wav2vec2-urdu-multitask-teacher

This model is a fine-tuned version of abidanoaman/wav2vec2-urdu-finetuned-ASR on an unknown dataset. It achieves the following results on the evaluation set:

Loss: 1.8960
Wer: 0.4799
Emotion F1: 0.4457
Gender Accuracy: 0.976
Combined Score: 0.6345

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: 8
seed: 42
gradient_accumulation_steps: 4
total_train_batch_size: 32
optimizer: Use OptimizerNames.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: 100
num_epochs: 40
mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Emotion F1	Gender Accuracy	Combined Score
3.0143	0.6849	50	2.5861	0.6759	0.0579	0.0	0.1470
2.8679	1.3699	100	2.5833	0.6569	0.0740	0.0	0.1594
2.8197	2.0548	150	2.5138	0.6460	0.1111	0.132	0.2145
2.6281	2.7397	200	2.3898	0.6089	0.1193	0.548	0.3566
2.51	3.4247	250	2.2804	0.5986	0.1349	0.674	0.4032
2.3366	4.1096	300	2.2034	0.5949	0.1134	0.74	0.4181
2.3195	4.7945	350	2.0848	0.5777	0.1867	0.794	0.4631
2.2542	5.4795	400	2.0140	0.5851	0.1676	0.892	0.4838
2.1615	6.1644	450	1.9212	0.5709	0.2389	0.91	0.5163
2.1168	6.8493	500	1.8401	0.5526	0.2426	0.944	0.5349
2.1297	7.5342	550	1.8193	0.5589	0.2636	0.914	0.5297
1.9481	8.2192	600	1.7621	0.5474	0.2876	0.946	0.5511
1.9068	8.9041	650	1.7353	0.5458	0.3476	0.964	0.5751
1.9135	9.5890	700	1.6984	0.5410	0.3363	0.988	0.5809
1.9114	10.2740	750	1.7621	0.5282	0.3521	0.948	0.5788
1.8333	10.9589	800	1.7001	0.5321	0.3749	0.96	0.5876
1.8406	11.6438	850	1.6963	0.5319	0.3657	0.984	0.5922
1.7488	12.3288	900	1.6972	0.5203	0.3938	0.974	0.6022
1.8025	13.0137	950	1.7500	0.5258	0.3553	0.952	0.5819
1.7811	13.6986	1000	1.7270	0.5196	0.4060	0.938	0.5953
1.6412	14.3836	1050	1.7763	0.5192	0.3617	0.95	0.5858
1.7068	15.0685	1100	1.6392	0.5196	0.4197	0.996	0.6169
1.6609	15.7534	1150	1.6699	0.5026	0.4418	0.986	0.6273
1.5145	16.4384	1200	1.6642	0.5103	0.4449	0.974	0.6216
1.5172	17.1233	1250	1.6888	0.5087	0.4730	0.958	0.6258
1.6041	17.8082	1300	1.7920	0.5002	0.4468	0.902	0.6045
1.5741	18.4932	1350	1.7422	0.5059	0.4607	0.928	0.6143
1.5694	19.1781	1400	1.7018	0.5007	0.4153	0.996	0.6231
1.4951	19.8630	1450	1.7225	0.5031	0.4469	0.966	0.6226
1.4128	20.5479	1500	1.7504	0.5111	0.4497	0.94	0.6125
1.3993	21.2329	1550	1.7292	0.4987	0.4536	0.982	0.6312
1.3689	21.9178	1600	1.7107	0.4952	0.4585	0.98	0.6335
1.3883	22.6027	1650	1.8204	0.5031	0.4295	0.97	0.6186
1.388	23.2877	1700	1.7624	0.4943	0.4713	0.966	0.6335
1.39	23.9726	1750	1.7373	0.5061	0.4780	0.988	0.6373
1.3144	24.6575	1800	1.7601	0.4941	0.4641	0.974	0.6338
1.308	25.3425	1850	1.8468	0.4906	0.4780	0.924	0.6243
1.1624	26.0274	1900	1.8270	0.4998	0.4624	0.98	0.6328
1.2308	26.7123	1950	1.8317	0.4921	0.4510	0.98	0.6325
1.2833	27.3973	2000	1.8082	0.4967	0.4462	0.984	0.6304
1.1986	28.0822	2050	1.7801	0.4854	0.4778	0.982	0.6438
1.2073	28.7671	2100	1.8085	0.4887	0.4900	0.978	0.6449
1.2382	29.4521	2150	1.8145	0.4884	0.4820	0.98	0.6432
1.2236	30.1370	2200	1.8277	0.4895	0.4563	0.98	0.6351
1.2668	30.8219	2250	1.8910	0.4965	0.4216	0.978	0.6213
1.0703	31.5068	2300	1.8507	0.4876	0.4637	0.958	0.6315
1.1168	32.1918	2350	1.8766	0.4867	0.4572	0.952	0.6281
1.2443	32.8767	2400	1.8841	0.4869	0.4609	0.962	0.6321
1.1923	33.5616	2450	1.9608	0.4856	0.4247	0.966	0.6230
1.155	34.2466	2500	1.9239	0.4830	0.4256	0.96	0.6225
1.153	34.9315	2550	1.8888	0.4838	0.4558	0.97	0.6342
1.1316	35.6164	2600	1.9003	0.4819	0.4482	0.978	0.6351
1.193	36.3014	2650	1.9069	0.4806	0.4527	0.982	0.6382
1.1364	36.9863	2700	1.8753	0.4808	0.4662	0.98	0.6415
1.142	37.6712	2750	1.8751	0.4819	0.4643	0.98	0.6405
1.0806	38.3562	2800	1.8894	0.4795	0.4568	0.982	0.6398
1.2624	39.0411	2850	1.8954	0.4797	0.4433	0.976	0.6339
1.0891	39.7260	2900	1.8960	0.4799	0.4457	0.976	0.6345