learning_rate: 0.0002
train_batch_size: 4
eval_batch_size: 4
seed: 42
gradient_accumulation_steps: 4
total_train_batch_size: 16
optimizer: Use adamw_torch_fused with betas=(0.9,0.95) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
lr_scheduler_type: linear
lr_scheduler_warmup_steps: 0.03
num_epochs: 2.0
mixed_precision_training: Native AMP

Training Loss	Epoch	Step	Validation Loss	Token Acc
0.6074	0.1114	1000	0.6064	0.8035
0.5380	0.2228	2000	0.5472	0.8191
0.5199	0.3343	3000	0.5312	0.8215
0.5344	0.4457	4000	0.5069	0.8283
0.4680	0.5571	5000	0.4901	0.8312
0.4942	0.6685	6000	0.4727	0.8360
0.4518	0.7800	7000	0.4644	0.8396
0.4598	0.8914	8000	0.4593	0.8425
0.4164	1.0028	9000	0.4526	0.8438
0.4103	1.1142	10000	0.4401	0.8479
0.3861	1.2256	11000	0.4390	0.8485
0.3911	1.3371	12000	0.4342	0.8513
0.4091	1.4485	13000	0.4259	0.8492
0.3802	1.5599	14000	0.4209	0.8530
0.3847	1.6713	15000	0.4190	0.8545
0.3460	1.7828	16000	0.4133	0.8554
0.3598	1.8942	17000	0.4109	0.8563
0.3611	2.0	17950	0.4089	0.8573

me1adi/Kvasir-VQA-x1-lora_260823-1829

Kvasir-VQA-x1-lora_260823-1829