learning_rate: 2e-05
train_batch_size: 4
eval_batch_size: 4
seed: 42
gradient_accumulation_steps: 4
total_train_batch_size: 16
optimizer: Use OptimizerNames.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: 1.0
mixed_precision_training: Native AMP

Training Loss	Epoch	Step	Validation Loss	Token Acc
0.6411	0.1114	1000	0.6324	0.7970
0.5704	0.2228	2000	0.5586	0.8126
0.5336	0.3343	3000	0.5287	0.8214
0.5451	0.4457	4000	0.5127	0.8242
0.4762	0.5571	5000	0.4923	0.8321
0.5142	0.6685	6000	0.4815	0.8363
0.4554	0.7800	7000	0.4701	0.8402
0.4807	0.8914	8000	0.4643	0.8406
0.4881	1.0	8975	0.4729	0.8371

mthsmtt/Kvasir-VQA-x1-lora_260504-1543

Kvasir-VQA-x1-lora_260428-1429