learning_rate: 5e-05
train_batch_size: 16
eval_batch_size: 16
seed: 42
optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
lr_scheduler_type: constant
lr_scheduler_warmup_steps: 0.2
num_epochs: 15

Training Loss	Epoch	Step	Average Per	Validation Loss	Per Ghonna	Per Hams Or Jahr	Per Istitala	Per Itbaq	Per Phonemes	Per Qalqla	Per Safeer	Per Shidda Or Rakhawa	Per Tafashie	Per Tafkheem Or Taqeeq	Per Tikraar
0.8080	2.0	24262	0.0281	0.1197	0.0203	0.0223	0.0142	0.0133	0.0457	0.0269	0.0140	0.0351	0.0099	0.0723	0.0351
0.0908	4.0	48524	0.0187	0.0665	0.0118	0.0126	0.0088	0.0073	0.0256	0.0210	0.0083	0.0206	0.0057	0.0661	0.0177
0.0547	6.0	72786	0.0124	0.0442	0.0077	0.0083	0.0055	0.0050	0.0170	0.0198	0.0052	0.0134	0.0036	0.0388	0.0124
0.0390	8.0	97048	0.0112	0.0344	0.0065	0.0065	0.0054	0.0040	0.0139	0.0173	0.0046	0.0108	0.0031	0.0365	0.0150
0.0302	10.0	121310	0.0090	0.0310	0.0054	0.0054	0.0045	0.0033	0.0117	0.0133	0.0037	0.0092	0.0026	0.0284	0.0119
0.0249	12.0	145572	0.0236	0.0091	0.0042	0.0074	0.0355	0.0026	0.0031	0.0161	0.0147	0.0020	0.0046	0.0044	0.0094
0.0210	14.0	169834	0.0227	0.0087	0.0040	0.0068	0.0256	0.0024	0.0028	0.0123	0.0127	0.0018	0.0038	0.0038	0.0077
0.0210	15.0	181965	0.0203	0.0078	0.0035	0.0060	0.0451	0.0026	0.0029	0.0188	0.0160	0.0022	0.0050	0.0043	0.0104

AbdullahHedeya/h4_f1024_l16_d256_no_kd_15_epochs_checkpoints

h4_f1024_l16_d256_no_kd_ours_new