learning_rate: 0.0002
train_batch_size: 1
eval_batch_size: 1
seed: 42
distributed_type: multi-GPU
num_devices: 2
gradient_accumulation_steps: 8
total_train_batch_size: 16
total_eval_batch_size: 2
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: cosine
lr_scheduler_warmup_steps: 0.1
num_epochs: 3.0
mixed_precision_training: Native AMP

Training Loss	Epoch	Step	Validation Loss
0.2388	0.1560	200	0.2548
0.2068	0.3121	400	0.2235
0.1739	0.4681	600	0.2106
0.1476	0.6241	800	0.2031
0.1615	0.7802	1000	0.1930
0.1282	0.9362	1200	0.1874
0.1202	1.0921	1400	0.1857
0.1229	1.2481	1600	0.1856
0.1028	1.4041	1800	0.1834
0.0872	1.5602	2000	0.1776
0.1098	1.7162	2200	0.1746
0.0964	1.8722	2400	0.1745
0.0720	2.0281	2600	0.1773
0.0899	2.1841	2800	0.1822
0.0750	2.3402	3000	0.1828
0.0674	2.4962	3200	0.1812
0.0721	2.6522	3400	0.1811
0.0703	2.8083	3600	0.1808
0.0687	2.9643	3800	0.1808

ThaoDuongDoingStuff/qwen-vi-rewriter-lora

qwen-vi-rewriter-lora