learning_rate: 8.555e-05
train_batch_size: 2
eval_batch_size: 8
seed: 42
distributed_type: multi-GPU
num_devices: 2
gradient_accumulation_steps: 8
total_train_batch_size: 32
total_eval_batch_size: 16
optimizer: Use OptimizerNames.PAGED_ADAMW with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
lr_scheduler_type: cosine
lr_scheduler_warmup_steps: 3
num_epochs: 2
mixed_precision_training: Native AMP

Training Loss	Epoch	Step	Validation Loss
0.7501	0.1006	50	0.6591
0.6413	0.2013	100	0.6200
0.6040	0.3019	150	0.6030
0.6158	0.4025	200	0.5909
0.5696	0.5031	250	0.5848
0.5841	0.6038	300	0.5780
0.5707	0.7044	350	0.5715
0.5729	0.8050	400	0.5650
0.5629	0.9057	450	0.5626
0.5403	1.0060	500	0.5587
0.5152	1.1067	550	0.5600
0.5616	1.2073	600	0.5585
0.5029	1.3079	650	0.5563
0.5153	1.4086	700	0.5540
0.4982	1.5092	750	0.5531
0.5019	1.6098	800	0.5528
0.4818	1.7104	850	0.5517
0.5055	1.8111	900	0.5504
0.4778	1.9117	950	0.5500

rohandhakal/llama2-7b-my-finetune

llama2_finetuning