whisper-small-CV_Fleurs_AMMI_ALFFA-sw-20hrs-v1

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: 0.7663
• Wer: 0.2445
• Cer: 0.0922

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-05
• train_batch_size: 8
• eval_batch_size: 4
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 16
• optimizer: Use adamw_hf with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 100
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
3.059	0.9993	752	0.7311	0.5301	0.2009
1.0691	2.0	1505	0.5315	0.3752	0.1488
0.6341	2.9993	2257	0.4776	0.2844	0.0992
0.3826	4.0	3010	0.4720	0.2725	0.1037
0.2525	4.9993	3762	0.4979	0.2961	0.1204
0.1966	6.0	4515	0.5199	0.2599	0.0924
0.1748	6.9993	5267	0.5406	0.3148	0.1430
0.1642	8.0	6020	0.5674	0.2654	0.0958
0.1639	8.9993	6772	0.5802	0.2621	0.1186
0.1601	10.0	7525	0.6058	0.2652	0.1026
0.155	10.9993	8277	0.6102	0.2733	0.1036
0.1289	12.0	9030	0.6181	0.2613	0.0996
0.1105	12.9993	9782	0.6233	0.2615	0.0984
0.0981	14.0	10535	0.6377	0.2493	0.0931
0.089	14.9993	11287	0.6352	0.2612	0.1007
0.0761	16.0	12040	0.6494	0.2518	0.0972
0.0722	16.9993	12792	0.6665	0.2470	0.0917
0.063	18.0	13545	0.6687	0.2428	0.0898
0.0584	18.9993	14297	0.6715	0.2550	0.0984
0.0551	20.0	15050	0.6856	0.2506	0.0935
0.0503	20.9993	15802	0.6928	0.2480	0.0969
0.0464	22.0	16555	0.6887	0.2432	0.0913
0.0426	22.9993	17307	0.7118	0.2457	0.0925
0.0376	24.0	18060	0.7240	0.2357	0.0882
0.0376	24.9993	18812	0.7268	0.2458	0.0946
0.0339	26.0	19565	0.7335	0.2492	0.0931
0.0313	26.9993	20317	0.7185	0.2419	0.0908
0.0322	28.0	21070	0.7345	0.2396	0.0919
0.0313	28.9993	21822	0.7401	0.2432	0.0937
0.0268	30.0	22575	0.7576	0.2474	0.0946
0.0267	30.9993	23327	0.7653	0.2432	0.0938
0.025	32.0	24080	0.7593	0.2432	0.0940
0.0241	32.9993	24832	0.7670	0.2443	0.0930
0.0238	34.0	25585	0.7663	0.2445	0.0922

Framework versions

• Transformers 4.46.1
• Pytorch 2.1.0+cu118
• Datasets 3.1.0
• Tokenizers 0.20.1