pashto-voice-modal

This model is a fine-tuned version of openai/whisper-small on the None dataset. It achieves the following results on the evaluation set:

Model description

More information needed

More information needed

More information needed

The following hyperparameters were used during training:

learning_rate: 1e-05
train_batch_size: 16
eval_batch_size: 16
seed: 42
gradient_accumulation_steps: 2
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: 500
num_epochs: 10
mixed_precision_training: Native AMP

Training Loss	Epoch	Step	Validation Loss	Wer
0.4989	3.3333	500	0.4834	0.4674
0.0944	6.6667	1000	0.5559	0.4611
0.0166	10.0	1500	0.6398	0.4512
0.6217	1.2797	2000	0.4274	0.4084
0.5870	1.5997	2500	0.3868	0.3689
0.5616	1.9197	3000	0.3637	0.3480
0.3566	2.2394	3500	0.3656	0.3371
0.3577	2.5594	4000	0.3569	0.3353
0.3836	2.8794	4500	0.3460	0.3294
0.2361	3.1990	5000	0.3615	0.3264
0.2615	3.5190	5500	0.3611	0.3269
0.2261	3.8390	6000	0.3613	0.3163
0.1367	4.1587	6500	0.3863	0.3313
0.1311	4.4787	7000	0.3953	0.3196
0.1461	4.7987	7500	0.3938	0.3207
0.0687	5.1184	8000	0.4159	0.3188
0.0712	5.4384	8500	0.4247	0.3218
0.0637	5.7584	9000	0.4341	0.3270
0.0272	6.0781	9500	0.4626	0.3214
0.0265	6.3981	10000	0.4743	0.3249
0.0286	6.7181	10500	0.4840	0.3283
0.0130	7.0378	11000	0.5048	0.3252
0.0140	7.3578	11500	0.5258	0.3232
0.0149	7.6778	12000	0.5335	0.3239
0.0099	7.9978	12500	0.5436	0.3253
0.0060	8.3174	13000	0.5647	0.3247
0.0042	8.6374	13500	0.5766	0.3242
0.0036	8.9574	14000	0.5812	0.3237
0.0030	9.2771	14500	0.5961	0.3236
0.0020	9.5971	15000	0.6007	0.3228
0.0028	9.9171	15500	0.6031	0.3232
0.0033	10.0	15630	0.6031	0.3234