learning_rate: 5e-05
train_batch_size: 4
eval_batch_size: 8
seed: 42
gradient_accumulation_steps: 4
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: 150
training_steps: 3000

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
0.7849	0.2651	250	0.8365	0.4657	0.2714
0.6954	0.5302	500	0.8200	0.4707	0.2849
0.6341	0.7953	750	0.7733	0.4279	0.2466
0.4331	1.0604	1000	0.7795	0.3988	0.2189
0.4099	1.3256	1250	0.7619	0.4037	0.2241
0.4453	1.5907	1500	0.7493	0.3886	0.2108
0.4158	1.8558	1750	0.7372	0.3948	0.2178
0.2548	2.1209	2000	0.7819	0.3904	0.2111
0.2695	2.3860	2250	0.7741	0.3886	0.2076
0.2658	2.6511	2500	0.7699	0.3832	0.2052
0.2761	2.9162	2750	0.7670	0.3812	0.2046
0.2514	3.1813	3000	0.7709	0.3797	0.2041

KasuleTrevor/cdli-whisper-ml-ugeng-lug-swa-sunbird-full-a40-lr5e5

cdli-whisper-ml-ugeng-lug-swa-sunbird-full-a40-lr5e5