whisper4

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

• Loss: 0.5409
• Wer: 28.2719

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: 128
• 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: 500
• training_steps: 300

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
3.8231	0.2778	10	3.7088	76.9377
3.1925	0.5556	20	2.9439	65.5654
2.1383	0.8333	30	1.7221	61.5311
1.0671	1.1111	40	0.8320	50.6989
0.6947	1.3889	50	0.6587	41.0102
0.6263	1.6667	60	0.5874	29.7967
0.5827	1.9444	70	0.5402	27.3825
0.4222	2.2222	80	0.5154	32.0521
0.4065	2.5	90	0.4997	25.6989
0.3959	2.7778	100	0.4804	23.8247
0.3081	3.0556	110	0.4670	24.8412
0.2497	3.3333	120	0.4687	23.2846
0.2535	3.6111	130	0.4594	23.0940
0.2428	3.8889	140	0.4545	23.5070
0.1627	4.1667	150	0.4651	24.4917
0.1224	4.4444	160	0.4686	23.6976
0.1326	4.7222	170	0.4653	23.6976
0.1334	5.0	180	0.4741	24.7459
0.0659	5.2778	190	0.4792	24.6823
0.0639	5.5556	200	0.4760	33.3863
0.0667	5.8333	210	0.4820	25.4765
0.042	6.1111	220	0.4933	29.4155
0.0325	6.3889	230	0.5066	29.9873
0.0333	6.6667	240	0.5126	26.0801
0.0333	6.9444	250	0.5073	24.6188
0.0187	7.2222	260	0.5129	27.3507
0.0214	7.5	270	0.5209	28.2084
0.0187	7.7778	280	0.5213	29.3202
0.0312	8.0556	290	0.5274	34.6569
0.0172	8.3333	300	0.5409	28.2719

Framework versions

• Transformers 4.40.1
• Pytorch 2.2.1+cu121
• Datasets 2.19.1.dev0
• Tokenizers 0.19.1