learning_rate: 0.0001
train_batch_size: 16
eval_batch_size: 16
seed: 42
gradient_accumulation_steps: 2
total_train_batch_size: 32
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: 2500

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
0.8768	0.8065	250	0.7848	0.5481	0.2873
0.7778	1.6129	500	0.7593	0.5299	0.2744
0.6779	2.4194	750	0.7492	0.5346	0.2761
0.631	3.2258	1000	0.7584	0.5326	0.2791
0.5983	4.0323	1250	0.7626	0.5579	0.2906
0.5382	4.8387	1500	0.7674	0.5495	0.2872
0.5291	5.6452	1750	0.7687	0.5450	0.2773
0.5007	6.4516	2000	0.7755	0.5478	0.2823
0.5065	7.2581	2250	0.7870	0.5530	0.2858
0.4862	8.0645	2500	0.7819	0.5514	0.2845

KasuleTrevor/cdli-whisper-ml-lug-swa-sunbird-lora

cdli-whisper-ml-lug-swa-sunbird-lora