Tejveer12/Indian-Accent-English-Whisper-Finetuned-epoch-15

Whisper Indian English Acccent

This model is a fine-tuned version of openai/whisper-large-v3-turbo on the Indian English Accent dataset. It achieves the following results on the evaluation set:

• Loss: 0.3230
• Wer: 7.9852

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: 1e-05
• train_batch_size: 16
• eval_batch_size: 8
• seed: 42
• optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• num_epochs: 15
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.0435	0.1943	1000	0.2509	9.7582
0.0449	0.3885	2000	0.2618	10.1957
0.0461	0.5828	3000	0.2575	10.0569
0.0586	0.7770	4000	0.2512	9.6972
0.0565	0.9713	5000	0.2559	9.7088
0.0375	1.1655	6000	0.2611	10.0012
0.0374	1.3598	7000	0.2623	9.9065
0.0405	1.5540	8000	0.2592	9.8424
0.0392	1.7483	9000	0.2551	9.6941
0.0473	1.9425	10000	0.2648	9.5111
0.0264	2.1368	11000	0.2720	9.6941
0.0292	2.3310	12000	0.2633	9.6373
0.0276	2.5253	13000	0.2666	9.5700
0.028	2.7195	14000	0.2687	9.5027
0.0252	2.9138	15000	0.2643	9.1788
0.0202	3.1080	16000	0.2715	9.1157
0.0195	3.3023	17000	0.2729	9.2766
0.0277	3.4965	18000	0.2733	9.3302
0.0196	3.6908	19000	0.2693	9.1504
0.017	3.8850	20000	0.2724	9.1619
0.0124	4.0793	21000	0.2786	9.0389
0.0191	4.2735	22000	0.2756	9.0168
0.0143	4.4678	23000	0.2796	9.0315
0.0135	4.6620	24000	0.2819	9.0620
0.0191	4.8563	25000	0.2761	8.9600
0.0071	5.0505	26000	0.2804	8.9085
0.0097	5.2448	27000	0.2842	8.7203
0.0088	5.4390	28000	0.2853	8.7318
0.0127	5.6333	29000	0.2833	8.8517
0.0125	5.8275	30000	0.2894	8.7918
0.004	6.0218	31000	0.2860	8.5604
0.0062	6.2160	32000	0.2881	8.5846
0.0074	6.4103	33000	0.2933	8.8212
0.0073	6.6045	34000	0.2964	8.6298
0.0079	6.7988	35000	0.2883	8.6698
0.0066	6.9930	36000	0.2887	8.5720
0.0051	7.1873	37000	0.2986	8.5593
0.004	7.3815	38000	0.2943	8.5583
0.0077	7.5758	39000	0.2885	8.5152
0.0075	7.7700	40000	0.3002	8.7623
0.0053	7.9643	41000	0.2955	8.5215
0.0021	8.1585	42000	0.2976	8.5057
0.0019	8.3528	43000	0.3001	8.5699
0.0076	8.5470	44000	0.3072	8.6582
0.0028	8.7413	45000	0.3007	8.6109
0.0027	8.9355	46000	0.2987	8.4237
0.0018	9.1298	47000	0.2980	8.4721
0.0022	9.3240	48000	0.3054	8.5068
0.0014	9.5183	49000	0.3022	8.4310
0.0023	9.7125	50000	0.3074	8.5352
0.0018	9.9068	51000	0.3072	8.4626
0.0004	10.1010	52000	0.3057	8.2386
0.001	10.2953	53000	0.3090	8.3595
0.003	10.4895	54000	0.3096	8.4216
0.0019	10.6838	55000	0.3011	8.2523
0.0007	10.8780	56000	0.3080	8.2765
0.0004	11.0723	57000	0.3098	8.2460
0.0006	11.2665	58000	0.3109	8.2775
0.0004	11.4608	59000	0.3099	8.2933
0.0007	11.6550	60000	0.3161	8.2996
0.0003	11.8493	61000	0.3107	8.1671
0.0002	12.0435	62000	0.3111	8.2628
0.0002	12.2378	63000	0.3133	8.2428
0.0002	12.4320	64000	0.3132	8.1324
0.0004	12.6263	65000	0.3173	8.1986
0.0002	12.8205	66000	0.3167	8.1313
0.0002	13.0148	67000	0.3151	8.1576
0.0001	13.2090	68000	0.3166	8.0304
0.0001	13.4033	69000	0.3177	8.0167
0.0002	13.5975	70000	0.3187	8.0514
0.0001	13.7918	71000	0.3161	8.0546
0.001	13.9860	72000	0.3180	8.0956
0.0001	14.1803	73000	0.3242	8.0367
0.0001	14.3745	74000	0.3217	7.9978
0.0001	14.5688	75000	0.3222	7.9999
0.0001	14.7630	76000	0.3231	7.9862
0.0001	14.9573	77000	0.3230	7.9852

Framework versions

• Transformers 4.49.0
• Pytorch 2.2.0a0+81ea7a4
• Datasets 3.3.2
• Tokenizers 0.21.0