learning_rate: 0.0003
train_batch_size: 8
eval_batch_size: 8
seed: 42
gradient_accumulation_steps: 2
total_train_batch_size: 16
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_ratio: 0.1
num_epochs: 15
mixed_precision_training: Native AMP

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
1668.8966	0.6944	800	1334.7141	1.0	0.7236
688.0655	1.3889	1600	550.3574	0.6826	0.2464
607.8135	2.0833	2400	490.8717	0.5944	0.2164
580.2456	2.7778	3200	465.0987	0.5503	0.2023
521.1551	3.4722	4000	458.1006	0.5283	0.2001
471.1814	4.1667	4800	452.3967	0.5102	0.1958
458.9287	4.8611	5600	428.9542	0.4913	0.1866
446.3736	5.5556	6400	423.3784	0.4809	0.1874
412.8066	6.25	7200	432.8220	0.4649	0.1832
394.0079	6.9444	8000	426.8859	0.4449	0.1766
402.78	7.6389	8800	414.7951	0.4431	0.1757
373.5881	8.3333	9600	421.6010	0.4541	0.1813
384.0306	9.0278	10400	418.7466	0.4320	0.1733
373.8867	9.7222	11200	436.5263	0.4216	0.1704
351.8441	10.4167	12000	432.9913	0.4211	0.1719
330.1864	11.1111	12800	444.9160	0.4136	0.1692
281.7933	11.8056	13600	443.5800	0.4095	0.1682
338.7603	12.5	14400	437.1237	0.4078	0.1691
320.0353	13.1944	15200	443.4001	0.4012	0.1666
284.5442	13.8889	16000	440.2469	0.4016	0.1660
276.7687	14.5833	16800	446.2755	0.3988	0.1659

nik1509/telugu_wav2vecmedium_basemodel

telugu_wav2vecmedium_basemodel