whisper-small-ken

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: 1.9940

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
1.3889	1.3889	100	0.2292	79.2622
0.2621	2.7778	200	0.1175	18.3450
0.1632	4.1667	300	0.0784	14.6062
0.1054	5.5556	400	0.0422	7.2782
0.0844	6.9444	500	0.0443	9.9202
0.0666	8.3333	600	0.0581	15.2044
0.0522	9.7222	700	0.0490	10.1695
0.0504	11.1111	800	0.0586	9.1725
0.0365	12.5	900	0.0386	12.7617
0.0336	13.8889	1000	0.0224	13.5593
0.0244	15.2778	1100	0.0138	7.3280
0.0177	16.6667	1200	0.0191	5.0349
0.0143	18.0556	1300	0.0050	5.7328
0.007	19.4444	1400	0.0014	1.9940
0.0018	20.8333	1500	0.0001	2.6421
0.0003	22.2222	1600	0.0000	1.9940
0.0	23.6111	1700	0.0000	1.9940
0.0	25.0	1800	0.0000	1.9940
0.0	26.3889	1900	0.0000	1.9940
0.0	27.7778	2000	0.0000	1.9940
0.0	29.1667	2100	0.0000	1.9940

Framework versions

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