.0001	70.0	70	0.2439	9.0909
0.0001	80.0	80	0.2441	11.3636
0.0001	90.0	90	0.2442	11.3636
0.0001	100.0	100	0.2442	11.3636
0.0001	110.0	110	0.2442	11.3636
0.0001	120.0	120	0.2443	11.3636
0.0001	130.0	130	0.2444	11.3636
0.0001	140.0	140	0.2445	11.3636
0.0001	150.0	150	0.2447	11.3636
0.0001	160.0	160	0.2448	11.3636
0.0001	170.0	170	0.2449	11.3636
0.0001	180.0	180	0.2451	11.3636
0.0001	190.0	190	0.2452	11.3636
0.0001	200.0	200	0.2453	11.3636
0.0001	210.0	210	0.2455	11.3636
0.0001	220.0	220	0.2456	11.3636
0.0001	230.0	230	0.2457	11.3636
0.0001	240.0	240	0.2458	11.3636
0.0001	250.0	250	0.2458	11.3636
0.0000	260.0	260	0.2459	11.3636
0.0000	270.0	270	0.2459	11.3636
0.0000	280.0	280	0.2460	11.3636
0.0000	290.0	290	0.2460	11.3636
0.0000	300.0	300	0.2460	11.3636

Framework versions

• Transformers 5.2.0
• Pytorch 2.9.0+cu126
• Datasets 4.6.0
• Tokenizers 0.22.2