whisper-small-kfn

This model is a fine-tuned version of openai/whisper-small on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.0000
• Wer: 3.6889

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: 0.0004
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 16
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 132
• num_epochs: 30
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
1.3799	1.0929	100	0.1761	28.4148
0.2765	2.1858	200	0.1491	30.0598
0.1829	3.2787	300	0.0823	17.1486
0.1281	4.3716	400	0.0763	20.0897
0.1134	5.4645	500	0.0662	13.1605
0.094	6.5574	600	0.0612	10.5184
0.079	7.6503	700	0.0671	23.4297
0.068	8.7432	800	0.0316	14.5563
0.0542	9.8361	900	0.0380	19.3420
0.0478	10.9290	1000	0.0665	15.4536
0.048	12.0219	1100	0.0144	42.9711
0.0363	13.1148	1200	0.0198	4.7358
0.0236	14.2077	1300	0.0144	8.8734
0.0223	15.3005	1400	0.0166	6.9292
0.0181	16.3934	1500	0.0124	5.2343
0.0138	17.4863	1600	0.0075	6.9292
0.01	18.5792	1700	0.0030	10.2193
0.0085	19.6721	1800	0.0154	7.7767
0.0056	20.7650	1900	0.0005	5.3838
0.0018	21.8579	2000	0.0002	4.5364
0.0002	22.9508	2100	0.0000	4.2373
0.0001	24.0437	2200	0.0000	3.9382
0.0	25.1366	2300	0.0000	3.9382
0.0	26.2295	2400	0.0000	3.8883
0.0	27.3224	2500	0.0000	3.6889
0.0	28.4153	2600	0.0000	3.6889
0.0	29.5082	2700	0.0000	3.6889

Framework versions

• Transformers 4.45.0.dev0
• Pytorch 2.4.0
• Datasets 2.21.0
• Tokenizers 0.19.1