Modelpublic

Adnan666/whisper-small-pashto-run11-cv24only

sourceHugging Faceapache-2.0updated 25d agoView on Hugging Face

0likes3.3kdownloads

Model Card

whisper-small-pashto-run11-cv24only

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

—Loss: 1.0145
—Wer: 49.0943

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
—trainbatchsize: 8
—evalbatchsize: 8
—seed: 42
—gradientaccumulationsteps: 2
—totaltrainbatch_size: 16
—optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
—lrschedulertype: linear
—lrschedulerwarmup_steps: 500
—training_steps: 40000
—mixedprecisiontraining: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.616	0.008	500	1.0120	59.9066
0.4933	0.016	1000	0.9658	55.7610
0.4126	0.024	1500	0.9368	57.8338
0.3876	0.0320	2000	0.9310	53.5014
0.3611	0.0400	2500	0.9143	52.1382
0.3649	0.0480	3000	0.9002	55.3688
0.326	0.0560	3500	0.9249	54.0990
0.3154	0.0640	4000	0.9160	53.8002
0.3176	0.0720	4500	0.9117	53.8562
0.2979	0.0800	5000	0.9307	53.9309
0.2921	0.0880	5500	0.9434	53.2213
0.2852	0.0960	6000	0.9478	53.6881
0.2574	0.1040	6500	0.9969	53.2960
0.2651	0.1120	7000	0.9148	54.0616
0.2686	0.1200	7500	0.9226	53.4827
0.2364	0.1280	8000	0.8766	53.0345
0.2544	0.1360	8500	0.8801	53.7068
0.2214	0.1440	9000	0.9277	54.2857
0.2286	0.1520	9500	0.9146	51.9701
0.2185	0.1600	10000	0.8898	50.1587
0.2263	0.1680	10500	0.8894	49.3744
0.2168	0.1760	11000	0.9393	50.3641
0.2237	0.1840	11500	0.9196	49.5612
0.2098	0.1920	12000	0.9052	51.0738
0.1978	0.2000	12500	0.9277	51.2979
0.19	0.2080	13000	0.9295	50.0280
0.2016	0.2160	13500	0.8830	48.7395
0.2106	0.2240	14000	0.8858	48.0299
0.1939	0.2320	14500	0.9448	49.8226
0.1958	0.2400	15000	0.9274	49.9720
0.1847	0.2480	15500	0.8702	49.0383
0.183	0.2560	16000	0.8794	48.3287
0.1777	0.2640	16500	0.8888	49.3371
0.1677	0.2720	17000	0.9206	49.8599
0.1731	0.2800	17500	0.9222	49.1130
0.1732	0.2880	18000	0.9107	47.9365
0.1908	0.2960	18500	0.8938	48.2540
0.1698	0.3040	19000	0.8937	47.3763
0.1565	0.3120	19500	0.9680	49.7106
0.167	0.3200	20000	0.9013	48.4220
0.1603	0.3280	20500	0.9219	48.5341
0.1688	0.3360	21000	0.8954	47.3016
0.157	0.3440	21500	0.9330	49.2250
0.148	0.3520	22000	0.9746	49.8973
0.157	0.3600	22500	1.0004	50.3828
0.1537	0.3680	23000	0.9720	50.4388
0.1548	0.3760	23500	0.9244	47.8618
0.1468	0.3840	24000	0.9394	49.6172
0.1443	0.3920	24500	0.9698	50.8870
0.133	0.4000	25000	0.9557	48.9076
0.1467	0.4080	25500	0.9332	47.5444
0.1478	0.4160	26000	0.9275	48.6088
0.1321	0.4240	26500	0.9067	48.2540
0.137	0.4320	27000	0.9913	50.7376
0.1268	0.4400	27500	0.9884	49.2063
0.1295	0.4480	28000	0.9654	49.0756
0.1273	0.4560	28500	0.9429	48.9449
0.1261	0.4640	29000	0.9121	46.1064
0.1357	0.4720	29500	0.9397	48.0112
0.1278	0.4800	30000	0.9671	48.8142
0.1306	0.4880	30500	0.9840	48.9262
0.1129	0.4960	31000	0.9419	48.6835
0.1168	0.5040	31500	0.9440	48.6088
0.1032	0.5120	32000	0.9889	50.2521
0.1054	0.5200	32500	0.9784	50.9430
0.1128	0.5280	33000	0.9684	49.0383
0.1022	0.5360	33500	0.9649	47.9552
0.1077	0.5440	34000	0.9855	49.8413
0.0972	0.5520	34500	0.9734	50.1027
0.1086	0.5600	35000	0.9814	49.4118
0.1087	0.5680	35500	1.0114	50.1214
0.1135	0.5760	36000	1.0149	50.4762
0.1017	0.5840	36500	1.0049	49.0570
0.094	0.5920	37000	1.0377	50.5696
0.0955	0.6000	37500	1.0082	49.5985
0.098	0.6080	38000	1.0034	49.0943
0.0945	0.6160	38500	1.0227	49.5238
0.0915	0.6240	39000	1.0142	49.1503
0.0954	0.6320	39500	1.0133	48.8142
0.0964	0.6400	40000	1.0145	49.0943

Framework versions

—Transformers 4.44.2
—Pytorch 2.4.0+cu118
—Datasets 2.19.0
—Tokenizers 0.19.1