learning_rate: 1e-05
train_batch_size: 4
eval_batch_size: 4
seed: 42
gradient_accumulation_steps: 8
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: constant_with_warmup
lr_scheduler_warmup_steps: 150
training_steps: 5000

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
0.5135	1.5445	250	0.7815	0.3083	0.2202
0.4341	3.0866	500	0.7594	0.2964	0.2108
0.3762	4.6311	750	0.7709	0.2986	0.2129
0.2942	6.1732	1000	0.7887	0.2968	0.2125
0.2888	7.7177	1250	0.7996	0.2984	0.2144
0.2544	9.2599	1500	0.8295	0.2948	0.2121
0.2429	10.8043	1750	0.8522	0.2992	0.2154
0.1883	12.3465	2000	0.8879	0.3013	0.2161
0.1878	13.8910	2250	0.9144	0.3011	0.2183
0.1709	15.4331	2500	0.9445	0.2997	0.2160
0.1521	16.9776	2750	0.9875	0.2991	0.2143
0.141	18.5197	3000	1.0080	0.3013	0.2165
0.1069	20.0619	3250	1.0838	0.3054	0.2173
0.1146	21.6063	3500	1.0732	0.3069	0.2196
0.0971	23.1485	3750	1.0980	0.3038	0.2180
0.0873	24.6930	4000	1.0909	0.3027	0.2169
0.0758	26.2351	4250	1.1415	0.3030	0.2164
0.075	27.7796	4500	1.1331	0.3044	0.2180
0.0592	29.3217	4750	1.1717	0.3040	0.2175
0.06	30.8662	5000	1.1884	0.3014	0.2162

KasuleTrevor/cdli-whisper-en-ug-openai-large-v3-full-a40-fixed-nogc-clean-5k

cdli-whisper-en-ug-openai-large-v3-full-a40-fixed-nogc-clean-5k