whisper-small-santali-sanlish-1934_lr2-lr2e5

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: 2e-05
train_batch_size: 8
eval_batch_size: 8
seed: 42
gradient_accumulation_steps: 2
total_train_batch_size: 16
optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED 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: 153
num_epochs: 25
mixed_precision_training: Native AMP

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
3.5011	1.0	99	0.7130	45.7772	10.6958
1.0163	2.0	198	0.5548	40.5141	9.2481
0.4113	3.0	297	0.5753	39.6573	9.2697
0.1843	4.0	396	0.6382	38.0661	8.5134
0.0998	5.0	495	0.6270	36.2301	8.1677
0.0677	6.0	594	0.6555	35.4957	7.9300
0.0511	7.0	693	0.6209	33.2925	7.5627
0.0339	8.0	792	0.6672	34.6389	7.8220
0.0233	9.0	891	0.6768	35.1285	7.8652
0.0107	10.0	990	0.6913	33.0477	7.3898
0.0035	11.0	1089	0.7290	33.9045	7.5194
0.0046	12.0	1188	0.7209	32.5581	7.2169
0.0010	13.0	1287	0.7212	32.0685	7.1953
0.0005	14.0	1386	0.7292	31.9461	7.0657
0.0004	15.0	1485	0.7344	32.1909	7.1089
0.0004	16.0	1584	0.7382	31.8237	7.0441
0.0004	17.0	1683	0.7414	31.8237	7.0441
0.0003	18.0	1782	0.7441	31.9461	7.0657
0.0003	19.0	1881	0.7463	31.7013	7.0225
0.0003	20.0	1980	0.7486	31.8237	7.0441
0.0003	21.0	2079	0.7504	31.8237	7.0657
0.0003	22.0	2178	0.7518	31.9461	7.0873
0.0003	23.0	2277	0.7528	31.8237	7.0657
0.0003	24.0	2376	0.7534	31.9461	7.0873
0.0002	25.0	2475	0.7537	32.1909	7.1305