learning_rate: 1e-05
train_batch_size: 2
eval_batch_size: 8
seed: 42
gradient_accumulation_steps: 4
total_train_batch_size: 8
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: 400
training_steps: 2000
mixed_precision_training: Native AMP

Training Loss	Epoch	Step	Cer	Validation Loss	Wer
0.5105	0.7634	100	24.4591	0.8714	65.0016
0.5478	1.5267	200	25.0686	0.8271	63.9583
0.3876	2.2901	300	24.8154	0.8435	63.0651
0.3008	3.0534	400	25.4704	0.8605	65.7683
0.2037	3.8168	500	26.1163	0.9008	63.9346
0.0794	4.5802	600	24.9943	0.9336	64.2586
0.0775	5.3435	700	23.7965	0.9494	62.8280
0.042	6.1069	800	23.1597	0.9661	61.2393
0.023	6.8702	900	22.0347	1.0048	60.1881
0.0172	7.6336	1000	21.7997	1.0287	60.0537
0.0107	8.3969	1100	22.5790	1.0427	60.2988
0.0067	9.1603	1200	21.2220	1.0573	59.6032
0.0065	9.9237	1300	21.2387	1.0705	59.5953
0.0065	10.6870	1400	1.0844	59.2475	21.1356
0.0042	11.4504	1500	1.1009	59.8957	21.4434
0.0022	12.2137	1600	1.1045	59.4135	21.4101
0.0024	12.9771	1700	1.1133	59.8245	21.4358
0.0018	13.7405	1800	1.1205	59.5953	21.5162
0.0014	14.5038	1900	1.1239	59.8561	21.6072
0.0014	15.2672	2000	1.1247	59.7692	21.5313

khier12/120min_whisper_small_FT_1e_5

120min_whisper_small_FT_1e_5