Hanhpt23/whisper-small-engmed-free_E0-8D3-11

openai/whisper-small

This model is a fine-tuned version of openai/whisper-small on the pphuc25/EngMed dataset. It achieves the following results on the evaluation set:

• Loss: 0.0004
• Wer: 24.9056
• Cer: 18.2542

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: 0.0001
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 100
• num_epochs: 20

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
6.4241	1.0	386	5.6109	97.7210	90.8039
0.6455	2.0	772	0.3786	41.8786	38.6418
0.4335	3.0	1158	0.1989	26.9336	23.3054
0.222	4.0	1544	0.1169	31.2215	29.6213
0.1897	5.0	1930	0.0712	31.2511	29.8822
0.0838	6.0	2316	0.0452	34.0730	30.9224
0.0829	7.0	2702	0.0311	30.9565	27.2632
0.0779	8.0	3088	0.0245	27.2090	23.4638
0.0334	9.0	3474	0.0125	29.9746	25.5428
0.0253	10.0	3860	0.0094	27.1884	21.9223
0.0106	11.0	4246	0.0069	27.7738	22.0527
0.0132	12.0	4632	0.0042	24.7527	19.2122
0.0076	13.0	5018	0.0030	20.6164	15.5505
0.0045	14.0	5404	0.0019	29.2774	21.6112
0.0031	15.0	5790	0.0010	26.0214	19.1235
0.004	16.0	6176	0.0006	28.0375	20.4247
0.0009	17.0	6562	0.0008	26.2828	18.6873
0.0007	18.0	6948	0.0005	25.0950	18.6785
0.0011	19.0	7334	0.0004	24.9328	18.0978
0.0004	20.0	7720	0.0004	24.9056	18.2542

Framework versions

• Transformers 4.41.1
• Pytorch 2.3.0
• Datasets 2.19.1
• Tokenizers 0.19.1