whisper-medium-SplitEndMovie-Sentences

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

• Loss: 0.7540
• Cer: 20.8488
• Wer: 32.7626

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: 5e-06
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• 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_steps: 1000
• num_epochs: 25
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Cer	Wer
2.559	1.0	1277	0.8053	35.8285	55.0781
0.9499	2.0	2554	0.7246	36.0860	54.7614
0.7357	3.0	3831	0.6841	34.7272	53.8555
0.6126	4.0	5108	0.6781	29.3859	46.3433
0.5156	5.0	6385	0.6672	30.6182	46.0451
0.4406	6.0	7662	0.6666	25.3170	39.2731
0.3765	7.0	8939	0.6689	27.9081	42.4179
0.3221	8.0	10216	0.6737	24.7555	37.9879
0.2758	9.0	11493	0.6827	24.3163	37.4466
0.2351	10.0	12770	0.6948	22.6004	35.6385
0.2023	11.0	14047	0.7071	21.7533	34.4970
0.1737	12.0	15324	0.7186	21.7414	34.1287
0.1502	13.0	16601	0.7267	22.0465	34.4528
0.1298	14.0	17878	0.7400	21.9167	33.8894
0.1126	15.0	19155	0.7540	20.8488	32.7626
0.0977	16.0	20432	0.7677	21.9437	33.9373
0.0862	17.0	21709	0.7844	21.2362	33.3186
0.0761	18.0	22986	0.7879	21.4320	33.2376
0.0682	19.0	24263	0.7964	20.8683	32.8767
0.0606	20.0	25540	0.8052	21.0014	32.8473

Framework versions

• Transformers 4.53.3
• Pytorch 2.7.1+cu118
• Datasets 3.6.0
• Tokenizers 0.21.2